TERRAFORMING MARS in a Climate of Existential Risk Keith Mansfield

Journal of the British Interplanetary Society

VOLUME 71 NO.9 SEPTEMBER 2018 General Issue

PRESERVING GEOSTATIONARY ORBIT: the next steps Mark Hempsell, Roger Longstaff & Sebastiane Alexandra FUTURE RENDEZVOUS AND DOCKING MISSIONS enabled by low-cost but safety compliant Guidance Navigation and Control (GNC) architectures Steve Eckersley et al TERRAFORMING MARS in a climate of existential risk Keith Mansfield

www.bis-space.com

ISSN 0007-084X PUBLICATION DATE: 10 JANUARY 2019 Submitting papers International Advisory Board to JBIS

JBIS welcomes the submission of technical Rachel Armstrong, Newcastle University, UK papers for publication dealing with technical Peter Bainum, Howard University, USA reviews, research, technology and engineering in astronautics and related fields. Stephen Baxter, Science & Science Fiction Writer, UK James Benford, Microwave Sciences, California, USA Text should be: James Biggs, The University of Strathclyde, UK ■ As concise as the content allows – typically 5,000 to 6,000 words. Shorter papers (Technical Notes) Anu Bowman, Foundation for Enterprise Development, California, USA will also be considered; longer papers will only Gerald Cleaver, Baylor University, USA be considered in exceptional circumstances – for Charles Cockell, University of Edinburgh, UK example, in the case of a major subject review. Ian A. Crawford, Birkbeck College London, UK ■ Source references should be inserted in the text in square brackets – [1] – and then listed at the Adam Crowl, Icarus Interstellar, Australia end of the paper. Eric W. Davis, Institute for Advanced Studies at Austin, USA ■ Illustration references should be cited in Kathryn Denning, York University, Toronto, Canada numerical order in the text; those not cited in the Martyn Fogg, Probability Research Group, UK text risk omission. Raghavan Gopalaswami, Aerospace Researcher, India ■ Captions must be labelled with their Fig. number and should be as short as possible. Lamartine Guimarães, Institute for Advanced Studies, Brazil Mark Hempsell, Hempsell Astronautics Ltd, UK Illustrations should be: Takuto Ishimatsu, Massachusetts Institute of Technology, USA ■ Colour or mono, but should be as close to print Les Johnson, Marshall Space Flight Center, USA resolution (300 dpi) as possible. Poor-quality illustrations may compromise the acceptance of Terry Kammash, University of Michigan, USA paper for publication. Images embedded in Word Kelvin F. Long, Initiative for Interstellar Studies documents may be acceptable, but JBIS reserves Inoue Makoto, Institute of Astronomy & Astrophysics Academia Sinica, Taiwan the right to request separate higher-resolution Gregory L. Matloff, City University New York, USA image files from the author prior to publication. Koichi Mori, Nagoya University, Japan ■ Responsibility for copyright clearance of images rests entirely with the author. Richard Obousy, Richard Obousy Consulting LLC, USA Robert Parkinson, BIS, Aylesbury, UK Submission of papers George Schmidt, NASA John H Glenn Research Center, Ohio, USA ■ Papers for consideration should be sent by Paul Schuch, The SETI League Inc, USA email to [email protected] as both a Word document and as a Word PDF file (in order to Tabitha Smith, Bifrost, USA check for font anomalies), together with any Andreas Tziolas, Variance Dynamical Corporation, USA separate image files. Chris Welch, The International Space University, Strasbourg, France ■ If a paper is accepted for publication, the Friedwardt Winterberg, University of Nevada, Reno, USA author will be asked to sign a License to Publish form. This can be downloaded at www.bis- space.com/wp-content/uploads/2012/08/ WebsiteLicense.pdf. ■ Authors will receive a complimentary copy of the issue in which their paper appears. Editor Roger Longstaff Deputy Editor Duncan Law-Green Associate Editors Stephen Ashworth, We respectfully ask authors to adhere to these Keith Cooper, Stephen Gamble, Paul Gilster, Rob Swinney, Production MP3 Media guidelines. Failure to do so will result in the Promotion Gill Norman JBIS Office British Interplanetary Society, Arthur C. Clarke House, delay of acceptable papers for publication. 27-29 South Lambeth Road, London, SW8 1SZ, United Kingdom tel +44 (0)20 7735 3160 email [email protected] www.bis-space.com

Our full Guidelines for Authors can be downloaded DISTRIBUTION from www.bis-space.com JBIS is distributed worldwide by mail and may be received by annual subscription or purchase of single copies. It is available through membership of the British Interplanetary Society at much reduced rates. Subscription details for members, non-members and libraries are available from the above address.

JBIS is a publication that promotes the mission of the British Interplanetary Society. Opinions expressed in signed articles are those of the contributors and do not necessarily reflect the views of the Editor or the Council of the British Interplanetary Society. Security clearance, where necessary, is the responsibility of the author.

Published by the British Interplanetary Society. Registered Company No: 402498. Registered Charity No: 250556. Printed by Latimer Trend & Company Ltd, Estover Road, Plymouth, PL6 7PY, England. © 2019 British Interplanetary Society. No part of this magazine may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying or recording by any information storage or retrieval system without prior permission from the Publishers. contents VOLUME 71 NO.9 SEPTEMBER 2018

314 PRESERVING GEOSTATIONARY ORBIT: the next steps Mark Hempsell, Roger Longstaff and Sebastiane Alexandra

323 FUTURE RENDEZVOUS AND DOCKING MISSIONS enabled by low-cost but safety compliant Guidance Navigation and Control (GNC) architectures Steve Eckersley et al

348 TERRAFORMING MARS in a climate of existential risk Keith Mansfield

OUR MISSION STATEMENT The British Interplanetary Society promotes the exploration and use of space for the benefit of humanity, connecting people to create, educate and inspire, and advance knowledge in all aspects of astronautics.

JBIS Vol 71 No.9 September 2018 313 JBIS VOLUME 71 2018 PAGES 314–322

PRESERVING GEOSTATIONARY ORBIT: the Next Steps

MARK HEMPSELL1, ROGER LONGSTAFF2 and SEBASTIANE ALEXANDRA3 1Hempsell Astronautics Ltd., 10 Silver Birch Avenue, Stotfold, Herts, SG5 4AR, UK; 2Guest Associates (Europe) Ltd., 2 Alyson Villas, Vincent Close, Ilford, IG6 7FE, UK; 3Barrister at Law, c/o Guest Associates (Europe) Ltd. email [email protected]

A 2016 study produced an initial feasibility assessment of a system, called “Necropolis”, to collect uncontrolled satellites in geosynchronous orbit and re-locate them at a long-term storage facility, kept under permanent control. The study identified three areas that required attention in subsequent work. The first area was improving understanding the collision risk in geostationary orbit and the status of the major debris that creates the risk. New work is required that properly models the concentrating effect of the libration points, and also evaluates the change in risk due recent events, in particular the AMC-9 incident. The second area is the revision of the Necropolis technical concept to incorporate the new understanding of the urgency to improve debris control measures. The third area is the requirements for evaluation of the current space law relating to use of geostationary orbit that can enable proper regulation of the environment. This needs to address the requirement to report incidents and their causes, and to enable more precise determination of liability.

Keywords: Geostationary orbit, Debris control, Necropolis

1 INTRODUCTION orbit. The study identified several areas where urgent work was required to increase the understanding of the real risks, The problem of anthropogenic debris and the threat it poses to develop the systems to control geostationary debris and to to operational space systems has been a growing concern. In establish the necessary regulatory environment. lower Earth orbits we already have a Kessler syndrome situation – that is “the number of space-debris objects in some 2 THE NECROPOLIS STUDY orbital regions grows even if mitigation measures are applied” [1]. With the concern over the urgent matter of low Earth or- The premise of the Necropolis study [2] was that the practice bit, much less attention has been paid to the geostationary or- of relocating geostationary satellites to an unregulated “grave- bit, but the situation is complex and there is a higher risk of yard orbit” (around 300 km above geostationary altitude) a Kessler syndrome there as well unless active measures are would in the long term be unsustainable, as the debris density, taken to prevent it. and therefore the collision risk, increases and debris from a graveyard collision could reach geostationary orbit, putting The problem of the growing debris environment around ge- operational satellites there at risk. It was assumed this would ostationary orbit was the subject of a four months ‘quick look’ be a long term problem – many decades away – so the specific study conducted in 2016 which was led by Guest Associates subject of the study was a technology demonstration mission (Europe) Ltd., supported by Hempsell Astronautics Ltd., with that would verify the means to provide a strategy that would financial support from a grant awarded by the UK Space Agen- permanently control the debris in the geostationary environ- cy’s National Space Technology Programme [2, 3]. Although ment. The Necropolis demonstration flight could then lead to not part of the funded study team both Airbus Defence and either: Space, at Stevenage, and QinetiQ at Farnborough, provided technical advice and support on the system concept. • An expansion of the demonstration system, or, • repeat builds of the demonstration system, or, The system studied was called “Necropolis” (a graveyard • the development of more capable operational system in away from centres of population). It demonstrated a system ca- light of lessons learned. pable of collecting and safely storing satellites was possible, but it also uncovered several unexpected insights that questioned As it is not practical to deorbit satellites at such a high alti- the basic consensus regarding the situation in geostationary tude, the study looked at an alternative, specifically to establish a controlled facility that could easily be reached from geostationary and graveyard orbits. Here satellites that had reached This paper was presented at 68th International Astronautical the end of their useful life could be permanently attached and Congress (IAC), Adelaide, Australia, 25-29 September 2017. removed as a collision risk.

314 Vol 71 No.9 September 2018 JBIS PRESERVING GEOSTATIONARY ORBIT: the Next Steps

Fig.1 The Necropolis system in operation. The Hunter (with a recovered satellite) approaches the Terminus.

The Necropolis system (Fig. 1) consisted of two independent spacecraft launched together on an Ariane 5 or 6 (Fig. 2) and placed above both the geostationary and graveyard orbits. Once on station, the “Hunter” spacecraft separates and, using ion propulsion, rendezvous with drifting satellites, and returns them to the “Terminus” storage facility, where they are permanently stored.

The Hunter performs the function of a “Chaser”, a term established for a spacecraft that can rendezvous with and capture uncooperative, derelict spacecraft. It was found that the Mer- cury Transfer Module (MTM), developed to propel the Bepi- Colombo spacecraft to Mercury, would provide the propulsion capability required and provide a suitable structure and power supply system for the Hunter. Indeed, in the experience of the study team, the apparent suitability of the MTM for this role is unprecedented for a system developed for such a different mission. The study showed that MTM based Hunter has sufficient propellant to recover at least six satellites.

To achieve the capture of the target satellites the Hunter has a conceptual stinger capture mechanism mounted on a spin table in order to capture spin stabilized derelicts (Fig. 3). It was loosely based on a series of MacDonald, Dettwiler and Associates patents [4, 5, 6]. It has a very demanding requirements to capture various types of satellite – both spinners and uncontrolled 3 axis stabilised – with one designed capture

Fig.2 The Necropolis system in launch configuration packaged within an Ariane fairing with Hunter stacked on top of Terminus. Fig.3 The Hunter’s capture mechanism concept

JBIS Vol 71 No.9 September 2018 315 MARK HEMPSELL et al point while being reliable and reusable. This mechanism was highlighted as the key area of technology uncertainty in the Necropolis concept.

The Terminus Satellite provides the long term controlled storage facility for the captured satellites in an orbit well clear of the Geostationary and graveyard environments. The study report suggests a general orbit about 600 km above geostationary altitude. When the study results were presented at the 7th European Conference on Space Debris [7], several delegates, including Dr Donald Kessler, suggested the use of a super-synchronous Laplace orbit, where perturbing forces stabilise the orbital elements, as the best location for the Terminus – as suggested by Rosengren, Scheeres and McMahon [8].

The Terminus satellite study concept was an all-new design, as shown in Figure 4 in its deployed configuration with a height of 18 metres. Following a circularisation burn – using its own conventional bipropellant propulsion system – it would then release the Hunter and deploy the tower structure that contains the 12 satellite attachment locations. This is more than the Fig.4 Terminus satellite with tower deployed. nominal six satellites expected to be recovered by the system as launched, but allows for mission designs which enable the GEO is orders of magnitude less that the problem in LEO, recovery of more than six targets, some failures in the location therefore ADR (active debris removal) efforts should concen- points and additional Hunter missions trate on LEO” [10].

The satellites are attached to the Terminus tower using a vari- An example of the studies leading to this perception would ation of the Airbus D&S space harpoon (Fig. 5). The Hunter will be Frey and Lemmens [9], which concluded that the annual rendezvous with Terminus and then hold station presenting the probability of collision in LEO was estimated at 1.5x10-1, while satellite as target. The harpoon then fires; permanently ensnar- the probability of collision in GEO was 3.2x10-4. This result is ing the satellite. The Hunter releases the capture mechanism consistent with the current consensus view that there is a two and the satellite is then drawn into the net walls of the tower. orders of magnitude difference in the risk. However, this evaluation was carried out using ESA’s “MASTER” tool, which per- 3 THE DEBRIS RISK IN GEOSTATIONARY ORBIT forms a stochastic calculation based on a population database. It does not take account of the longitude, or right ascension, of 3.1 Analytical Studies the objects, which clearly influence the probability of collision, nor does it define the altitude bandwidth adopted for the anal- The initial assumption of the Necropolis study that the geo- ysis. Indeed, based a narrowly defined spacial ring density (4.5 stationary orbit debris problem was long term, and likely to km) the situation in a very narrow band around GEO is similar be centred on the graveyard orbit population, was following a to that in LEO, as shown in Fig. 6 [11]. This figure should not widely-held perception of the problem. Only two thirds of op- be taken as a precise numerical comparison of collision risk as erational geostationary objects successfully reach a graveyard complicating factors work in both directions. For example the orbit at the end of their life, around a quarter fail after attempt- "shell volume" at GEO is over 100 times greater than that for ing the manoeuvre and a tenth do not even try [9]. Thus the LEO. However the figure does illustrate the need to determinis- graveyard orbit strategy only slows the developing problem of tically and precisely capture real satellite locations to accurately debris creation in GEO it does not solve it, and it creates a sep- assess collision risks in orbits where satellites are concentrated arate problem in another location. for operational or orbit mechanics reasons.

However, the perception that the risk of collision in geosta- The modelling approach of MASTER, and similar stochas- tionary orbit is very low increasingly appears to be incorrect. tic tools, produce good statistical results in most cases. How- The UK Space Agency, in a communication with the authors, ever, in geostationary orbit there are high order perturbations summed the consensus view up as “the problem of debris in (which can be ignored at other altitudes) giving an important effect, as evidenced by the need for east/west and north/south station keeping for active satellites. Once this orbit control is lost the lack of north/south station keeping means a slow change in inclination which oscillates with an amplitude of 15 degrees and a period of 53 years. As a consequence the debris velocity relative to the geostationary ring can be up to 800 m/s.

Of more consequence is the loss of east/west station keeping – without this the object drifts towards one of the two libration points at 105 west and 75 east. The subsequent motion means that all drifting satellites pass through these libration points twice a year (and some satellites are permanently resident there) which causes a concentrating effect that is not cap- Fig.5 Airbus D&S Space Harpoon in Terminus tower mounting. tured in stochastic debris modelling.

316 Vol 71 No.9 September 2018 JBIS PRESERVING GEOSTATIONARY ORBIT: the Next Steps COURTESY OF AGI Fig.6 Orbital Ring Density.

The only analysis of collision probability in GEO (that the key metric in determining ADR requirements in GEO is the authors are aware of) which does fully account for these effects probability of debris-debris collisions. However the indications has been performed by the University of Colorado [12]. These are that stochastic debris modelling is not providing anything calculations estimate the probability of collision for active sat- like a realistic assessment of the collision risks, being some- ellites in various locations on the GEO arc and simulates “near thing like an order of magnitude optimistic. It is the contention miss events” over a 5 year period (see Fig. 7). The method of of the authors that this probability can only be evaluated using calculation is to simulate a 50 km radius torus around GEO fully deterministic calculations, and the authors are not aware and calculate intersections at various longitudes with derelict of any modelling work addressing the specific and unique risks satellites. It showed that the concentrating factors increase the of debris impacts in GEO, and as a consequence the policy and number of near misses by a factor of 7 at, or near to, the li- operational response to these risks could be seriously flawed. bration points. The analysis concludes with nominating the 10 most dangerous derelict objects, whose removal to a safe loca- 3.2 2017 Events tion would reduce the probability of collision in GEO by 50%. The concern over the geostationary environment must be in- The purpose of the research was to assist in better predic- creased by the series of events occurring in 2017. “A minimum tions of collision avoidance contributions to the orbit control of four aging geostationary satellites have unexpectedly mal- propellant budget of operational satellites. It was not exploring functioned this summer. In addition to Telkom-1 and AMC-9, the overall collision risk and so it did not assess debris to debris the 20-year-old EchoStar-3 failed in late July — right around collisions, and thus provides only an indication of the likely the same time another SES satellite, the 19-year-old NSS-806, errors in the stochastic modelling of debris growth and the im- lost roughly a third of its transponders to an unexplained pact of various active debris removal strategies. However the glitch” [13]. Three of these incidents (AMC-9, Echostar 3 and basic tools created by University of Colorado could be used to Telkom-1) have led to satellites losing orbit control and two consider these subjects. (AMC-9 and Telkom-1) have resulted in fragmentation and hence a significant increase in the debris population in GEO. Such modelling is urgently required, because if one assumes that all active satellites receive debris conjunction warnings and On 17th June 2017 SES, the operators of the AMC-9 satellite successfully manoeuvre in order to avoid physical contact, the stationed at 83W, were reported as having lost contact with the UNIVERSITY OF COLORADO

Fig.7 Near-miss events per day at 50 km occurring during 5 year forecasting period [y].

JBIS Vol 71 No.9 September 2018 317 MARK HEMPSELL et al spacecraft and requesting tracking data as they thought that its data to help establish causes of incidents orbit may have been perturbed by an energetic event [13]. The • no collision warning body with legal liabilities authors calculated from public domain orbit data [14] that the • no consistency in responses to collision warnings momentum vector of the spacecraft had changed by about 15 • no independent and open investigation of incident to es- m/s. So this was a comparatively high momentum event but tablish lessons to be learnt also a comparatively low energy event as the satellite was ultimately left with sufficient functionality for communications to If any measure of control is to be established in GEO there be re-established which rules out an explosion or high speed needs to be the legal and technical means to create reliable situ- collision with a small object (say a 0.5 kg meteorite). This leaves ational awareness of the status of satellites and their operations. an unintended thruster firing, which the operators would be aware of, or a low speed impact by a large body (speed in the 4 REVISED NECROPOLIS PROGRAMME order 10- 30 m/s and mass in the order 500-2000 kg). The problem with this conclusion is such an object would be large 4.1 Background enough to be tracked and a proximity warning given and this did not happen. In light of the revised opinion on the real collision risk in the geostationary environment the original study outline pro- If the initiating event were a low speed collision, the impact- gramme has been revised. It is proposed that a scout mission ing object may have got entangled with AMC-9. This would ex- and an early, single target chaser is added, as a quickly deploy- plain the apparent break up on the 29 July when the tumbling able short term action which provides a temporary means to object created several large pieces of debris, an event captured reduce the risk, while also acting as a precursor to the main by ExoAnalytic Solutions geostationary monitoring network Necropolis system, which is still seen as the long term solution. [13]. This could be the colliding body separating, which would explain why after this major “break-up” contact with AMC-9 4.2 “Scout” Mission was re-established and some measure of control regained. SES seem to have an explanation along these lines as an option as During the course of the study two issues arose that suggested they “cannot say where the debris comes from and can definite- the need for some form of in-orbit reconnaissance of the target ly not confirm that it comes from the spacecraft” [15]. satellites before undertaking a capture and move mission, in order to establish their status and condition. Although communication with the spacecraft was resumed on 1 July, as of late August it has not been re-orbited and re- The first issue is a lack of knowledge of the state of the target mains in a GEO altitude crossing orbit, with a perigee 145 km satellites after half a century in the space environment and dec- below GEO and an apogee 181 km above, with a slightly su- ades in an uncontrolled state. Key concerns are the spin behav- per-synchronous semi-major axis and therefore drifting slowly iour of targets, do spin stabilised satellites retain their spin or westward in relation to GEO. Two months after the anomaly de-spin over time, or even tumble? Do 3 axis satellites develop the satellite had drifted by about 12 degrees to the west, and its a spin or tumble? Another concern is the state of the thermal inclination had increased by 0.1 degrees (indicating expected blankets and other parts that will be impacted by the recovery luni-solar perturbation). process.

If it is not re-orbited this satellite is clearly a hazard to navi- In addition to the specific information required for the de- gation in the region, and as there have been reports of it shed- tailed design of the Necropolis system the mission would also ding debris its configuration remains unknown (even though provide valuable general information for the design of satellites it clearly retained some power and communications ability, as for the geostationary environment – in particular the alteration evidenced by the contact on 1st July). of thermal finishes over time.

On 25 August another spacecraft, Telkom-1, experienced a The second issue is the desirability of demonstrating rendez- sudden event that created a significant amount of debris and vous with non-functioning satellites at geostationary altitudes this was also captured by ExoAnalytic Solutions geostationary and proving the procedures for safe approach to target satel- monitoring network. The company suggested that preliminary lites. This mission would also confirm and refine the accuracy indications was that this was not due to an impact event [13] of the ground tracking of spacecraft in geosynchronous orbit. but also notes “at GEO there are a lot of untracked pieces of de- It was also thought to be a valuable opportunity to assess the bris that the air force does not publish” [16]. Regardless of the capability of the laser range finder and its ability to support the initiating cause the incident has clearly generated a significant capture process. amount of uncontrolled debris. There are three potential routes to meeting this objective; These 2017 events, considered in the context of the 2010 uncontrolled drift of Galaxy 15, highlight the problem of inad- i. make civilian use of existing military geostationary in- equately regulated operations in geostationary orbit. While all spection assets, created uncontrolled debris, none of the events can unequiv- ii. use geostationary satellite serving systems if they are de- ocally be linked to an impact event as the initiator, although ployed, or AMC-9 is highly suggestive of this. However this is due to a iii. undertake a dedicated mission with a dedicated space- lack of knowledge and reporting rather than a good under- craft standing of what happened. While there is no evidence of the overtly irresponsible use of GEO, all operators’ actions regard- The only known system that could be used for Option i is ing collision safety are voluntary and standards vary. For ex- the USAF’s Geosynchronous Space Situational Awareness ample there is; (GSSA) system [17]. The programme has launched four space- • no mandatory telemetry standard to provide “black box” craft, (two in 2014 and two replenishment satellites in 2016)

318 Vol 71 No.9 September 2018 JBIS PRESERVING GEOSTATIONARY ORBIT: the Next Steps into near geostationary orbits. Their purpose is to “collect space electric propulsion system could be included to rehearse the situational awareness data allowing for more accurate tracking rendezvous Hunter’s mission profile but it would not be neces- and characterization of man-made orbiting objects” and they sary for Scout to match orbits will all its targets as in many case have the capability to perform rendezvous and proximity oper- a flyby will be sufficient to get the required information. ations with geostationary objects of interest. However it is achieved it is felt important that the provision The payload and capabilities of the GSSA satellites are clas- of status and condition data from direct in-orbit observation is sified, although a precursor satellite SBSS (Space Based Space on the critical path for successful development of any interven- Surveillance) satellite launched in 2010 had a 30 cm telescope tion system, and thus a matter of urgency. on a two axis gimbal with a 2.4 megapixel image sensor. This was capable of detecting a 1 m object at 22,000 kilometres and 4.3 I-Hunter could make an average of 15,000 observations a day, intended to examine every spacecraft in geosynchronous orbit at least once Given the desirability of early intervention to remove the high- a day [18]. It is reasonable to assume that the GSSA satellites est risk systems it is suggested that an early chaser system, a have an equivalent or superior capability, and would be able to single target rendezvous, capture and removal spacecraft, called provide the sort of information required as input to a removal “I-Hunter” for Interim-Hunter, be deployed. This is foreseen as system design. However releasing useful images and data into a simpler chaser system which would be launched in clusters the public domain would reveal at least part of the GSSA sys- and each I-Hunter will only deal with one object. This differs tem’s classified capability and therefore this is not thought to from the Necropolis proposal and ESA ROGER study concepts represent a realistic option for a civilian disposal system. [23] which assumed the casher spacecraft would handle several satellites in succession. Once I-Hunter has attached to its target Option ii would rely on one of the several ventures to pro- object it would re-locate it to a temporary graveyard orbit and duce a Geostationary Satellite servicing system becoming op- maintain attitude control and some orbit control, maybe for erational. Examples are DARPA’s Robotic Servicing of Geo- more than a decade, until the Hunter arrives to collect it for synchronous Satellites (RSGS) programme [19], MDA’s Space delivery to the Terminus. Infrastructure Services (SIS) venture [20], Vivisat [21] and Effective Space [22]. To perform the servicing task the ability Because each I-Hunter spacecraft will only be dealing with to image and externally assess the spacecraft status will be a one object the capture system can be both simpler and also tai- necessary capability. So in addition to the servicing of active lored to match the target. This allows the exploitation of capture geostationary satellites, these commercial ventures could also systems that are currently under development (nets and har- add income by scouting removal targets. poons) rather than a new and complex reusable system, both reducing the development time and increasing the chances of a Option iii would be a fast track mission separate from the successful capture. It also means that it is feasible that the sys- main Necropolis development, but with the purpose to acquire tem can undertake its mission using only chemical propulsion. the required information during the earlier stages of the Ne- cropolis detailed design phase. It is currently envisaged as a Figure 8 show a concept outline for the I-Hunter spacecraft. small spacecraft with a payload of a couple of cameras and a It employs a conventional MMH/NTO chemical propulsion laser range finder. Using SBSS as a model it should be possible system with 350 kg of propellant contained in 4 tanks. The to keep the launch mass to around a tonne. It possible that an main propulsion is provided by 4 Airbus Safran 200N thrusters

Fig.8 I-Hunter approaching target satellite.

JBIS Vol 71 No.9 September 2018 319 MARK HEMPSELL et al

[24]. This thruster has a relatively poor specific impulse of 2650 Ns/kg which provides a mission velocity of around 2200 m/ sec. The specific impulse is actually worse than the assumed 4N attitude control thrusters from the same company, and an im- provement would have significant benefit to mission capability.

The concept design encapsulated a philosophy of simplicity, mostly to ensure maximum life expectancy during the holding phase. This evident in the use of body fixed arrays which can provide 500 Watts power when correctly orientated to the Sun but once attached to satellite gives a continuous power of 200 Watts assuming the blockage from the target satellite gives effectively an 8 hour eclipse each day, during which two lithium ion batteries with a combined capacity of 3600 watt hours provide power. These batteries were also expected to cover peak demand during the rendezvous and capture processes.

The structure has a centre cylinder, with a Universal Space Interface System (USIS) connection at either end enabling them to be stacked during launch, for example up to 11 on an Ariane 5 or 6 launch vehicle as shown in Figure 9. The USIS concept [25] was also widely used in the original Necropolis system as a multi-purpose connection, minimising mass and complexity while maximising the utilisation of the few surfaces suitable for system to system connection. On the I-Hunter it is the launch system connection, the passive docking port for connection to the Hunter and the berthing port for final permanent connection for the Terminus.

The capture mechanism, which could be different for different targets, is placed in the centre of this cylinder. In the concept artwork it is the Airbus D&S space harpoon. The propulsion and other service equipment is mounted in the external box structure.

The mission sequence for the I-Hunter is, after separating from the stack in GTO, to circularise into a near geostationary orbit optimised to reach its target satellite. It then rendezvous with and captures its target, bringing it under control. Each target will then be moved to best place for subsequent capture by Fig.9 I-Hunter spacecraft in Ariane fairing the Hunter, and held there until it arrives (that maybe decade or so later). The main Hunter then docks with the rear of the USIS on the I-Hunter, take control of the stack and take it to The Hunter no longer needs a capture mechanism as that the Terminus. An RMS at the Terminus captures the I-Hunter is now on the I-Hunter. The Hunter uses the docking USIS on using the USS-RMS grapple point on the north face. The Hunt- the upper face and so now only needs to perform a controlled er separates to collect another satellite and the Terminus berths docking. The removal of the Hunter’s complex reusable cap- the I-Hunter and its target satellite permanently on its tower. ture mechanism saves 100 kg and considerable development expense and risk. 4.4 Revisions to Necropolis Concept The I-Hunter can also be fitted with manipulator grapple The proposal for a system like the I-Hunter is to remove key points and hard attachment points, so the harpoon attachment risks urgently, and a response to the reassessment of the con- that was assumed by the study to be the means by which the sensus perception of the real debris problem in geostationary derelict spacecraft would be attached to the Terminus can be orbit. It adds a new space system development and increases replaced by an ISS type manipulator which can take the object the launches per debris item by 50%. However it means the full to a more certain, permanent attachment system. Necropolis could be delayed. However; if addressing the maintenance of the safety of geostationary environment is treated as These strategies and concept designs are not proposed as -fi an annual overhead rather than a one off project; then this new nal solutions rather they are early feasibility concepts to scope system would not alter the required budget corridor. the scale of the potential solution and initiate discussion. The systems that are eventually deployed must be the subject of In this context it is important that the short term response to an extensive, system-level study, and the concepts described the issue (I-Hunter) does not create problems for the implemen- here are a “first look” to verify that a technical solution may be tation of the longer term solution that remains a permanent con- found that could mitigate both the short term, and longer term trolled storage facility (Hunter and Terminus). In the proposed threat to GEO. Other solutions are possible, for example, an al- concept design this new strategy simplifies and de-risks both the ternative concept could be to launch the single capture I-Hunt- Hunter and Terminus spacecraft as originally proposed. er spacecraft on a satellite “bus” that circularises its orbit close

320 Vol 71 No.9 September 2018 JBIS PRESERVING GEOSTATIONARY ORBIT: the Next Steps to GEO and manoeuvres the individual I-Hunter spacecraft avoidance measures, and be legally bound to do so. This would to their individual targets. This would have the advantage of ensure greater transparency and is crucial to keeping the GEO deleting the requirement for the significant mass of propellant clear of fragments and the creation of a space debris belt. required by each I-Hunter spacecraft in order to circularise its orbit from a GTO, and hence making it a simpler and cheaper This legal liability must include a responsibility for incidents system. This could be thought of as analogous to the operation such as those experienced by Galaxy 15, AMC-9 and Telkom-1 of multiple, independent warhead ICBMs. to be independently investigated and, so far as possible, causes established and publically reported, so that lessons can be learnt and recommendations for improved orbit safety made. 5 LEGAL AND REGULATORY ISSUES The authors recommend the subject of existing debris re- During the Necropolis study it became clear the regulatory removal (ADR) in the GEO region be seriously addressed as gime regarding the creation and control of space debris was soon as possible. When quantitative, deterministic analysis has inadequate, both generally and for geostationary orbit in par- established the magnitude of the threat to the GEO environ- ticular. Lessons should be learnt from the failure to prevent a ment an internationally agreed programme should be insti- Kessler syndrome in low Earth orbit in order to prevent a re- tuted to mitigate the threat, starting with the most dangerous peat situation occurring in geostationary orbit, where a debris objects. The owners of derelict satellites that pose a threat to the field will remain indefinitely as there is no air drag in GEO. environment must share the responsibility, and cost, of ADR measures, on a pro-rata basis. Space debris was not specifically mentioned in the 1967 Outer Space treaty [26] because it was not deemed sufficient- What is often missed when considering liability in such ly important at the time. The authors however aver, there now cases, is that the liability is not a constant, fixed in time. As is a dire need for a regulatory regime that mandates incident time progresses new technology and infrastructure capabilities enquiries and releases data to ensure that satellite operators create more options for debris control and thus change what are able to circumvent and take orbit avoidance manoeuvres, could be judged as reasonable efforts to deal with the risk. Thus avoiding collisions within the GEO environment, and mini- satellite owners, which under the UN Outer Space treaty means mizing the probability of potentially catastrophic debris show- ultimately nations, need to actively pursue methods that can ers, which could render the orbit unusable indefinitely. reduce the risk and where practical enact them. The internationally agreed programme outlined here would be a way to In the authors’ opinion there is an existing duty on States to ensure nations continue to meet their liability obligations as protect the orbit environment. Current space mitigation guide- they alter. lines, which indicate launch vehicles and payloads should either be de-orbited, or their orbits relocated to a “graveyard orbit”, do not assist us in respect of many satellites already in orbit 6 CONCLUSIONS without that capacity built into their systems, or have ceased to obey commands while resident in their operational orbit. As The current practice of relocating end of life satellites is not identifiable objects they remain under the responsibility of the sustainable, as a third of objects fail to reach a graveyard orbit state of registry and this should be the case until re-entry or it only slows the rate of debris creation, and as more satellites positioning to a graveyard orbit. However, the difficulty aris- are placed in graveyard orbits and are then uncontrolled in the es when a collision or explosion occurs, or a meteor strikes a longer term the collision risks in the graveyard orbit will grow space object. What was once an identifiable object now dis- until they reach unacceptable levels. Collisions in these grave- perses and forms debris. These fragments can then collide with yard orbits will create dangerous debris that intersects GEO. other debris and a cascade begins creating a belt of debris that A new approach to end of life disposal will be required if this is dangerous to other spacecraft. It then becomes exceptionally unique environment is to be preserved in perpetuity. difficult to determine who is then responsible for these fragments, or indeed liable for the damage they cause. The Necropolis study and subsequent work have concluded the collision risks in geostationary orbit, while not well under- Current liability in Art III requires fault to be established, stood, are clearly significantly worse than the consensus view. which in the above circumstances would be difficult, or impos- Because the conventional modelling of debris environment sible, to do. does not adequately address the unique behaviour of objects in geostationary orbit we cannot with any confidence know the Accordingly, and in order to circumvent such a cascade from collision risks and hence the urgency of implementing any new taking place, a nominated space situational awareness organi- strategy. A deterministic calculation of debris/debris collisions zation should be mandated to monitor satellites in the GEO is urgently required in order to quantify the threat to the GEO region and publically release conjunction warnings to ensure environment. any perilous satellite/object does not create harmful interference with other satellites. As we know, under the Outer Space The events of 2017 are suggestive of the lack of knowledge Treaty 1967, there is a legal duty to use space for the benefit of we have of satellites operating in GEO. Not only do we not have all mankind, having regard to the interest of other states, the a good grasp of the real risks, but we have no knowledge of Treaty specifically mentions the avoidance of ‘harmful inter- the actual current debris environment or the status of derelict ference & contamination’ with the activities of others (Art IX). satellites and rocket stages that would be the target of an active Signatories to the Treaty agree to notify the UN, the scientif- debris removal programme. Thus we need systems to estab- ic community and public of the nature, conduct, location and lish this missing information to reliably determine an effective results of their space activities. This is clearly not happening. debris control methodology. This could be gained from the It is therefore necessary to mandate the release of conjunction release of data from existing, military assets, or by launching warnings so that operators of active satellites can take collision bespoke missions to survey the environment.

JBIS Vol 71 No.9 September 2018 321 MARK HEMPSELL et al

More technology development and system concept evalua- body should also ensure the publication of investigations into tion work (such as the Necropolis study) is needed that focuses incidents and anomalies such as recently befell AMC-9 and on the geostationary problem, as solutions that work in LEO Telkom-1, so that lessons can be learnt and implemented by all are not directly applicable to GEO. This is important as any de- satellite operators. cisions of appropriate responses to the real risks must be based on a firm technology base which we do not currently have. This Once the modelling, scouting, legal and technology assess- level of technical understand will take many years to acquire ments are completed a programme of ADR in the GEO should and so cannot be left until after the new modelling and scouting be internationally agreed and threat mitigation measures, for activities are complete, but must be done in parallel with them. existing derelict satellites and debris in this region, be under- taken before the environment is damaged forever. The UN must look carefully at a new regulatory structure to regulate the commercial activities in space and in particu- The conclusions of the Necropolis study are, by necessity, lar focus upon the avoidance of harmful interference with the very early and provisional. But they are also very strong. They activities of others. An internationally recognized Space Situa- suggest that the only practical long term solution is the active tional Awareness organization should be instituted to publical- relocation of debris objects by chaser spacecraft, combined with ly issue conjunction alerts to operators of active satellites, and end of life manoeuvres of active satellites, to get all non-active operators should publically announce the measures that they objects relocated to a permanently controlled storage facility at take (and maybe even mandate the owners of active satellites to an inherently safe location, such as a super-synchronous Lap- act on conjunction warnings, as currently some do not). This lace orbit.

REFERENCES 1. H. Krag. “Guest Editorial – 7th European Conference on Space Debris”, (accessed 11 August 2017) Journal of the British Interplanetary Society, 70, pp 43, 2017 14. N2YO.com http://www.n2yo.com/satellite/?s=27820#results (accessed 2. R. Longstaff, & M. Hempsell,Sustainable Disposal of End of Life GEO 11 August 2017) Satellites. National Space Technology Programme report to the UK 15. C. Henry, “SES trying to retire AMC-9, uncertain on debris origin,” Space Agency, 2016. (Downloadable from www.hempsellastro.com) Space News Website, August 4 2017. http://spacenews.com/ses-trying- 3. R. Longstaff, M. Hempsell, and S. Alexandra “A Study into the to-retire-amc-9-uncertain-on-debris-origin/ (accessed 11 August 2017) Sustainable Disposal of End of Life GEO Satellites”, Journal of the British 16. E. Beger, “It looks like yet another satellite is breaking apart at GEO,” Interplanetary Society, 69, pp 429-438, 2016 Ars Technica website, 30 August, 2017, https://arstechnica.com/ 4. US 9399295 B2 - Spacecraft capture mechanism (using Marman Clamp science/2017/08/it-looks-like-yet-another-satellite-is-breaking-apart-at- ring), 2013, Paul Roberts et al, MacDonald, Dettwiler and Associates geo/ (accessed 11 August 2017) 5. US 20130249229 A1 Spacecraft capture mechanism (using Marman 17. http://www.afspc.af.mil/About-Us/FactSheets/Article/ 730802/ Clamp ring), 2103 – Paul Roberts et al, MacDonald, Dettwiler and geosynchronous-space-situational-awareness-program-gssap/ (accessed Associates 11 August 2017) 6. US 6969030 B1 Spacecraft docking mechanism (using the Apogee 18. Space Based Space Surveillance, Ball Aerospace Fact sheet D1910, 2016 Motor), 2005, Howard Martin Jones et al, MacDonald, Dettwiler and 19. G. Roesler, “Robotic Servicing of Geosynchronous Satellites (RSGS)” Associates DARPA Website, www.darpa.mil/program/robotic-servicing-of-geo 7. R. Longstaff, M. Hempsell, “A Mission to Demonstrate the Preservation synchronous-satellites (accessed 11 August 2017) of the Geostationary Orbit”. 7th European Conference on Space Debris, 20. MDA announces On-Orbit Satellite Servicing business formation and Darmstadt, April 2017 contract awards for spacecraft and first life extension customer, http:// 8. A.J. Rosengren, D.J. Scheeres and J.W. McMahon, “The classical Laplace mdacorporation.com/news/pr/ pr2017062803.html. (accessed 11 plane as a stable disposal orbit for geostationary satellites,” Advances in August 2017) Space Research 53. pp.1219-1228, 2014 21. http://www.vivisat.com (accessed 11 August 2017) 9. S. Frey and S. Lemmens, “Status of the Space Debris Environment: 22. https://www.effective-space.com/ (accessed 11 August 2017) Current Level of Adherence to the Space Debris Mitigation.” Journal of the British Interplanetary Society, 70, pp118-124, 2017 23. Robotic Geostationary Orbit Remover (ROGER), ESA, 2002, http:// robotics.estec.esa.int/Xcel_export/TEC/Robotics/SEMTWLKKKSE_2. 10. Email to authors from the UK Space Agency, dated 9 March 2017. html (accessed 11 August 2017) 11. http://www.space-data.org/sda/wp-content/uploads/ 24. Chemical Bi-propellant Family, Airbus Safran brochure, 2016 downloads/2012/03/20120312_SDA_Users_Mtg_4_General_Session. pdf 25. M Hempsell. “Creating a Universal Space Interface Standard”, Presented at the 66th International Astronautical Congress, Jerusalem, October 12. P.V. Anderson, H. Schaub, “Methodology for Characterizing High-risk 2015, Paper No. IAC-15.D3.2.6, later published in the Journal of the Orbital Debris in the Geostationary Orbit Regime,” Advances in Space British Interplanetary Society, 69, pp 163-174, 2016 Research 57 (2016) 604–619 26. Treaty on Principles Governing the Activities of States in the Exploration 13. C. Henry, “ExoAnalytic Video Shows Telkom-1 Satellite Erupting and Use of Outer Space, including the Moon and Other Celestial Bodies. Debris” Space.com website, 3 September, 2017, https://www.space. Opened for signature 10 October 1967 com/38017-exoanalytic-video-telkom-1-satellite-erupting-debris.html

Received 5 June 2018 Approved 12 October 2018

322 Vol 71 No.9 September 2018 JBIS JBIS VOLUME 71 2018 PAGES 323-347

FUTURE RENDEZVOUS AND DOCKING MISSIONS enabled by low-cost but safety compliant Guidance Navigation and Control (GNC) architectures

STEVE ECKERSLEY1 (corresponding author), CHRIS SAUNDERS1, DAN LOBB1, GAVIN JOHNSTON1, TIM BAUD1, MARTIN SWEETING1, CRAIG I. UNDERWOOD2, CHRISTOPHER P. BRIDGES2, RUNQI CHEN2 1Surrey Satellite Technology Ltd. (SSTL), Tycho House, Guildford, Surrey, GU2 7YE; 2Surrey Space Centre (SSC), University of Surrey, Guildford GU2 7XH, United Kingdom email [email protected]

Proximity flight systems for rendezvous-and-docking are traditionally the domain of large, costly institutional manned missions, which require extremely robust and expensive Guidance Navigation and Control (GNC) solutions. By developing a low-cost and safety compliant GNC architecture and design methodology, low cost GNC solutions needed for future missions with proximity flight phases will have reduced development risk, and more rapid development schedules. This will enable a plethora of on-orbit services to be realised using low cost satellite technologies, and lower the cost of the services to a point where they can be offered to commercial as well as institutional entities and thereby dramatically grow the market for on-orbit construction, in-orbit servicing and active debris removal. The forthcoming AAReST mission, will demonstrate some key aspects of low cost close proximity “co-operative” rendezvous and docking. However this is only a very small scale academic mission demonstration using cubesat technology, and is limited to very close range demonstrations. A study has examined the industrialisation of this existing research, culminating in a representative model that can be used to develop low-cost GNC solutions for many different mission applications that involve proximity activities, such as formation flying, and rendezvous and docking. The final results show that such a GNC model and mission demonstrator is feasible, and in line with anticipated UK regulatory constraints that may apply to the mission.

Keywords: Rendezvous and Docking, GNC, Mission Analysis

1 INTRODUCTION/MOTIVATION space telescopes) – Repair of satellites that are damaged or failed Proximity flight systems for rendezvous-and-docking, have – Re-fuelling of satellites traditionally been the domain of large, costly institutional – Movement of satellites to new orbital locations (i.e. a manned missions which require extremely robust and expen- “space tug”) sive GNC solutions. – Capture and disposal of space debris • Involving close proximity orbital operations without phys- However, we are now entering into a new and exciting era of ical contact: space exploitation, with a significant number of new mission – Inspection missions to observe and document the applications on the horizon (and in some cases already being physical status of other satellites (e.g. for insurance or investigated) which will require close proximity rendezvous intelligence purposes) and docking and/or formation flying, to enable the creation of – Distributed science or Earth observation missions new space services, and the generation of new commercial and (e.g. bi-static or multi-static radar) institutional markets on-orbit. These missions in turn will require lower cost GNC approaches, in order to be commercially All of these applications require the ability for a spacecraft to competitive, whilst still being safety compliant. safely operate in close proximity to other assets in space, something that is not normally part of most current space missions. The potential to bring two or more spacecraft into close Here ‘close proximity’ is taken to mean down to separation dis- proximity in a safe manner, has a number of future applications: tances of metres or closer, bearing in mind that a docking of • Involving docking or controlled physical contact of space- two satellites is actually a controlled and managed “collision” craft: (i.e. down to zero separation distance with a controlled “colli- – Assembly of structures in space that are too large or sion” velocity). massive to launch as a homogeneous structure (e.g. An important differentiator between types of mission is whether the ‘target object’ (i.e. the object being observed, cap- This paper was presented at the 15th Reinventing Space Conference, tured, or docked onto) is cooperative or uncooperative. In a Glasgow, 24-26 October 2017. cooperative scenario the target is under control and is opera-

JBIS Vol 71 No.9 September 2018 323 STEVE ECKERSLEY et al tional. It can therefore itself be maneuvered if required, and can adopt different attitudes if needed (e.g. to align a docking port with the approaching satellite). In the uncooperative scenario, the target is uncontrolled and may have any arbitrary attitude. The latter has the potential to be a much harder situation in terms of RDV as the target cannot assist in the close proximity operations. On the other hand co-operative scenarios require both spacecraft to work together, and in the case of large telescopes will eventually require multiple elements to coalesce. Additionally the dynamics of the Hub would be constantly changing as it grows in size. Fig.1 The AAReST Industrialised Mission Concept (left), leading For extremely large orbital structures such as sparse aper- to larger in-orbit construction missions (right), which require ture telescopes, next-generation communication antennas and detailed GNC modelling and design solutions. space tourism assets, In-orbit construction is considered a lower cost method than launching carefully stowed and elaborate- ly deployed monolithic structures, due to the reduced level of of 25m or greater), such as the GOAT [2] (Giant Orbiting As- structural analysis necessary and the relaxed requirements on tronomical Telescope) concept – see also Fig. 2. the materials used for the structure. In brief, it is much easier to construct a large structure in space when one does not have As such the information shown in this paper concentrates to also consider how to make it survive launch in one piece and on cases where the attitude and position of the target object fit within a launch vehicle fairing. It is for this reason that SSTL can be controlled. However the GNC model is intended to be and SSC consider in-orbit construction the more appropriate generic and can be utilized for other “non-co-operative” ren- future technical route for large in-orbit structures than mono- dezvous and docking missions. lithic deployable structures, and the techniques developed and demonstrated by a GNC validation simulator will go a long way This paper provides a concise summary of the findings and to achieve that ultimate end goal. final results from the study, particularly the mission analysis, sensors, key mission and spacecraft systems aspects, and GNC Indeed SSC are already actively involved the forthcoming simulation and modelling. AAReST (Fig.1) mission, which will demonstrate some key aspects of low cost in-orbit assembly (including close proximity 2 STUDY OVERVIEW rendezvous and docking) and reconfiguration of a space telescope based on multiple mirror elements [1]. The purpose of this UK NSTP-2 project, in the context of long term SSTL roadmaps, was to jump-start the industrialisation However this is only a very small scale academic mission of existing research - building on the AAReST mission where demonstration using three cubesats (a “Fixed Core NanoSat” appropriate, culminating in a representative model that can be plus 2 separable “MirrorSats”), and is only limited to very close used to develop GNC solutions for many different mission ap- ranges (the spacecraft are initially joined together). plications that involve formation flying, rendezvous and cooperative or uncooperative docking – known also as “proximity For this study, a cooperative two-spacecraft rendezvous and activities”. The study was led by SSTL, with SSC providing tech- docking mission demonstrator using microsatellites (an active nical support on proximity sensor development. The study was Chaser and a passive Target), is assumed as the reference mis- co-funded by the UK Space Agency (UKSA). sion, as this will be a natural follow-on to the AAReST telescope scenario, and a landmark demonstrator mission for larg- The main objectives and scope of this project are the fol- er multi-spacecraft demonstrations and ultimately even larger lowing: mission concepts such as persistent surveillance from GEO and • Definition of a reference mission design (based on a sce- very large astronomical telescopes (with large primary mirrors nario that SSTL considers credible as a realistic scenario

Fig.2 Conceptual 25m primary diameter modular GOAT telescope2.

324 Vol 71 No.9 September 2018 JBIS FUTURE RENDEZVOUS AND DOCKING MISSIONS enabled by low-cost but safety compliant Guidance Navigation and Control (GNC) architectures

for future RDV&D missions) and mission/system GNC TABLE 1 Longer term reference mission – EO telescope requirements. demonstrator in LEO • Develop a GNC architectural design for low cost missions applications that involve close proximity formation fly- Parameter Characteristics ing, RDV&D - i.e. “proximity activities” • Develop a low cost sensor suite suitable for use on prox- Large optical telescope demonstrator in LEO for Earth imity missions Mission Application • Consider possible regulatory constraints that may apply Observation in the visible wavelength range. to the mission Eight spacecraft (~150kg each): The early phases of the study involved the definition of a • 6 Primary Mirror Segment Spacecraft preliminary reference mission and requirements, a review of Number of regulatory aspects, and a trade-off of sensors and GNC archi- spacecraft • 1 Primary Detector Spacecraft tectures. The high level “User Requirements” are defined as follows: • 1 Secondary Mirror Spacecraft • Define a reference mission (co-operative or non-co-operative) • Instrument: Optical telescope (Visible Range) • Be compatible with a low-cost launcher • Perform safe rendezvous and docking and generate no • Ground sample distance: 0.15m, PAN debris Payload • Overall Primary mirror diameter: 1.75m • Comply with the UKSA legal, licensing and regulatory framework in order to be eligible for a UK space licence • Primary to Secondary separation: 3.25m (inc. UK Space debris mitigation standards) • Assume a launch for the reference mission before 2025 • LEO Altitude: 500 km, Sun-Synchronous • Develop a flexible GNC design which can (if possible) ap- Orbit ply to both co-operative and non-co-operative missions • 10:30am Ascending Node (note that this should not drive the design if there is a clear much lower cost option for the selected mission) RDV Co-operative Mission • Implement "low cost" GNC equipment for the RDV&D, where possible

These User Requirements were then used to derive a more detailed set of mission and GNC requirements. Some examples of key driving requirements are: • To be one failure tolerant to a single failure and be able to continue the mission, and two-failure tolerant to avoiding a catastrophic situation (i.e. a collision between the constituent spacecraft). This is common for RDV&D missions but not normally applicable for normal missions. • To start homing from 10km range and perform the relative navigation from at least 1km Fig.3 Longer term reference mission EO telescope demonstrator in • To perform proximity operations (inc. pose estimation) LEO using 8 modular spacecraft. from 100m and the close approach from at least 10m • To complete docking for all spacecraft in 6 months In parallel a thorough review of regulatory aspects (par- Two reference missions have been defined as part of the ticularly UK as the assumption is that the design would need study rather than one, in a more logical sequential two-step to be UK-licensing compliant) was carried out. A dedicated approach: regulatory aspects meeting with UKSA’s licensing department • A longer term “co-operative” Earth Observation (EO) was held, and they agreed that at least one simpler precursor telescope demonstrator in LEO” using 8 modular space- demonstrator mission would likely be required to de-risk the craft – see Table 1 and Fig. 3). This is intended to define RDV&D. They also provided a draft list of licensing questions the longer term direction of the application and develop- that would be applicable for future missions involving rendez- ments. vous and docking, and it was clear that mission safety and ro- • A shorter term lower complexity but safety compliant bustness will be a major driver of the spacecraft design. “co-operative” two-spacecraft RDV&D precursor mission demonstrator using microsatellites (an active Chas- The second phase of the study was focused on the following er and a passive Target) (see Table 2 and Fig. 4). This is using the baseline reference mission: the baseline reference mission and the focus of the study • Detailed modelling of the GNC architecture and the ref- (and this paper). erence mission scenario to define the system performance and behaviour (with respect to the requirements). Three Note that, whilst the focus in on “co-operative” missions, the main topics have been analysed: GNC model is intended to be generic and can be utilized for – Mission Analysis other “non-co-operative” rendezvous and docking missions. – Detailed GNC simulation and modelling – Systems related to the GNC and CONOPS This was then followed by a trade-off of the rendezvous sen- • Testing and bread-boarding of proximity sensors. sors and GNC/propulsion systems up until the mid-term review. • Development plans.

JBIS Vol 71 No.9 September 2018 325 STEVE ECKERSLEY et al

3 BASELINE REFERENCE MISSION(S) TABLE 2 Baseline Reference mission – Co-operative RDV&D Precursor Demonstration Mission Table 2 summarises the updated mission parameters for the precursor demonstration mission, with two spacecraft (i.e. Parameter Characteristics a target and chaser). This is envisaged to be a maximum six Mission Co-operative RDV&D Precursor Demonstration Mission month mission in line with the requirement to carry out RD- Application V&D within 6 months. Mission Lifetime Maximum 6 months As the RDV&D precursor demonstrator is primarily aimed Two spacecraft: Number of at demonstrating co-operative RDV&D for a modular optical • 1 Active “Chaser” Spacecraft spacecraft telescope demonstrator mission, the broad platform concept • 1 Passive “Target” Spacecraft for the telescope demonstrator mission is retained which uses Chaser: Relative GPS, Proximity Camera, COTS LIDAR a hexagonal structure. However this is largely arbitrary, so long Rendezvous as the diameter is at least the same as for Optical telescope sensors Target: Relative GPS, Glyph/LED panel demonstrator. In fact a pure RDV&D demonstration does not even need to be hexagonal and it may be more cost effective to Other sensors Video Camera on each spacecraft consider more standard SSTL platforms for the first mission. Nevertheless it provides some basic parameters for mission Docking system Variant of AAReST’s Electromagnetic System analysis and GNC purposes (i.e. drag, control etc). Orbit LEO Altitude: 500 km It is also sensible to try to make the Target very similar to the chaser in terms of hardware, and thus lower the cost. It is Launch PSLV with an arbitrary ejection scenario generally cheaper to build two identical or very similar spacecraft than it is to build two different spacecraft. Thus the Target Hexagonal Cylinders: Same as the Optical telescope Spacecraft size is assumed to be the same design as the Chaser, except where demonstrator there are payload equipment differences.

150kg assumed for mission analysis and design 4 MISSION ANALYSIS Spacecraft mass estimates 4.1 Introduction • Xenon (Warm and Cold Gas) Propulsion SSTL have carried out substantial mission analyses throughout • Thrust: 100mN (2x 50mN thrusters) the study. This has been iterated as the study has evolved. For Inter-spacecraft this study, the scenario is assumed to be a cooperative RDV (as S-Band 2-way Intersatellite Link (ISL) communications per future modular telescopes such as in RD2), and as such the mission analysis concentrated on cases where the attitude and position of the target object can be controlled. The focus of this paper also reflects this. However, a range of other potential sce- 4.2 Background narios were also analysed for broader use and are also briefly covered here. The orbit of the target object is assumed to be in Low Earth Orbit (LEO) near circular orbit. For the analysis shown herein an eccentricity of e~0.001 is used for most of the analysis as this is typical of most LEO missions. For a satellite at ~500 km altitude this corresponds to a difference between the perigee and apogee radius of ~13.5 km.

When two space objects share a similar near-circular orbit, and are in relatively close proximity to each other, it is convenient to describe the relative motion of the objects with respect to another, in a coordinate frame that is centred on – and moves with – one of the two objects. A natural reference frame is one centred on the Target vehicle, which acts as the coordinate origin.

As long as the orbit is near circular, and the distance between the two objects is small compared to the orbital radius, then a linearized set of equations can be used to describe the motion of the Chaser with respect to the Target. These equations are commonly known as the “Clohessy Wiltshire” (CW) equations. As described in [3], the CW equations can be used to derive useful – and simple – equations that can be used to describe the motion of the Target with respect to the Chaser when an impulse or force is applied to the Chaser spacecraft. These can then be used as inputs into more detailed numerical modelling. Fig.4 Baseline Reference mission – Co-operative RDV&D In general, with the coordinate system centred on the Tar- Precursor Demonstration Mission. get, a series of ‘bars’ are defined which are centred on the Target

326 Vol 71 No.9 September 2018 JBIS FUTURE RENDEZVOUS AND DOCKING MISSIONS enabled by low-cost but safety compliant Guidance Navigation and Control (GNC) architectures

4.3 Indicative Mission Scenarios

As part of the study, a number of different scenarios, with differing levels of mission complexity, have been analysed and examined. All assume a two-satellite demonstration mission, with both satellites launched on the same launch vehicle. The scenarios cover: • The initial launch and separation of the two satellites • Different approach, rendezvous, proximity operations, and docking options. • Un-docking of the two satellites Fig.5 Definition of V-Bar and R-Bar (an additional H-bar • Possible collision avoidance scenarios completes the coordinate axes, and points out of the plane of the paper). In the example the Chaser has a relative position of –x on The focus of this paper is on the mission analysis for the ref- the Target V-bar, and +y on the Target R bar. erence mission (co-operative RDV) which utilizes a “straight V-bar” approach.

(see Fig. 5). This system rotates with the spacecraft along its Both the Target and Chaser are initially assumed to have orbit at a rate equal to 1/P where P is the orbit period. identical physical characteristics and orbit: • Semi-major axis: 6878 km (altitude ~500 km) The baseline GNC design envisages firing pairs of Xenon • Eccentricity: 0.001 thrusters for ΔV translation, which would give a total thrust to • Mass: 150 kg the spacecraft of 100 mN. For the size and mass of spacecraft • Maximum cross-sectional area: 0.9m2 envisaged in this study (~150 kg) this brings up an important • Propulsion: Xenon aspect relating to how “impulsive” on-orbit manoeuvres can be – Isp: 48 sec (Resistojet), 30s (Cold Gas) with this kind of system. Typically most GNC and RDV litera- – Thrust: 0.1 N ture assumes impulsive manoeuvres as this significantly simplifies the analysis. In a true impulsive manoeuvre the spacecraft’s All orbits are propagated numerically including the effects velocity is changed instantaneously whilst its position vector of the Earth’s J2 oblateness term. Atmospheric drag perturba- remains the same. In practice of course any ΔV is generated tions are also modelled. by a finite thruster force, T providing an acceleration, a, to the spacecraft (mass M) which is integrated over some time Δt. 4.3.1 Initial Launch and Early Operations

A manoeuvre is typically defined as ‘impulsive’ if Δt is much For the purposes of this study, the Indian PSLV launcher is as- less than the orbit period of the spacecraft, such that in this sumed. This launcher has the ability to carry multiple vehicles time the position vector of the spacecraft does not change ap- on a variety of different carrying structures and adapters. preciably. Taking 1% of the orbit period as a reasonable approximation of ‘much less than the orbit period’, for a LEO satellite A number of different possibilities exist for how the satellites with ~6000 s period, burn times of ≤ 60 sec are required to could actually be deployed. One example is given here as an keep with the impulsive approximation. indicative scenario. In this case, the two satellites are assumed to be ejected simultaneously from an upper stage both with dif- If short impulsive manoeuvres of no more than 60 sec are fering ejection velocity magnitudes and directions. desired, then the maximum ΔV available is 0.04 m/s (using radially directed burns), which would correspond to a displace- As shown in the diagram in Fig. 6, the two satellites are as- ment of only ~144 m on the V-bar. sumed to be ejected from the upper stage at an angle to the velocity vector of the stage. The ejection velocity relative to the up- As shown later in this paper, ΔV’s of 0.04 m/s or lower are per stage, is tuned so that V1>V2. The net effect of both of these required during parts of typical RDV missions, and as such can effects is that the two satellites drift apart in all three dimensions be treated impulsively. However, much larger ΔV’s are ideally (radially, along-track and cross-track). A similar approach was needed during the early parts of the mission (e.g. phasing and used in the launch of the three SSTL DMC-3 satellites on PSLV long range rendezvous). Theoretically these could be delivered impulsively with a larger propulsion systems. However, for example, the mission scenario calls for an approach to 10km and then down to 1km as part of the rendezvous phase. To move ~10 km distance along the V-bar using ~0.1 km sized hops (i.e. with impulsive manoeuvres) would take ~100 separate manoeuvres on the spacecraft. Each hop (if using radial burns) takes half an orbit (e.g. ~50 min) which implies that even if there was no waiting period between each hop, then to traverse this 10 km would take 5000 minutes or 3.5 days. This however could be an acceptable trade-off considering the cost-benefit of using low thrust resistojet technologies compared to higher thrust systems. Furthermore in many mission scenarios there is no particular advantage in ‘going quickly’ and a ‘slow and steady’ approach is perfectly acceptable. Fig.6 Sketch of upper stage deployment.

JBIS Vol 71 No.9 September 2018 327 STEVE ECKERSLEY et al for example. The ejection angle, β, is taken as 5°, and the differ- are considerably more efficient than radial impulses in terms ential between the velocities is 0.05 m/s. Tuning of ejection sys- of ΔV. Furthermore, the large separation range at this point tems (e.g. clamp bands and push-off springs) to this level of fidel- means that even if a tangential manoeuvre was missed or there ity is well within the capabilities of state of the art mechanisms. was some other anomaly causing an undesired drift towards the Target, there would be sufficient time to either a) correct The initial baseline is to assume a 10 day period for initial the anomaly on the Chaser and re-attempt the manoeuvre, or LEOP and commissioning, and in this period it is assumed that b) manoeuvre the Target to avoid any risk (cooperative scenar- no manoeuvres take place on the spacecraft. In this time the io). This is only acceptable due to the large range and small two spacecraft drift apart by ~130 km. manoeuvre sizes that are needed. As shown later in this paper, when at close range, radial impulses are preferred for their pas- Although both spacecraft are assumed to initially be identi- sive safety characteristics. cal, the Target spacecraft will use less propellant than the Chas- er, and as such it will experience less drag than the Chaser. As The approach modelled here is for a series of small Hohmann a result, it is advantageous for passive safety reasons [3], for transfer manoeuvres to be initiated by the Chaser spacecraft. the chaser to approach the target from ‘ahead’, i.e. the chaser The objective is to raise the orbital altitude of the Chaser rela- approaches in an anti-velocity direction from the viewpoint of tive to the Target, so that it drifts back towards the Target. Be- the Target. Therefore in terms of the launch, we would like the cause of the limited thrust and ΔV capability on the spacecraft, spacecraft to naturally drift apart such that the Chaser ends up the actual transfer to this higher orbit is split into 2 smaller lying ahead of the Target satellite. Counter initiatively, this re- manoeuvres. In the scenario modelled here, two Hohmann quires the Target satellite to have the larger of the two ejection transfers are initiated by the Chaser, each one requiring two velocities from the launcher. manoeuvres each of 0.02 m/s magnitude (four burns in total for an aggregate ΔV of 0.08 m/s). Naively it would be assumed that this would cause the Tar- get to drift away ahead of Chaser which of course is not the de- Each 0.02 m/s manoeuvre requires an impulse of ~30 sec- sired outcome. However, the larger ejection ΔV actually causes ond duration from the spacecraft. This raises the Chaser orbit the satellite to be thrown into a slightly higher orbit, which has sufficiently for it to slowly drift back towards the Target at a a lower mean motion that the Chaser, and hence then causes a rate of ~8 km/day. At this rate it takes ~15 days for the Chas- relatively backward drift of the Target compared to the Chaser. er to start to approach the 10km far rendezvous point. When Initially the Target moves ahead of the Chaser, but the effect close to the 10km range point, a small braking manoeuvre is of the larger ΔV is to loop the spacecraft back over the Chaser initiated. This is again a Hohmann transfer to lower the alti- from where it then drifts backward (i.e. the Chaser drifts for- tude of the spacecraft, bringing its mean motion closer to the ward in a series of loops along the Target V-bar). The net result Target, and slowing the drift rate. In the case here this is initi- is a passively safe trajectory that causes the Chaser to drift for- ated when at ~13 km range, although it could be applied at any ward along the V-bar whilst ‘orbiting’ around the V-bar due to point. A Hohmann manoeuvre of 0.02 m/s in total (2 burns of the small (~20m) cross-track difference imparted by the differ- 0.01 m/s each) slows the drift rate to ~3 km/day, meaning the ential ejection angles. Chaser then has ~24 hours to drift in towards the 10 km far rendezvous point. After ten days the Chaser spacecraft reaches a point where it is ~130 km along the Target V-bar and ~1.4 km radially be- When the range reaches 10km along the V-bar a stop ma- low the Target. At this point the spacecraft has to start moving noeuvre is applied. This is a radially directed impulse, that tar- back towards the Target to initiate a rendezvous. In this first gets minimising the differential semi-major axis and eccentric- phase after launch, the objective is to bring the Chaser back to ity between the two spacecraft (slowing the relative drift to as a position 10 km away from the Target, as a ‘far rendezvous’. close to zero as possible). In this case the ΔV needed is 0.084 From there the Chaser will hold before initiating the close m/s. This is a relatively large manoeuvre (~120 s burn time), range rendezvous and proximity operations. but should be achievable with the given thruster layout and the proposed thruster-controlled attitude mode on the spacecraft. Because of the large separation between the two satellites The trajectory up to 10 km range is shown in Fig. 7: after 10 days, there is minimal risk in using tangential impulses to initiate a return trajectory. As shown previously these In total therefore the time from launch to get back to 10km

Fig.7 Relative motion of the Chaser relative to the Target in the R-bar, V-bar plane.

328 Vol 71 No.9 September 2018 JBIS FUTURE RENDEZVOUS AND DOCKING MISSIONS enabled by low-cost but safety compliant Guidance Navigation and Control (GNC) architectures

Fig.8 Total Chaser-Target range following launch. range is ~27 days as shown in Fig 8. The total ΔV is 0.184 m/s, option represents. and so is not expected to be a major mission driver. In addition, when at 10km at the beginning of this phase, Following the initial LEOP and far rendezvous phase de- the Chaser spacecraft has an out of plane (H-bar) motion with scribed in the preceding section, the chaser will be ~10 km respect to the Target. This is due to the relative ejection angle distant from the Hub on the V-bar. From this point a number of the two spacecraft from the upper stage as described earlier. of different approach strategies can be adopted. A number of Although at ~10 km range between the spacecraft, this cross- different options have been explored as part of the study, with track motion does not cause any particular operational or safety differing levels of complexity and orbital manoeuvring needed. issue, it is convenient to reduce the magnitude of the out-of- plane motion at this point. In terms of the H-bar motion of the 4.3.2.1 Strategy to move from 10km to 1km Chaser just prior to arrival at 10km V-bar range, the spacecraft has an out-of-plane oscillation of ~±60m (having grown from Once the spacecraft has manoeuvred back to a point 10km 20m at launch vehicle ejection due to J2 induced drift of the two distant from the Target, then it has achieved a ‘far rendezvous’. (Chaser and Target) orbit planes due to their slightly different Now the Chaser moves to ~1 km distant from the Target, from orbital elements during the preceding mission phases). where a close rendezvous, and the true ‘proximity’ operations start. To cover this movement from 10km to 1km, two options When the spacecraft crosses the V-bar/R-bar plane (i.e. are available: • Use a series of small radial hops to ‘nibble’ along the V-bar towards the Target. This is completely passively safe, but requires a high number of manoeuvres on the spacecraft, and so could be operationally intensive. • Use a single impulse tangential manoeuvre, to set up a drift trajectory that naturally moves towards the Target. This does result in motion that causes the Chaser to continu- ously move towards the Target, and this does raise safety issues. However, as shown below, with a sufficiently small impulse, the drift rate is relatively slow taking. Even if a stop manoeuvre was missed for any reason, there would still be many hours for either the Chaser or Target to attempt corrective or evasive manoeuvres.

If we wish the spacecraft manoeuvres quasi-impulsive, then it is assumed that the magnitude of the radial hop ΔV is assumed to be limited to 0.04 m/s, as discussed earlier. Each hop therefore takes 0.08 m/s, and moves the Chaser ~150 m along the V-bar. It therefore takes 62 separate hops (124 manoeuvres in total) to reach 1km range (taking 5.2 days in total). The total ΔV is 4.96 m/s. Conversely the tangential drift uses a single starting impulse of 0.02 m/s and drifts across the 9 km in 1.75 days. As the drift is uncontrolled in this case, a dedicated braking manoeuvre is needed in this case to bring the spacecraft to a stop just prior to reaching 1 km (0.02 m/s), followed by a final ‘trim’ burn using a passively safe radial burn to reach 1 km.

The trajectory for both options is shown in Fig. 9. Both Fig.9 Relative motion of the Chaser relative to the Target when options are a possibility for the mission, and the time and ΔV moving from 10km to 1km range. The bottom plot shows the final must be traded against the risk (real and perceived) that each elements of the trajectory when close to 1km range.

JBIS Vol 71 No.9 September 2018 329 STEVE ECKERSLEY et al when the H-bar range is zero) an appropriately directed ΔV can be used to remove some or all of the H-bar rate.

4.3.2.2 Approach Strategy to move from 1km to 100m Range

From 1 km to 100 m range represents the next element of the rendezvous phase. As the separation between the satellites is now much closer than the previous case, radial hop manoeuvres are used exclusively to transit along the V-bar. Fig. 10 shows the trajectory in the V-bar/R-bar plane in this case. Initially it is assumed that the spacecraft sits at 1km for one day prior to initiating the approach. This can be seen in Fig. 10 as the thicker line oval at ~1 km range. This is the spacecraft experiencing a slow backward drift due to the differential area-to-mass of the two spacecraft. As shown, the spacecraft only drifts ~10m Fig.10 Relative motion of the Chaser relative to the Target when backwards in one day, thus effectively remaining stationary moving from 1km to 100m range. with respect to the Chaser, and thus affecting propellant-free station keeping. As the spacecraft then crosses the V-bar it then initiates a series of 12 radial hop manoeuvres (each consisting TABLE 3 Components of the ΔV vectors needed to arrive at of 2-off 0.02 m/s manoeuvres) to bring the spacecraft to ~150 100m Chaser-Target range m range. From here a final stop and trim manoeuvre can be Start manoeuvre used to put the spacecraft onto the V-bar at 100 m range. X (Velocity): 0.00451 m/s Up until this point, it is acceptable to allow the Chaser spacecraft some degree of ‘oscillation’ in both V-bar and R-bar Y (Normal): 0.00000 m/s when moving closer to the Target, and this can be seen in the ‘looping’ nature of the spacecraft with each V-bar hop. This Z (Co-Normal): 0.00286 m/s arises if the accuracy of any targeting sequences are relaxed and Stop manoeuvre the manoeuvres are applied ‘open-loop’ with fixed ΔV. This is perfectly acceptable when at longer distances from the Target, X (Velocity): -0.00446 m/s and from a planning and operations point of view is a simpler procedure. In this case the spacecraft is simply instructed to ex- Y (Normal): 0.00000 m/s ecute a ΔV manoeuvre of a fixed size and fixed orientation at a certain time epoch. As long as the orbit knowledge of the both Z (Co-Normal): spacecraft is accurate (which should be achievable with Rela- 0.01496 m/s tive GPS) then this approach yields satisfactory results. How- ever, at ~100 m, the Chaser GNC sensors (optical cameras) will also start to acquire the pose of the Target and the system can the drift is extremely slow. For example, the drift of the Chaser transition to closed-loop manoeuvres (using both the relative over 24 hours after arrival at the 100m point, is only ~2m back- GPS navigation and optical camera data). The spacecraft can wards. The radial oscillation has also be reduced to ~±0.5 m. then target a stop manoeuvre on the V-bar at 100m range. We also wish in this case to null out the relative radial velocity, so 4.3.2.3 Final Approach to move from 100m to Docking that the spacecraft comes to rest at 100m range. From the 100m holding point, the spacecraft can initiate the This can be achieved with a two impulse manoeuvre, but in concluding approach, the close rendezvous and inspection, this case the burns are not identical. Instead of open-loop ma- and the final docking sequence. noeuvres, in this case a differential correction targeting routine is used to solve for the components of the ΔV vector that will When at 100m, although the spacecraft has very small R-bar result in the desired end state for the spacecraft. In the example and V-bar errors, it still has residual cross-track motion at this here, we wish to arrive at 100m range with minimal relative point, and this must also be nulled-out for a successful dock- semi-major axis and eccentricity between the two spacecraft ing. Similar to the approach when at 10km, an out-of-plane (to arrest any V-bar drift) as well as reducing the radial-rate manoeuvre applied at the V-bar crossing can be used to re- to as close to zero as possible (removing the radial oscillatory move the out of plane motion. A ΔV of 0.0065 m/s applied in movement of the spacecraft). This can be numerically solved as the direct of the orbit-normal is sufficient to effectively scrub- there are three control variables (the three components of the out the H-bar oscillation. Immediately after this ΔV, a series ΔV vector) being used to target three parameters (semi-major of very small radial hop manoeuvres can be used to bring the axis, eccentricity and radial rate). Initially a burn with mag- spacecraft to 10m range from the Target. This sequence has 4 nitude 0.0053 m/s is applied (~10 sec burn) which puts the hops, each using 2-off manoeuvres of 0.005 m/s, followed by a spacecraft on a trajectory that will intersect the V-bar at the slightly smaller radial hop to target 10m range with a ΔV’s of next crossing at 100m range. At the V-bar crossing a second 0.0024 m/s and 0.0026 m/s. The trajectory is shown at Fig. 11. manoeuvre, this time with magnitude 0.0156 m/s is then ap- Note that by this point, due to the nulling manoeuvre carried plied. The components of the two burns are shown in Table 3. out earlier at 100m, the H-bar error has now been reduced to a very small value (~±4 cm). The Chaser can safely sit on the V-bar at this point, and as with the example above at 1km, it will experience a slow back- The final approach to a docking with the Target, starts at 10m wards drift. In this case because of the ‘closed-loop’ targeting range. At this point the range is low enough for the GNC LIDAR

330 Vol 71 No.9 September 2018 JBIS FUTURE RENDEZVOUS AND DOCKING MISSIONS enabled by low-cost but safety compliant Guidance Navigation and Control (GNC) architectures

Fig.11 Relative motion of the Chaser relative to the Target over Fig.13 The along-track, radial and cross-track ranges between moving from 100 m to 10 m range. Chaser and Target during the final forced motion (straight-line) approach trajectory from 10m to docking. R-bar and H-bar ranges are shown on the secondary axis on the right of the plot. to acquire the target, and this can be used alongside the optical navigation sensors for closed-loop trajectory control. In the simulation software used here, a true closed-loop GNC model is The actual radial, along-track and cross-track ranges be- not possible, but as for the previous examples, a numerical dif- tween the Chaser and Target are shown in Fig. 13. The duration ferential correction routine can be used to solve for the desired of the forced motion is 890 sec, giving an actual closing rate of end-state given initial conditions and control variables. 0.011 m/s, which is acceptably close to the nominal 0.01 m/s for this early study phase. Note that in this approximation of a Based on available data on the proposed docking mecha- true closed-loop forced trajectory, the V-bar motion is actually nism a relative velocity of 0.01 m/s at contact is within the ac- a slightly curved path. ceptable capture criteria for the mechanism. The forced motion approach therefore starts when the Chaser arrives at 10m range, Fig. 14 shows the final motions of the forced motion ap- where the radial stop manoeuvre of 0.0026 m/s is applied, be- proach in more detail. As noted above the assumed magnetic fore a tangential impulse of 0.01 m/s is immediately applied docking mechanism has an activation boundary of 0.5m, and in the anti-velocity direction. Then immediately following this a capture cone of 45°, and these are illustrated in Fig. 14. Also tangential impulse, a radial forcing thrust is applied to hold shown on the figure on the spacecraft is an assumed order of the spacecraft on a straight line trajectory. The differential -cor magnitude boundary for the Target spacecraft itself with as- rector is then used to solve for the components of the forcing sumed dimensions of 0.6 x 0.6 x 1.2 m (with the coordinate thrust to drive the Chaser towards the Target spacecraft. The system centred nominally on the geometric centre of the Tar- scenario is illustrated in Fig. 12. get spacecraft, this means the actual docking mechanism exists +0.3m ‘along’ the Target V-bar). Although the trajectory visually looks like a straight line, in actuality the spacecraft is moving on a slightly curved path. However, as long as the spacecraft arrives at the Target within a 45° cone centred on the docking mechanism, and with a relative velocity of 0.01 m/s when at 0.5m range then the proposed magnetic docking mechanism can (for the purposes of this initial study) be assumed to be “activated” and the spacecraft captured into a docked state.

Fig.14 Schematic of the final moments of the forced motion approach. The top plot shows the motion in the R-bar/V-bar Fig.12 The final forced motion (straight-line) approach trajectory plane, whilst the lower plot shows the motion in the V-bar/H-bar from 10m to docking. The red sphere represents a volume of plane. The Target spacecraft and the docking capture cone and radius 10m around the Target spacecraft. boundary conditions are also shown.

JBIS Vol 71 No.9 September 2018 331 STEVE ECKERSLEY et al

Fig. 14 shows that the Chaser easily meets the required arrival conditions, arriving at the docking mechanism boundary (50cm from the edge of the spacecraft, 80cm along the V-bar) with an R-bar error of -12.9 cm, and an H-bar error of -3.2 cm.

A more complex final approach scenario can be constructed in which the Chaser performs a fly-around manoeuvre prior to initiating the docking manoeuvre. This type of trajectory allows the Chaser to ‘orbit’ around the Target, which could be useful for a number of different mission scenarios including those not explicitly related to docking (e.g. inspection-type missions). A range of these scenarios have been covered in the study but are not covered here in any detail to keep the paper as concise as possible.

4.3.2.4 Total ΔV Fig.15 Cumulative ΔV from launch vehicle ejection to docking, for the three different scenarios discussed in this document. This Fig. 15 shows the cumulative ΔV expended by the spacecraft assumes the use of tangential free drift from 10 km to 1 km. from launch vehicle ejection to final forced motion docking, for the three cases discussed above. This assumes the use of the lower ΔV tangential free-drift trajectory to move from 10km The relative motion of the Chaser with respect to the Tar- to 1km as discussed earlier in this section (If radial hops were get after an undocking with an effective ΔV of 0.01 m/s shows used to move from 10km to 1km, then an approximately an that the spacecraft initially moves in the –V-bar direction, but additional 60 manoeuvres would be needed, and the total ΔV then loops underneath the Target. Half an orbit later, when the would increase by ~5m/s). In total ~50-55 separate manoeu- spacecraft reaches the V-bar a stop ΔV of 0.01 m/s is then ap- vres are needed on the Chaser (dependent on the scenario in plied, which puts the spacecraft on to the V-bar in a very slow question), but even in the worst case the ΔV is very low, with drift away from the Target. From this point any other manoeu- the spacecraft able to achieve the whole sequence from launch vre sequence could then be started. to docking for a total ΔV of < 1.1 m/s. Note that this value does not include any additional effective ΔV needed for attitude Analysis of the long term relative motion between the space- control, which is handled at spacecraft system level. craft following undocking shows that if the undocking occurs in the –V-bar direction as above then the chaser permanently 4.3.3 Undocking of the Spacecraft accelerates away from the Target due to drag perturbations and an increased relative mean motion. If the undocking occurs For many mission scenarios involving on-orbit assembly there in the opposite direction however, then initially the spacecraft may be a requirement for undocking of two spacecraft that are drifts away on the –V-bar side. However, after ~13 days, the physically coupled. In these scenarios the two spacecraft must increased drag perturbation on the Chaser slows its drift rate separate and move apart safely (no collision risk), ideally with and it then starts drifting back towards the Target reaching a minimal ΔV. zero-along track separation after ~25 days. In this particular case there would not be a collision as there is a radial separa- The Target is assumed to have a mass of 150 kg, whilst the tion of ~20m when the along-track distance crosses through chaser has a mass of 147 kg when undocking (i.e. propellant zero. However it is clearly undesirable to have spacecraft drift- has been used in the preceding elements of the mission). If ing – potentially in an uncontrolled manner - so close to other the two objects have such a differential area-to-mass ratio (as vehicles. Thus it can be seen that in most cases it will prudent above), then it is advantageous to perform the undocking in for undocking manoeuvres to occur with a geometry that en- the anti-velocity (-V-bar) direction. Although the initial mo- sures no possibility of uncontrolled re-contact between the two tion is along the –V-bar, if the undocking imparts an effective spacecraft. impulse to the Chaser, it will loop around the Target and ultimately end up on the +V-bar side of the Target. If the un- 4.3.4 Collision Avoidance docking occurs at End-of-Mission, and the two spacecraft are to be permanently separated, then with the differential area-to- A key requirement for any close proximity mission is the abil- mass ratio the impact of atmospheric drag on the Chaser will ity for the approaching spacecraft to abort its manoeuvre or continually pull it away from the Target. If on the other hand, trajectory if an anomaly or other error means it is placed onto the undocking does not occur at the end of the mission, then a trajectory that has an unacceptable collision risk. In such cir- the same approach can be used to recover back on the +V-bar, cumstances a Collision Avoidance Manoeuvre (CAM) may be where any subsequent manoeuvres can take place (e.g. another needed. The primary aim of the CAM is to avoid a collision, approach or fly-around etc.) with a secondary aim of moving the spacecraft to a location whereby the mission can be continued at a later date (following In the process of un-docking an effective ΔV is imparted to any necessary recovery procedures). the Chaser. This can either be from a mechanism in the docking receptacle (in the configuration used in this study, this would Two CAM scenarios have been modelled as part of the study: be from the magnetic docking system), or from thrusters on • A CAM is needed following a radial hop towards the 10m the Chaser, or from some combination of both. If sufficient ΔV boundary at which forced motion would nominally be was available from the docking mechanism, this could also al- started low ejection of the Chaser even if its propulsion system was • A CAM is needed during the final forced motion phase, unavailable or had suffered some other failure. when the spacecraft separation is very low (~few m).

332 Vol 71 No.9 September 2018 JBIS FUTURE RENDEZVOUS AND DOCKING MISSIONS enabled by low-cost but safety compliant Guidance Navigation and Control (GNC) architectures

Depending on the failure or anomaly in question, it may not be possible to use certain thrusters on the spacecraft. Therefore CAM’s have been modelled using both radially and tangential- ly applied impulses. The results of these analyses are not shown here to keep the paper concise, but they do show that very modest total ΔV’s of <0.1m/s (and in some cases considerably less) are sufficient to provide CAM protection.

5 GNC SIMULATION AND MODELLING

This section of the paper will focus on the modelling and simulation of the GNC system during the rendezvous final approach i.e. prior to docking.

The objectives of this phase are the reduction of the chaser range to the target and the acquisition of the final approach corridor leading finally to a successful dock. Different acqui- Fig.16 Local Orbital Reference Frame. sition strategies for V-bar and R-bar approaches can be implemented. However we will assume the final approach is made along the V-bar with a nominal initial chaser/target separation GNC and environment on which a closed loop trajectory con- of 10 m. trol scheme can then be implemented. The simulator, which is based on Matlab/Simulink, can then be used to verify the The goal of the final approach phase is to achieve docking performance of the closed-loop controlled quasi-straight line capture conditions in terms of relative positions and veloci- trajectory approach. ties and also of relative attitude and angular rates. To meet the docking requirements, the relative attitude between the dock- Relative motion is expressed in the local orbital reference ing ports of the two spacecraft must be reduced to close to frame as shown in Fig. 16. zero at the end of the manoeuver. For observability and safety reasons a cone-shaped approach corridor will usually be de- The origin of this frame is based on the target centre of mass fined, within which the approach trajectory has to remain. with the Z-axis (“R-bar”) pointing to the centre of the Earth, We assume this corridor has a half cone angle of 10° [3]. A the Y-axis (“H-bar”) pointing in the opposite direction to the quasi-straight line trajectory is used in this manoeuver phase angular momentum vector of the orbit and the X-axis (“V-bar”) meaning that the chaser GNC follows the direction of the tar- completing the right-handed system and in the direction of the get docking axes. orbital velocity vector.

This GNC modelling and simulation will therefore cover The GNC simulator, illustrated in Fig 17, covers the follow- the GNC design during final phase of rendezvous and docking models/functions: ing when both relative position and attitude are required to be • Orbital dynamics controlled closed loop. For the purposes of this study we will • Sensor model assume ‘standard’ SSTL attitude control will be sufficient to • Thruster model achieve the rotational requirements for a successful docking. • Control algorithms Attitude control is therefore not implicitly modelled in order • Guidance to avoid unnecessary detail and work at this phase of the pro- • Navigation filter ject. However position (trajectory) control is required in order to realize a safe controlled approach. The design approach The underlying simulation of the spacecraft environment is is therefore to develop a relative position simulation of the contained in the Orbit Model. The non-linear dynamics of the

Fig.17 Simulator Overview.

JBIS Vol 71 No.9 September 2018 333 STEVE ECKERSLEY et al target and chaser spacecraft can be derived from Kepler’s laws. For LQR control we wish to minimize the cost function: However to make the problem understandable and tractable we will linearize the relative motion using the well-known Clo- (3) hessy-Wiltshire-Hill (CWH) equation [4]: where Q is the state gain matrix, R is the control effort gain matrix (as usual, for simplicity, the gain matrix N is assumed to be zero and is not included here)

(1) The optimal feedback control is given by: (4)

where the optimal state feedback is:

Where F is the perturbing force (primarily thrusters), mc is the mass of the chaser spacecraft and ω is the target orbit angular velocity. The CWH equations assume circular orbits for the S is the solution of the associated Riccati equation. The Mat- target and chaser spacecraft lab lqr function is used to solve for the state feedback gain K and S given the weighting matrices Q and R. The CWH equations are frequently used to approximate the relative motion dynamics of spacecraft at close range, which is The choice ofQ and R follow the Bryson methodology [5], precisely the design case chosen for our analysis/simulations. from which the gain matrices are selected for efficient control effort coupled with relatively short manoeuvre durations. The The state-space form of these time-invariant equations is: weighting matrices are chosen to trade-off state convergence and control effort efficiency. The weighting matrices are there- (2) fore taken to be diagonal with values chosen to normalize each of the state (x) and input (u) variables: where x is the state vector = and u is the control vector in the usual formulation. (5)

The denominators of these expressions refer to the largest desired values. For the input variables u this will be given by the maximum acceleration available from the thrusters. The proposed control algorithm uses the robust Linear Quadratic Regulator (LQR) method as a means of providing For the state variables x, assuming the approach is along optimal control effort serving as an attractive force towards the V-bar, we take max(x) = R max (initial starting range of the relative position goals. The close proximity LQR controller is approach, e.g. 10m for the nominal case). For the cross track based on the linear dynamics described above. axes, we assume that the approach is restricted to an approach

Fig.18 Thruster Layout.

334 Vol 71 No.9 September 2018 JBIS FUTURE RENDEZVOUS AND DOCKING MISSIONS enabled by low-cost but safety compliant Guidance Navigation and Control (GNC) architectures

corridor given by a half cone angle Ɵ = 10°, so max(y,z) = Rmax TABLE 4 Thrusters used for 6 DOF Control tan(Ɵ). The velocity terms are given by the maximum velocity Rotation Thrusters Translation Thrusters allowed e.g. 1 cm/s. Main Red Main Red +X 3&4 13&14 6&7 18&19 As the chaser and target spacecraft converge the cost slope tends to flatten because of the small state values being consid- -X 1&2 15&16 5&8 17&20 ered. This levelling of the cost can be avoided by re-calculating +Y 7&8 17&18 1&3 13&15 the gains as the range decreases. However the solution of the -Y 5&6 19&20 2&4 14&16 feedback gains is not trivial and computationally expensive so +Z 2&3 14&15 11&12 23&24 an on-board solution is best achieved by pre-calculating the -Z 1&4 13&16 9&10 21&22 gains and storing as a table lookup.

It should be noted that although the LQR technique is very robust any resulting controller from the above method does not guarantee the constraints are never violated and further verification by simulation will be required.

The thruster model is based on an SSTL Resistojet propulsion system (discussed in more detail later in this paper), which features 24 thrusters to meet the failure tolerance requirements (see Fig. 18).

The thrusters are arranged as two redundant sets of 12 thrusters each of which can provide full six DOF control as shown in Fig.19 Camera Experimental Results. Table 4. For the final closed loop approach phase Resistojets in cold gas mode are employed with assumed thrust capability of 50mN each and a Minimum Impulse Bit (MIB) of 10mS. In a final system the measurements from all sensors would be combined via a navigation filter but this is outside the scope In order to avoid plume impingement effects as much as pos- of this present work. However a simple Kalman filter has been sible, the target facing thrusters (-X) are canted at a small angle. implemented in the simulator to alleviate the effects of sensor This will have a very small impact on the force available (< 5%) noise. This is particularly important in order to avoid excessive but could lead to parasitic torques when used in attitude con- sporadic thruster firings triggered by noise. trol mode. In order to alleviate this the Pulse Width Modulator (PWM) shall be designed to compensate where possible oth- A nominal closed-loop simulation case considers a V-bar erwise the parasitic torques will be seen as small disturbances. approach from 10m with zero initial cross track dispersion. Realistic sensor and actuator models as described above are The thrusters are inherently non-linear on/off devices which included. The choice of sensor error parameters mean that only offer a fixed thrust or acceleration over a fixed on-time. control can either be assumed to be the result of relative GPS However they can be employed in a quasi-linear mode by using measurements or optical/LIDAR measurements but taken to a Pulse Width Modulator (PWM) to modulate the width of the be at the maximum 10m range. It is expected that optical/LI- activated reaction pulse proportionally to the acceleration (or DAR measurements will improve in accuracy as a function of torque for attitude control) command input to the controller. range so the results could be considered worst case for optical/ An update rate of 1 Hz is assumed in the simulator. LIDAR output.

Optical imaging cameras and LIDAR are probably the most Sensor noise means that estimated relative position differs appropriate types of sensors for close range. Fehse [3] gives from the real value but as can be seen in Fig. 20 with the Kal- typical performance figures for range measurement by optical rendezvous sensors as follows: <10m range => accuracy better than 0.01m <30m range => accuracy better than 0.1m <100m range => accuracy better than 1m

The PRISMA project [6] also found that range measurement accuracies of decimetres could be achieved using relative GPS combined with a suitably designed navigation filter.

Typically, therefore, we expect better than 1% range accuracy with optical sensors. SSC are looking to achieve better than 5-10% range accuracy for COTS based optical RvD sensors. Experimental results shown in Fig. 19 indicate accuracies of better than ~2% can be achieved for an RGB camera with NIR filter and active LEDs:

No experimental results are available for range rate, however a range rate requirement accuracy of 1 cm/s is appropriate to this study although a figure of 1 mm/s is preferred as a goal. Fig.20 Simulation Results – Relative Position (5% sensor error).

JBIS Vol 71 No.9 September 2018 335 STEVE ECKERSLEY et al

Fig.21 Simulation Results – Relative Velocity (5% sensor error). Fig.22 Simulation Results – Acceleration (5% sensor error). man filter in place the estimated results broadly follow the real amount occurs on the H-bar axis as we would expect as motion values. Final cross-track position errors at docking (range = in this axis is decoupled from the other axes. Thruster usage 1m) are the order of mm. on this axis is most likely triggered by noise etc. Motion along the V-bar requires an initial build-up in order to provide the In Fig. 21 it can be seen that the controller applies close to slew manoeuvre. On the R-bar the ΔV ramps up to compen- the maximum slew velocity of 0.01 m/s allowed in order to sate for the Coriolis acceleration. The total ΔV on all axes sums manoeuvre to the target. The velocity is gradually backed off to about 3 cm/s. Thruster non-linearity plus noise and error as the target is approached, and would approach zero as the sources will tend to increase this figure. docking distance tends to zero. However docking is permitted for relative velocities less than 0.01 m/s so this would be count- As an observation it can be seen that the initial manoeuvre ed as a successful dock. Cross-track velocity errors in Fig. 21 on the X-axis is constant so that for closed-loop manoeuvres are the order of a few mm/s. In Z there appears to be a very from longer distances (e.g. 100m) this ΔV will remain much small steady-state error owing most likely to the Coriolis dis- the same. However on the Z-axis compensation for the Coriolis turbance. If necessary this could be removed by, for example, acceleration means that this part of the ΔV budget is increasing introducing an integral control term to the state vector. linearly.

As can be seen in Fig. 22 introducing a non-linear thruster In conclusion, the purpose of this section of the paper has model, albeit with a quasi-linear overall effect due to the PWM, been to look at the final approach of the close range rendez- results in fixed magnitude thruster pulses with variable on- vous phase leading to the spacecraft docking conditions and times. A positive acceleration profile at the start of the run starts to define and perform the GNC modelling and simulation to the slew on the X-axis. Before docking a negative acceleration is cover the GNC design during the final phase of rendezvous and applied. The Z-axis has a constant thruster pulse stream to offset docking. the Coriolis acceleration. On the Y-axis sporadic pulses can be seen, most likely caused by noise and cross-coupling effects. A suitable control scheme has been designed and a simulator based on Matlab/Simulink has been built and tested. The The total ΔV for each axis is depicted in Fig. 23. A small simulator has then been used to perform various simulations for a close range rendezvous with a closed-loop controlled quasi-straight line trajectory on the V-bar axis from 10 m range.

Orbital dynamics based on the CWH equations has been derived in order to provide a plant model for the simulator and linearized input to the controller design. The CWH equations are more than adequate for the very short range considered here as their inaccuracy does not become significant until a few tens of kilometres. A possible enhancement to the simulator would be to include non-linear dynamics equations so that differences with CWH could be assessed along with non-circular orbits, perturbations etc.

The Linear Quadratic Regulator (LQR) is a favoured approach for optimal controllers with the benefits of fuel efficiency and guaranteed robustness in terms of stability margins. For the trajectory control system considered in this section it provides a known, well-structured and safe design principle whose Fig.23 Simulation Results – Total ΔV (5% sensor error). X-axis is at main disadvantage is the solution of the Riccati equation which the top, Y-axis in the middle, and Z-axis at the bottom. is both computationally difficult and expensive. A solution has

336 Vol 71 No.9 September 2018 JBIS FUTURE RENDEZVOUS AND DOCKING MISSIONS enabled by low-cost but safety compliant Guidance Navigation and Control (GNC) architectures been found to pre-generate the state gain matrix required and load on-board the spacecraft using look-up tables. This greatly reduces the on-board computation and removes the risk of cod- ing complicated algorithms with associated safety implications.

Sensor and actuator models have been developed with regard to the existing literature and experimental results obtained by SSC and information provided by SSTL engineers. Particularly for the sensors any performance is indicative and in most cases either worst-case or requirement figures are used in the simulator. Therefore it is recommended that further work to understand and quantify sensor characteristics be performed. Some thruster calibration work would also provide more knowledge concerning their performance, particularly in Fig.24 Camera and optical filters. cold-gas mode.

The navigation filter is an important entity for combining 6.2 Passive Optical Machine Vision System various sensor sources and producing the best estimate of the current state vector. Some information is available but it is A HP WebCam HD 2300 with a 1280×720 pixel sensor and recommended that further work be performed concentrating a 90° field of view lens was tested both in the laboratory, and on the requirements for the final approach of rendezvous and outside in sunlight. The results can be applied to longer range docking. It is particularly pertinent to consider the safety ben- operations by means of scaling using the camera focal lengths efits of using multiple sensors for spacecraft in close proximity. (or equivalently different fields of view). For machine vision systems (MVS), the Perspective-n-Point (PnP) and POSIT al- Finally simulation results were obtained for the final ap- gorithms are used and avoid the need for an initial pose es- proach scenario in order to assess the suitability of the pro- timates or computationally expensive iteration loops. Glyph posed control approach and of the GNC sensors and actuators recognition is popular in pose estimation augmented reality considered. With the current 5% SSC proximity sensor errors, systems and robotic navigation. A similar process is followed excellent positional control was demonstrated to mm level and for LED-based pose detection embedded in the glyph grid. a final closing velocity less than the required 1 cm/s. Variations in the guidance profile can be made in order to tweak the re- The sensor IC detects wavelengths from around 400 nm to sulting terminal position and velocity if required. Fuel usage is 1100 nm and cannot block near-infrared at ~800 nm. However, also very good with a total ΔV of ~3 cm/s for a 10 m starting for business applications, the webcam manufacturer has added point. The fuel usage can also be scaled to longer distances ow- an infrared cut-off filter to improve the camera’s performance ing to the linear consumption on the radial axis. Care must be in visible light. Hence, when no additional filter is added, the exercised in the design of any navigation filter to ensure sensor most of the light intensity detected comes from the visible light. error effects, particularly noise, do not cause excessive sporadic If applying a near-infrared filter, most of the visible light would thruster firings causing an increase in fuel usage. be sharply decreased, with a remaining interval around 850 nm. A particular wavelength pass-band filter with the central 6 PROXIMITY SENSOR PACKAGE wavelength of λ, cannot block light with a wavelength of λ⁄2^n (n = 0, 1, 2…) so the final wavelengths passed would be at both 6.1 Introduction 425 nm and 850 nm (Fig. 24).

We propose a hybrid approach, where the close-in operations Typical edge detection was use to capture lines, and then are guided by and active COTS LIDAR sensor overlapping shape matching located the glyph pattern (Fig. 25). The LEDs with the passive optical sensors (cameras observing glyphs and are Vishay TSHG6400 IR LEDs with a peak wavelength of 850 LEDs) operating out to a target 100m to provide proximity sen- nm with half intensity angle is 22º. The typical forward current sor provides relative range, range-rate, pose and pose-rate in- and voltage is 100mA and 1.5V but can draw up to 1A if higher formation to a “chaser” spacecraft as it attempts to rendezvous illumination is necessary. and dock (RvD) with a “target” spacecraft. However, over a short distance (~3m), active LIDAR (SoftKinetic DS325) could The maximum detection range for non-illumination of also provide good pose and range information. glyphs was 3800 mm indoors and 1600 mm outdoors, while the illumination method of LEDs was 2500 mm indoors and A key concern which emerged in the system trade-offs, is that the PSP must be capable of operating both under full sunlight and eclipsed conditions as both extremes may be experienced in Earth orbit. To this end, particular focus was made in establishing a “solar blind” optical system using. In this Sec- tion, we present the recent PSP test experiments.

By means of the LEDs, the passive optical system could operate under all lighting conditions (day or night). The LIDAR is known to be blinded by sunlight, and so further mitigation is needed for this – and we are currently investigating the use of narrow-band near-IF optical interference filters. Fig.25 Image Capture to Pattern Detection.

JBIS Vol 71 No.9 September 2018 337 STEVE ECKERSLEY et al

Fig.28 DepthSense SDK Output.

Fig.27 Greyscale without the Sun in view (top) and with Sun in view (above). Fig.29 Refined Distance Output.

1600 mm outdoors. No matter the case, the detection range for range, which was the same as the indoor test. LEDs in sunlight was tested to a maximum of 1500 mm. The LED intensity, even at low power, was able to pass the effective 6.3 LIDAR Test Results optical pass band filter. The accurate range after testing was 300 mm to 1600 mm for Sun not in camera view and 300 mm to Our previous work is targeted on the AAReST mission and uti- 1400 mm for Sun within the camera view. Though the sunlight lises the Raspberry Pi Compute modules. For initial tests, data had been filtered, it seemed that solar blind issues still affected from the DepthSense SDK was initially setup and analysed in the detection range. MATLAB to detect the centre depth frames (Fig. 28).

If the Sun was not in view, the detected range increases from At tested distances between 10 cm and 2.6 m, there is a sig- 900 mm to 1400 mm. The grey level transformation of image nificant range to the results due to the centre detected covering processing would enhance the contrast of image by rising the a large area of the target. Filtering the data and combining with brighter part to an even brighter part, and decreasing the grey knowledge of the target allowed us to further refine the result- level of the darker ones. It would result in a detectable back- ant range and pose measurements. ground because the darker LEDs may appear similar to other objects in sunshine (Fig.27). Fig. 29 shows that below 1m, the result is shown to close- ly follow the measured value. At most distances the range is If the Sun was within the camera viewing, it was different within ± 10 cm of the measured value, equivalent to a 5% error. that the average brightness of the image was large, so that the Investigating the SoftKinetic driver found an offset in readings, background remained dark and undetectable. The detection compared to the measured data. For most tests, we found a po- error rate of was 5% to 10% in the middle of the detection sitional error between 5% for any given reading. Filtering of the data provides the data in Fig. 29, where the accuracy can be improved. To detect rotational position, we note high pose accuracies at < 1% error from truth. Range and pose measurements operated at up to 40 Hz for state vector output with no debugging information.

To further investigate the flexibility of the PSP, the LIDAR was also operated as a camera stream. The main problem with find- ing the target object is its similarity to the surrounding RGB values with silver and white being the most prevalent colours in the lab. As such the target was modified and a binary mask was ap- Fig.30 Test model coloured & masked. plied using a threshold on the colour values is shown in Fig. 30.

338 Vol 71 No.9 September 2018 JBIS FUTURE RENDEZVOUS AND DOCKING MISSIONS enabled by low-cost but safety compliant Guidance Navigation and Control (GNC) architectures

chapter. There has also been a substantial preliminary analysis of Mission Safety and Robustness to collisions, which is covered in a separate chapter in this paper.

7.2 Mission Architecture/CONOPS

7.2.1 Mission Architecture Overview

The precursor demo mission is focussed on the Rendezvous and Docking of an Active/Master Chaser spacecraft and a Pas- sive Co-operative Target spacecraft, and will be a 6 month mission. The Target is very similar to the chaser in terms of hardware to minimise costs. Fig. 32 shows the Baseline Reference Mission Architecture. Fig.31 The distance output from the depth algorithm. The launch assumes a PSLV launch of both spacecraft into a 500km SSO, as was originally foreseen for the telescope To verify the accuracy of the optical algorithm, tests were demonstrator mission. carried out moving the target to different known distances from the LIDAR and the algorithm was used to find the distance to The Space Segment consists of the Chaser and Target space- the object. For the following plot (Fig. 31), the target was moved craft, which broadcast their telemetry and status information at 20 cm intervals from the LIDAR up to 1m, and at 2m and 3m. almost continually from the two spacecraft via the BGAN (Broadband Global Area Network) data relay asset, or direct We observed that most of the results, at a distance below 1m, to ground (via KSAT’s ground network at Svalbard or Troll) show the measured distance but with many outlying results. when in view of the mission ground segment. This will al- For example, one result was 9.4m when the measured distance low ground operators to continually monitor the behaviour was 3m. of both spacecraft, and to command an abort manoeuvre if considered necessary (if time permits). Downlink and Up- 7 MISSION AND SPACECRAFT SYSTEMS link data rates assumed for spacecraft telemetry and status are 10kbit/s via BGAN or S-Band TTC (Telemetry, Telecommand 7.1 Introduction and Control), though up to 200kbits/s is available for S-Band downlink and uplink via Core-DHS (Data Handling System). Whilst this is not a systems study, it has required a certain and Additionally BGAN can provide typical average data rates of significant level of mission and spacecraft systems and sub- 240kbits/s, albeit with greater power requirements. There are systems analysis and trade-offs key to the GNC architecture, a small number of short outages when the spacecraft have no in order to provide a sensible mission scenario and CONOPS BGAN coverage or no ground station visible. An S-Band ISL (Concept of Operations). These are briefly described in this between the spacecraft is also used to transmit Relative GPS

Fig.32 Baseline Reference Mission Architecture.

JBIS Vol 71 No.9 September 2018 339 STEVE ECKERSLEY et al measurements from the target to the chaser and to transmit Target sends a status message or ‘flag’ to the Chaser over the health flags between the two spacecraft (so the ISL is two- ISL. The Chaser likewise sends a status message to the Target as way). A standard SSTL Spacecraft Control Centre and Mission it computes its relative trajectory. Operations Centre are foreseen, with the Spacecraft Control Centre able to communicate 2-way to each spacecraft via the The Target will ‘sit and wait’ whilst the Chaser approaches KSAT (Kongsberg Satellite Services) or BGAN Ground Net- the docking port. If the Chaser is on a safe approach trajectory works. then the docking mechanism on both spacecraft will be activated when within 1m (TBC), and the final docking approach will 7.2.2 CONOPS of the 2 spacecraft work together be automatically commanded (i.e. command from the Chaser to the Target for a final ‘go for docking’). Telemetry and status The Chaser spacecraft will be the Master Spacecraft in the -for information will be almost continually broadcast from the two mation, and will be ‘in charge’ of the on-orbit relative manoeu- spacecraft (including status messages) via the BGAN data relay vres, including the phasing, rendezvous, and formation flying asset, or direct to ground (via Svalbard or Troll) when in view manoeuvres (i.e. the Chaser approaches the Target spacecraft of the mission ground segment. This will allow ground opera- and not the other way around). tors to continually monitor the behaviour of both spacecraft, and to command an abort manoeuvre if considered necessary For Far and Medium range relative navigation, the Target (if time permits). There are a small number of short outages will transmit its GPS measurements across a redundant two- when the spacecraft has no BGAN coverage or no ground sta- way S-Band ISL, enabling the Chaser to perform Relative GPS tion visible, but these can be planned in advance to allow a long measurements and derive position, range and range rate esti- uninterrupted approach. It should be noted that BGAN is not mation. A small internally redundant optical camera payload necessarily required for both spacecraft if a 2-way ISL is used, will also provide line of sight imaging and limited range/range as the key telemetry could be sent via the ISL. However double rate estimation as a backup. Relative GPS and Camera meas- use of BGAN reduces risk which is important for a preliminary urements will start from 10km inwards as they home in on the demonstration mission. Target spacecraft, resulting in both of them able to track the Target by the time they reach 1km range. Fig. 33 shows a high level block diagram of the Mission level GNC architecture which summarises how the Target and The camera will then also provide range, range rate and pose Chaser work. Note that the spacecraft FDIR (Failure Detection measurements of the Target spacecraft from ranges of 100m Isolation and Recovery) details are not shown here, just the and closer, with Relative GPS being used as the backup sen- high level baseline mission FDIR hierarchy, i.e. that the Chaser sor. Visual identifiers (e.g. an array of Glyphs and LED’s (Light is the “Master” spacecraft and can act on the FDIR from both Emitting Diodes) on the Target) are also used to help the Chas- itself and Target, as well as the Ground (which has the ability to er camera to acquire and determine the attitude of the Target, be in overall command of both spacecraft). in any lighting (eclipse and full sunlight). 7.2.3 Mission Phases and Timeline The Chaser spacecraft also carries a small COTS LIDAR for independent range, range rate and pose measurements of the The main mission phases are as follows: Target from ranges of 10m and closer as part of the final ap- • Launch and Commissioning proach and docking. Periodically (at the OBC update rate) the • Phasing

Fig.33 Mission level GNC architecture.

340 Vol 71 No.9 September 2018 JBIS FUTURE RENDEZVOUS AND DOCKING MISSIONS enabled by low-cost but safety compliant Guidance Navigation and Control (GNC) architectures

Fig.34 Nominal Mission Timeline.

• Rendezvous and Docking with the control knowing we have a backup system in place. • TBC Experiments • Final Un-Dock and De-Orbit In this scenario, cold gas thrusters are used for the entire Phasing and RDV approach to docking (inc. CAM’s), as well The launch assumes a single PSLV launch into a 500km SSO as AOCS throughout the mission. Resistojets are used for drag and then the spacecraft are assumed to be ejected with a small maintenance (after docking). differential ΔV. The Chaser drifts apart from the Target for 10- days (i.e. the Commissioning Phase) to a maximum distance 7.3 Spacecraft Systems and Ground Operations of about 130km, before Phasing starts with active ΔV control of the spacecraft to bring the Chaser back towards the Target. 7.3.1. Processing architecture

Phasing lasts for ~17 days (i.e. 27 days from launch) until a During the study, various trade-offs have been performed with relative separation of 10km between the Target and Chaser is SSC also actively involved, in order to agree and define the reached, at which point the Rendezvous and Docking Phase com- baseline processing architecture: mences and Relative Navigation is initiated (The Homing Phase). • A distributed processing architecture (as already envis- This Phase is nominally 4 days (i.e. up to 31 days from launch). aged on AAReST), i.e. separate rendezvous sensor processors rather than a single spacecraft computer for all It should be noted that the phases above are largely arbitrary, tasks (sensor processing, spacecraft GNC, other space- because the ‘real’ timeline may need to take into account a) or- craft OBC functions) bit determination and planning, b) check points by ground, c) • A double fault tolerant and highly responsive OBC solu- waiting for illumination conditions, d) comms coverage etc. tion using SSTL’s Core-DHS OBC (spacecraft GNC, other spacecraft OBC functions, relative GPS) After a successful Docking (including stabilisation, checks • Separate standalone redundant processors for each sensor etc) phase, additional TBC RDV&D experiments are envisaged to provide further demonstration within the constraints of the The proposed solution is to use three Core-DHS units, two 6 month mission duration and ΔV allocation. The final phase is units will be nominally powered with the third powered down to un-dock the spacecraft and perform a passivation manoeu- ready to boot if required. By using three units we can (a) cope vre so that the spacecraft will not collide on the de-orbit. with a unit failure and still continue the mission, and (b) cope with two unit failures and still be able to operate safely and have Fig. 34 shows the nominal mission timeline with further the ability to perform CAM’s. details on the Rendezvous and Docking Phase, including the different rendezvous and docking phases, key approach points, The use of Core-DHS impacts the number of equipment and baseline sensors used as a function of range. used and interfaces, as the Core-DHS OBC’s have some units which are specific to each OBC (and not cross-strapped). This The Magnetorquers provide coarse attitude control during means we need separate Sun Sensors, GPS, Magnetometers, phasing, station keeping and during the RDV phases, with at- and S-Band TM/TC Rx and Tx for each Core-DHS OBC. titude control accuracy improved by the thrusters. However Magnetorquer control is highly experimental, so a set of four 7.3.2 Propulsion small SP-10 wheels for both target and chaser are implemented as part of the GNC kit for the demonstration mission. This The propulsion architecture has been substantially developed would remove the risk and unknowns and simplify the design in close co-operation with SSTL’s propulsion group. This has at least for this initial phase. Therefore we can then experiment included significant iteration and one concurrent session to re-

JBIS Vol 71 No.9 September 2018 341 STEVE ECKERSLEY et al view potential failure modes with respect to failure tolerance enhance safety. By “critical phases” we mean phases where there requirements, to evaluate the risk and robustness to avoid col- is an enhanced risk of collision compared to normal missions, lision. The emphasis has been on the Chaser design, as the Tar- i.e. where passive safety is no longer guaranteed. A long com- get only has very limited manoeuvres to carry out. munication window of 20 minutes (1200s) or more is required, to allow for supervision of the synchronized flight phase, the In order to try and keep the overall system costs as low as docking and the first minutes of the stabilization phase. possible, a decision was made earlier in the project (as part of the GNC trade-offs) to try and keep the mission envelope with- As a result of this requirement, several planned RDV&D in that possible using simple Xenon cold or warm gas thrust- missions assume the use of a ‘chain’ of expensive ESTRACK ers (note also that the latter, i.e. resistojets, can operate in both (and partner) ground stations, which can provide continuous “cold” or “warm” modes). The main Xenon propulsion system coverage of up to around 30 minutes (~1800s). However this parameters assumed are summarised as follows: approach also has additional constraints on timing and lighting • Thrust: 100mN (two thrusters) (if this is an issue), as some of the passes will be in eclipse or • Isp: with no/poor lighting from the Sun, possibly further limiting – 48s in resistojet mode (60W for two thrusters) the approach timing. A string of ground stations is therefore – 30s in cold gas mode (0W) expensive, limited in continuous duration and very difficult to • Tank: SSTL-150 tank (mass 4.59kg capacity 12kg) plan/synchronise. • Minimum Impulse Bit: 50mN for 10 msec As an alternative, AddValue’s BGAN Inter-Satellite Data Xenon propulsion systems have been one of the heritage Relay System (IDRS) can provide quasi continuous real-time approaches for SSTL missions requiring relatively low ΔV’s, 2-way contact with the ground, via the InmarSat 4 constella- and the unit cost of the thrusters and tanks is considerably less tion in Geostationary Orbit. With IDRS, the mission would than that typically found for chemical systems such as hydra- not be reliant on costly ground station services for commu- zine (fuelling costs at the launch site are also much lower for nicating with their satellites. Additionally as BGAN is only gas systems). needed for low data rates and during short RDV phases (<1 month) and possibly only for the closer proximity phases, the The Xenon propulsion system has to be able to provide tan- operations cost is very reasonable for RDV missions. AddVal- gential and radial burns, retro/CAM manoeuvres, AOCS (along ue’s BGAN terminal is also small (only a few kg) and reason- with the Magnetorquers which provide coarse attitude control) ably low power when the uplink/downlink data rate is low, during the initial acquisition, burns, station-keeping and the which is all that is required in this mission. SSTL have been RDV phases, and 6 DOF control during the final approach. liaising substantially with AddValue to further investigate the technical details, develop the CONOPS, and assess the cov- It is useful to note that not all thrusters need to be resisto- erage in more detail. This has included provision of “day in jets. In fact cold gas thrusters are lower mass (0.6kg less than the life” simulation results by AddValue, including a detailed resistojets) so resistojets are only justified if the propellant mass timeline of gaps and coverage. savings exceed the additional thruster mass. Additionally the power required needs to be taken into account in the CONOPS It was found that BGAN can provide substantially longer when using resistojets as 60W for the combined warm-up time contact times than possible with a ‘chain’ of ground stations. (of 8 to 10 minutes) and burn time. No power is needed in cold To some extent this can be improved by improving the anten- gas mode. na solution (e.g. number of switched antennas). However there are still short breaks in the coverage during every orbit espe- The resistojets are therefore only foreseen for thrusters that cially at higher latitudes, where the terminal’s link margin is provide substantial phasing or larger tangential/radial burns. insufficient or there are simply gaps between the InmarSat-4 The other thrusters can be cold gas thrusters. For this reference footprints. Nevertheless this can be further augmented by us- mission, the Phasing and Rendezvous phases can be performed ing KSAT’s low-cost network with the Svalbard and Troll po- using the resistojets in cold gas mode, as the ΔV’s are very lar S-Band TM/TC stations. Indeed SSTL already uses KSAT’s small. Additionally CAM’s (which need operating immediate- Svalbard ground station. By combining the coverage of BGAN ly) must be operated in cold gas mode as there will be no time and Svalbard/Troll, much longer periods of unbroken coverage for warm up. Nevertheless warm gas mode is foreseen for drag of up to ~8400s (140min) can be achieved between ground and maintenance and the nominal collision avoidance allowance. spacecraft, using the example “day-in-the-life” data. Further- Also there is added flexibility for other missions may need more it was noted that after this long period of coverage, there higher ΔV’s for Phasing and Far Range Rendezvous. is only a short 10s gap, followed by another long pass of 2580s. Such short gaps may in fact be acceptable unless the Chaser is The baseline 24-thruster architecture uses both resistojets within few metres of the Target, thus potentially allowing even and cold gas only thrusters, and is completely redundant (see greater flexibility. earlier in Fig. 18). Plumes area avoided on the other spacecraft when in very close proximity, by using canted retro thrusters. Therefore the baseline TTC during the RDV&D is to use BGAN and the polar KSAT stations at Svalbard and Troll - 7.3.3 Operations and Ground Station coverage thereby switching between the two to maximize the length of coverage. Whilst fully autonomous “lights out” operations for RDV&D is desired and expected in the longer term to drive down cost, this 7.3.4 Relative GPS and Preliminary ISL is highly likely to be considered much too risky for nearer term demonstration missions. As a result, a major driver for any near- 7.3.4.1 Overview term RDV&D missions is likely to be a requirement for permanent Ground Station contact during critical RDV&D phases to Relative GPS was selected as part of the baseline sensor payload

342 Vol 71 No.9 September 2018 JBIS FUTURE RENDEZVOUS AND DOCKING MISSIONS enabled by low-cost but safety compliant Guidance Navigation and Control (GNC) architectures

Fig.35 ISL assumptions for Nominal V-Bar Approach Mode (<10km range). complement (for co-operative missions) during the trade-off phase. Relative GPS requires an ISL (Inter-Satellite Link) sys- This preliminary ISL analysis and design shows that the link tem to transmit the GPS signal from the Target to the Chaser. budget closes for all the reference options (i.e. up to 10km for A 2-way ISL is required to also transmit health/FDIR flags be- the V-Bar approach, and up to 50m fly-arounds, although the tween both the target spacecraft and chaser spacecraft. The ISL latter would need further analysis/testing to confirm). Also the is required to function from a range separation of 10km to very Power Flux Density (PFD) is okay even in worst cases (during close rendezvous approach (down to a few m). fly-arounds), with 9.7dB Margin.

A preliminary Relative GPS and ISL solution has been de- 7.3.5 Reference Video Camera veloped in this second phase using expertise from SSTL’s GNSS and RF teams. A small video camera is assumed on both the Target and Chas- er spacecraft, in order to provide the following: 7.3.4.2 Relative GPS • Visual validation of the close approach • PR (Public Relations) The required data rate for Relative GPS measurements across the ISL (Target to Chaser) is estimated to be less than 9kbits/s This has the advantage of providing simultaneous vantage at 1Hz. With the additional data rate for flags/health, a data rate points from both the Chaser and the Target, although the of ~10kbits/s would be required for the ISL. lighting conditions may be non-optimal on one or both of the spacecraft if the final approach and docking is performed in Unlike Optical sensors (which improve with reducing poor lighting or eclipse. This may be alleviated by using an ar- range) the position and velocity accuracy is generally constant tificial light from the Chaser or the Target, so long as power is with Relative GPS (so it gets less accurate in terms of % as you available and it does not affect the LIDAR or Close Proximity approach), although with single frequency GPS, the greater the Camera. This is not in the study baseline but has been investi- separation the more errors this will introduce, as a greater dis- gated briefly and shown to be potentially feasible. tance equates to less cancellation of the ionospheric effects if the receivers are separate. The reference Video Camera (Fig. 36) is the Supervision Camera on RemoveSat (i.e. SSTL’s spacecraft within the Re- Better than 1m 3D RMS position and 2-3 cm/s 3D RMS ve- move Debris mission). locity could be achieved, via implementation of a real-time relative navigation filter. This would therefore achieve better than 2% accuracy on position/range at 50m range.

No detailed investigation has been performed into the impact on GPS reception as a function of orbit latitude or the attitude of the spacecraft. Neither of these is expected to be an issue for the standard V-Bar approach, as the spacecraft are in LEO, and the GPS antenna is always zenith pointing. However there may be reduced reception for fly-arounds (if required) or R-Bar approaches, which may require an additional GPS antenna. This should be investigated in future studies.

7.3.4.3 Preliminary Intersatellite Link (ISL) Design

A preliminary low resource dual redundant S-Band ISL solution has been developed with the help of RF expertise at SSTL, building on current ISL developments on other missions (albeit for much longer ranges). This is summarized as follows: • <5W total system power • <2.3kg total system mass

Two main ISL modes have also been investigated in line with the mission analysis: • Nominal V-Bar Approach Mode (<10km range), see Fig. 35 • Additional Fly-around Modes where the Chaser is always Fig.36 RemoveSat Supervision camera showing FoV (Field-Of- pointing at the target. View).

JBIS Vol 71 No.9 September 2018 343 STEVE ECKERSLEY et al

TABLE 5 Payload Resource Budget (exc. margins) Total Mass Total Power Unit Spacecraft No. (kg) (W) SSC Proximity Chaser 1 1 2.5 Camera Processors for Chaser 2 0.5 7 Proximity Camera SSC COTS LIDAR Chaser 1 0* 2.5 Processors for Chaser 2 0.5 7 COTS LIDAR Glyph Panel (GP) 1x GP Target 0* 2-20** LED strings 2 x LEDs

Docking System Both 2 8 100***

Inspection Camera Both 1 0.65 4.1

*The mass for the LIDAR and the Glyph panel/LED’s is included in the estimate for the Fig.37 Surrey Space Centre Magnetic Docking System Concept and docking system which is much heavier Breadboard hardware. **~2W (may need to be up to 20W at 100m but non-continuous so less than 100% duty cycle) ***20-100W (TBC) for the Electromagnets during docking mode, but only for a few minutes (from 1m inwards). The baseline assumption for the Video Camera is to send basic video after docking in non real-time, and that it does not drive any operational requirements, only that there is sufficient TABLE 6 Payload Dimensions mass and power to accommodate it. The advantage of this approach is that a high power X-Band system is avoided which Unit Spacecraft No. would drive power, mass and cost. SSC Proximity 1 20x5x15cm Camera Interestingly, a preliminary investigation has shown that it Glyph Panel (GP) 1 GP2 30x30cm may be feasible to send some limited real time/near real-time LED Strings LED strings low data rate video (by reducing the frame rates and resolu- Docking System/ 1 GP 15x15x15cm tion) over the S-Band system or the BGAN system, to observe LIDAR Box the final stages of the RDV&D (<10m range) for the limited pe- Docking System 2 7.5x15x15cm riod during close approach. Note that an additional constraint Boxes here is the substantially increased power required for BGAN at Inspection Camera 1 6.3x 6.3x4.7cm high data rates, though this is only transient power case. Fur- ther work would be needed to ascertain whether these low data rate videos would be acceptable. SSC, which show the capture cone extends some 30cm from 7.3.6 Docking System Assumptions the port’s surface, with a half-cone angle of approximately 45º.

As a core docking technology and reference for the study, it is As the docking system on AAReST is for much smaller proposed to employ a derivative of the systems being developed spacecraft, a more powerful variant is required for this mis- by the Surrey Space Centre for the AAReST mission [7, 8]. This sion and some scaling has been estimated by SSC in order to is an innovative magnetic system, which exploits both fixed investigate whether the preliminary system design is feasible. A and electro-magnets to enable docking between two spacecraft. wider spacing of the ports is required to provide greater stabil- The Electro-Magnetic Kelvin Clamp Docking System, com- ity (which would be needed ultimately for a real modular tele- prises four pulse-width modulated (PWM), H-bridge-driven, scope). The ports should also form an equilateral triangle and dual polarity electro-magnets, each of over 900 A-turns. These should be centred on the spacecraft panel, to reduce the risk of are coupled to three “probe and drogue” (60° cone and 45° cup) contact between the Chaser and Target due to any residual tilt. type mechanical docking ports, arranged to form and extended area docking surface. Kinematic constraint is established using The pull in force may also not be so concentrated over such the Kelvin Clamp principle (3 spheres slotting into 3 V-grooves a wide separation, but this could be compensated with bigger arranged at 120º). (more power hungry) electro-magnets. The assumption is that the power would be up to 100W but only for a short operation The concept behind the magnetic system is to use on-board time from less than 1m inwards (which is about <100 seconds propulsion to manoeuvre the docking-satellite into the mag- at 1cm/s approach velocity). netic “capture cone” of the item being docked against at an appropriate relative velocity (< 1cm/s). By then pulsing and Additionally, for greater robustness, it is assumed that the changing the polarity of the electro-magnets on-both sides of docking system for this mission will be redundant rather a the docking interface the final docking manoeuvre can be elec- single point failure as it is on AAReST. The assumption is that tromagnetically controlled, which ‘pulls’ the spacecraft into a the redundant ports are cold redundant and would be pulled final docked position. Items can be easily undocked by revers- in and automatically connected when the Prime port also ing the polarity of the electro-magnetic. This has been simu- connects. Also with a prime system failure we could still have lated and experimentally verified on air bearing table tests at an identical 3 point mount with the redundant set. This straw-

344 Vol 71 No.9 September 2018 JBIS FUTURE RENDEZVOUS AND DOCKING MISSIONS enabled by low-cost but safety compliant Guidance Navigation and Control (GNC) architectures man docking concept would however need further investiga- TABLE 7 Summary of the baseline spacecraft parameters for tion after the study, as it has not yet been analysed properly. the reference mission design

7.3.7 Payload Resource Budgets and Dimensions Parameter Characteristics Spacecraft Heritage baseline is the SSTL-42 range with some Table 5 summarises the Sensor Resource Budgets. Strictly speak- heritage additional changes (e.g. PIU) ing, Relative GPS is also a payload but covered under the OBDH Redundancy/ Fully Redundant (extra 3rd units in critical areas) (Core-DHS GPS) and communications (ISL) subsystems. Table Reliability 6 summarises the preliminary payload dimensions. Spacecraft Hexagonal Cylinders: Dimensions • Height 1.16m (exc. launcher 7.3.8 Spacecraft parameters for the reference mission design • Diameter: 0.6m (diameter across flats) interface) Dry Mass: 134.9kg (inc. system margin) Table 7 summarises the preliminary spacecraft parameters for Max Spacecraft Launch Mass: 146.9kg (inc. 12kg propellant) the precursor reference mission design. This also shows that Mass the overall preliminary mass and worst case power budgets are c.f. 150kg assumed for mission analysis and design estimates feasible in line with the assumptions made. Several Chaser and • Chaser: Camera, COTS LIDAR, Payload Processors, Target “day-in-the-life” scenarios have been analysed with the Video Camera, Docking System worst case being a steady state mode for the Chaser at 10m, Payload • Target: Glyph/LED panel, Video Camera, Docking with the LIDAR “on”, however this may be pessimistic as we are System highly unlikely to need the LIDAR and its processor on for so long prior to approach. Xenon Cold Gas/Resistojet System providing redundant 6 Propulsion DOF control, and 100mN thrust. It can be seen from this table that the mass is in very close Isp: 48s (Warm Gas), 30s (Cold Gas) agreement with our assumption of 150kg after margins. Magnetorquers and Thrusters AOCS Actuators Reaction Wheels as backup The power budget could easily be improved by making plat- Star-trackers, GPS, Magnetometers, Sun-Sensors, form a little wider. This is not an issue because original design AOCS Sensors was arbitrary (a minimum width to accommodate 0.6m mir- Accelerometers rors) and widening platform would have no impact on original mirror configuration. Power System Body-Mounted Solar Arrays Power 82.8W on Chaser It should be noted that an optional artificial LED light (from Generated either or both spacecraft) could also be used in conjunction Orbit Average 79.1W (inc system margin) on Chaser with the Video Cameras, to provide lighting during the final Power Required approach. This is not in the current baseline, but a brief inves- • S-Band TTC to Svalbard and Troll (nominal 10kbps, max tigation of the technology and system model shows that such a 200kbps) solution seems feasible even with a required power of ~60-70 watts. This is because the required power is only transient (e.g. RF • L-Band TTC using BGAN via I4 (nominal 10kbps, max for ~1000s) rather than steady state which would not be fea- Communications 250kbps) sible. Further work would be required to investigate artificial • S-Band 2-way Intersatellite Link (ISL), (nominal lighting in greater detail. 10kbps)

7.3.9 Spacecraft Configuration Data Handling Double Fault Tolerant Core-DHS OBC

Fig. 38 show the preliminary external spacecraft design config- Payload Storage PIU from RemoveDebris urations for the Chaser and Target.

Fig.38 Image of the Chaser and Target (left) and both spacecraft after docking (right).

JBIS Vol 71 No.9 September 2018 345 STEVE ECKERSLEY et al

The dispenser attachments are currently undefined but a developments and road-mapping exercise to define the need- could be either on a side face or the +Z face for this precursor ed Development Plans for flight implementation of both the demonstration as we have no mirror/detector unit (the latter is GNC and Proximity Sensors. This included the GNC Design/ assumed in the mission analysis). Development/Verification Approach, Proximity Sensor Devel- opment/Breadboarding, Model Philosophy, and scheduling of 8 MISSION SAFETY AND ROBUSTNESS AGAINST required GNC and Proximity Sensor activities such as Integra- COLLISION tion and Testing. However this work is not detailed here in order to keep the paper concise. SSTL normally deals with missions/spacecraft where failure tolerance is addressed by simple redundancy (where possible) 10 CONCLUSIONS allowing 1 failure tolerance (i.e. single point failure tolerance), although in places it is much more tolerant. RDV&D missions This paper has provided a brief overview of this study. A are however, much more complex because they usually must baseline reference mission has been defined and developed in have faster reactions and additional failure tolerance to a sec- the study, which is a precursor “co-operative” RDV&D demon- ond failure during the RDV&D phases, i.e. the following re- strator consisting only of 2 spacecraft – a target and chaser. The quirements approach: following main topics have been analysed in detail for the base- • 1 failure tolerance to complete mission line reference mission: • 2 failure tolerance to avoid colliding with the other • Mission Analysis spacecraft • Systems • GNC simulation and modelling This means that any single unit/subsystem credible failure • Sensor Breadboarding and Testing must not prevent the mission from proceeding (implying the • Development Plans need for redundancy), whilst the second requirement means that a second credible failure (either in the same subsystem or The mission analysis included analysis of a complete end a different subsystem) should not result in a collision with the to end trajectory, including a close proximity forced approach other spacecraft (implying the need for the ability to perform a from the baseline 10 m. However a range of other approach Collision Avoidance Manouevre even after two credible failures, scenarios were analysed to reflect the fact that other applica- which in turn implies the need for extra redundancy in critical tions (e.g. non co-operative missions or inspection missions) areas, e.g. the need for a third OBC). It should also be noted may need different approaches. Other aspects such as undock- that a single failure close to the other spacecraft could also pose ing, CAM’s and lighting conditions were also investigated. a collision risk, so a rapid FDIR system is required to switch quickly to a redundant unit or perform a CAM, should passive At the heart of the study was the GNC simulation and mod- trajectory safety not provide sufficient protection. elling which looked at the final approach of the close range rendezvous phase leading to the spacecraft docking condi- Therefore, a substantial investigation into mission safety has tions. This included defining and performing the GNC mod- been carried out, with particular emphasis on designing both elling and simulation to cover the GNC design during the final spacecraft to provide a failure tolerant mission system which phase of rendezvous and docking. A suitable control scheme can avoid collisions between the Chaser and Target spacecraft. was designed and a simulator based on Matlab/Simulink has This has also included inputs from the mission analysis on tra- been built and tested, including sensor and actuator models jectory safety and CAM’s. Particularly emphasis has been on developed with regard to the existing literature and experi- the following: mental results obtained by SSC and information provided by • Redundancy strategy and allowable non-credible single SSTL engineers. point failures and allowable non-credible double failures • Preliminary FMECA and fault tree analysis (FTA) for ro- The simulator has then been used to perform various sim- bustness against collision ulations for a close range rendezvous with a closed-loop con- • Failure detection, isolation and recovery (FDIR) architec- trolled quasi-straight line trajectory on the V-bar axis from ture for the chaser and target spacecraft 10 m range. Simulation results were obtained for the final approach scenario in order to assess the suitability of the pro- Potential causes of Catastrophic Failure Modes which could posed control approach and of the GNC sensors and actuators result in collision of the chaser and target spacecraft, have been considered. Excellent positional control was demonstrated to analysed via system level FMEA/FMECA and Fault-Tree Anal- mm level and a final closing velocity less than the required 1 yses, in order to design the system to be sufficiently robust. This cm/s. Variations in the guidance profile can be made in order to analysis has been used to iterate and define the system archi- tweak the resulting terminal position and velocity if required. tecture, levels of redundancy and preliminary FDIR strategy. Fuel usage is also very good with a total ΔV of ~3 cm/s for a Particular focus has been on subsystems directly related to the 10 m starting point. The fuel usage can also be scaled to longer GNC architecture and related CONOPS (Propulsion, GNC/ distances owing to the linear consumption on the radial axis. AOCS, OBDH, Rendezvous Sensors, and TT&C). For the proximity sensor package, the passive optical sys- The combination of all these analyses in combination with tem using LEDs provides good results and could operate “solar the expected failure tolerance requirements, has been a major blind” under all lighting conditions (day or night). The LIDAR design driver with the result being a highly robust and failure performs well over short distances (up to 3m) but is known tolerant spacecraft to be blinded by sunlight, and so further mitigation is needed for this. Both have an error in positioning of 5%, although the 9 DEVELOPMENT PLANS GNC simulator shows that this is acceptable. Nevertheless, SSC are currently investigating the use of narrow-band near-IF op- In the final part of the study, SSTL and SSC jointly conducted tical interference filters and new algorithms in image process-

346 Vol 71 No.9 September 2018 JBIS FUTURE RENDEZVOUS AND DOCKING MISSIONS enabled by low-cost but safety compliant Guidance Navigation and Control (GNC) architectures ing data fusion techniques. Further work is needed to develop RDV&D and provides both parties with the core GNC/sensor the LIDAR to work to the required range of 10 m. capability to be actively involved in future missions in this area.

Whilst this is not a systems study, a very wide range systems Acknowledgements tasks related to the GNC were investigated and developed to achieve a sensible mission scenario including processing archi- The authors would also like to acknowledge Charles Mc- tectures, relative GPS/ISL development, propulsion architec- Causland (NSTP Portfolio Manager at UKSA), UKSA for their tures, collision avoidance and mission safety/robustness, com- financial support, and the following people from SSTL for their munications and high level spacecraft system design. technical support: M. Temple (OBDH and processing), D. Gib- bon (propulsion), P. Garner (ISL Communications), S. Dun- Finally, Development plans and Roadmapping exercises can/A. Palfreyman (Relative GPS), and J. Friend (Spacecraft were carried out for the GNC and Sensors. Design).

All of these parallel investigations has shown that a robust The authors would also like to thank R. Crowther, R. Blay- yet low cost GNC architecture using the sensor baseline for ber and R. Jeffreys at UKSA for their inputs and advice on UK RDV&D is feasible and can be achieved via a mission using mi- licensing for close proximity missions, and Addvalue for their crosatellites. This has developed SSTL and SSC’s capability in inputs and advice regarding their BGAN terminal.

REFERENCES 1. http://www.pellegrino.caltech.edu/aarest1/ Control, CRC Press, 1975 2. C. Saunders, D. Lobb, M. Sweeting, and Y. “Building large telescopes in 6. J. S. Ardaens, S. D’Amico, J, Sommer, “GPS Navigation System for orbit using small satellites” IAC-16.B4.2.4x34408, Guadalajara, Mexico, Challenging Close Proximity Formation-Flight”, ISSFD, 2014 26-30 September 2016 7. C. Underwood, S. Pellegrino, V.J. Lappas, C. P. Bridges, J. Baker 'Using 3. W. Fehse, Automated Rendezvous and Docking of Spacecraft, Cambridge CubeSat/micro-satellite technology to demonstrate the Autonomous University Press. 2003 Assembly of a Reconfigurable Space Telescope (AAReST)'. Acta 4. G. Arantes, L. Martins-Filho, “Guidance and Control of Position and Astronautica, 114, pp. 112-122, 2015 Attitude for Rendezvous and Dock/Berthing with a Noncooperative/ 8. C. Underwood, et. al., Autonomous Assembly of a Reconfigurable Space Target Spacecraft”, Mathematical Problems in Engineering, 2014, Telescope (AAReST) – A CubeSat/Microsatellite Based Technology 5. A. E. Bryson, Applied Optimal Control: Optimization, Estimation and Demonstrator, SSC-VI-5, 27th AIAA/USU Conference on Small Satellites

Received 22 November 2018 Approved 26 November 2018

JBIS Vol 71 No.9 September 2018 347 JBIS VOLUME 71 2018 PAGES 348-354

TERRAFORMING MARS in a climate of existential risk

KEITH MANSFIELD Founder and CEO, Herschel Publishing, Suite 801, 1 Pepys St, London, EC3N 2NU, United Kingdom. email [email protected]

Technology has led us to a point where we are close to being able to begin the colonization of the solar system and ultimately beyond; it has also created the means to generate our own extinction, just as every formerly dominant species on Earth has gone extinct before us. While that remains a significant risk, the Outer Space Treaty should be revised to allow a Mars colonization and terraforming programme designed to create a second viable biosphere for humans – an important step in securing the future survival of our species. Mindful of Robin Hanson’s “Great Filter”, this paper will argue that an aggressive terraforming programme should take priority over preserving the red planet for the long-term scientific investigation of life on Mars.

Keywords: Great filter, Outer Space Treaty, Colonization, Extraterrestrial life, Planetary protection, Bostrom, Hanson, NASA, SpaceX, Blue Origin

1 INTRODUCTION process of terraforming the red planet, changing it from a cold dry world to a warmer, wetter one, with a thicker breathable Here in the early twenty-first century we have reached a point atmosphere. where we can give serious consideration to sending humans to Mars. The question we need to ask ourselves is why. What are The practical means for embarking on such an enormous our objectives? engineering project on a planet-wide scale are beyond the scope of this paper, but we do make the point that most experts For such an epic quest it is not enough to echo the words on the matter believe for a duration of hundreds of millions of of George Mallory, “Because it’s there”. The costs are so great, years, early Mars was both warm and wet [1, 2]. that we cannot treat the red planet as a vanity project for na- tion states or even for all mankind, at the behest of politicians. The water is still there. Not only is it at the poles, starkly There is little point in simply planting the American, Chinese, at the North Pole and beneath a layer of carbon dioxide ice at Indian or even the UN flag on some red soil, say 150 million the South Pole, but it appears to be at mid-latitudes. Recent kilometres away. work shows there appear to be gigantic ice sheets close to the surface that are sometimes revealed when cliffs are eroded and What, though, of scientific curiosity? Many would think it huge glaciers are exposed, a hundred metres thick [3]. There a noble purpose for Mars to become a research base like the are estimates of more than five million cubic kilometres of wa- Antarctic stations. ter which, if all melted, would cover the entire surface of Mars to a depth of 35m [4]. The purpose of this paper is to say we need to be very clear about our goals, because a lot is at stake. What is a lot? Noth- Timescales and feasibility are certainly contentious, and ing less than the future of humanity. If our survival as a spe- will not be resolved without at least the first stage of a coloni- cies is the most important thing, then scientific questions re- zation programme with significant boots on the ground. Paul garding the existence of life on Mars can only be a secondary Birch suggests the task could be achieved in as little as half consideration. a century, with others looking at many millennia to take full effect [5]. 2 TERRAFORMING MARS Given there is such good evidence that Mars was in a warm The first thing you see upon entering the SpaceX rocket fac- and wet state for hundreds of millions of years, there are at least tory in Hawthorne, California is the image of four worlds in some grounds to believe the planet might be returned to this Fig. 1 The inspirational rendering takes the viewer through a past state. Unless we embark on such a task, Mars will remain permanently unwelcoming to human life. Nowhere in the solar system is as hospitable as Earth’s South Pole or the summit of This paper was prepared for the “Mars in the Age of New Space Mount Everest, so if we do not terraform, we could ask what is Launchers Symposium”, London, 28 February – 1 March 2018. the point of a Martian colony.

348 Vol 71 No.9 September 2018 JBIS TERRAFORMING MARS in a climate of existential risk COURTESY OF DAEIN BALLARD OF DAEIN COURTESY

Fig.1 Possible stages of transition of Mars during the terraforming process.

3 CURIOSITY AND MARTIAN WATER “States Parties to the Treaty shall pursue studies of outer space, including the moon and other celestial bod- Let us fast forward from ancient wet Mars to the present. In ies, and conduct exploration of them so as to avoid their 2011 McEwen et al. announced the discovery of “recurring harmful contamination and also adverse changes in the slope lineae” at equatorial latitudes [6]. While there is ongoing environment of the earth resulting from the introduc- disagreement over the cause of these flows down Martian cra- tion of extraterrestrial matter and, where necessary, shall ter walls and mountain sides, the initial explanation and still adopt appropriate measures for this purpose.” a strong contender is that this is evidence of salty liquid water flowing on the surface today. Other examples were soon found We are not sending Curiosity to take a look at the most scien- across the red planet (see Fig. 2). As more potential sites were tifically interesting sites within its compass for fear of contam- discovered, it was noted that Mount Sharp, situated in Gale inating a pristine planet. At first sight this might appear to be a Crater where the Mars Science Laboratory Curiosity Rover was laudable aim, but this paper will argue that in order to survive, by now exploring, showed signs of such flows. humanity will have to adopt a more “species-ist” approach, putting humanity first. Our existence as a species is precarious and For those who see scientific enquiry as our prime motiva- our establishing a new human-friendly beachhead in the solar tion for solar system exploration, such a potentially important system should be a moral imperative. discovery would surely dictate that the robotic laboratory be directed to investigate the nearby site? Instead, the existing legislation has the result of putting the survival of possible Martian microbes ahead of our own. Iron- The then US Planetary Protection Officer, Cassie Conley, ically, the rationale for the original treaty was the prevention intervened to forbid any possibility of Curiosity investigating of existential risk, with an understandable desire to prevent any potential water discoveries within the Gale Crater. In a nuclear weapons being positioned in space, yet over time it meeting called with the NASA project team, it was agreed between Conley and the Mars Science Laboratory Project Scien- tist, Ashwin Vasavada, that the Curiosity team would scrutinize images from the lander on a daily basis for signs of water, to be able to avoid them for fear of contamination [7].

A Martian probe, sent at great expense to perform scientific experiments on the red planet was not to be allowed to perform arguably the most important experiment within its capabilities for fear of contamination of this single site. If such a robotic explorer, which has been through the NASA decontamination process, is prevented from such examinations, what will the attitude of NASA and the Office of Planetary Protection be -to wards sending humans, teeming with microbes, to these sites in future?

4 THE OUTER SPACE TREATIES COURTESY OF NASA/JPL-CALTECH/UNIVERSITY OF ARIZONA OF OF NASA/JPL-CALTECH/UNIVERSITY COURTESY

It is efforts to enforce the 1967 Outer Space Treaty and its suc- cessors that prevent the most interesting science being under- taken on Mars. The fifth and last of the treaties was drawn up in 1979 and refers to the Moon. While 106 nations have ratified the original “Treaty A”, at a time of the height of the Cold War, the numbers agreeing to the successor treaties dwindled to only 16 by the release of Treaty E, none of which are space-far- ing nations.

Article IX of the original 1967 treaty contains the text [8]: Fig.2 Oblique view of warm season flows in Newton Crater.

JBIS Vol 71 No.9 September 2018 349 KEITH MANSFIELD has taken on the opposite effect of placing humanity’s future which to sustain future humans and allow them to live good in the balance. and productive lives.

If the treaty stands, it jeopardizes a colonization programme Earth can only sustain a limited human population for a rel- and effectively kills off all prospect of terraforming Mars. The atively short period (compared with the potential lifespan of consensus from space lawyers appears to be that contamina- the Universe). Over trillions of years the Universe will be able tion cannot be allowed until all the science has been completed; to sustain unimaginably large numbers of humans should we for scientists, the science will never be over. successfully manage a programme of expansion and colonization, meaning far more good can be achieved. The window is 5 EXISTENTIAL RISK currently open, but the great filter will show us that it may not remain so for long. It should therefore be a moral imperative Swedish philosopher Nick Bostrom popularized the concept to expand into space now, while we can, and not risk the astro- of “Existential Risk” [9]. It is something that threatens our nomical cost of not seizing the opportunity. very existence as a species – any event that could cause human extinction or permanently and drastically curtail humanity's 7 KARDASHEV CIVILIZATIONS potential. Taking these first steps further out places us on the pathway Humans are the dominant species here on planet Earth, but to becoming, first a Kardashev Type I, then II and ultimately so were the dinosaurs or the trilobites or the Ediacaran “mat- a Kardashev Type III civilization. Proposed by Russian astro- tresses” that appear to have been the first such species on our physicist Nikolai Kardashev in 1964 [12], this logarithmic scale planet. The stark truth is every other species that once ruled the indicates the energy usage and signatures of advanced extra- world is nowadays consigned to the fossil record, unless it has terrestrials. disappeared even from there. A Type I civilization would use all the energy available on There are many potential forms of existential risk, from a its home planet. By comparison, cosmologist Carl Sagan posi- giant asteroid or comet crashing into Earth, to nuclear Arma- tioned humans at 0.7 on this scale in the 1970s [13]. geddon, to runaway artificial intelligence, to more subtle risks such as the political disruption that will be caused by climate Type II is when a civilization is so advanced it is able to har- change in the second half of the twenty-first century when the ness all the energy available from a star, an example of which population will be over 10 billion people, but areas of our plan- would be a Dyson swarm, material surrounding a star in a cus- et will no longer be habitable and those people have got to live tomizable configuration to extract its energy. And Type III is somewhere else. Which in turn could begin the chain reaction so far advanced it uses and controls all the energy output of a that leads to war, with nuclear, biological or chemical weapons, galaxy, a simple example being perhaps in the form of billions or through AI. of Dyson swarms.

It is a straightforward argument that if we remain earth- Could such civilizations be possible in terms of what we bound as a species, extinction is a certainty. If nothing else, know of the laws of physics today? Given the answer is an we could wait for the Sun to engulf Earth in five billion years’ emphatic “yes”, this begins to point to the urgency of humans time. But given that is such a long timeframe the conclusion becoming multiplanetary. The mathematics of galactic colo- might be considered trivial or irrelevant. Oxford philosopher nization was created by John Von Neumann (and published Toby Ord makes a quantitative argument that the next two- posthumously) as part of his work on cellular automata [14]. to-three centuries will prove the most dangerous epoch for It shows that even our own Milky Way galaxy, a hundred thou- our species [10]. The claim is that there will be a significant sand light years across, can easily be colonized in a million per-century risk of extinction because we have created the years or so. The term “von Neumann probe” has entered the means to destroy ourselves without the mechanisms to pre- scientific consciousness around this idea, and readers might vent this happening. Long term, such a high level of risk is think these have to be self-replicating automatons. However, unsustainable so either we find a way of reducing it or we shall the name is but a mathematical shorthand and it does not mat- have failed and humanity’s race will be done. ter if we are talking about self-replicating probes or if we mean organic creatures, such as human or alien colonists who arrive The timescales for terraforming and for when existential in a new star system and lay down roots, before building more risk will be greatest make the argument for the urgent need to ships that will go on to spread to neighbouring systems. The colonize the solar system. The sooner we begin a terraforming mathematics is the same. process, the more valuable it will prove as an insurance policy. By comparison, the Milky Way is very old: 10 billion years 6 ASTRONOMICAL WASTE old. Ours is a youngish solar system, Earthlike worlds will have existed in our galaxy for billions of years before ours and The short-term goal might be to colonize the red planet, but yet it only takes a million years for life to spread through the one reason for that is it will be a stepping stone, our first of whole galaxy. many, to becoming fully multiplanetary, ultimately having humanity’s descendants becoming intergalactic civilizations. Where is everyone?

The Universe is expanding. Every second, an average of But what is so very important to this discussion is that it twenty thousand stars pass beyond human influence, even were is not just about our galaxy. Indeed, sometimes it is harder to we capable of travelling at the speed of light [11]. The Hubble observe the Milky Way as effectively as other galaxies because Sphere that humanity can influence will continue to diminish we are inside it, but if there were civilizations operating on a and it is only within this that we can acquire the resources with galactic scale elsewhere in the Universe, we should be able to

350 Vol 71 No.9 September 2018 JBIS TERRAFORMING MARS in a climate of existential risk detect some of their engineering projects today. and makes no assumption about advanced extraterrestrials being able to communicate with us. Instead the assumption is that 8 THE GREAT FILTER if there were intergalactic civilizations we would see them. For instance, we would witness the products of their galactic-wide The Great Filter is a concept devised by polymath Robin Han- engineering projects. Because we see nothing, we can say: son, who argues that every single thing we observe beyond Earth in the entire observable Universe appears explicable by N × pK 1 inert processes. This is in stark contrast to what we find on Earth, where life appears to adapt to and occupy every conceiv- where the symbol means “not much greater than”. Maybe able niche [15]. there are 5 or 20 or zero or a thousand, but because N is so incredibly large, one thing we can definitely say is that: The idea is that there is a Great Filter – or filters – that always prevents life expanding to fill every nook and cranny out there. pK ≈ 0 But is this really the case or is it that our powers of observation are limited? (or pK is very close to zero, effectively vanishingly small). But we can split pK into two factors, the probability of a civilization In 2015, Jason Wright’s team at Penn State used the Wide- developing to the equivalent of our current level of technology Field Infrared Survey Explorer (or WISE) telescope data to an- and sophistication here on Earth, which we will call pE, and alyse 100,000 nearby large galaxies in search of systematic use the probability of a civilization at that level being able to go on of Dyson swarms [16]. Discussing the results he concludes: and form an intergalactic civilization, which we will call pIG. So we have: “On Kardashev’s scale, a type 3 civilization uses energy equal to all the starlight produced by one galaxy pE × pIG ≈ 0 … We looked at the nearest, largest 100,000 galaxies we could find in the WISE catalogue and we never saw which tells us that either pE or pIG or both are vanishingly close that. One hundred thousand galaxies and not one had to zero. that signature. We didn’t find any type 3s in our sample because there aren’t any.” [17] Let us think about what that means for a moment. Either the Great Filter, that thing which prevents intergalactic civili- Some might think it is just the case that extraterrestrials have zations from occurring, is in our future. Or we are the luckiest, become very good at covering their tracks and they do not have most successful civilization in the Universe and have overcome what we call “waste heat” for us to detect. And of course they it without realizing. might be right, but only if the laws of physics and especially something as fundamental as conservation of energy does not 10 IS THERE LIFE ON MARS? apply. If the laws of physics hold we would be able to see the infrared signatures. What would be the consequences for the Great Filter is we discovered life on Mars? Nick Bostrom writes: Scott’s team dug deep into these galaxies, and are continuing to lower the limit to the point it appears we will soon be ruling “I hope that our Mars probes will discover nothing. out Type II civilizations. Michael Garrett at Leiden has done It would be good news if we find Mars to be completely similar work and come to the same conclusion [18]. sterile. Dead rocks and lifeless sands would lift my spirit.

9 MATHEMATICS OF THE GREAT FILTER “Conversely, if we discovered traces of some simple extinct life form—some bacteria, some algae—it would The Great Filter is based on the ideas of Bayesian probability, be bad news. If we found fossils of something more whereby we use observational evidence to refine our probabili- advanced, perhaps something looking like the remnants ty estimates. And because we observe no evidence of advanced of a trilobite or even the skeleton of a small mammal, intelligent extraterrestrial civilizations anywhere in the Uni- it would be very bad news. The more complex the life verse, Bayesians infer that the probability of a civilization going we found, the more depressing the news of its existence on to colonize the universe, going on to be Kardashev Type II would be. Scientifically interesting, certainly, but a bad or III, must be vanishingly small. omen for the future of the human race.” [20]

The mathematical formulation is implicit in Hanson’s origi- Many people including scientists would be thrilled to find, nal paper [15], and is spelt out by Häggström [19]. say fossilized remains of large creatures on Mars, but if complex life can form on two (or more – we do not know about Let N be the number of planets in the entire observable Uni- other places yet) different worlds just in our solar system then verse and pK the probability that one of those chosen at ran- it seems very unlikely that pE, the probability of a society an- dom will go on to develop an intergalactic civilization. Then ywhere in the universe reaching our level of advancement, is we have: vanishingly small. Suggesting it is what happens from hereon that is the problem. But if the Great Filter is in our future, there N × pK = the number of intergalactic civilizations in the is a further argument that it has to catch up with us soon. observable universe Once a society is established on two or more different plan- For those familiar with traditional SETI, it is important to etary bodies the immediate risk of extinction reduces dramat- bear in mind that this is not a number like in the Drake equa- ically. For instance, it is not just that an asteroid might bash tion. This number covers the entire Universe (not the galaxy) into one of those bodies but not the other, but in developing

JBIS Vol 71 No.9 September 2018 351 KEITH MANSFIELD the capability to colonize other worlds we also develop the ca- Option 3 There is no life on Mars today, but there is evidence of pability to prevent asteroid impacts. A nuclear war might wipe complex multicellular life in the planet’s past. humanity from the face of the Earth, but if there were humans living on Mars or in the asteroid belt, or on the Jovian or Sat- This would be positively alarming in terms of the Great Fil- urnian moons, then we would go on, hopefully having learnt a ter, as Bostrom points out. The fact that complex life can devel- salutary lesson. op, only for a whole planet to go extinct, should cause us to re- double our efforts to ensure we safeguard the life that survives The same goes for biological warfare, or some terrible pan- on Earth. Again we should aggressively terraform to mitigate demic that we cannot control. Two looks not just better than against the filter(s) that lie in our future. one but a lot better. If there were two independent biospheres in our own solar system then the size of the pIG term, the prob- Option 4 Microbial life exists on Mars in the here and now, per- ability of us going on from here to become an intergalactic civ- haps being very difficult to detect. ilization, is not doubled – it is increased by orders of magnitude because of the very significant reduction in our existential risk. To safeguard the future of our species, we have to be reso- And also because of the platform it gives us to extend beyond. lute, but acknowledge this will be the most controversial of our assertions. This is the reason why we should have an aggressive policy of terraforming Mars to begin the process as soon as we are able. We suggest that if life is so common in the Universe that there are two neighbouring biospheres just within our solar Many will believe that even if we get to Mars there is still an system, then we should be concerned the Great Filter remains awfully long way to go, but the pace at which we are moving ahead of us. We should still have no qualms about initiating a towards becoming a spacefaring species means we shall soon programme of aggressive terraforming. Martian life will evolve be able to cover a great deal of ground. As an example we only and some can be preserved in the equivalent of “national parks” have to consider the Breakthrough Starshot initiative, planning as well as laboratories. Given the high levels of radiation at the to send probes to arrive at Alpha Centauri on decadal times- Martian surface, it is considered a near certainty that if life ex- cales [21]. ists today it will be below ground and in caves so relatively easy to isolate. Alpha Centauri is in our stellar backyard, but 2013 work by Stuart Armstrong and Anders Sandberg shows how in as little A major advantage of this stance is that it does not require as two centuries humans may be able to begin the early process decades or even centuries of painstaking scientific evaluation of sending probes throughout the visible Universe [22]. And to determine the answer. We go right ahead while the scientists they argue that any Kardashev Stage II civilization would have among the colonists can and would continue their research far the ability to do that routinely. more effectively than could ever be achieved by either remote rovers or a minute number of NASA explorers. If the Great Filter lies in our future, it may be running out of time to take effect because we are close to becoming a whole lot There is a further possibility, albeit remote. more secure instead of metaphorically keeping all our eggs in one basket. The natural means of engineering our own extinc- Option 5 Complex multicellular life exists on Mars today. tion pale in comparison to the ways we have devised ourselves, yet our science and technology can also be the means to our Already this is highly unlikely, given current conditions on salvation. Now is the crucial time for the future of humanity the red planet and our exploration to date. It is reasonable to where we must make the right decisions. assume this will be far easier to determine than any of the other four options. In this case we shall be very happy to leave With that in mind, we present a decision tree to show how Mars to the Martians, at the same time noting that it makes it and why we should act regarding the red planet. extremely likely there’s one or more filters ahead of us so we should turn our attention immediately to safeguarding human- 11 THE LIFE ON MARS DECISION TREE ity’s future elsewhere in the solar system.

In relation to the Great Filter and what our priorities should 12 THE BATTLEGROUND TO COME be regarding Mars, human colonization and terraforming, we propose that there are five basic scenarios to consider. Led by the commercialization of space technology with com- panies such as SpaceX, Bigelow and Blue Origin as examples, Option 1 There is not and never has been life on Mars, with the things are coming to a head. planet apparently always sterile. Many, perhaps most, scientists appear happy to treat Mars We suggest that Mars is humanity’s to develop to do what and the rest of the solar system like Antarctica, with a hand- we want with it, terraforming rapidly in order to safeguard the ful of privileged human explorers following on from robotic future of our species and possibly with it the future of intelli- probes to observe but not interfere. To study. To answer the gence in the Universe. great question of life on other solar system worlds. This appears to be a view held by governments and their space agen- Option 2 There is no life on Mars today, but there is evidence of cies, backed up by space lawyers. past microbial life. NASA’s Space Launch System will be able to carry a whole We argue that this should alert us to the possibility of the half dozen astronauts to Mars at a time on a glorified science Great Filter being in our future. We can study the extinct mi- fieldtrip, and every launch will cost billions of dollars (re- crobes as we colonize and aggressively terraform the planet. stricting the programme to no more than two a year) [23],

352 Vol 71 No.9 September 2018 JBIS TERRAFORMING MARS in a climate of existential risk COURTESY OF SPACEX COURTESY

Fig.3 Twin landing of the inaugural Falcon Heavy stage 1 boosters, to demonstrate reuse capability. let alone the development costs, which continue to spiral. actors and the power-output this will provide will also enable Spaceflight is a risky proposition and, by their nature, gov- a much faster colonization programme. Because of that, gov- ernmental organizations are risk-averse, placing the empha- ernments have to buy into this, yet the question of colonization sis on tried and tested technologies above innovation, and on remains far from the mainstream agenda. avoiding at all costs the public relations disaster of leaving astronauts dead or stranded on the red planet. Any mission Acting NASA Planetary Protection Officer Lisa Pratt states: will bring the humans back to Earth as soon as is not regarded as inappropriately fast. “As a federal agency, we don’t talk about Mars exploration in terms of colonization. NASA’s goal is human Let us consider SpaceX as an example of the commercial exploration for scientific discovery. There was a NASA view. The company’s entire raison d’etre is to build a self-sus- hosted workshop a few years ago that talks about land- taining human colony on Mars to mitigate against existential ing zones but not terraforming.” [25] risk. This is why the terraforming of Mars image is there in the entrance to the Hawthorne factory (Fig. 1). It is why the launch Once upon a time what NASA proclaimed would hold sway, vehicle under development (known as the Starship) will take but the balance of power may be shifting. The richest human 100 astronauts at a time with reuse built-in, and the hoped for alive, Jeff Bezos (the founder of Amazon) is entering the race to cost to be $250,000 a ticket to encourage settlers for the new Mars with his company Blue Origin as a rival to SpaceX. And frontier [24]. history suggests that rivalry is good as there is some evidence it drives innovation and lowers costs [26]. Given NASA’s SLS development costs, can SpaceX deliver its promises on a far smaller budget? Currently the world’s most The private and public space powers appear to have mutual- powerful launch vehicle is the SpaceX Falcon Heavy which cost ly exclusive aims: human colonization on the one hand; scien- ~$500m to develop and, crucially, has small launch costs due to tific discovery for its own sake (and in which humans cannot the great innovation of reuse (Fig. 3). The first SpaceX Starships interfere) on the other. At this crucial stage in humanity’s de- are due to begin “hop” landing tests in 2019. Even if NASA velopment, we have big decisions to make. reaches Mars first which is becoming increasingly doubtful, in a matter of decades the number of humans they are able to send We ask, from an unashamedly species-ist perspective, what will likely be dwarfed by those travelling via SpaceX (and other is the point of life on Mars or elsewhere in the universe if we commercial) rockets. cannot appreciate it? And what if, as Robin Hanson suggests, the rest of the Universe really is dead and we are its only chance? While it is inevitable for the first colonists to spend the rest of their lives indoors or in spacesuits, those people will not 13 CONCLUSION want their far descendants to be destined to do the same. And what if the technology fails? What if civilization on Earth col- The Outer Space Treaty was written at a time when only na- lapses? Only if Mars has a breathable atmosphere can we count tions could conceivably send humans into space and the idea it as making humanity significantly safer. of colonizing other worlds was a distant dream. It came about to help prevent existential risk, yet its opposition to terraform- NASA holds a trump card, also governed by the Outer Space ing is now creating further such risks. It must be revised as a Treaties: nuclear power. It would be difficult to power an ear- matter of urgency, if only to account for crewed exploration of ly Mars civilization solely with solar energy because there are Mars, let alone the solar system. This revision process should year-long planet-wide dust-storms. We should send nuclear re- take into account the new astrobiological knowledge that we

JBIS Vol 71 No.9 September 2018 353 KEITH MANSFIELD have gained since 1967. The Great Filter tells us that, at best, Acknowledgements intelligent life is the rarest and most precious of things. It helps The author wishes to thank the British Interplanetary Society justify the argument that this is a most dangerous time for the for the opportunity to present these views at the 2018 “Mars future of humanity and perhaps even the future of intelligence in the Age of New Space Launchers Symposium”, and also in the Universe. Therefore it is becoming urgent to confront thanks the anonymous reviewers for their helpful comments. these issues and recognize that the essential task of our time is Then, as a scientific publisher, the author expresses thanks to to preserve the future of the intelligent life that we know about, his own authors, whose writing and interactions have helped even if this is at the expense of the scientific quest to discover shape the views in this paper, while acknowledging all errors life that is, as yet, unknown. are his own.

REFERENCES 1. R.A. Craddock and A.D. Howard, “The case for rainfall on a warm, wet Extraterrestrials, Oxford University Press, Oxford, p. 148. early Mars”, J. Geophys. Res., 107, 5111, doi:10.1029/2001JE0011505, 14. J. Von Neumann (ed. A.W. Burks), Theory of Self-Reproducing 2002. Automata, University of Illinois Press, Urbana and London, 1966. 2. J.M. Davis, M. Balme, P.M. Grindrod, R.M.E. Williams and S. Gupta, 15. R. Hanson, “The Great Filter and are we almost past it?”, 1998, http:// “Extensive Noachian fluvial systems in Arabia Terra: Implications for mason.gmu.edu/~rhanson/greatfilter.html [Last accessed 25 April early Martian climate”, Geology, 44(10), pp. 847–850, doi: 10.1130/ 2018]. G38247.1, 2016. 16. R.L. Griffith, J.T. Wright, J. Maldonado, M.S. Povich, S. SigurĐsson and 3. C.M. Dundas, A.M. Bramson, L. Ojha, J.J. Wray, M.T. Mellon, S. Byrne, B. Mullan, “The Ĝ Infrared Search For Extraterrestrial Civilizations A.S. McEwen, N.E. Putzig, D. Viola, S. Sutton, E. Clark and J.W. Holt, With Large Energy Supplies. III The Reddest Extended Sources In Wise”, “Exposed subsurface ice sheets in the Martian mid-latitudes”, Science, Astrophys J. Suppl. Ser., 217(25), 34pp doi:10.1088/0067-0049/217/2/25, 359(6372), pp. 199–201, doi: 10.1126/science.aao1619, 2018. 2015. 4. P.R. Christensen, “Water at the Poles and in Permafrost Regions 17. L. Billings, “Alien Civilizations Absent from 100,000 Nearby Galaxies”, of Mars”, Geo. Sci. Elements, 2(3), 151–155, doi: 10.2113/ Scientific American, 17 April 2015, https://www.scientificamerican. gselements.2.3.151, 2006. com/article/alien-supercivilizations-absent-from-100-000-nearby- 5. P. Birch, “Terraforming Mars Quickly”, JBIS, 45, 331–40, 1992. galaxies/ [Last accessed 25 April 2018]. 6. A.S. McEwen, L. Ojha, C.M. Dundas, S.S. Mattson, S. Byrne, J.J. Wray, 18. M.A. Garrett, “Application of the mid-IR radio correlation to the S.C. Cull, S.L. Murchie, N. Thomas and V.C. Gulick, “Seasonal Flows on Ĝ sample and the search for advanced extraterrestrial civilisations”, Warm Martian Slopes”, Science 333(6043), pp. 740–743, doi: 10.1126/ Astronomy & Astrophysics, 581(L5), 6pp., doi: 10.1051/0004- science.1204816, 2011. 6361/201526687, 2015. 7. K. Carey, “Martians might be real. That makes Mars exploration 19. O. Häggström, Here Be Dragons: Science, Technology and the Future of way more complicated”, Wired, 7 August 2016, https://www.wired. Humanity, Oxford University Press, Oxford, 2016. com/2016/08/shouldnt-go-mars-might-decimate-martians/ [Last 20. N. Bostrom, “Where are they? Why I hope the search for extraterrestrial accessed 25 April 2018]. life finds nothing.” MIT Technology Review, May/June 2008, pp. 72–77. 8. Treaty on Principles Governing the Activities of States in the 21. Breakthrough Starshot, https://breakthroughinitiatives.org/initiative/3 Exploration and Use of Outer Space, including the Moon and Other [Last accessed 25 April 2018]. Celestial Bodies, United Nations Office for Disarmament Affairs, New York, 1967. http://disarmament.un.org/treaties/t/outer_space [Last 22. S. Armstrong and A. Sandberg, “Eternity in six hours: Intergalactic accessed 25 April 2018]. spreading of intelligent life and sharpening of the Fermi paradox”, Acta Astronautica, 89, 1–13, doi: 10.1016/j.actaastro.2013.04.002, 2015. 9. N. Bostrom, “Existential risks: Analyzing human extinction scenarios and related hazards”, J. Evolution & Tech., 9(1), 37pp., 2002. 23. E. Berger, “NASA is trying to make the Space Launch System Rocket more Affordable”, Ars Technica, 15 December 2017, https://arstechnica. 10. T. Ord, The Precipice, Hachette, London, in preparation. com/science/2017/12/nasa-is-trying-to-make-the-space-launch-system- 11. E. Siegel, “The Universe is disappearing and there’s nothing we can rocket-more-affordable/ [Last accessed 27 April 2018]. do to stop it”, Forbes, 17 August 2018, https://www.forbes.com/sites/ 24. K. Mansfield, The Real Martian: Why and How Elon Musk Plans to startswithabang/2018/08/17/the-universe-is-disappearing-and-theres- Colonize Mars, Herschel, London, in preparation. nothing-we-can-do-to-stop-it/ [Last accessed 26 November 2018] 25. L. M. Pratt, Planetary Protection Officer, NASA, private 12. N. Kardashev, “Transmission of Information by Extraterrestrial communication, 27 February 2018. Civilizations”, Soviet Astronomy, 8, p. 217, 1964. 26. S.J. Nickell, “Competition and Corporate Performance”, J. Political 13. G. Basalla, Civilized Life in the Universe: Scientists on Intelligent Economy, 104(4), pp. 724–746, doi: 10.1086/262040, 1996.

Received 27 April 2018 Approved 25 August 2018

354 Vol 71 No.9 September 2018 JBIS JBIS Vol 71 No.9 September 2018 355 356 Vol 71 No.9 September 2018 JBIS DIARY FORTHCOMING LECTURES & MEETINGS OF THE BIS

THE TOOLS OF APOLLO 22 January 2019, 7.00pm VENUE: BIS, 27/29 South Lambeth Road, London, SW8 1SZ Mark Yates looks at three artefacts from the Apollo programme, each with a fascinating story behind them. APOLLO MISSIONS: THE MECHANICS OF RENDEZVOUS & DOCKING BY DAVID BAKER 20 February 2019, 7.00pm VENUE: BIS, 27/29 South Lambeth Road, London, SW8 1SZ Starting with Apollo 9 launched on 3 March 1969, a key feature of the Apollo missions was the ability to rendezvous and dock in orbit – a capability that NASA had evolved over the preceding four years. SpaceFlight Editor David Baker describes the process in detail and casts an expert eye over the different options considered by mission planners in the run-up to the lunar landing missions. APOLLO 9 – TESTING THE LUNAR MODULE 6 March 2019, 7.00pm VENUE: BIS, 27/29 South Lambeth Road, London, SW8 1SZ Jerry Stone continues his series of talks to celebrate the 50th anniversary of the Apollo missions with a uniquely personal take on the story of Apollo 9 – the first test of the full lunar landing package and only the second outing of the Lunar Module. WEST MIDLANDS BRANCH: A NEW SPACE RACE? & PROJECT CHEVALINE 16 March 2019, 1.45pm VENUE: BIS, 27/29 South Lambeth Road, London SW8 1SZ Gurbir Singh posits the beginning of a new space race between India and China, while John Harlow and Paul Jackman look back to the days of Project Chevaline and the famed Twin Chamber Propulsion Unit. ARTISTS IN SPACE: THE EARLY YEARS 3 April 2019, 7.00pm VENUE: BIS, 27/29 South Lambeth Road, London SW8 1SZ David A. Hardy FBIS, the “longest established astronomical artist”, uses art from Lucian Rudaux, Chesley Bonestell and our own R.A.Smith, plus other ‘lesser-known’ artists (and of course his own!) to trace the genre of space art from its inception in 1874. APOLLO 10 – DRESS REHEARSAL FOR THE MOON LANDING 22 May 2019, 7.00pm VENUE: BIS, 27/29 South Lambeth Road, London SW8 1SZ Jerry Stone continues his coverage of Apollo with the first flight to carry both the Apollo spacecraft and the Lunar Module on a full dress rehearsal of a landing. Call for Papers RUSSIAN-SINO FORUM 1-2 June 2019, 9.30 am to 5pm (tbc) VENUE: BIS, 27/29 South Lambeth Road, London SW8 1SZ The BIS has now scheduled its 39th annual Russian-Sino Forum – one of the most popular and longest running events in the Society's history. Papers are invited. Watch this space for further details. APOLLO MISSIONS: LANDING ON THE MOON BY DAVID BAKER 12 June 2019, 7.00pm VENUE: BIS, 27/29 South Lambeth Road, London, SW8 1SZ SpaceFlight's editor looks at the systems evolved by NASA for calculating optimum lunar landing trajectories, and at the descent procedures needed to achieve the maximum chance of success while preserving emergency abort and safety considerations. Journal of the British Interplanetary Society

VOLUME 71 NO.9 SEPTEMBER 2018

www.bis-space.com

ISSN 0007-084X PUBLICATION DATE: 10 JANUARY 2019