Investigating a Demand Access Scheduling Paradigm for NASA's
Total Page:16
File Type:pdf, Size:1020Kb
Investigating a Demand Access Scheduling Paradigm for NASA’s Deep Space Network Timothy M. Hackett and Sven G. Bilen´ Mark D. Johnston School of Electrical Engineering Jet Propulsion Laboratory/ and Computer Science California Institute of Technology The Pennsylvania State University 4800 Oak Grove Drive 101 Electrical Engineering East Pasadena, California 91109 University Park, Pennsylvania 16802 Abstract view of a spacecraft at any given time between the three main ground complexes (Imbriale 2002). The Morehead an- NASA’s Deep Space Network supports the communica- tenna has been performing technology demonstrations in tions to and from spacecraft, rovers, and landers across our solar system and beyond. The weekly tracking re- support of becoming an operational capability for future quirements for these spacecraft are scheduled by mis- deep-space SmallSat missions (Martinez 2018). sion representatives at least eight weeks before the start Currently, the DSN supports a few dozen missions, which of the track through a combination of automated al- are scheduled through a process that starts roughly four gorithms and peer-to-peer negotiations. This process months prior to the start of the track start time. This process has worked well for traditional users with determinis- starts with mission representatives submitting their track- tic science collections, such as those doing mapping ing requirements into the Service Scheduling Software (S3). and imaging. But, this process will not scale well to These tracking requirements include constraints and pref- accommodate a new class of users whose data collec- tions are not completely predictable (i.e., event-driven erences, such as the number of tracks during the schedule science), which includes both traditional, determinis- week, the duration of these tracks, and timing and depen- tic users with unexpected discoveries as well as au- dence between tracks. The DSN Scheduling Engine (DSE) tonomous users exploring and monitoring for certain compiles all of the individual mission requests into one mas- events, such as solar flares. In this paper, we propose ter schedule and then systematically deconflicts the sched- a “demand access” scheduling approach in which the ule. The level at which the schedule can be deconflicted spacecraft, rovers, and landers themselves request track is dependent on both the flexibility of the submitted re- time on the network using a beacon-tone system and are quirements as well as the trajectories of the spacecraft. The scheduled track time “on-the-fly” using pre-scheduled schedule is then further deconflicted by a human scheduler shared-user block tracks. We show through simulation named the “Builder of Proposal” (BOP), who uses her ex- that this demand-access approach can both decrease the mean duration between the time of data collection to pert knowledge on DSN scheduling and the context of the the start of the downlink and the number of tracks re- various missions and their current mission phases. This pro- quired compared to the traditional scheduling method cess generally results in about 10–20 remaining conflicts for an example mission concept of autonomous Small- that need to then be resolved using a peer-to-peer negotiation Sat explorers at near-Earth asteroids. We also show how process facilitated in S3. Once the remaining conflicts have this demand-access approach can be used in combina- been negotiated and resolved, the master schedule is base- tion with the traditional scheduling method to support lined. For a typical schedule week, the schedule is baselined legacy users. eight weeks or more ahead of the start of the schedule week. This process is very labor and time intensive and requires Introduction a couple dozen full-time mission representatives working on multiple schedule weeks in parallel (Carruth et al. 2010; NASA’s Deep Space Network (DSN) supports the commu- Pinover, Johnston, and Lee 2017). nications and tracking for spacecraft, rovers, and landers This scheduling process has worked well so far partially across our solar system and beyond for a variety of space due to the relatively low number of missions supported by agencies around the world. The network consists of three the DSN at any given time. This process will not scale main ground complexes in Goldstone, CA, USA; Madrid, well to support future mission concepts deploying groups Spain; and Canberra, Australia with an experimental sta- of spacecraft as secondary payloads, such as the SmallSats tion in Morehead, KY, USA. Each main complex consists on Exploration Mission 1 (EM-1). Furthermore, this pro- of one 70-m and three or four 34-m parabolic dish antennas; cess has worked well so far because currently supported the Morehead antenna has one 21-m antenna (Nelson 2018; missions generally have deterministic data collection pro- Malphrus et al. 2018). At least one of the complexes is in cesses, such as those doing routine mapping and imaging. Copyright c 2019, California Institute of Technology. Govern- The amount of data collected in a time period is entirely ment sponsorship acknowledged. predictable or constrained by mission representatives on the ground, so the tracking requirements can be pre-scheduled craft to worsen. An urgent tone indicates a spacecraft emer- months in advance. This does not allow for last-minute flex- gency for which ground intervention is required because the ibility in the case of unexpected discoveries. For example, spacecraft could not recover on its own. Lastly, if no tone is when images from OSIRIS-REx revealed dust plumes emi- received from the spacecraft, this could indicate many differ- nating from Bennu’s surface in March 2019 (Lauretta et al. ent scenarios including the spacecraft antenna is not point- 2019), the tracking requirements for the following weeks ing towards Earth or an anomaly is stopping the beacon from could not be easily adapted based on this discovery as they being transmitted (DeCoste et al. 2004). were already set into the master schedule weeks before the Using the beacon tone monitoring service instead of discovery. For missions based on event-driven science, such scheduling traditional telemetry tracks can reduce mission as heliophysics missions monitoring for major solar activity, risk, mission cost, and network loading. Because of cost and pre-scheduling tracks weeks ahead of time could result in a scheduling constraints, missions are limited to the amount of number of those scheduled tracks to yield no useful, new in- track time allotted per week. This service could allow mis- formation as no new activity had occurred. The wasted DSN sions to have more timely health status reports because the resources could have been used to support other missions spacecraft will notify the ground immediately that there is instead. an anomaly, as opposed to finding out during the next sched- The science community has expressed a need for on- uled telemetry track. Mission cost is decreased by reducing demand communications, where spacecraft and missions track time and reducing telemetry data, which corresponds can request communications and tracking time in “near real- to staffing fewer operations analysts. Network loading will time”. Providing this ability enables a new class of science be decreased by spacecraft using fewer tracks than in a typ- missions: event-driven science (Shaw et al. 2018). The tra- ical week (DeCoste et al. 2004). ditional static pre-scheduling method described above does With the validation of the beacon monitoring service, the not suit these types of missions well due to its lack of New Horizons mission to Pluto baselined it into their com- adaptability and latency to changing events. In response munications design. The mission estimated that, by using the to this need, on-demand concepts termed “demand access” beacon system, they would be able to decrease their 70-m scheduling and “user-initiated services” (UIS) have been antenna time by approximately 2274 hours over its mission proposed (Johnston and Wyatt 2017; Shaw et al. 2018; lifetime (Bowman et al. 2004) for two reasons. First, for the Israel et al. 2018). The main idea of these proposed con- data rates required, the 70-m antenna typically would be re- cepts is that the spacecraft itself directly requests time on quired for all telemetry tracks beyond Jupiter (about 4 AU) the network at the moment when it needs tracking time, from Earth. The beacon service allowed the 34-m antenna rather than mission representatives requesting time on be- to be used until 25 AU. Second, beacon tracks take signifi- half of the spacecraft in advance. For spacecraft relatively cantly less time than typical telemetry tracks (1.5 hours vs. close to Earth (i.e., LEO and MEO users), spacecraft could 8 hours). In 2016, it was estimated that New Horizons was use a low-data rate, high availability link to request time able to save about 80% of its track time using the beacon on the network through a packetized handshake protocol. service (Wyatt et al. 2016). This could be accomplished by leveraging the Demand Ac- In addition to previous work using the beacon-tone ser- cess Service (DAS) using the multiple access antennas on vice for signaling the spacecraft’s status, it will now be used NASA’s Tracking and Data Relay Satellite System (TDRSS) to indicate when the spacecraft has new data to downlink. A fleet (Gitlin and Horne 2012). For deep space spacecraft, spacecraft will use the beacon-tone service to indicate that it the narrow beamwidths of the larger aperture antennas re- will be using a followup track to downlink its new science quired to close links along with the long round-trip light data. When the ground receives the beacon, it will sched- times makes this impractical. For these types of missions, ule the spacecraft into the next reserved-for-demand-access, a “beacon-mode” contact mechanism has been proposed, in unallocated track, if possible (Johnston and Wyatt 2017).