CNES OPERATIONAL FEEDBACKS IN COLLISION AVOIDANCE FOR LEO SATELLITES

Grégory BEAUMET

CNES DCT/OP/MO, 18 av. Edouard Belin, 31401 Toulouse cedex 9, France Phone: +33-5-61-27-43-91 Email: [email protected]

ABSTRACT

Because of the ever-increasing number of orbital debris, the possibility of a satellite collision with a space debris or another satellite is becoming more and more likely. This phenomenon particularly concerns the LEO altitude regime, the most frequently used region in space, where the environment has reached a point in which collisions will become the dominant debris generating mechanism in the future. Therefore CNES performs collision risk monitoring for the 15 LEO satellites it controls. Most of the operational Collision Avoidance activities are gathered at OCC (Orbit Computation Center) in its Operations Sub-directorate.

First this paper briefly presents the current CNES Collision Avoidance Process which includes OCC contingency operational procedure for collision risk management, procedure which is divided in five stages, each successive stage representing an escalation in the contingency from collision risk screening to dangerous conjunctions fine analysis and risk mitigation. The different criteria, actors, types of avoidance manoeuvres will be presented with the information flow settled between different actors. The analysis of known collisions will be presented in order to experiment the efficiency of the process.

Then this paper tries to give feedbacks of two years of operational Collision Avoidance assessment for 15 LEO satellites in order to underline the preference actions which should be taken to improve global efficiency of the process. Therefore, key figures for the operational activities are given: number of alerts, number of risks, number of avoidance manoeuvres, etc... The evolution of the operational process within the past two-year experience is presented, especially the need for more and more automation. In order to give an idea of the variety of situations which need to be handled in terms of Collision Avoidance, first a real typical case example is introduced. Then the case of a close approach between two controlled satellites is discussed and the differences are illustrated with real cases encountered. Finally the case of a close approach between an operational satellite JASON 1 with a non-operational one TOPEX which was (and still is) on a close orbit for mission purpose is presented with its operational mitigation of close encountered close approaches.

1. INTRODUCTION

There are eight recorded collisions and four of which with catalogued space debris. In December 1991, the Russian satellite COSMOS 1934 and a COSMOS 926 debris collided. Five years later, on July 24 th 1996, the French satellite CERISE was hit by a debris. On January 17 th 2005 there was a collision between THOR BURNER and a debris from Long March. And recently, on February 10 th 2009, an operational satellite IRIDIUM 33 and a non operational one COSMOS 2251 collided in orbit.

One difficulty of the collision risk management is the limited number of tracked objects: only 13,000 out of 100,000 potentially dangerous debris. Another difficulty is the unavailability of the accuracy of the publicly available data needed to properly monitor collision risks. CNES uses different data sources: • Space-Track database: the main Two Lines Elements (TLE) database. Due to its breadth and update frequency, it has been adopted as the primary source for the screening phase. Nevertheless, its accuracy and orbit propagation models are not fully available. Therefore, complex estimations are required to estimate and improve consistency.

• GRAVES database: the French Air Force Space Surveillance System (built by ONERA).

• Specific orbital data : radar measurements from French Defence facilities, FGAN radar and CNES Guyana Launch Center. These sources produce very accurate measurements but they are only used to improve the orbit determination of dangerous objects due to their limited availability.

2. CURRENT CNES COLLISION AVOIDANCE PROCESS

2.1 OCC contingency operational procedure for collision risk management

The CNES Collision Avoidance (CA) process is operational since July 2007 on all LEO satellites controlled by CNES. In order to be capable of handling a collision risk whenever it occurs, on-call teams in each speciality are in place. Two on-call teams of flight dynamics engineers are available 24/7: one in charge of CA analyses (CA team) and one in charge of the station-keeping of the LEO satellites controlled at CNES. Obviously, when an avoidance manoeuvre is considered, the CA procedure also relies on all the other on-call teams (vehicle, payload, ground segment, station network …). The CNES procedure is divided in 5 stages as presented in Fig. 1, each stage being an escalation in the contingency.

Fig. 1. Five stages procedure (PoC = Probability of Collision)

The automated screening stage consists of a fully automated process which is performed daily at OCC. Collision risk predictions are made within an horizon of 7 days. The orbits of the CNES satellites are given by ephemeris provided by the dedicated control centers. The secondary object information comes from the public NORAD catalogue (Space-Track database) and is eventually completed with data from the GRAVES catalogue. The estimation of the orbital position uncertainty is performed using in-house tools as described in [1].

The manual risk assessment stage consists in the manual analysis of the risks detected by the previous stage. This stage cannot be completely automated since each conjunction characteristic (collision probability, geometry and date, orbit uncertainties, miss distance on each axis, trending miss distance and object size) is important and has to be analysed and taken into account to evaluate the risk. The aim of the dangerous conjunctions fine assessment stage is to determine the conjunctions that need to be mitigated. This is also a manual stage and it involves requesting radar data in order to improve the knowledge of the object orbit. At this stage, three kinds of risks can lead to conjunction mitigation: - if the secondary object orbit is confirmed by radar measurements and the operational collision probability exceeds 10 -3, - if radar measurements are not available and the operational collision probability exceeds 10 -2, - if the risk is confirmed by NASA (when the conjunction concerns one of the CNES satellites involved in the A-Train).

During the conjunction mitigation stage Collision Avoidance team keeps on analysing the risk so that any change on the collision risk can be detected. At the same time, possible avoidance manoeuvres are computed in cooperation with the station-keeping team taking into account satellite mission, platform and operational constraints, as well as potentially new high-risk conjunction events in the post manoeuvre trajectory. This stage is coordinated by the on-call mission representative through the exceptional Operational Coordination Group (OCG) which decides if an avoidance manoeuvre has to be performed.

The final stage is the formal end of the conjunction mitigation. At this stage, the CA team tries to analyse the lessons learned in order to improve the process. Post Collision Avoidance reports summarizing the risk management are produced and presentations are performed in annual exploitation reviews.

2.2 Actors and information flow settled between different actors

The actors mentioned in the previous section are summarized in the Fig. 2 with the information flow settled between each of them.

CNES OCC

S GRAVES c TLEs r CONTROL e e CONTROLCENTER n M A i Alert CENTER n i s CONTROL CENTER NORAD g t s TLEs i e g s a s t m One team per mission: i e o n Radar n t DV - Mission coordination measurements analysis - Station-keeping

- Vehicle

NASA - Ground segment

Fig. 2. CNES Collision Avoidance information flow

3. REAL CASE EXAMPLES

This section describes real cases handled by CNES CA team since the beginning of its operational activities. The objective is to illustrate the variety of situations encountered, firstly describing a typical case and then presenting two special cases for which the involvement of several CNES entities was required. 3.1 Typical close approach

On July 18 th 2007 the automatic screening provided the following information concerning a new collision risk [1] between SPOT5 satellite and a debris of the Chinese FENGYUN satellite:

• time of closest approach (TCA): July 22 nd 2007 at 05:49; • collision maximum probability exceeding 10 -4; • conjunction geometry: o miss distance = 125 m; o in-track separation = 90 m; o radial separation = 50 m; o cross-track separation = 70 m.

1 13 0.8 0.6 8 0.4 0.2 3 0 -0.2 -2 -0.4 I differencesI (km) I differences (km) -0.6 -7 -0.8 -1 -12 24/1 13/2 5/3 25/3 14/4 4/5 24/5 13/6 3/7 0 1 2 3 4 5 6 7 8 910 Date Number of extrapolation days

0.2 0.3

0.15 0.2

0.1 0.1

0.05 0

0 -0.1

-0.05 -0.2 -0.1 R differences (km) -0.3 Rdifferences (km) -0.15 -0.4

-0.2 -0.5 24/1 13/2 5/3 25/3 14/4 4/5 24/5 13/6 3/7 0 1 2 3 4 5 6 7 8 910 Date Number of extrapolation days

0.2 0.4

0.15 0.2

0.1 0 0.05 -0.2 0 -0.4 -0.05 -0.6 -0.1 -0.15 -0.8

C differences (km) -0.2 -1 Cdifferences (km) -0.25 -1.2

-0.3 -1.4 24/1 13/2 5/3 25/3 14/4 4/5 24/5 13/6 3/7 0 1 2 3 4 5 6 7 8 910 Date Number of extrapolation days Fig. 3. FENGYUN debris TLEs consistency Fig. 4. FENGYUN debris TLEs expected along In-Track (I), Radial (R), and Cross-Track dispersion along In-Track (I), Radial (R), and (C) axes Cross-Track (C) axes

The first task was to estimate FENGYUN debris orbit consistency and extrapolation accuracy from NORAD TLEs. The orbit consistency evaluation is performed by extrapolating each TLE of the concerned object to the following one and then comparing them at the on-orbit-position of the conjunction. The results of this phase are the differences between contiguous TLEs in the three axes of the local reference frame over a given time interval. The analysis of these differences is used to identify and eliminate anomalous TLEs, to filter manoeuvres (for in activity satellite) and to define the best period to used for the next CA procedure stage. Fig. 3 shows the result of this treatment applied to FENGYUN debris in I (In-Track), R (Radial) and C (Cross-Track) axes. In the graphs, Y axis represents the differences between TLEs and X axis represents the TLE date.

The TLE extrapolation accuracy estimation is performed by comparing at the on-orbit-position of the conjunction each TLE with the following ones that are within a given extrapolation horizon. The results of this computation represent the expected dispersions when extrapolating a TLE. Fig. 4 is the graphical representation of this estimation in the three local reference frame axes, using the NORAD TLEs for the FENGYUN debris. In the graphs, Y axis represents the differences between TLEs and X axis represents the extrapolation duration. For instance, when extrapolating during 4 days a debris TLE, we expect a dispersion of 4 kilometres in In-Track direction, around 400 meters in Radial direction and 750 meters in Cross-Track direction.

The risk was manually analysed (stage 2 of the CNES CA procedure). CA team determined the evolution of the probability of collision, the object‘s relative position in the collision plane 1 and the statistical distribution of the debris position as shown in Fig. 5. It can be observed that on July 20th (2 days before the time of closest approach) the probability of collision exceeded 10 -3. The risk went to the stage 3 analysis and the CA team requested for radar tracking data.

(m) Statistical distribution of the debris position in the collision plane 600 500 400 300 200 100 0 -100 -200 -300 -400 (m) -1500 -1000 -500 0 500 1000 1500 2000

(m) Debris TLEs evolution in the collision plane Typical cloase approach on 7/22/2007 50 (OCC automated screening results) Debris Satellite -0.7 -0.3 1.00E-01 0 0.7 P robability of Collision OCC probability alert threshold (1.00E-4) 1.2 2.8 -50 3.8 1.00E-02 1.7 3.8 4.7 -100 5.2 5.7 1.00E-03 6.3 -150 7.3 8.8 6.4 1.00E-04

-200 Maximum probability 9.3 1.00E-05 (m) -250 10 8 6 4 2 0 -2 -800 -600 -400 -200 0 200 400 Days before TCA

Fig. 5. Evolution of the probability of collision, relative position in collision plane and distribution of the debris position

1 This is the plan determinate by the relative velocity vector between both considered objects at the time of close approach.

One of the French Defence radars provided two passes: the first one on July 20 th at 16:27 and the second one on July 21 st at 6:20. Using these data the debris orbit could be affined giving a quite different conjunction:

• in-track separation = 290 m; • radial separation = 360 m; • cross-track separation = 225 m.

With this new data CA team computed the statistical distribution of the debris position. The result is shown in Fig. 6.

(m) Statistical distribution of the debris position in the collision plane 600 500 400 300 200 100 0 -100 -200 -300 -400 (m) -1500 -1000 -500 0 500 1000 1500 2000

Fig. 6. Distribution of the debris position after the orbit determination

The new probability of collision was 7 ⋅10 -5, therefore the uploaded avoidance manoeuvre was cancelled before its execution. Complete information about the management of this risk can be found in [1].

3.2 Close approach between controlled satellites

In the beginning of year 2006 the orbit of CNES satellite DEMETER was decreased for mission management purpose but unfortunately created a periodical dangerous configuration with ESSAIM constellation (also controlled by CNES). The semi-major axes of DEMETER and ESSAIM differed only by 1.7 km causing a relative drift in argument of latitude. This fact, added to the variations of eccentricities, produced both DEMETER and ESSAIM constellation orbits to intersect twice a year.

CNES managed this risk by considering the precision of the orbits and the operational constraints related to the effective realization of possible avoidance manoeuvres [2]. The difference with a classical case was that the concerned satellites were both controlled by CNES and their orbits were well-known, therefore risky periods could be identified long time before the close approach (with the limitation of orbit propagation uncertainties) and then gradually characterised. The priority was to prepare an avoidance strategy but only perform a DEMETER avoidance manoeuvre if it was really needed.

While for typical close approaches (conjunctions with non-operational satellites) the criterion is the probability of collision, in this case CNES preferred to consider geometrical criterion and define exclusion volumes around each satellite. The idea was to guarantee that collision was not possible (no intersection of the defined exclusion volumes). Detailed information about the management of these close approaches and how the exclusion volumes were estimated can be found in [2]. CNES managed four close approaches (every 6 months) from January 2006 to January 2008. The first close approach, on June 2006, needed an avoidance manoeuvre on DEMETER, for the other close approaches the risk was managed using the autonomous orbit control of DEMETER.

The periodical risk was finally cancelled in early 2008 with the decision to decrease the orbit of DEMETER by 1.7 km which froze the relative drift between the satellites.

3.3 Close approach between a controlled satellite and a non-controlled one which are in a close orbit (JASON 1 - TOPEX)

Launched in 1992, TOPEX was a common project of NASA and CNES to measure the surface of the global ocean. This project was followed-on by JASON 1 satellite, launched in 2001 and placed on the same orbit.

TOPEX mission ended in October 2005 due to attitude control troubles and the satellite was finally decommissioned in January 2006 losing manoeuvre capability. As both TOPEX and JASON shared the same low-drag orbit, the non-controlled TOPEX became a dangerous debris for JASON. Some studies were carried out in order to predict the period of close approach and the relative position of the satellites was periodically monitored by CNES and NASA teams.

0.2 1

0 0

-1 -0.2 -2 -0.4 -3

-0.6 -4 I differences (km) I differences I differences (km) I

-0.8 -5

-1 -6 31/12 19/2 10/4 30/5 19/7 7/9 27/10 16/12 4/2 26/3 0 1 2 3 4 5 6 7 8 910 Date Number of extrapolation days

0.2 1

0.15 0.8 0.1 0.6 0.05

0 0.4

-0.05 0.2

-0.1

R differences (km) differences R differencesR (km) 0 -0.15

-0.2 -0.2 31/12 19/2 10/4 30/5 19/7 7/9 27/10 16/12 4/2 26/3 0 1 2 3 4 5 6 7 8 910 Date Number of extrapolation days

0.6 1.5

0.4 1 0.2

0.5 0

-0.2 0

-0.4 C differences (km) differences C C differencesC (km) -0.5 -0.6

-0.8 -1 31/12 19/2 10/4 30/5 19/7 7/9 27/10 16/12 4/2 26/3 0 1 2 3 4 5 6 7 8 910 Date Number of extrapolation days Fig. 7. TOPEX NORAD TLEs consistency Fig. 8. TOPEX NORAD TLEs expected dispersion

CNES CA team operational activity began one month before the conjunction when the period of close approach (from May 5 th to May 10 th 2008) was precisely identified. The first task was to estimate TOPEX orbit consistency and extrapolation accuracy from NORAD and GRAVES TLEs as described in section 3.1 (see Fig. 7 and Fig. 8).

Fig. 9. Classical avoidance strategy

Whereas in a classical close approach the avoidance manoeuvre is chosen to mitigate a well-defined conjunction (in terms of time) in this case the avoidance manoeuvre should have eliminated several risks since the close approach period lasted around 5 days. It implies that a come back manoeuvre would have been necessary, at least, 5 days after the avoidance manoeuvre. Due to mission constraints, a typical semi-major axis avoidance manoeuvre would have implied to perform two come back manoeuvres in order to come back to the same on-orbit-position than the one before the avoidance manoeuvre (see Fig. 9). It would have meant a mission interruption of 10 days. Other manoeuvre types, like eccentricity manoeuvres, were analysed too but, in all cases, a mission interruption of at least 1 week would have been necessary.

Taking into account all this information the CA team chosen strategy was the one disturbing the less possible JASON mission: to compute a precise TOPEX orbit in order to perform the avoidance manoeuvre only if it was absolutely necessary.

Therefore two weeks before the identified risky period a measurement campaign was organized with the support of German TIRA (FGAN) and DGA (Monge vessel) radars. In total 10 radar passes were received:

- 5 FGAN passes from 04/23 to 04/25, - 1 Monge pass on 04/30, - 3 FGAN passes from 05/06 to 05/07, - 1 Monge pass on 05/07.

The first orbit computation was made on April 28 th . The minimal radial miss distance obtained using this orbit was around 400 meters in the evening of May 8 th . The orbits determined the following days gave confirmed those results (minimal radial miss distance and closest approach date). Taking into account the estimated orbit dispersions of both TOPEX and JASON satellites, CNES considered that this radial miss distance ensured a sufficient radial separation. The last orbit determination on May 7 th again confirmed the minimal radial miss distance around 400 meters on May 9 th at 4:32. Decision was made between CNES and NASA JPL that no avoidance manoeuvre was required.

4. ANALYSIS OF KNOWN COLLISIONS

There are four recorded collisions with space debris appearing in the Space-Track catalogue. In December 1991, the Russian satellite COSMOS 1934 and a COSMOS 926 debris collided. Five years later, on July 24 th 1996, the French satellite CERISE was hit by a debris. On January 17 th 2005 there was a collision between THOR BURNER and a debris from Long March. And recently, on February 10 th 2009, an operational US satellite IRIDIUM 33 collided with a non operational Russian one COSMOS 2251. In order to experiment the efficiency of the CNES process THOR BURNER and IRIDIUM collision cases were applied to the automated screening stage considering these two objects as they were CNES satellites. However this experiment is mainly constrained by the lack of knowledge on the THOR BURNER and IRIDIUM accurate orbits. It was performed using Space-Track TLEs as accurate orbits of both objects. Therefore this experiment deals only with the first stage of the operational procedure described in section 2.1, stages 2 to 5 not being representative of a real case in which the accurate orbit and its associated dispersion would have been known.

4.1 THOR BURNER œ Long March collision

A debris of the American launch vehicle THOR BURNER, launched in 1974, and a debris of the Chinese launch vehicle Long March collided in orbit on January 17 th 2005 at 02:14 UT.

The experiment was performed using the Space-Track TLE available before the collision for THOR BURNER debris orbit and all the TLEs available up to seven days before the collision for the Long March debris. Fig. 10 shows the evolution of the debris relative position in the collision plane. The values represent the ages (in days) of the Long March TLEs before the collision date. One can notice that two days and a half before the collision date, two Long March TLEs (-2.6 and -2.5) indicated a very close approach between both objects.

(m) 1.00E+00 150 P robability of Collision Long M arch OCC probability alert threshold (1.00E-4) 100 -0.1 THOR B URNER -0.5 1.00E-01 50 -1.5 0 1.00E-02 -2.5 -2.6 -50 -3.5 1.00E-03 -4 -7.5 -100 -4.5 -6.5 Maximum probability -150 1.00E-04 -5.1 -200 1.00E-05 -5.5 -250 (m) 8 6 4 2 0 -800 -600 -400 -200 0 200 400 600 Days before TCA

Fig. 10. Long March TLEs evolution in the Fig. 11. Evolution of the probability of collision collision plan between THOR BURNER and Long March debris (OCC automated screening results)

As shown in Fig. 11 the probability of collision computed at the OCC is almost always higher than 10 -4 (OCC probability alert threshold) and reach a peak around two days and a half before the collision. This means that this conjunction would have been monitor during the seven days before the collision date.

4.2 IRIDIUM - COSMOS collision

The operational US satellite IRIDIUM 33 and the non-operational Russian satellite COSMOS 2251 collided in orbit on February 10 th 2009 at 16:56 UT. Around 350 debris of IRIDIUM and 800 of COSMOS are catalogued in the Space-Track database.

Since IRIDIUM accurate orbit is not available for us, the experiment was performed using the Space-Track TLE available before the collision. For COSMOS, all the TLEs available up to seven days before the collision were considered. Orbit dispersion associated to available TLEs of both IRIDIUM and COSMOS were estimated with the same method than the one described in [1] and used in daily operational activities. (m) 1.00E-02 250 Probability of Collision 11.3 OCC probability alert thresho ld (1.00E-4) 9.4 200 10.7 1.00E-03

10.2 150 8.8 7.8 8.3 COSM OS 7.3 7.2 IRIDIUM 6.2 100 1.00E-04 5.8 4.8 3.8 3.6 Maximum probability 2.3 2.8 50 2.2 1.2 1.00E-05 0 (m) 12 10 8 6 4 2 0 -2500 -1500 -500 500 1500 2500 Days before TCA

Fig. 12. COSMOS TLEs evolution in the Fig. 13. Evolution of the probability of collision collision plan between IRIDIUM and COSMOS (OCC automated screening results)

Fig. 12 represents the relative position between IRIDIUM satellite and COSMOS satellite predicted by COSMOS TLEs at the collision date. One can clearly see the COSMOS TLEs getting dangerously more and more close to the IRIDIUM satellite position.

In Fig. 13, one can see that maximum probability of collision (OCC detection criterion) is higher than the alert threshold each day within the screening horizon which means that the risk would have been daily analysed.

5. FEEDBACKS OF TWO YEARS OF OPERATIONAL COLLISION AVOIDANCE ASSESSMENT FOR 15 LEO SATELLITES

This section aims to present the main operational results since the operational kick-off of the Collision Avoidance process in July 2007. CA team considers a collision risk as a close approach between a satellite controlled by CNES and an object when the minimal distance is lower than 10 km and the maximum probability of collision is higher than 10 -4. A single risk generally gives place to several successive alerts in the daily report of the automatic screening.

5.1 Key figures for the operational activities

5.1.1. Alert Number per day over 2 years of collision risk assessment

Mean number in 2 weeks 7 Mean alert number Mean risk number 6

5

4

3

2

1

0 Jul-07 Sep-07 Nov-07 Jan-08 Mar-08 May-08 Jul-08 Sep-08 Nov-08 Jan-09 Mar-09 May-09 Date COSMOS – IRIDIUM collision Fig. 14. Mean daily alert and risk number

Fig. 14. Mean daily alert and risk number represents the mean number of alerts and risks per day detected by the OCC Collision Avoidance process since July 2007 on all LEO satellites controlled by CNES. During the 6 months of operational activities of 2007 CA team handled 428 alerts corresponding to 179 risks in stage 2. Three cases needed a stage 3 analysis with a request for radar data and 1 risk needed to be mitigated by programming an avoidance manoeuvre that was finally cancelled few hours before its execution thanks to radar measurements.

In 2008, 881 alerts corresponding to 344 risks were analysed in stage 2. These values represent almost the same average number of alerts (2.4) and risks (1) per day than in 2007. During 2008, 13 cases were analysed on the stage 3, 4 of them triggered by NASA and 2 confirmed risks were mitigated by anticipating a coming station keeping manoeuvre.

During the 6 months of operational activities of 2009 CA team handled 498 alerts corresponding to 189 risks in stage 2. Ten cases needed a stage 3 analysis, 3 of them triggered by NASA and no avoidance manoeuvre was needed thanks to radar measurements. In Fig. 14 we can see that the COSMOS œ IRIDIUM collision on February 2009 increased the number of alerts and risks.

5.1.2. Alert number per satellite

DEMETER (660 km) ESSAIM (660 km) A-TRAIN (700 km) HELIOS (750 km) SPOT (820 km) COROT (910 km) JASON1 (1340 km)

100.00 49.3 42.0 25.7 23.3 16.8 16.7

10.00 8.0 7.33 4.67 3.33 3.33 2.83 2.22 1.78

1.00 0.67 0.67 0.28

0.10 > 1.e-02 > 1.e-03 > 1.e-04 Probability of Collision

Fig. 15. Annual collision risk number per satellite and probability threshold

Risk distribution by satellite family can be represented versus the chosen threshold for the probability of collision. In Fig. 15 we can find the number of collision risks per year and per satellite for each family depending on the probability threshold considered for the screening. SPOT, ESSAIM and A-TRAIN (CALIPSO and PARASOL) families are the most dangerous ones which is consistent with the risks treated at stage 3 and the avoidance or advanced station-keeping manoeuvres performed.

5.1.3. Risk distribution by satellite family

Fig. 16 shows the most dangerous debris for each CNES orbit family (700km, 800km and 900 km). For orbits between 800 and 900 km Fengyun 1C debris represent the largest proportion of alerts. In 700 km orbits, alerts are mostly generated by Sea-Lunch, CZ-4 and Fengyun debris in equal shares. SCOUT X-4 COSMOS 886 100% ORBVIEW 1 OV2-1 COSMOS 2251 DELTA 1 R/B DMSP 5D-2 F11 THOR ABLESTAR 90% BEIJING 1 PSLV R/B CZ-4 80% CZ-4 SL-18 THOR ABLESTAR 70% SL-8 R/B SL-8 R/B SL-18 Aqua-Train 60% SL-8 SL-8 (699 km) SL-8 50% FENGYUN 1C FENGYUN 1C 40% FENGYUN 1C 30% OTHER OTHER 20% OTHER 10% 0% 2007 2008 2009

METEOR 2-9 SCOUT D-1 R/B OPS 7323 IRIDIUM 33 SL-3 OPS 4078 100% METEOR 2-5 COSMOS 2251 SL-14 SL-3 R/B COSMOS 1275 SL 16 R/B 90% THORAD AGENA D CZ-4 METEOR 2-5 DMSP 5D-2 F11 THORAD AGENA D SL-3 R/B DELTA 1 80% DMSP 5D-2 F11 CZ-4 SL-16 DELTA 1 70% DMSP 5D-2 F11 SL-16 DELTA 1 SPOT 60% SL-16 (822 km) 50% FENGYUN 1C FENGYUN 1C 40% FENGYUN 1C 30% OTHER 20% OTHER OTHER 10% 0% 2007 2008 2009

SCOUT X-4 COSMOS 886 100% LANDSAT 2 COSMOS 2251 TIROS 9 SL-16 COSMOS 1266 90% THOR ABLESTAR DELTA 1 THORAD AGENA D SCOUT X-4 COSMOS 1579 COOLANT 80% SL-16 PSLV R/B COSMOS 1275 70% THORAD AGENA D THORAD AGENA D COROT 60% COSMOS 1275 COSMOS 1275 (909 km) 50% FENGYUN 1C 40% FENGYUN 1C FENGYUN 1C 30%

20% OTHER OTHER 10% OTHER 0% 2007 2008 2009

Fig. 16. Most dangerous debris families by orbit altitude

We can see that the debris of the Chinese satellite FENGYUN 1C are responsible for a large part of the alerts. Since 2009, the debris of the IRIDIUM and COSMOS 2251 satellite cause around 10 % of the collision alerts on the three kinds of orbit.

5.2 Impact of the IRIDIUM - COSMOS collision on the OCC Collision Avoidance process

900 COSM OS debris number 800 IRIDIUM debris number 700

600

500

400

300

200

100

0 Feb-09 Mar-09 Apr-09 May-09

Fig. 17. Evolution of the number of the IRIDIUM œ COSMOS debris

Fig. 17 shows that the number of IRIDIUM 33 and COSMOS 2251 debris catalogued in the Space- Track database grows up until 809 for COSMOS and 349 for IRIDIUM in June 2009.

Since the collision on February 10 th 2009, 21 collision risks corresponding to 56 alerts triggered by the CNES Collision Avoidance process, involved a debris of one of both satellites. It results in increasing the alert number by 15 %. These risks are listed in Tab. 1. Two of them required radar measurements to be mitigated.

Tab. 1. Risks due to either IRIDIUM 33 or COSMOS 2251 debris

NORAD Time of closest Probability of Alert In-track Radial Cross-track Debris origin number approach collision number separation (m) separation (m) separation (m) COSMOS 2251 33901 2/28/09 16:58 2,27E-04 1 374 -193 111 COSMOS 2251 33843 3/12/09 0:04 1,08E-04 1 311 -320 -236 COSMOS 2251 34449 3/17/09 10:36 1,20E-03 6 -49 -2 332 COSMOS 2251 34291 3/21/09 3:48 3,55E-04 1 -268 194 -287 IRIDIUM 33 34522 3/26/09 15:10 1,30E-04 4 1020 258 565 COSMOS 2251 34019 4/1/09 23:07 8,45E-04 2 -57 235 -90 COSMOS 2251 34727 4/19/09 9:03 4,32E-04 5 26 -27 -492 COSMOS 2251 33989 4/19/09 23:37 3,29E-04 2 -116 49 624 COSMOS 2251 34734 4/21/09 20:27 1,04E-04 1 -46 -405 -470 COSMOS 2251 34423 4/25/09 6:02 3,67E-04 6 26 -27 492 COSMOS 2251 34029 4/26/09 5:35 1,96E-04 6 -31 375 383 COSMOS 2251 34385 4/29/09 8:00 1,71E-04 2 -156 262 -564 COSMOS 2251 34436 5/2/09 23:41 1,19E-04 1 111 -427 -767 COSMOS 2251 34756 5/14/09 11:22 4,23E-04 1 -5 268 -49 COSMOS 2251 33811 5/18/09 22:25 3,75E-04 2 -437 -175 -480 COSMOS 2251 34617 6/15/09 10:14 6,22E-04 2 -360 61 -98 IRIDIUM 33 33951 6/14/09 6:41 1,90E-03 6 -28 -69 -203 COSMOS 2251 34058 6/18/09 6:58 1,47E-04 2 -164 -383 -193 IRIDIUM 33 34771 6/25/09 10:29 3,68E-04 2 -9 440 13 IRIDIUM 33 34074 6/25/09 12:07 1,02E-04 1 45 -78 740 IRIDIUM 33 33880 6/25/09 21:46 1,14E-04 2 -44 -120 -690 5.3 Feedbacks of two years of operational activities

Because of the increasing alert and risk numbers, it became necessary to optimize the time dedicated to each manual analysis of collision risk. In this way, after each daily screening process, a risk summary is sent by email to the on-call CA team member. This email summarizes the main information about the detected collision risks (concerned objects, time of closest approach, geometry of the conjunction, orbit uncertainties, miss distance and probability of collision) and allows the CA team member to easily estimate the dangerousness of the in-orbit collision risks.

In order to minimize the time spent for each manual risk assessment (stage 2 of the CNES Collision Avoidance procedure) CNES introduced a unique software which sequences the various tasks of a manual analysis of a collision risk: collect of the freshest information about both concerned objects, computation of their orbit uncertainties, of the conjunction geometry, miss distance and time of closest approach, computation of the debris position evolution and of its statistical distribution in the collision plan. This software makes analysis faster and avoids possible typing or use errors.

Statistics can be useful in the decision-making process. According to the alert and risk numbers, statistics are useful in tuning the alert threshold value used in the Collision Avoidance process. It can help Collision Avoidance team to identify the most dangerous debris families or the less safe orbit for operational satellite. Since two years of operational activities, CNES stores the features of each collision risk and alert triggered for its 15 LEO satellites. Moreover, each confirmed collision risk results in writing a High Interest Event report summarizing the risk management, in order to improve the current Collision Avoidance process.

6. CONCLUSION

Today, the CNES Collision Avoidance process has been operational since more than two years for the 15 LEO satellites controlled by CNES.

The process depends on:

- completeness and reliability of catalogues for detection. The US catalogue is the more complete and publicly available. The accuracy of the GRAVES one is available at OCC; - capacity to determine the orbit of the threatening secondary object for mitigation. Therefore, tracking radar access improvement is the priority to improve the process. In this perspective, several possibilities are currently being considered such as update of existing radars, new radar purchase, radar measurements purchase or international cooperations.

With a controlled knowledge of orbits, avoidance manoeuvre will be rare and small and only when absolutely needed. The CNES Collision Avoidance process cannot guarantee there will be no collision for the satellites of interest, it tries to guarantee that the detectable ones will be mitigated.

7. REFERENCES [1] Laporte, F. Operational management of collision risks for LEO satellites at CNES. March 2008 Space OPS 08, Darmstadt. [2] Delavault, S., Lorda, L., Lamy, A. Analysis and management of 2006 and 2007 periodic conjunctions between CNES satellites. January 2008 AAS/AIAA Space Flight Mechanics Meeting, Galveston.

BACK TO SESSION DETAILED CONTENTS BACK TO HIGHER LEVEL CONTENTS