LESSONS LEARNED FROM SPACE-SI EXPERIMENTS ON PRISMA MISSION Drago Matko (1), Tomaž Rodič (1,3), Sašo Blažič (1,2), Aleš Marsetič (1,4), Robin Larsson (5), Eric Clacey (5), Thomas Karlsson (5), Krištof Oštir (1,4), Gašper Mušič (1,2), Luka Teslić (1), Gregor Klančar (1,2) (1) Space-SI, Aškerčeva cesta 12, 1000 Ljubljana, Slovenia (2) University of Ljubljana, Faculty of Electrical Engineering (3) University of Ljubljana, Faculty of Natural Sciences and Engineering (4) Scientific Research Centre of the Slovenian Academy of Sciences and Arts (5) OHB Sweden AB 2013 1
OUTLINE 1. INTRODUCTION 2. IN FLIGHT SIMULATED RADAR INTERFEROMETRY REMOTE SENSING 3. OBSERVATION OF NON-CO-OPERATIVE OBJECTS 4. IN FLIGHT SIMULATED DISTRIBUTED INSTRUMENT REMOTE SENSING 5. CRITICAL EVALUATION OF FORMATION FLYING MODELS 6. CONCLUSION 2
1. INTRODUCTION To investigate newly emerging technologies SPACE-SI and OHB Sweden performed a set of formation flying experiments in September 2011 with Prisma satellites. Mango and Tango were launched into a sun synchronous orbit with 725 km altitude and 06.00h ascending node in June 2010. In the SPACE-SI formation flying experiments the critical maneuvers for three types of missions were investigated with respect to in-orbit performances: In-flight simulated radar interferometry Observation of non-co-operative objects - space debris In-flight simulated distributed instrument 3
2. IN FLIGHT SIMULATED RADAR INTERFEROMETRY REMOTE SENSING The objective of the experiment was to examine whether a noncontrolled flight in a nearly circular orbit provides satisfactory orbital stability and accuracy for SAR applications. In the along-track synthetic aperture radar interferometry one of the satellites acts as the SAR transmitter and receiver, while the other is a receiver only This technique is currently not used widely since the only system capable of providing it is the German TanDEM-X constellation. It offers, however entirely new opportunities; this application uses two separate radar antennas arranged longitudinally along the direction of flight. The method is used particularly in oceanography, glaciology, and traffic research. 4
2.1. The experiment Mango and Tango were flown one behind the other (along-track) separated by a distance of approximately 200 m for three consecutive orbits. No thrust; only natural forces have influenced the relative position of the satellites. The appplicability of the noncontrolled fligh to SAR applications was assesed from the relative distances of the satellites. 4 3 2 Position SLO x-200 y z 1 [m] 0-1 -2-3 -4 23 24 25 26 27 28 29 time [h] 5
2.2. Lessons learned Since small drifts can be detected and corrected in InSAR processing, the Prisma constellation flight has proved to have the stability that is required for interferometric data processing. All the drifting effects are small and removable during the InSAR processing steps. We have estimated that the measurement of water current speeds accuracy can reach 0.05 m/s. In a similar across-track constellation with similar orbital parameters and orbital accuracy the vertical accuracy of the digital elevation model production could reach 1 m. 6
3. OBSERVATION OF NON-CO-OPERATIVE OBJECTS It is expected that non-co-operative objects such as space debris will become a serious problem in the near future. The orbits of debris often overlap with trajectories of operational spacecrafts, and represent a potential collision risk. In order to remove the debris, it must be identified. Two experiments were performed to simulate the required procedures: Orbit identification Close observation 7
3.1 Orbit identification On the basis of the space debris Two Line Elements (TLE) the Mango s VBS camera was directed in the direction of the point of the closest approach and several images were taken in a sequence. The Simulation toolkit (AGI - STK) and the newest TLE database were used to simulate the trajectories of the Mango and debris. The criteria for choosing the objet to be observed with Mango vision based camera (VBS), was the distance and the vicinity period. Also additional constraints were considered: the camera should not be pointing neither towards the Sun nor towards the Earth. 8
STK simulation (upper) and VBS camera shots (lower) of: Geosat (ID 15595) right 3 pictures and SL-14 R/B (ID 22237) left 4 pictures Timeframe: 2011 09 20, 09:25:57-09:26:56 Animation next slide 9
File name: SpaceDebris.wmv and Technologies, Munich,May 29-31, 2013 10
3.2. Close observation 1 The satellites were flown in the (in-track) distance of 5 m, Mango was pointing with its Digital Video System (DVS) camera towards Tango, Tango was rotating around (with a bit of wobbling) its cross-track axis, pointing all times with its solar panels toward the sun. Several pictures of Tango (which simulated the debris) were taken in order to make a 3D model of the observed object. Animation (real data) 5th International Conference on Spacecraft Formation Flying Missions and Technologies, Munich,May 29-31, 2013 File name: Tangoiz5m_ok.wmv 11
Pictures of Tango by the Mango s DVS camera (left) and the 3D model (right) Reconstructed model from close sideviews (only) of Tango (animation ) File data: Tango3D_2.avi 12
The timing of imaging (during encircling) was adjusted to have some areas of interest on the Earth (Kuwait - left, Djibouti middle Crete - right) in the background Mango was first circumvoluting Tango on an ellipse (20m x 10 m) Mango was pointing with its Digital Video System (DVS) camera towards Tango Then a cross-track impulse was given to Mango and Mango was encircling the Tango in a relative 60 degrees inclined orbit on a circle with radius 20 m Animation 3.3. Close observation 2 File name: Kuw_Dji_in_obkrozanje_fullHD.wmv 13
STK verification (upper) and DVS shots (lower) of Kuwait Djibouti Kreta Spacecraft Formation Flying Missions 2013 14
Pictures of Tango by the Mango s DVS camera (left) and the 3D model (right) Reconstructed model from 20 m distant all around views of Tango (animation ) File name: Tango_3D_6.avi 15
3.3. Lessons learned The main lesson learned is that it is possible to predict close encounter between two satellites (including debris) from the TLE data. The second lesson learned from the close observation experiment is, that the reconstruction can be done successfully under satisfactory illumination conditions. However, such conditions are very hard to be met. The 3D model, reconstructed from the DVS camera shots with adequate resolution (from 5 m distance) is satisfactory, while the reconstruction of the parts reconstructed from a lower resolution images (taken from the 20 m distance), was degraded. 16
4. IN FLIGHT SIMULATED DISTRIBUTED INSTRUMENT REMOTE SENSING The idea of distributed instrument is to form a telescope by using of two small satellites instead of one big and more expensive satellite. In such a instrument one of the satellites holds the optical system with lenses and/or mirrors and the other one the detectors (sensors). The objective of the experiment was to investigate whether the positioning of the satellites is adequte to form a telescope that can acquire high-resolution multispectral images of the Earth s surface. In this experiment the Tango was simulating the holder of the optical system with lenses and/or mirrors while Mango, simulating the holder of detectors (sensors), was driven to an appropriate position. This experiment was performed in two different versions: In-track displacement (satellites flying one after the other) Radial and cross-track displacement (satellites flying one above the other) 17
4.1. In-track displacement This formation is preferable as the consumption of propellant is very small. One of the satellites must carry a mirror at an angle of approximately 45 that reflects the beam to the detectors on the other satellite. In the experiment the satellites flu 5 m apart on the same orbit; this formation was kept for two orbits. The absolute and relative positions and the attitude data were used to asses the capability of the formation for high-resolution imaging. Changes of attitude for all Mango axes during the two consecutive orbits: Relative distance variations between the satellites during two consecutive orbits: 0.4 Position SLO 0.3 0.2 0.1 [m] 0-0.1-0.2-0.4-0.3 14 14.5 15 15.5 16 16.5 17 time [h] x-5 y z 18
4.2. Radial and cross-track displacement The satellites were aligned with predefined target locations on the Earth, such as Piran, Cape Town, Piran Punta Arenas Animation (real data) File name: Piran_CapeTown_Piran_ZRC_HDTV7205s.wmv 19
Cape Town 20
Piran 21
Next day: Punta Arenas Spacecraft Formation Flying Missions and Technologies, Munich,May 29-31, 2013 22
Piran May 9,2012 Spacecraft Formation Flying Missions 2013 23
4.3. Lessons learned The lesson learned with the In-track formation was that very precise orbit maneuvering and navigation, required for exact focusing (in micrometer scale), cannot be achieved by the implemented thruster-control and relative GPS navigation as used in this experiment. It can be concluded that accuracies in the regulation of relative distances and attitudes are inadequate for a distributed platforms instrument where relative positions and attitude shifts should be kept within millimeters. The experiment of the radial and cross-track displaced instrument remote sensing simulation has shown that the positioning of the satellites in order to align them with predefined targets was satisfactory as demonstrated by attractive pictures despite very unsuitable (for taking images of the Earth s surface) dusk down orbit, autumnal equinox time and bad weather. 24
5. CRITICAL EVALUATION OF FORMATION FLYING MODELS Follower dynamics in the Leader RIC co-ordinate system. μ( R+ x) μ && 2 & ϕ& && ϕ & ϕ = + + 3 2 (( R+ x) + y + z ) R 2 x y y x a 2 2 2 2 x μ y && + 2 && + && & = + 3 2 (( R+ x) + y + z ) 2 y ϕx ϕx ϕ y a 2 2 2 y μz && z = + a 3 2 R+ x + y + z 2 2 2 (( ) ) z Slika z oznakami Leader follower Leader orbit: 2 R&& μ = R& ϕ R 2 R 2Rϕ R && ϕ R = & &. R Spacecraft Formation Flying Missions 2013 μ - Earth gravitational constant φr - True anomaly a - Accelerations 25
R1 (0) = a R& (0) = 0 1 & ϕ (0) = 2n 1 Application of the Method of perturbations to leader s orbit equations: R= a+δr & ϕ = n +Δ& ϕ Δ R = ε R Δ & ϕ = εϕ& Expanding Leader orbit eq. into Taylor series and collecting terms with respect to ε yields: 1 1 μ εr&& μ ε 2an n R 2 R = + & ϕ + ε & ϕ + ε & ϕ + a a 2 2 1 2 3 2 1 an 2 1 + 1+ 2 Rn 3 1 1 R1 1... 2Rn & 1 2 2εRR & 1 1n 2 2εR& & 1ϕ1 3 ε&& ϕ1 = ε ε ε ε (...)... 2 a a a R&& = 2an & ϕ + 3n R 2 1 1 1 2n && ϕ = R & a 1 1 HCW R = a & ϕ = n ε HCW = Hill-Clohessy-Wiltshire Linear model - method of perturbation Higher order terms Spacecraft Formation Flying Missions 2013 n μ 3 a = = mean motion Linear model method of perturbation 26
Application of the Method of perturbations to the Follower s dynamics equations x () t = x () t +Δx() t p c y () t = y () t +Δy() t p c z () t = z () t +Δz() t p c Δ xt () = ε x() t 1 Δ yt () = ε y() t 1 Δ zt () = ε z() t 1 Hill-Clohessy-Wiltshire Linear model - method of perturbation Higher order terms Expanding Follower dynamics eq. into Taylor series and collecting terms with respect to ε yields: x& ny& n x && & & 2 2 2 2 2 3 c 2 c c + ε( x1 4nyccosnt 2ny1+ 2n ycsinnt n x1 4n xccos nt) + ε (...) + ε (...) +... = = + + + 2 2 2 2 3 2 nxc ε(2nx1 + 6nxccos nt) ε (...) ε (...)... && y + 2nx& n y 2 c c c + + a 2 2 2 2 3 ( y1 4nxc cosnt+ 2nx1 2 n xcsin nt n y1 4n yccos nt) (...) (...) 2 2 2 2 3 c ε( ny1 3n c c t ε (...) ε (... x + ε && & & + ε + ε +... = = ny+ y o s n ) + + ) +... + a y && z + ε && z = n z ε( nz+ 3nzcos nt) + ε (...) + ε (...) +... + a 2 2 2 2 3 c 1 c 1 z Spacecraft Formation Flying Missions 2013 27
HCW && x 2ny& 3n2x = a && y + 2nx& = c c c x c c a y && z + n2z = a. c c z Linear model method of perturbation x 1(0) = z1(0) = 0; y1(0) = y 0 x & (0) = ny ; y & (0) = z & (0) = 0 1 0 1 1 && x 2ny& 3 nx= (10nx+ 4 ny& )cosnt 2n y sinnt 2 2 2 1 1 1 c c c && y nx& n y nx& nt n y nt 2 2 1+ 2 1 = ( c 4 c)cos + 2 csin && z n z n z nt 2 2 1+ 1 = 3 c cos. ε 28
5.1. Evaluation of the models POD impoved data Initial conditions Model + - Optimization Optimization cost function: ( ( ) ( ) ( ) ) 2 2 2 i i i i i i 1 N m m m D= x x + y y + z z N i = 0 Hill-Clohessy-Wiltshire Linear model - method of perturbation STK J2 & HPOP 1.5 0.4 0.4 1 0.3 0.3 HCW Nonlin,Lin-Pert, STK EMP STK J2,HPOP, HPOPa 0.2 0.2 x [m] 0.5 0-0.5 measured HCW Nonlin,Lin-Pert, STK EMP STK J2,HPOP, HPOPa Δ x [m] 0.1 0-0.1 Δ y [m] 0.1 0-0.1-1 -1.5 195 196 197 198 199 200 201 202 203 204 y [m] -0.2-0.3 HCW Nonlin,Lin-Pert, STK EMP STK J2,HPOP, HPOPa -0.4 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 t [s] Spacecraft Formation Flying Missions 2013-0.2-0.3-0.4 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 t [s] 29
5.3. Lessons learned The models were fitted to the data from a POD product, which is already the output of a reduced dynamics orbit determination (employing rigorous force models, which however were not known to us). The question arising was, whether our successive model comparison did provide the accuracy of a kind of an orbit determination rather than the true net propagation error of the models, intended to be used for the orbit prediction. Our intention was not to compare the models with respect to their ability to predict the relative propagation but with the respect to their ability to cope various with phenomena, such as e.g. Earth oblateness etc. If the POD data is obtained by a kind of a Kalman filter (including a model) then the POD data is the best estimate of the real position of the satellites and this is, what is in our interest. Our desire was to fit the responses of the model to the real (relative) position of the satellites, not to fit them to the faulty measurements. The lesson learned is that while preprocessing significantly reduces the noise, it also introduces some filtering artifacts and due to these artifacts the evaluation of the models becomes extremely difficult. It will be part of our future work. 2013 30
5. CONCLUSION A set of experiments performed by SPACE-SI and OHB Sweden in September 2011 with Prisma satellites was reviewed. The in flight simulated radar interferometry remote sensing experiment demonstrated that speed estimation accuracy can reach 0.05 m/s, while in a similar cross-track constellation with similar orbital parameters the vertical accuracy of the digital elevation model production could reach 1 m. The observation of non-co-operative objects - space debris experiment has demonstrated that space debris, can be identified on the basis of the TLE data and optically tracked by a narrow angle camera. The in-flight simulated distributed instrument experiment, where one of the satellites holds the optical system with lenses and/or mirrors and the other one the detectors (sensors), has proven that very precise orbit manoeuvring and navigation, required for exact focusing, cannot be achieved by the implemented thruster-control and GPS navigation technology. However the positioning of the satellites in order to align with predefined targets (Piran, Cape Town, Punta Arenas) was satisfactory as demonstrated by attractive pictures 31