PLATFORM: a Test-Bench to Test GNC Algorthms and Sensors for Formaton Flyng, RvD and Robotc Applcatons Pablo Colmenarejo, Fernando Ganda, Valentn Barrena, Angelo Tomassn GMV S.A. c/ Isaac Newton 11, P.T.M., 2876 Tres Cantos, Madrd, SPAIN Tel.: +34 918721, Fax: +34 91872199 e-mal: pcolmena@gmv.es, fganda@gmv.es,vbarrena@gmv.es; atomassn@gmv.es Despte the practcal dffcultes to be overcome, scentfc space mssons, as well as exploraton mssons are becomng everyday more and more challengng, ths beng true both from scentfc and techncal pont of vew. New ways to desgn satelltes, as well as new technques to dstrbute the functonalty of a sngle spacecraft among several satelltes flyng n formaton or to Rendezvous and sometme Dockng (RvD) satelltes launched wth dfferent space vehcles, need to be nvestgated at the scope of fulfllng the severe requrements mposed by these new msson scenaro, n partcular n terms of postonng and maneuverng accuraces. In order to reduce the rsks assocated by flyng multple satelltes n formaton, or to RvD n space two or more S/Cs, t s sutable to test as much as possble on ground the requred GNC algorthms, sensors and actuators. At ths scope, t s essental to have avalable low-cost and hgh-flexble test benches allowng the on-ground valdaton of most of the nvolved technologes, so to reduce the number of new technology chan elements to be tested and valdated through the costly space demonstraton mssons. Ths paper descrbes the steps tll now conducted to set up and explot the GMV s PLATFORM test bench, whch has as major objectve to provde a cost-effectve soluton for: 1) On-ground Valdaton and Testng of Autonomous Navgaton Algorthms for mn-satelltes, FF, RvD and Robotc (f.. cooperatve robotc operatons on the Mars surface) applcatons. 2) Beng the frst step toward the valdaton of navgaton sensors and actuators. Wth respect to other ground test benches, PLATFORM ncludes the partcularty to allow the use of real sensor measurements (ncludng most of the error sources present n a space scenaro, as transmsson delays) obtaned through the recreaton of a real dynamc profle of spacecraft mockups by usng an accurate numercally controlled robotc arm. The correct and accurate calbraton of the test bench s a key drver durng the test bench settng up, beng t n drect relaton to the valdaton level that wll be acheved by the test bench tself. Snce the calbraton resdual error wll be drectly added over the performance fgure evaluated for the GNC functons, t s fundamental to keep the test bench calbraton resdual as low as possble.
NOMENCLATURE ANT = Helx ANTenna. AOCS = Atttude and Orbt Control System. C/A = Coarse Acquston. CDTI = Centro para el Desarrollo Tecnológco e Industral. CPU = Central Processng Unt. DKE = Dynamcs, Knematcs and Envronment. DoF = Degree of Freedom. ESA = European Space Agency. FDIR = Falure Detecton Isolaton and Recovery. FF = Formaton Flyng. GMV = Grupo M ecánca de Vuelo. GNC = Gudance, Navgaton and Control. GPS = Global Postonng System. GNSS = Global Navgaton Satellte System. L1/L2 = Carrer Phase of GPS frequences L1 and L2 respectvely. MCS = Master Control Staton. PC = Personal Computer. PL = PseudoLte. PLATFORM = PLATform for FORMaton Flyng. RC = Reference ReCever. RF = Rado Frequency. RM = Rado M odem. RRF = Robotc Reference Frame. RvD = Rendez-vous and Dockng. S/C = SpaceCraft. I. PLATFORM TEST BENCH CONCEPTUAL DESIGN PLATFORM test bench faclty s not conceved to dsplace or replace current used test benched. The purpose s to provde a complementary faclty to be matched wth other avalable test benches. In partcular, PLATFORM s able ether to make use of valdated developments over other test benches to ncrement the valdaton level of the developments or to provde the means to perform a full development from the scratch. Most usual current ground testng test benches are based on SW smulators, dstngushng: 1) Complete SW smulators (f.. based on Matlab/Smulnk), wth dedcated functons to the smulaton of the DKE and the smulaton of the sensors measurements and actuators behavour and, separately, dedcated blocks to the on-board functons: GNC, FDIR, Msson and Vehcle Management, Communcatons, 2) Complete SW smulators embedded n an envronment allowng the use of real hardware n the loop (as f.. the European envronment EuroSm), where the use of a n on-board processor ERC32 emulator allows to host all the on-board functons n a realstc space processor separated from the DKE, sensors and actuators through a communcaton layer that reproduces the real nterfaces. Real sensors may also be used, although the nput of the sensors shall be convenently smulated.
PLATFORM test-bench represents a step ahead, snce t allows generatng a real relatve DKE between the spacecraft constellaton though the use of the robotc arm together wth the use of real sensors where the provded measurements are based upon real sgnal transmsson and/or real observatons (GPS-lke recevers/emtters, optcal cameras or others). Actuators behavour can be reproduced through the answer of the robotc arm to control commands produced by the on-board functons. Ths approach allows complementng the EuroSm-based test benches, snce t reproduces the real dynamc of the constellaton and provdes real sensors measurements and keepng, at the same tme, the advantage of usng on-board processor emulators f desred (through the EuroSm envronment) or usng a PC-based on-board processor emulator as frst step. Three man objectves are consdered: 1) The development of a hardware test-bed that actually mmcs the relatve moton of two or more satelltes n dfferent space scenaros. 2) Implementaton upon the test-bed of real navgaton sensors, such as GPS recevers and pseudoltes so as to test GNC algorthms under condtons as close as possble to real space condtons. 3) Development and/or ntegraton upon the test-bench of the most advanced gudance, navgaton and control algorthms conceved to solve formaton flyng ssues, such as those navgaton algorthms developed by GMV for IRSI/DARWIN msson. The followng fgure presents the three concepts of test benches, showng the natural evoluton from one to the others as consequence of the ncreasng valdaton and demonstraton level before beng ready for a demo flght. Fgure 1. Robotc arm calbraton scheme II. PLATFORM TEST BENCH SETTING In order to reduce as much as possble the rsk assocated to fly multple satelltes n formaton, whch bascally act, from the pont of vew of carred nstruments, as a unque platform, t s worthwhle to test as much as possble on ground the requred GNC algorthms, and sensors and actuators. Ths, together wth the GMV nvolvement wthn the desgn and mplementaton of the navgaton algorthms for dfferent ESA s FF msson (.e. SMART-2/3, IRSI-DARWIN), brought to the dea of settng up the PLATFORM test-bench able to reproduce as much as possble onground space flyng condtons, ncludng sensors and actuators. The test bench wll allow nvestgatng and testng GNC, sensng, communcaton and msson management ssues assocated wth precse formaton flyng.
PLATFORM, developed n the frame of the Spansh Space Programme, and co-funded by the Spansh Scence Mnstry (va CDTI) and GMV S.A. tself, has as major target to provde the more cost-effectve soluton for: 1) On-ground Valdatng and Testng of Autonomous Navgaton Algorthms for mn-satelltes, FF, RvD and Robotc applcatons. 2) Beng the frst step toward the valdaton of navgaton sensors and actuators. PLATFORM ntal settng s specfcally adapted to a DARWIN-type scenaro (RF-based sensors are the man ones), although t s easly extensble to other scenaros. Current settng s composed by: 1) A 6 DoF robotc arm for accurate reproducton of the constellaton DKE. 2) Two S/C mock-ups, that wll host all the sensng equpments. They shall be representatve n shape and structure of the real spacecraft, snce the external structure wll mpact on the accuracy of the sensors measurements (multpath effect over GPS-lke measurements, mage processng n camera-based navgaton). 3) Four GPS-lke pseudoltes, for creatng a vrtual constellaton of multple spacecraft (only two mock-ups are used up to now). 4) Two poston-atttude GNSS recevers each wth 3 antennas, for provdng the navgaton flters wth measurements of relatve poston, velocty, atttude and atttude rate. 5) One navgaton camera, for acqurng relatve navgaton observatons n case of scenaros wth uncooperatve spacecraft. 6) One GPS constellaton sgnal outdoor-ndoor repeater. 7) Several PCs, for controllng the robotc arm, the GPS-lke pseudoltes, hostng the on-board processor functons and others. and provdes the followng major features: 1) DKE computer-based generaton. 2) Hgh level of DKE accuracy knowledge through numercally controlled robotc arm. 3) Real sensng (radated RF sgnals). 4) Real on-board relatve navgaton algorthms (DARWIN-based as startng pont). 5) Very accurate performance assessment thanks to the accurately known robotc DKE. 6) Possblty of feedng-back the robotc DKE wth a control law. One of the S/C mock-up s statcally placed, whle the second mock-up s placed on the robotc arm, smulatng the formaton flyng S/C wrt the frst statc one. The moton of the robot s gven from one sde by the DKE ncludng all actng perturbaton, and from the other by the S/C AOCS tendng to fulfll the formaton accuracy requrements. The followng sectons wll ntroduce n detal the man components of PLATFORM. A. ROBOTIC ARM FOR DKE MOTION GENERATION The PA1-6CE Robot s a 6 degrees of freedom manpulator wth the followng characterstcs: the arm unt weght s 38 Kg and can lft an artcle of 1 kg weght. Sx jonts compose the vertcal jont type archtecture of the man body: S1, S2, E1, E2, W1, W2 from robot mountng base ( S stands for shoulder jont, E for elbow jont and W for wrst jont). The robot arm has the reach of 1m, a postonal repeatablty of ±.1 mm and s controlled by a personal computer. The followng table shows jont operatng lmtatons:
Jont operatng range and maxmum operatng speed Lmt (degree) Maxmum Name of axes Mechancal Servo Software operatng speed lmt lmt lmt (rad/sec) S1 (Rotaton) ±18 ±178 ±177 ±1 S2 (Swng) +127,-67 +125-65 +124,-64 ±1 E1 (Swng) +164,-113 +159,-18 +158,-17 ±2 E2 (Rotaton) ±27 ±256 ±255 ±2 π W1 (Swng) ±18 ±166 ±165 ±2π W2 (Rotaton) ±27 ±256 ±255 ±2π Table 1. Jont operaton range and speed lmts The robotc arm characterstcs allow the desgn and smulaton of several trajectores as lnear, arcs, crcles, ellpses and others, down dfferent control modes: velocty and trajectory. These control possbltes shall defne dfferent trajectores, atttudes, veloctes and postons n order to smulate spacecraft trajectores, poston, atttude, navgaton, relatve navgaton, camera navgaton and others. B. SPACECRAFT MOCK-UPS Spacecraft mock-ups have been manufactured usng an external structure on Alumnum alloy and recovered wth a thermal coat. The followng fgure shows the mock-up n an ntermedate assembly step. Characterstc sze of the central body s 4 cm and the wngspan of 1.5 meters. C. RF SIGNAL GENERATION THROUGH PSEUDOLITES Fgure 2. Manufactured spacecraft mock-up The GPS-lke sgnal s generated by usng the NAVndoor system manufactured by Space Systems Fnland, and t s composed by fve prncpal components: 1) 4 Pseudoltes (PL) 2) 5 Rado modems 3) 4 Helx antenna 4) 1 Reference recever (wth patch antenna) 5) 1 Master control unt The full assembly dagram s as n the followng fgure. Fgure 3. GPS-lke emttng pseudoltes assembly dagram
D. NAVIGATION SENSORS 1. GPS RECEIVERS The selected GPS recever s the PolaRx2 by Septentro. PolaRx2 s a versatle hgh-end dualfrequency GNSS recever for precse postonng and tmng applcatons. Among all the characterstcs of the PolaRx2 recever, there are some specfc features that make ths equpment to be the most adapted one for the test bench: 1) The PolaRx2 provdes raw measurements: C/A and P1/P2 pseudo ranges (wth.15 and.1 meter measurement nose respectvely), L1/L2 carrer phases (wth.2 and.4 mm measurement nose respectvely) and Doppler (.5 mm/s nose). The raw measurements are requred to feed the RF-based relatve navgaton algorthms. 2) The PolaRx2 s capable of trackng satelltes from up to 3 dfferent antennas, thus, allowng to provde rangng and atttude measurements. 3) The hgh number of trackng channels (48) that wll allow usng several channels to calbraton purposes. 2. CAMERA A commercal hgh-resoluton dgtal camera (Sony DFW-X7) has been selected for camerabased navgaton development. The mage processng algorthms wll be nternally developed by GMV. E. INTEGRATED TEST BENCH The result after the test bench ntegraton s shown n the followng fgures. Fgure 4. Statc mock-up close to the robotc arm and ts own mock-up Fgure 5. S/C mock-up mounted at the top of the robotc arm
III. PLATFORM TEST BENCH CALIBRATION The correct and accurate calbraton of the test bench s a key drver on the valdaton level that wll be acheved by the test bench. Snce the calbraton resdual error wll be drectly added over the performance fgure evaluated for the GNC functons, t s fundamental to keep the test bench calbraton resdual as low as possble (target fgure s a few mllmeters). The test bench set-up s calbrated n sequental steps accordng to an ncremental approach by ncludng the dfferent equpments on the test bench. The set-up sequence s as follows: 1) Robotc arm placement. 2) Satellte mock-up ntegraton at the tp of the robotc arm. 3) Hostng of the mock-up on-board equpments: GPS antennas and camera for vson-based navgaton. 4) Statc mock-up and assocated on-board equpment set up. 5) Emttng GPS pseudoltes set up on the bay upper structure. The test bench calbraton procedure s, then, defned accordng to the equpment set up sequence: 1) Laser calbraton staton reference frame defnton: orgn + reference frame axes drecton. 2) Robotc arm calbraton ncludng the satellte mock-up, that wll be consdered as part of the robotc arm for calbraton purposes. Ths calbraton shall consder ether the robotc arm mountng errors as the robotc arm nternal artculaton mountng and control errors (bases and scale factor over the commanded artculatons rotatons) as the mock-up mountng errors. 3) Mock-up GPS recevng antennas phase center calbraton. 4) Emttng GPS pseudoltes antennas phase center. 5) Camera for vson-based navgaton: calbraton of the focal pont poston and related reference frame. 6) Statc mock-up and assocated on-board equpment placement calbraton: the procedure wll be analogous to the mock-up mounted at the tp of the robotc arm. In the followng sectons, some detals on the calbraton procedures and performed tasks are ncluded. A. ROBOTIC ARM CALIBRATION The robotc arm calbraton procedure s splt n two phases: the calbraton error measurement and the compensaton of ths calbraton error. The purpose of the calbraton operatons s to reduce the gap exstng between the robot smulaton world and the real world. In fact, the robot controller works wth a nomnal model that, wthout the calbraton, can be qute dfferent from the real robot. To provde a lnk between smulaton and the real world the calbraton routnes are performed both on the robot arm and on the possble tools and the work envronment. The pose term s commonly used to refer the poston and orentaton of an object. In three dmensons, the pose s gven by the sx-tuple [x, y, z, roll, ptch, yaw]. The robot calbraton may be manly performed n two ways: 1) Pose error measurement: through the controller commands, the robot s requested to move to a partcular poston. External contact-less measurement tools are used to measure the true pose. The dfference between the measured and commanded locaton s the pose error.
2) Pose matchng: the robot s drven to a known locaton and the pose calculated by the robot controller s recorded. The dfference between the known pose and that calculated by the controller leads to the pose error. The two poses are dfferent because the controller uses the nomnal robot model and measured jont angles at the encoders to calculate the pose, both dfferent from the real robot world. The frst method s selected for calbraton of the robotc arm, where the external contact-less measurement tool selected s a laser staton composed by a rangng and a theodolte devce. The procedure for the robotc arm calbraton s shown n Fgure 6 and s composed by the followng phases: 1) Set the laser calbraton system reference frame. 2) Take a reference pont on the robot or the attached satellte mock-up. 3) Wth the manufacturer provded (un-calbrated) robot commandng model, command a few number of robot arm confguratons. 4) Retreve the poston coordnates (n Robot-related Reference Frame or RRF) of the reference pont for each commanded statc confguraton through the robotc arm control model. Ths model s based on the descrpton provded n Table 1 by sequentally applyng rotatons around the sx artculatons. 5) Use an ndependent measurement equpment (laser rangng and theodoltes) to measure the poston coordnates (n Laser-related Reference Frame) of the reference pont poston for each robotc arm commanded statc confguraton. 6) Estmate through mnmum resdual flters technques, the rotaton matrx between the robotc-related reference frame (RRF) and the laser measurement system related reference frame [ A ] est RRF Laser. 7) The estmated rotaton matrx allows to pass laser measured coordnates to RRF frame or vce versa (poston coordnates retreved from the robotc arm controllng model n RRF to be translated nto Laser reference frame) to be able to compare the predcted (through robotc arm control model) and measured coordnates, so to be able to detect and estmate the nternal robotc arm mechansms calbraton errors. 8) Wth the manufacturer provded (uncalbrated) robot-commandng model, command a hgh number of robot arm confguratons. 9) Retreve the poston coordnates (n Robot-related Reference Frame or RRF) of the reference pont for each commanded statc confguraton. 1) Estmate the robotc arm calbraton parameters usng optmzaton-flterng technques. 11) Update the robotc arm-controllng model wth the estmated calbraton parameters. 12) Repeat the steps 5, 6, 1 and 11 untl convergence of the resdual n the dfference between the robotc arm control-model predcted coordnates and the laser measured coordnates. The consdered robotc arm related calbraton parameters are (18 total consdered parameters): 1) RRF orgn devaton n Laser-related reference frame (3 coordnates). 2) RRF atttude devaton wth respect to the deal mountng one. real bas θ = θ + 1 + f 3) Sx robotc arm artculatons angular bas errors. = 1,..6 4) Sx robotc arm artculatons commanded angle scale factor errors. 5) Reference pont devaton n RRF (3 coordnates) scale ( ) θ commanded
Laser-based Independent Measurements Uncalbrated Robot Arm Moton Model robot { x, y, z} = f ( θ ) = 1..6 1 robot { ϕ, θ, ψ} = f ( θ ) = 1.. 6 2 Frst set of robot moton commands: 5 samples Estmaton of Rotaton Matrx between Robot and Laser references [ A ] est RRF RRF Laser est Iteraton Second set of robot moton commands: 4-6 samples Laser-based Independent Measurements Estmaton of Robot Arm Calbraton Paramenters robot bas scale r cal { x, y, z} = f ( θ, θ, f, δ ) = 1..6 1 RRF robot bas scale r cal { ϕ, θ, ψ} = f ( θ, θ, f, δ ) = 1.. 6 2 Calbrated Robot Arm Moton Model RRF Fgure 6. Robotc arm calbraton scheme Some smulatons have been performed n order to evaluate the accuracy level achevable wth ths knd of scheme wth the followng characterstcs: 1) RRF orgn devaton ntroduced: 5 cm (1σ) on each coordnate. 2) RRF atttude devaton wth respect to the deal mountng one: 5 (1σ) on each axs. 3) Sx robotc arm artculatons angular bas errors: 1 (1σ) on each artculaton. 4) Sx robotc arm artculatons commanded angle scale factor errors: 1% (1σ) on each one. 5) Reference pont devaton n RRF: 5 cm (1σ) on each coordnate. 6) Laser system measurements accuracy:.2 mm (1σ). The results have shown that usng 5 confguraton samples for rotaton matrx estmaton between RRF and Laser-related frames and between 4 and 6 confguraton samples for the dfferent consdered robotc arm calbraton parameters, t s acheved an optmum estmaton (optmzaton methods provded wthn Matlab tool have been used) wthout overchargng the processor CPU. The convergence s acheved untl 15-2 teraton steps. The accuraces achevable for the calbraton parameters accordng to the performed smulatons are shown n the followng fgures.
1.8 SET OF ROBOT TIP POINTS USED FOR CALIBRATION CONVERGENCE OF THE ESTIMATION OF THE ANGULAR BIASES IN THE ROBOT ARTICULATIONS 4 1 st Artculaton 2 Artculaton nd 3 3 rd Artculaton 4 th Artculaton 5 th Artculaton 2 Z RRF (m).6.4.2 -.2 Resdual error (deg) 1-1 -.4 1-2.5 Y RRF (m) -.5-1 -.8 -.6 -.4 -.2 X RRF (m).2.4.6.8-3 -4 5 1 15 2 25 3 35 4 Number of teratons Fgure 7. Set of reference pont postons commanded durng step 8 Fgure 8. Error n the estmaton of the angular bases of the commanded artculaton rotatons.15.1 ERROR IN THE KNOWLEDGE OF THE POSITION OF THE REQUIRED MOCK-UP POINT CONVERGENCE OF THE ESTIMATION OF THE ANGULAR SCALE FACTOR IN THE ROBOT ARTICULATIONS 1 1 s t Artculaton 2 n d Artculaton 3 Artculaton r d 8 4 Artculaton t h 5 t h Artculaton 6 t h Artculaton.5 6 Z RRF (m) -.5 -.1 Resdual error (%) 4 2 -.15.3.2.1 Y RRF (m) -.1 -.2 -.3 -.2 -.1 X RRF (m).1.2-2 5 1 15 2 25 3 35 4 Number of teratons Fgure 9. Error n the reference pont poston estmaton at the end of the calbraton process Fgure 1. Error n the estmaton of the scale factor applcable to the commanded rotatons The statstcal results correspondng to the error n the reference pont poston estmaton at the end of the calbraton process (Fgure 9) are (n meters): Mean Standard Devaton X coordnate -.4874 ±.9996 Y coordnate -.61 ±.9278 Z coordnate.1576 ±.7466 Table 2. Expected calbraton accuracy (meters) for the reference pont at the tp of the robotc arm B. RF PSEUDOLITES AND GPS RECEIVERS The calbraton of the GPS-lke pseudoltes and the recevers s made together n a common calbraton process. The reason s that each of the equpments s necessary to calbrate the other one and, then, the collected measurements contan both pseudoltes and recevers errors. The man elements that are calbrated are: 1) Center of phase of the pseudoltes emttng antennas. 2) Delay tme between the GPS-lke sgnal generaton and the antenna emsson.
3) Center of phase of the mock-up recevng antennas. 4) Delay tme between the GPS sgnal recepton at the antenna and the nserton nto the recever processor. 5) Multpath effects over the receved sgnals at the mock-up antennas. Snce the test bench s hosted ndoor, and although the clear avalable envronment (buldng room of 25x25x8 meters wthout any nternal metallc structure) n terms of obstacles that could be the source of the multpath reflectons and dffractons, the multpath effect s expected as one of the major error sources n the calbraton process. Snce the envronment s frozen and wll mnmally change durng testng campagns, a multpath calbraton campagn based on the repeatablty of the multpath s currently beng carred out. 6) Desync hronsaton between pseudoltes clock and recevers clock. In ths case, ths parameter s part of the relatve navgaton state vector (as t would be n a real msson wth multple vehcles hostng RF emttng/recevng devces) and wll be estmated through the relatve navgaton algorthms. All above elements wll be estmated and calbrated through the use of the measurements collected from the recevng antennas wth the help of the laser theodolte tool that wll allow accurately measurng the real poston of the dfferent equpment elements and comparng aganst the collected sensors measurements. Dfferent measurements combnatons wll be created to solate some effects from the others. C. CAMERA FOR VISION -BASED NAVIGATION The calbraton of the camera s performed wth the help of the laser theodolte measurement tool by matchng the estmated poston and atttude of the camera through the mage processng wth the real poston and atttude measured by the theodolte. IV. PLATFORM TEST BENCH CUR RENT STATUS PLATFORM test bench s not stll fully operatve. Last phases of ntegraton and calbraton campagn are on gong and t s expected to be fully operatve, wth capacty to nsert GNC algorthms able to nterface wth the sensors and the robotc arm control, by md-25. Collaboraton wth other companes regardng, manly, gudance and control algorthms development and valdaton, sensng and/or actuators testng and valdaton, wll be very welcome by GMV, that wll allow nterested companes to use PLATFORM test bench wth no more constrants than calendar schedule and personnel supportng. For further nformaton, please get n contact trough the followng e-mal address: platform-testbench@gmv.es. V. CONCLUSIONS AND FUTURE TEST BENCH DEVELOPMENTS Formaton flyng and Rendez-vous and Dockng (RvD) technologes wll have a predomnant role n the comng space mssons, partcularly focused on planetary exploraton. To demonstrate the requred technologes n space before ntense ground testng s expensve and rsky. Some way of ground smulaton under close to real space condtons s needed. The proposed test bench tres to fll ths gap n the path from concept to real msson demonstraton. The proposed test bench s at the same tme flexble and relable. Flexble because t allows to test dfferent types of mssons under dfferent envronment condtons, and relable because t nvolves
real hardware equpment under quas-real physcal confguraton. Regardng the flexblty, t shall be hghlghted that, although the drect applcaton scenaro (the one that has been kept n mnd durng the orgnal proposal) s a short-range (1-25 meters) formaton flyng scenaro composed by two platforms, our calbrated scenaro offers the nvaluable feature of beng a repeatable scenaro. Ths means that a multple vehcle scenaro could be generated by superposng several test bench confguratons. Specal care shall be taken nto account on how the spurous effects are amplfed by the superposton. Other scenaros, as medum and long-range formaton flyng scenaros may be recreated by usng scalable scenaros. In the same way, the medum and long-range phases of a RvD scenaro (e.g. probe RvD wth return vehcle n the Mars Sample Return scenaro) may be acheved through scalable scenaros, whle the short-range RvD phase (ncludng the contact) s drectly acheved by placng the statc (target) spacecraft close enough to the dynamc controlled (chaser) spacecraft. Fnally, the test bench has been conceved to be able to ncrease ts complexty and capabltes n the future wth the ncluson of actuators, new GNC algorthms, new sensors or more smulated satelltes. Future extensons of PLATFORM test bench foresee: 1) Cooperatve robotc operatons on a planetary surface exploraton wll requre for a relatve navgaton system that can be based on the same concept as the formaton flyng, ths s emttng/recevng RF devces coordnated by a reference devce hosted on the landng platform. Addng real robotc prototypes to the test bench, the robotc arm may reproduce the landng sequence and platform aperture and the on-board robots may descend nto a model of the planetary surface and test and valdate the proposed autonomous navgaton and robot fleet management algorthms. 2) Landng scenaro recreaton through the modelng of the e.g. Mars surface. In ths case, the robotc arm (wth extended dynamc through the use of a track moton) wll create the dynamc approxmaton to a surface model. Landng navgaton based on optcal camera observaton may be tested n such scenaro. VI. REFERENCES Mguel Angel Molna, Carmelo Carrascosa, Pablo Colmenarejo, Fernando Ganda, Valentín Barrena, Alberto García-Casas. PLATFORM: the GMV s Test-bench for Formaton Flyng, RvD and Robotc Valdaton. 2nd Internatonal Formaton Flyng Symposum. 24. Washngton, Unted States. Fernando Gandía, Alberto García Casas. Fast Spacecraft Pose Estmaton based on Zernke moments. 7th Internatonal Symposum on Artfcal Intellgence Robotcs and Automaton n Space (ISAIRAS). 23. Nara, Japan. C. Bourga, C. Mehlen, P. Colmenarejo et al. Autonomous Formaton Flyng RF Rangng Subsystem. ION GPS/GNSS 23. September 23. Portland, Oregon. F. Chavez and D.K. Schmdt, Usng Nanosatelltes for Research n Formaton Flyng and Relatve Navgaton. Unversty of Colorado. H.F. Messnger et al, Low-Cost, Mnmum-Sze Satelltes for Demonstraton of Formaton Flyng Modes at Small, Klometer-sze Dstances. 13th AIAA/USU Conf. on Small Satelltes. J. Adams, Technologes for Spacecraft Formaton Flyng. Stanford Unversty. J.M. Stone et al, GPS Pseudolte Transcevers and ther Applcatons. ION-Natonal Techncal Meetng. 1999. K. Zmmerman y R. Cannon, Expermental Demonstraton of GPS for RV between two Prototype Space Vehcles. Stanford Unversty.