The DLR On-Orbit Servicing Simulator: Reproducing Free-Floating Dynamics with Robotic Facilities

Transcription

1 Workshop on Gravity Offload Testbed for Space Robotic Mission Simulation September 24, 2017 The DLR On-Orbit Servicing Simulator: Reproducing Free-Floating Dynamics with Robotic Facilities Marco De Stefano and Jordi Artigas DLR (German Aerospace Center) Institute of Robotics and Mechatronics, Germany

2 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 2 Contents 1. Motivation: On-orbit servicing 2. DLR On-ground facilities for simulating free-floating dynamics Light-weight-robots based OOS-SIM facility Main control modes 3. Factors that affect the free-floating dynamics simulation with a robot Time delay Discretization 4. Reproducing free-floating dynamics: An Energy-based approach

3 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 3 Contents 1. Motivation: On-orbit servicing 2. DLR On-ground facilities for simulating free-floating dynamics Light-weight-robots based OOS-SIM facility Main control modes 3. Factors that affect the free-floating dynamics simulation with a robot Time delay Discretization 4. Reproducing free-floating dynamics: An Energy-based approach

4 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 4 On-Orbit Servicing Maintenance and life extension of existing satellite Active space debris removal Servicer satellite equipped with a robotic arm (left) approaching a client satellite (right) Density of Space Debris in LEO and GEO orbit

5 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 5 Why Robotic Facilities? 6 dof dynamics simulation Mass/inertia pars. can be easily changed Reproduction of microgravity Large workspace Space Scenario: Servicer satellite with manipulation arm and client satellite On-ground scenario: The robot simulates the dynamics of the satellite

6 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 6 On-ground Robotic Simulator in DLR-RM LWR-2 based LWR-3 based OOS-SIM OOS-SIM+ Servicer LWR-2 Client LWR-2 Fixed-base No free-floating Servicer LWR-3 Client LWR-3 Fixed-base Free-floating em Servicer KR-120 Client KR-120 Arm LWR4+ Free-floating dynamics Servicer KR-120 Client KR-120 Arm LWR4+ Stereocamera Haptics interface End2End

7 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 7 On-ground Robotic Simulator in DLR (2011) Maintenance and life extension of existing satellite Active space debris removal

8 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 8 On-ground Robotic Simulator in DLR-RM: the OOS-SIM Light-Weight Robot (LWR) Gripper and stereocamera system Force-Torque Sensor (FTS) Force-Torque Sensor (FTS) Servicer satellite: KUKA KR-120 Client satellite: KUKA KR-120

9 On-ground Robotic Simulator in DLR-RM: the OOS-SIM Light-Weight Robot (LWR) Gripper and stereocamera system Force-Torque Sensor (FTS) Force-Torque Sensor (FTS)

10 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 10 The DLR OOS-SIM q Data flow of the target simulator Force-torque sensor Simulated Rigid-body Dynamics Industrial Robot q Data flow of the free-floating robot simulator Force-torque Sensor Light-Weight Robot Free-Floating robot Dynamics Industrial Robot Control modes

11 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 11 Control modes The LWR can be controller in position or torque mode for different operations: q Semi-autonomy Stereo camera at the end-effector Visual servoing q Telepresence Remote human operator with haptic interface Teleoperation with forcefeedback q Shared control torque input from the visual-servoing and telepresence

12 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 12 Space link experiment Passive bilateral controllers have been developed to cope with time delay and was tested on a space link infrastructure Targeted scenario Implemented scenario

13 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 13 The OOS-Sim facility

14 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 14 Contents 1. Motivation: On-orbit servicing 2. DLR On-ground facilities for simulating free-floating dynamics Light-weight-robots based OOS-SIM facility Main control modes 3. Factors that affect the free-floating dynamics simulation with a robot Time delay Discretization 4. Reproducing free-floating dynamics: An Energy-based approach

15 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 15 Reproducing free-floating dynamics with Robotic Facilities 3. Factor that affect the free-floating dynamics simulation on a robot 4. Reproducing free-floating dynamics: An Energy-based approach Time delay Time delay between measured force-torque and command to the robot causes system instability Virtual energy is generated due to intrinsic latencies Discretization Standard Euler Integrator leads to generation of energy and position drifts Implicit integration methods require a numerical and iterative solution, Iterative solutions can be prohibitive for real-time determinism.

16 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 16 Time Delay: Problem Statement

17 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 17 Time Delay: Problem Statement Energy without time delay, with time delay and angular velocity comparison

18 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 18 Time Delay: Proposed Approach Control goals 1. Ensure a stable system with the passivity condition 2. Guarantee the performance of the simulated dynamics on a robot Passivity Condition: E(m)=E(0)+ m F T (k)v(k) T 0, k=0 If E(m) < 0, then the system is producing energy and such a regenerative effect can destabilize the system

19 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 19 Modeling: Making the system passive 1. Block diagram F E v 1 (t) v 1 (t T d ) SAT.DY N. TD R E v 1 (t) Z R 2. Electrical and Network scheme F Z M v 1 (t) v is F C Z C F Z E N 0 N 1 N 3 N 4 v 1 (t) v 1 (t T d ) v 2 (t) Z R 3. Energy leaks can be seen in electrical network F + + v 1 (t) Z M TD F C Z PC v is Z C + + F Z E N 0 N 1 N T N 2 N 3 N 4

20 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 20 Ensure the stability of the robot with the passivity criteria Passivity Controller (PC) acts if passivity condition is violated v 2 (k)=v 1 (k µ) β (k)f c (k) Passivity Observer (PO) Electrical and Network scheme E obs (m)= m Fc T (k)(v 1(k) v 1 (k µ)) T = k=0 v 1 (t) v 1 (t T d ) v 2 (t) Z R 3. Energy leaks can be seen in electrical network F + + v 1 (t) Z M TD F C Z PC v is Z C + + F Z E N 0 N 1 N T N 2 N 3 N 4

21 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 21 Ensure the stability of the robot with the Passivity criteria

22 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 22 Time Delay: Proposed Approach Control goals 1. Ensure a stable system with the passivity condition 2. Guarantee the performance of the simulated dynamics on a robot v 1 (k) TD v 1 (k µ) PC v 2 (k) R E F E SAT.DY N v 1 (k) PO E obs (n) β i MIN F c Γ FC F E F c

23 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 23 Guarantee performance with a designed problem optimization Performance is guaranteed through a multidimensional optimal damping as a result of a minimization problem: min v 1(k µ) β (k)f c (k) v 1 (k) 2 β (k) The minimization problem will force the velocity to stay as close as possible to the ideal value. The constraints to satisfy are: m k=0 F c(k) T β (k)f c (k) T = 0 if E obs (m) 0 E obs (m) if E obs (m) < 0. The minimization problem generates a β (k) such that the active energy (produced by the delay) is dissipated and the velocity transmitted to the robot is as close as possible to the ideal target velocity.

24 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 24 Time Delay: Experiment Results The observed active energy is dissipated with the optimal damping and the system results to be stable. [J] [J] [J] Observed Energy Energy of the Passivity controller x Net Energy: Passivity proof time [s] Observed Energy, Energy of the passivity controller and Passivity proof

25 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 25 Time Delay: Experiment

26 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 26 Reproducing free-floating dynamics with Robotic Facilities 3. Factor that affect the free-floating dynamics simulation on a robot 4. Reproducing free-floating dynamics: An Energy-based approach Time delay Time delay between measured force-torque and command to the robot causes system instability Virtual energy is generated due to intrinsic latencies Discretization Standard Euler Integrator leads to generation of energy and position drifts Implicit integration methods require a numerical and iterative solution, Iterative solutions can be prohibitive for real-time determinism.

27 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 27 Discretization P[W] A Admittance architecture of the robot simulator: time [s]/samples [T] A B F(k) Des. Sim.Dyn Dyn. v (k) T v(k) R E a b P(k 1) e d P(t) a T P(k 1) e P(t T) c b P(t T) d The desired dynamics is subjected to external forces, T c P(t) Explicit and Discrete Integrator strategy by modifying the output of the Euler integrator. Discretization Standard Euler Integrator leads to generation of energy and position drifts Implicit integration methods require a numerical and iterative solution, Iterative solutions can be prohibitive for real-time determinism.

28 20 with the10euler integration 0 explicit passive integrator. 10 of the 20 paper is twofold. F(k) F [N] 0.15 tion100 for adjusting the velocity output of the Euler integrator behavior of a rigid body using the TDPA. An analysis of the energy behavior of the continuous and discrete time systems is presented and the discrete-system 5 0 sampling passivity-based integrator scheme is introduced asily implemented in15 stane show Fig. 2: Admittance architecture with the desired dynamic A. Energy produced 0 5by the 10 Euler 15 integration method that x the proposed R is the robot, E is the environment, Des. Dyn. is the forceacceleration model Consider the dynamics (1) discretized by means of the 10 desired performance on a Euler of the method dynamics and reported toin implement, (5). The discretetkinetic Σ isenergy the H(k) is given by: IM in Fig. 1. discrete integrator with time step T drift w T1 [m] drift w T2 [m] time [s] in simulating the desired dynamics with a discrete integrator that is usually implemented for rendering a desired dynamics with a robot. The robot will receive position commands accordingly but, as it has been shown, the energy properties of the simulated mass will be not preserved. Such a drift may lead the robot to interact to unforeseen objects that produce new (drifted) behaviors leading to a deteriorated performance Ḣ of the = system. F T M 1 p = F T v, (3) The goal of this work is to design a controlled Euler integration method that preserves the energetic balance in (3) in the discrete case. In this way, it will be possible to reproduce by the robotic simulator the behavior (1) while preserving its energetic properties independently of the sample time. roduced and then we will where x R 3 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, represents the Cartesian configuration and ẋ = step. The integrated value v. 0,I R 3 3 J. Artigas Sept. 24, 2017 Slide 28 are the null and the identity matrix respectively. to dissipate Discretization: the produced As for any port-hamiltonian system without damping, we Problem Statement tem a passive integrator. In have that the following balance holds [9]: integrator, we will exploit Considering the Hamiltonian system: oach (TDPA) proposed in or damping out the excess which represents the fundamental energetic property of any ironment. Recently, Standard TDPA Euler method undamped causes mechanical drift in position system, and namely energy that inconsistency the power due to the impedance that can be the interaction with the environment is energetically stored in tic interfaces [19]. In our III. THE PASSIVITY-BASED INTEGRATION METHOD ed explicit integrator that Des. Dyn. Force profile, drift in position due to the discretization with different sampling time H [J] 160 As shown in Sec. II, the extra energy due to the discrete integration makes the energy behavior of the discrete dynamic system different from that of its continuous counterpart. In this 120 v section, (k) we formally identify the extra energy introduced by the discrete integrator and we exploit this 15 v(k) informa- T H(k)= 1 2 p(k)t M 1 p(k)= time [s] H c H T1 H T2 1 2 [p(k Energy 1)+TF(k drift 1)]T for M 1 different [p(k 1)+TF(k sampling 1)], (7) time and comparison with the continuous case H c. R E

29 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 29 Energy produced by the Euler Integrator The energy variation should be due only to the energy provided through the port (F,v), The Energy due to Euler method is: } H(k)=H(k 1)+Tv(k 1) T F(k 1) T 2 F(k 1) T M 1 F(k 1) }{{} H H is the active energy introduced at each time step T, The active energy will cause a non-physical and a non-passive behaviour.

30 Similarly, m in part B of Fig. 6, E is given by the sum of the +β (k 1)F(k 1) T, IROS 2017 areas Workshop of rectangle The DLR On-Orbit abde and Servicing triangle Simulator bcd, whichresultsin: 1) 2 (11) where value of β is E a time-varying damper later discussed. We can d.consideringthe2-portasshowninfig.7,the M. De Stefano, J. Artigas Sept. 24, 2017 Slide 30 E H represent = energy F(k 1)T[v(k observer the discretization 1) then v(t T)]. becomes, (13) problem with a network analogy proposed in Fig. 7 where E Equation (12) and (13) represent analytically the error between E(12) and H continuous (shown integration scheme c can be seen as the energy in ) v(t)]. Passivity-based E Fig. time 5). Thewhich, equationsdue clearly to the discretization, assume a show that (11) obs (k)=e obs (k 1)+E c (k) E d (k) as the sampling rate increases, E gets closer to by the H sumsince of v(k the value of E d.consideringthe2-portasshowninfig.7,the 1) gets closer to v(t) +β and(k v(t 1)F(k T). Inthe 1) graphical The analysis, energy energy this turnsobserver to be a smaller then checks areabecomes, 2 T,,whichresultsin: of rectangle the energy flows: abde. = E obs (k 1) E(k)+β (k 1)F(k 1) 2 T, causes H). v(t)]. T)]. This analysis (12) (13) shows that there is always a difference between E what we can estimate in obs (k)=e E real time, namely obs (k 1)+E H(k)+β H, andthereal c E d (k) 1)F(k 1) 2 T, 0 ically y the the sum difference error of Thus, the beehichresultsin: equations clearly (16), as follows: between if activethenergy energetic is detected behavior(i.e. E The passivity controller +β (kin acts the 1)F(k obs discrete < 0), the corrected 2 velocity is sent to the robot as follows: in admittance 1) 2 T, (16) case and in the real case. This error is due to the loss of configuration: 4 information related to the{ discretization = E obs process (k 1) and it E(k)+β can (k 1)F(k 1) 2 T,, E gets 6 notcloser be avoided. to { Such a difference v(k) gets β (k)f(k) smaller asethe obs sample < 0 v c (k)= (19) 8 E obs (k) (t T)]. (13) where E(k) has been approximated to H(k) as per E E time gets lower. β (k)= F(k) v(k) else. 2 T obs (k) < 0 obs (k 1) H(k)+β (k 1)F(k 1) 2 equation (11). Notice that the energy observer will measure d v(t T). Inthe T, an 0 else. er llyarea theof However, error rectangle beequations 0.02 adjusting the output velocity for recovering active energy as soon as there is an external 0.04 force (which (16) passivity where of the v(k) discrete achieved model has by (6). several Therefore, advantages. thefirst, observed energy aphysicalbehaviorofthediscretedynamicsisensured.the clearly will E causes obs (k) 0makingthenetwork,i.e.theintegrator, H). quantity: 0 difference E getsevolution closer between passive. will to be Aclose where schematic At toeach the E(k) ideal of integration the one has integration the been limits stepapproximated scheme reported E obs must is in Fig. betogreater 8. H(k) than as per zeroequa- tion F(k) ensuring damping (11). βthe Notice is modulated passivity. that v(k) by the Therefore, thenergy observed it observer v(k) is active possible will for F(k) 0.02 lyv(t H, T). andthereal in (12). Inthe Second, a stable interactive behavior is achieved The variable Sim. Sim.Dyn Dyn. T PC v 0.04 tomeasure define thanks to the passivity of the discrete dynamics. c(k) thean ior areain ofthe rectangle discrete energy andactive the time-varying applied energy force damper aswhich soonwill β as (k), activate there function the is an PC ofexternal (in theobserved force energy (which R E C. Passive 15 (18), Integration (19)) to provide scheme due to the loss of the corrected velocity v c to the robot. x 10 5 causes (16), as H). follows: v(t) v(k) 10 TDPA is a passivity ensuring tool widely applied in ifference process β 5 the and between fields it can of haptics At and eachtime-delayed integration teleoperation. { step E obs Themust be greater E c F(t) Discretization than zero for F(k) F(k) 0 H, aller andthereal as underlying sample principle ensuring of TDPAthe is topassivity. observe the Therefore, input E obs (k) and E it is possible to define F(k) the F(k) v(k) v(k) Sim. Dyn. T E obs PC v c(k) 5 output energy flow (with the Passivity Observer) of a singleport network, (thetime-varying virtual environment, r in the discrete β (k)= F(k) 2 T obs (k) < 0 R E (17) v(k) damper in case0β of (k), haptics) function else. oftheobservedenergy ue fortorecovering the Discrete Syst. or aloss 2-portof the network (16), (the communication as follows: channel with β delay, F(k) locess advantages. and in case it of First, can teleoperation) [20]. The The passivity passivity-based condition for a integration scheme F(k) two-port network isthe givenvelocity by: E ler icsisensured.the as the sample obs corrected { by the PC is given by the following nt E obs (k) quantity: E 0 Fig. 8: The passivity-based integration scheme. (F k=0 The 1 (k)v energy 1 (k)t + F 2 (k)v observer 2 (k)t β (k)= F(k) 2 T obs (k) < 0 the limits reported )+E(0) 0, (14) (17) havior (E0 obs ) measures else. the active 0.2 r recovering is achieved the v pc (k)=β(k)f(k). energy (18) 0.4 advantages. namics. First, IV. SIMULATIONS 0.6 The velocity corrected by the PC is given by the following sisensured.the 0.8 The method quantity: is firstly verified in simulation where the widely e limitsapplied reported v(t) v(k) proposed in integration scheme is applied. We consider the = E obs (k 1) E(k)+β (k 1)F(k 1) 2 T, E obs (k 1) H(k)+β (k 1)F(k 1) 2 T, v c (k) time [s] The passivity control (PC) corrects the velocity with a variable damper β vpc [m/s] Eobs w/o PC[J] Eobs wpc[j] Fig. 9: Sim. 1 - E obs without (w/o) PC, velocity correcte by the PC and E obs with β (E(w) PC for T 1 = 0.1s. obs ) Eobs w/o PC[J] x 10 3 (16) where E(k) has been approximated to H(k) as per equation (11). Notice that the energy observer will measure an active energy as soon as there is an external force (which At each integration step E obs must be greater than zero for ensuring the passivity. Therefore, it is possible to define the time-varying damper β (k), functionoftheobservedenergy (17) The velocity corrected by the PC is given by the following v pc (k)=β (k)f(k). (18) Fig. 7: Continuous energy (E c )anddiscreteenergy(e d ). Analogue discrete system with the designed variable damper. E d

31 middle). Also for the case with T 2,thedriftappears.Sincethe sampling time is smaller, it results in a drift 10 times lower, as shown in Fig. 3 bottom. This drift causes inconsistency IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 31 Results: Energy comparison The energy is preserved for the simulated rigid body H c H T1 H T H c H T1 H T H [J] H [J] time [s] Fig. Energy 4: Problem drift statement: considering mechanical different energy sampling considering time: different before sampling applying time the (H T1 method and H T2 )andcomparisonwith the continuous case H c time [s] Energy considering different sampling time: with the proposed method

32 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 32 Experiments on the robot simulator

33 H c H T 1 H T 2 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 33 Passive integrator: extention to coupled dynamics 0 Energy generated by the Euler integrator: H[J] 3: E obs (k)=e(k 1) (k)+tχ(k 1) T where β (k H(k 1)χ(k 1)= 1) 1 2 ω(k 1)T Iω(k 1). Thus,itis 10 to identify the total energy generated during th 4: if E obs (k) < 0 then integration for the rotational dynamics as follows: 5: β (k)= E obs(k) T χ(k) 2 H(k)=H(k 1)+Tω(k 6: else β (k)=0 1) T τ(k 1) H(k)=H(k 1)+Tω(k 1) T τ(k 1) T 2 ω(k 1) T S(Iω(k Fig. 7: 1)) Sim. T I 1 1S(Iω(k with T 1)) 1 - }{{ correction and passivity T 2 ω(k 1) T S(Iω(k 1)) T I 1 S(Iω(k 1))ω(k 1) + 1 }{{} 2 T 2 τ(k 1) T I 1 τ(k 1). 7: ω c (k)=ω(k) β (k)χ(k) }{{} H 1 H Notice that (8) does not correspond 0.15 to a physical beha 0.2 fact, the energy variation should be due only to the 0.25 time[s] provided through the power port, i.e. Tω(k 1) T τ Rob. τ Fig. 4: Problem Statement: inv.kin. energy E in continuous Rot. Dyn. ω ω time (H c ) T and discrete times (H T 1, H T 2 ). 1 The terms on the even lines of (6) will result to be equal and ω in sign. c 0.2 PC χ E obs β Fig. 6: Admittance scheme (dashed line) with the new elements (dotted line). 4 The dependency of Iω is omitted in the skew matrix S(Iω) for brevity. Eobs wpc[j] Eobs w/o PC[J] ωpc [deg/s] 20 x 10 5 Eobs wpc[j] H x M. De Stefano, J. Artigas, C. Secchi, " A passive Integration Strategy for Rendering Rotational Dynamics on a Robotic Simulator, IROS TuAT12, Room 208 Fig. 8: Sim. 2 with T 2 - correction and passivity

34 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 34 Passive integrator extended to coupled dynamics Method extended for rotational dynamics M. De Stefano, J. Artigas, C. Secchi, " A passive Integration Strategy for Rendering Rotational Dynamics on a Robotic Simulator, IROS TuAT12, Room 208

35 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 35 Summary The OOS-SIM was presented as an on-ground facility for testing on-orbit servicing tasks Autonomy, Telepresence and Shared-control algorithms can be tested Main issues in rendering passive free-floating dynamics are addressed Time delay and discretization can lead to a non-physical behaviour of the simulated dynamics Energy-based methods have been develop to realistically simulate physical dynamics

36 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 36 Acknowledgment The researchers made this possible! Roberto Lampariello Ribin Balachandran Wolfgang Rackl Phillip Schmidt Giorgio Panin Alessandro Giordano Michael Panzirsch Martin Stelzer Bernhard Brunner Robert Burger Florian Schmidt Wieland Bertleff Cristian Secchi Christian Ott Alin Albu-Schaeffer

37 IROS 2017 Workshop The DLR On-Orbit Servicing Simulator M. De Stefano, J. Artigas Sept. 24, 2017 Slide 37 References [1] J. Artigas, M. De Stefano, W. Rackl, R. Lampariello, B. Brunner, W.Bertleff, R.Burger, O. Porges, A. Giordano, C.Borst, and A. Albu-Schäffer, The OOS-SIM: An On-ground Simulation Facility For On-Orbit Servicing Robotic Operations. in Proc. of the 2015 IEEE International Conference on Robotics and Automation (ICRA), Seattle, USA, pp , May [2] Marco De Stefano, Jordi Artigas, Wolfgang Rackl and Alin Albu-Schäffer " Passivity of Virtual Free-Floating Dynamics Rendered on Robotic Facilities", in Proc. of the 2015 IEEE International Conference on Robotics and Automation (ICRA), Seattle, USA, pp , May [3] Marco De Stefano, Jordi Artigas, Alessandro Giordano, Roberto Lampariello and Alin Albu-Schäffer On- ground experimental verification of a torque controlled free-floating robot. 13th Symposium on Advanced Space Technologies in Robotics and Automation 2015 (ASTRA 2015), ESA/ESTEC Noordwijk, Netherlands, May 2015, [4] Phillip Schmidt, Jordi Artigas, Marco De Stefano, Ribin Balachandran, Christian Ott. Increasing the Performance of Torquebased Visual Servoing by applying Time Domain Passivity, in Proc. of the 11th IFAC Symposium on Robot Control SYROCO 2015, Salvador, Brazil, vol.48 pp , August [5] Jordi Artigas, Ribin Balachandran, Marco De Stefano, Michael Panzirsch, Jan Harder, Juergen Letschnik, Roberto Lampariello, Alin Albu-Schaeffer, Teleoperation for On-Orbit Servicing Missions through the ASTRA Geostationary Satellite, in Proc. of the 2016 IEEE Aerospace Conference, Big Sky, Montana, USA, pp. 1-12, March 2016 [6] Marco De Stefano, Jordi Artigas, Cristian Secchi, " An optimized passivity-based method for simulating satellite dynamics on a position controlled robot in presence of latencies", in Proc.of the 2016 IEEE/RSJ International Conference on Intelligent Robot and System (IROS), Daejeon, Korea, pp , October [7] Marco De Stefano, Ribin Balachandran, Jordi Artigas, Cristian Secchi " Reproducing Physical Dynamics with Hardware-in-theloop Simulators: A Passive and Explicit Discrete Integrator ", in Proc. of the 2017 IEEE International Conference on Robotics and Automation (ICRA), Singapore, pp , May [8] Marco De Stefano, Jordi Artigas, Cristian Secchi, " A passive Integration Strategy for Rendering Rotational Dynamics on a Robotic Simulator", accepted to 2017 IEEE/RSJ International Conference on Intelligent Robot and System (IROS), Vancouver, Canada, pp. -, September 2017.