Space Robotics - Towards Advanced Mechatronic Components and ...

Space robotics – towards advanced mechatroniccomponents and powerful telerobotic systems

G. Hirzinger, B. Brunner, J. Butterfaß, M. Fischer, M. Grebenstein,

K. Landzettel, M. Schedl

DLR (German Aerospace Center), Institute of Robotics and MechatronicsOberpfaffenhofen; D-82234 Wessling/GermanyTel. +49 8153 28-2401; Fax: +49 8153 28-1134

[email protected]; www.robotic.dlr.de

AbstractThe paper describes recent design and developmentefforts in DLR`s robotics lab towards a new generationof “mechatronic” ultra-light weight robots witharticulated hands and high-level remoteprogrammability. The design of fully sensorized robotjoints with complete state feedback and the underlyingmechanisms are outlined. The second light-weight armgeneration is available now the third is in preparation; inthe same way the second generation of a most highlyintegrated 4 finger-hand is near completion. And thetask-level programming system MARCO has proven itspotential in ETSVII, NASDA`s free-flying space robotprojects. Together with the Canadian Space Agency weprepare its use for the space station robot arm’s groundcontrol. Thus it is hoped that big steps towards a newgeneration of space as well as service and personalrobots have been achieved.

1. IntroductionOur first big experience with space robotics has been

ROTEX (fig.1) the first remotely controlled robot inspace [ 1 ]. It flew with Spacelab-Mission D2 insideshuttle COLUMBIA in April ’93 and performed severalprototype tasks (e. g. assembly and catching a free-floating object) in different operational modes, e. g.remotely programmed, but also on-line teleoperated byman and machine intelligence. Its success wasessentially based on• multisensory gripper technologies• local autonomy using the above sensory feedback

capabilities

• predictive graphics simulation compensating for 5 –7 seconds delay

We gained our second big space robot experience withNASDA’s ETS VII project, the first free-flying spacerobot, who was operable for around two years. In April’99 we got the permission by our Japanese friends toremotely program and control their robot from Tsukuba/ Japan [ 3 and [ 9 ]. The project called GETEX(German Technology Experiment) was again verysuccessful (as was the whole ETS VII mission); ourgoals in particular had been:• To verify the performance of the MARCO

telerobotic concept (see below), in particularconcerning the implicit task level programmingcapabilities as well as the sensor-based autonomyand world model update features.

• To verify 6 dof dynamic models for the interactionbetween a robot and its free-flying carrier satellite.

2. Preparing the futureWhen comparing human skills with those of present-dayrobots of course human beings in general are by farsuperior, but when comparing the skill of an astronaut ina clumsy space-suit with that of the best available robottechnology, then the differences are already going todisappear, the more if there is a remote control andmonitoring capability on ground with arbitrarily highcomputational and human brain power. For IVAactivities a robot basically would have to compare withthe full human skill and mobility; to be honest, many ofthe manual operations to be done in a space-laboratoryenvironment are fairly simple standard operations, likehandling parts, opening and closing doors, pullingdrawers, pushing buttons etc. which have to be done just

Proceeding of the 6th International Symposium on Artificial Intelligence and Robotics & Automation in Space: i-SAIRAS 2001, Canadian Space Agency, St-Hubert, Quebec, Canada, June 18-22, 2001.

by stepping through extensive, written procedures. Realintuition and manual skill is particularly requested innon-nominal situation and repair situations. Although itis not clear today when a multi-fingered robot handmight be as skilled as the human hand and when (ifever) a robot might show up real intelligence andautonomy, it nevertheless is obvious that even withtoday’s technology and the available teleroboticconcepts based on close co-operation between man (e.g.the ground operator) and machine there are many tasksin space, where robots can replace or at least augmenthuman activities with reduced cost from a mid-termperspective.

Fig. 1 DLR´s Robonaut concept for a free-flying robotsatellite with two arms and two articulated hands

What we definitely need for space (as a technologydriver) but also for the wide variety of future terrestrialservice robot applications, are sensor-controlled light-weight arms (in contrast to the stiff and heavy industrialsolutions) and articulated, multifingered hands, whichcome closer and closer to the delicate humanperformance. Two of these arms combined with anarrangement of a stereo camera pair tends to providesuch a system with humanoid appearance and thusprovokes the “robonaut” terminology (Fig. 1). NASAhas recently presented remarkable results in this context.

3. LIGHT-WEIGHT ARMSDesign AspectsThe design-philosophy of DLR’s light-weight-robots isto achieve a type of manipulator similar to the kinematicredundancy of the human arm , i.e. with seven degreesof freedom, a load to weight ratio of better than 1:2(industrial robots ≈ 1:20), a total system-weight of lessthan 20 kg for arms with a reach space of up to 1,5 m,no bulky wiring on the robot (and no electronics cabinetas it comes with every industrial robot), and a highdynamic performance. As all modern robot controlapproaches are based on commanding joint torques,

joint torque control (allowing programmableimpedance, stiffness and damping) was a must for us.Another must for us has been the use of precise motorposition sensing, and link angular sensing.

Taking these requirements into account, we started akinematic simulation to determine the best jointconfiguration. Based on the kinematics and desiredpayload, the required dynamic joint torques werecalculated. Suitable motors and gears were chosen toobtain the desired joint torques.The design of the mechanical parts and the integrationof all electronic components was done by using modern3D-CAD techniques. This resulted in a virtual prototypeof the new light-weight robot (II), which could be easilymanufactured and assembled (Fig. 2).

Fig. 2 DLR´s second lightweight robot generation

Based on the number of joints (7), the payload (up to8kg while the arm should not weigh more than 17kg),the overall length (1m) and the tasks to be performed,the kinematic simulation should provide an optimaljoint configuration. In our simulation we have chosen atrajectory consisting of three perpendicular spirals (Fig.3). As a result, it was suggested to equalize the distancesbetween joint 2-4 and to use 3 perpendicular pitch jointsfor joint 2-4.

Fig. 3 kinematic simulation Fig. 4 robot kinematicsThe light-weight robot was designed using a modern 3DCAD system. All mechanical parts as well as the

electronic components are shaped in detail and have adensity assigned to them forming a so-called virtualprototype.

Fig. 5 mechanicalstructure

Fig. 6 electronics

Thus it is possible to retrieve masses, centers of gravityand mass matrices for single components as well as forthe whole robot. The possibility of getting importantmodel parameters from the model simplifies theidentification of the arm significantly. The geometricaldata of the electronic components have been obtainedvia the interface between the mechanical and electronicCAD tool.The electronics was effectively integrated in the wholedesign. In addition, the internal structure of the virtualprototype allows an easy animation and visualization.

Mechatronic JointsEach joint contains a torque sensor, a link positionsensor and a motor position sensor (Fig. 8 and Fig. 9).Furthermore, all joints are equipped withelectromagnetic brakes. All these components,including motor and gear are placed inside the housingto be as space-saving as possible. We use brushless DC(Kollmorgen) motors without housing, special short andlight-weight gears and modified electromagnetic brakeswith reduced power consumption and decreased weight.

Fig. 7 Cross section of joint 2

The light weight gears have been developed by thecompany Harmonic drive in close co-operation with us,they are provided with aluminium crafted wavegenerators and circular splines. All housings are madeof aluminium (saving 40 % weight) and are designed totransfer thermal energy from the motor to thesurrounding air. All joints are equipped with hollowshafts for the internal cabling.

Fig. 8 Joint Components of the light-weight robot

Fig. 9 The intelligent joint

ElectronicsFor setup and maintenance reasons we have decided touse a backplane concept for the main electronic boards(Fig. 9).One backplane is designed for carrying electronics fortwo joints. Joint 7 and 6, Joint 5 and 4, Joint 3 and 2share each one electronic block, which is integrated in

the robot structure. The electronics for joint 1 is locatedin the base of the robot.The electronic block is built up with a backplane, asupply board, two DSP boards and two analoginterfaces. The electronics of the whole robot consumesless than 80W. The 20 KHz/100 V power supply(galvanically decoupled) had successfully been usedalready in the ROTEX missionthe first remotely controlled space robot.The robot joints communicate via a fiber optical bussystem. The standardized SERCOS protocol, which is areal time bus system, is used. The desired and actualmotor position, link torque and link position aretransmitted every millisecond. Status and supervisorysignals like temperatures, voltages and error messagesare transferred by means of the acyclic channel, whichis defined by SERCOS as well.On a joint-integrated DSP, the joint torque controlalgorithms run with 3 KHz and artificial jointimpedances may be commanded from the external PC-based cartesian controller.

The power electronics basically drives three-phasemotors. The MOS-fets are thermally coupled with thestructure of the joint in an ideal manner (Fig. 10).

Fig. 10 Power converter Joint 2

To get a high performance drive unit it is essential thatthe motor position can be exactly controlled.Two analog hallsensors are integrated into the motor tomeasure the magnetic field of the rotor. Thus we wereable to meet the challenging space restrictions. The twosensors have a displacement so that their outputscorrespond to a sine and cosine signal. With the sineand cosine the motor position is calculated.

Fig. 11 motor with integrated position sensor

Each joint is equipped with a safety brake. Anintelligent drive electronics reduces the powerdissipation of the brake by the factor of 10. As a resultthe brake could be redesigned in collaboration with themanufacturer. The total mass went down from 281g to155g.

Fig. 12 Original and redesigned safety brake

The control electronics and a limit switch circuit aredirectly mounted on the brake - another good examplefor the mechatronics approach.Joint torque sensing is based on the deformation ofradial beams as measured by strain gauges. Theresistance variation of the strain gauges is proportionalto the applied torque. By using eight strain gauges

temperature effects and transverse forces can becompensated.

Fig. 13 FEM calculation and real torque sensor

Special efforts were made to find the optimal shape ofthe beam, the placement of the strain gauges and tomake the sensor insensitive to overloads. Four differentsensors have been designed with load ranges from30Nm up to 200Nm. The boards for preamplifying thestrain gauge signals are integrated in the sensormechanics. The mechatronics concept saved space andweight, while the performance of the sensor could beincreased.

The link position sensor is able to measure the off-drive position with an accuracy of 0,01°. As theabsolute position is measured no reference sequenceshave to be performed during the power up of the robot.The sensor has a flat shape and allows the use of a hugehallow shaft. The analog joint position sensor signal aswell as the torque signal are digitized and transferredvia a serial, high speed, differential communication tothe DSP-board.

Fig. 14 link position sensor

Robot Control (impedance control)Considering the application fields for which this robotwas designed, a main focus obviously had to be theability to perform compliant manipulation in contact

with an unknown environment. The robot should beable to guarantee the safety of humans interacting withit not only at the TCP, but also along the entire robotstructure. This was one of the main motivations forintroducing torque sensors in each joint, allowing notonly gravity compensation (thus emulating outer spaceconditions), but also stiffness and impedance control.Another challenging control problem results from thelight-weight design of the robot, which inherently leadsto increased joint elasticity. Since the links can beregarded as rigid compared to the joints, a flexible jointrobot model can be assumed. This implies a fourth ordermodel for each joint. Therefore, by measuring the motorposition qm and the joint torque τ and by computing thenumerical derivatives dqm and dτ, the complete jointstate can be obtained. Our control strategy is to use theavailable sensors to implement the desired taskbehaviour as well as to compensate the effects of jointelasticity.

The first stage in the controller development was a jointstate feedback controller with compensation of gravityand friction (Fig. 15).

Fig. 15 state feedback controller with gravitycompensation

In our opinion, this structure represents the minimalconfiguration that can provide both positioning accuracyand effective oscillation damping. The controllerconstitutes a direct extension of the PD controller stillused in most industrial robots to the case ofmanipulators with joint elasticity. It is proven to bepassive and global asymptotically stable for a widerange of parameters.

An important feature of this controller structure (Fig.16) is that, by a suitable parameterization, it can be usedto provide a position, a torque, as well as an impedancecontroller. In fact, the position and the torque controllersare implemented as special cases of the impedancecontroller, for maximal and zero stiffness respectively.The gains are computed online, to provide the desiredjoint stiffness and damping. The state feedbackcontroller is implemented in a decentralized manner on

the signal processors in each joint, with a high samplingrate (3kHz).

Using joint parameter identification and CAD modelingof the rigid robot components, an accurate robot modelwas obtained, which was subsequently used for thedesign and simulation of model based controllers.

The Cartesian position control uses a singularity robustinverse kinematics module for redundant manipulators.It enables collision avoidance and the optimization ofadditional task dependent criterions.

Impedancecontrol

Torquecontrol

Positioncontrol

Impedancecontrol

Forcecontrol

motor position joint torque link position

Force-torque sensor

stiff robotdynamics

adaptive gainadjustment

4-th order joint state feedback: qm, dqm, τ, dτ

kÔ0kÔmax

inversekinematics

serial bus1kHz

cartesian level

Joint level

3kHz joint7 Njoint1

Fig. 16 controller architecture

Towards lightweight arm IIIOur next steps try to continuously reduce weight byusing a self-developed, specialized external rotor motor,a new piezo brake (weighing only 30g, see Fig. 17),more carbon fiber technology and higher jointmodularity. As a central goal we try to further optimizethe joint performance index as defined and proposed in[ 2 ].

Fig. 17 Piezo brake (integrable into new DLR motor)weighs only 30g

Presumable the biggest step we have made in the recentpast is a new high energy motor RoboDrive, which wespecially have designed for lightweight robotapplications. It has been optimized over more than two

years using all kind of concurrent engineeringtechnologies (Fig. 18).

Fig. 18 Magnetic flow optimization during the designprocess

Indeed it seems that robot manufactures so far havealways just used the best available motors, but whatrobotics (especially service and space robotics) reallyneeds are lightweight motors which are optimized interms of permanent, high-dynamic reversion around azero position and which have minimal losses (inaddition to hollow shafts for cabling). DLR`s newRoboDrive motor indeed weighs only half of the (toour knowledge) best available motors and shows uponly half of the usual losses.

Fig. 19 shows the amazing new design (pareto) curvesand in Fig. 20 the CAD design and realization areshown next to a first link of the coming modular carbonfiber arm III.

Fig. 19 Comparing DLR`s RoboDrive motor with thebest available motors

Fig. 20 3D CAD-design and realisation of the newRoboDrive, to be integrated into a carbon fiber link

Indeed DLR`s light-weight robot III will be fullymodular, with a limited set of link types for the basicarm structure (typically 5 joints), and a two dof wristjoint.

Fig. 21 shows the modular concept of a joint betweentwo links; in Fig. 22 the link types are displayed for asymmetric robot, a special type of which has beenassembled in Fig. 23. Indeed we have now simulationtools by which we pick up single links (with adaptablelengths) from a basic library as shown in Fig. 22,assemble the robot kinematics as we want, and then weimmediately run an interactive (e.g. space mouseoperated) kinematic-dynamic simulation displaying allrelevant forces and torques in the joints.

Fig. 21 3D CAD design of modular joint

Fig. 22 Arm-Segments “Symmetrical Robot”

Fig. 23 Configuration Example “Symmetric Robot”

Our great hope is that with this next lightweightrobot generation we will be able to approach the 1:1ratio between arm (12-13kg) and load and evenfurther reduce the extremely low powerconsumption of our light weight arms (typically only200W or less).

4. ARTICULATED HANDSIn 1997 DLR developed one of the first articulatedhands with completely integrated actuators andelectronics (Fig. 24).

Fig. 24. DLR's Hand I.

This well known hand has been in use for several yearsand has been a very useful tool for research anddevelopment of grasping. The main problems remainingwere maintenance and the many cables (400) leadingout from the Hand. The experiences with Hand Iaccumulated to a level that enabled us to design a newhand (Hand II) according to a fully integratedmechatronics concept which yields a reasonably betterperformance in grasping and manipulation and thereforeaccelerates further developments.

The DLR HAND II

Design issuesDue to maintenance problems with Hand I and in orderto reduce weight and production costs the fingers andbase joints of Hand II were realized as an open skeleton-structure. The open structure is covered by 4 semi shellsand one 2-component fingertip housing realized instereolitography and vacuum mold (Fig. 25 and Fig.26). This enables us to test the influence of differentshapes of the outer surfaces on grasping tasks withoutredesigning finger parts.

The main target developing Hand II from the beginninghas been the improvement of the grasping- performancein case of precision- and power-grasp.

Soon it turned out very important to adapt the positionof the 4th finger and the thumb as well. To performpower-grasps it is absolutely necessary to have a nearlyparallel position of the second, third and fourth finger asseen in Fig. 25

Fig. 25 Hand II in power grasp

Fig. 26 Hand II in fine manipulation configuration

On the other hand performing precision-grasps/fine-manipulation requires huge regions of intersection ofthe ranges of motion and the opposition of thumb andring-finger (Fig. 26). Therefore Hand II was designedwith an additional minor degree of freedom whichenables to use the hand in 2 different configurations.This degree of freedom is a slow motion type to reduceweight and complexity of the system. The motion of thefirst and the fourth finger are both realized with just onebrushed dc actuator using a spindle gear. The realizedfinger positions for both types of grasping weredesigned virtually. Optimized configurations werefound using interactive grasping tools as seen in Fig. 27and mapped to each other using the positions of 2nd and3rd finger.

Fig. 27 Interactive grasp and manipulability simulationtools.

Realizable kinematics were calculated and imported tothe two virtually found configurations and optimizeduntil the actual configuration with an overall number of13 DOF was found.

ActuatorsThe three independent joints (there is one additionalcoupled joint) of each finger are equipped withappropriate actuators. The actuation systems essentiallyconsist of brushless dc-motors, tooth belts, harmonicdrive gears and bevel gears in the base joint. Theconfiguration differs between the different joints. Thebase joint with its two degrees of freedom is ofdifferential bevel gear type, the harmonic drive gears forgeometric reasons being directly coupled to the motors.The differential type of joint (Fig. 28) allows to use thefull power of the two actuators for flexion or extension.

Fig. 28 Differential bevel gear of the new basejoint.

Since this is the motion where most of the availabletorque has to be applied, it allows to use the torque ofboth actuators jointly for most of the time. This meansthat we can utilize smaller motors. The actuation systemin the medial joint is designed to meet the conditions inthe base joint when the finger is in stretched positionand can apply a force of up to 30 N on the finger tip.Here the motor is linked to the gear by the transmission.The motor in the medial joint has less power than themotors in the base joint, however there is an additional

reduction of 2:1 by the transmission belt. Thus weachieve the torque which corresponds to the torquecreated by the two motors in the base joint for anexternal force of 30 N on the finger tip. The harmonicdrives used are of the same type for all joints, since thesmallest appropriate type can stand the torque for bothtypes of actuation.

Sensor EquipmentA dextrous robot hand for teleoperation andautonomous operation needs (as a minimum) a set offorce and position sensors. Various other sensors add tothis basic scheme. Each joint is equipped with straingauge based joint torque sensors and specially designedpotentiometers based on conductive plastic. Besides thetorque sensors in each joint we designed a tiny sixdimensional force torque sensor for each finger tipwhich will be explained more precisely in 0. Thepotentiometers, each with an analogous filter of thirdorder, would not be absolutely necessary, since one maycalculate the joint position from the motor position,however they provide us with a more accurateinformation of joint position, and they can by the wayeliminate the necessity of referencing the fingers afterpower up. In case of not using the potentiometers onewould have to consider the elasticity of the transmissionbelt and the harmonic drive. With the potentiometer weachieve a resolution for the joint angles of 1/10°, thismeans approximately 10 bits for the joint.

Force Torque Fingertip Sensor A tiny six dimensional force torque sensor (20 mm indiameter and 16 mm in height) as shown in Fig. 29 withfull digital output has been developed for the fingertip.The force and torque measure ranges are 10 N for Fx

and Fy, 40 N for Fz, 150 Nmm for Mx, My and Mzrespectively. Also a 200 % mechanical overloadprotection is provided in the structure.

base element

cantilever beam

z

y x

ε5

3ε

7ε

6ε

1ε

2ε

4ε 10ε9ε

8ε

Fig. 29 Six dimensional force torque sensor infingertip.

Communication ArchitectureAll electronics needed locally is integrated into thehand. However the control of the fingers and the handis done by an external computer. In order to use thehand freely on different manipulators and to reducecables and the possibility of noise in the sensor signals,we decided to design a fully integrated serialcommunication system. Each finger holds onecommunication controller in its base unit (see Fig. 30).

communicationcontroller

ADC16 ch

ADC8 ch

sensorcircuitry

sensorcircuitry

sensorcircuitry

ADC8 ch

sensorcircuitry

ADC8 ch

Motor

powerconverter

MotorMotor

powerconverter

powerconverter

sensorcircuitry

power supply

serialto hand base

Fig. 30 Electronics and communication in a finger.

This controller is responsible for the collection anddistribution of all information of interest. Furthermore itdoes some reasonable signal processing.

CommunicationController(FPGA)

Motor

hand dof

fingers 1-4

serial

serial/differentialIEEE 1355-1995

Fig. 31 The communication controller in the hand baselinks the fingers to external computers.

It collects the data of all five ADCs per finger withtogether 40 channels of 12 bit resolution each andtransmits these data to the communication controller inthe hand base (see Fig. 31). On the other hand itdistributes the data from the control scheme to theactuators for finger control. The communicationcontroller in the hand base links the serial data stream ofeach finger to the data stream of the external controlcomputer. By this hardware architecture we are able tolimit the number of external cables of DLR's Hand II toa four line power supply and an eight linecommunication interface since the data is transmittedvia differential lines. This interface even provides thepossibility of using a quick-lock adaptor forautonomous tool exchange. Reducing external cablingfrom 400 (in Hand I) to 12 here, is one of the majorsteps forward in our new hand.

Cartesian Impedance ControlWhen a robot hand performs any fine manipulation,there is always need that the fingertip should be soft inthe direction normal to the contact surface and hardtangential to the contact surface. Thus the impedanceshould be adaptable to the orientation of the fingertip.Therefore, a cartesian impedance controller has beendeveloped, where the fingertips behave like aprogrammable spring.

5. MARCO – DLR’s Task-DirectedSensor-based Teleprogramming System

Based on the ROTEX experience (Fig. 32), we havefocused our work in telerobotics on the design of a high-level task-directed robot programming system MARCO,which may be characterized as learning by showing in avirtual environment and which is applicable to theprogramming of terrestrial robots as well. The goal wasto develop a unified concept for• a flexible, highly interactive, on-line teleoperation

station based on predictive ground simulation(including sensorbased local autonomy) see Fig. 32as well as

• an off-line programming environment, whichincludes all the sensor-based control and localautonomy features as tested already in ROTEX, butin addition provides the possibility to program arobot system on an implicit, task-oriented level.

Stereo TV

Board

SensorbasedRobot-Control

GroundT

Predictive Simulation in"virtual" Robot world

T

Model-Update

SensorbasedRobot-Control

R

R

trmctrle.vsd

Fig. 32 ROTEX telerobotic control concept

A non-specialist user - e.g. a payload expert - should beable to remotely control the robot system in case ofinternal servicing in a space station (i.e. in a well-defined environment). However, for external servicing(e.g. the repair of a defect satellite) high interactivitybetween man and machine is requested.

To fulfil the requirements of both application fields, wehave developed a 2in2-layer-model, which representsthe programming hierarchy from the executive to theplanning level.

Task

Operation

Elemental Operation

Sensor Control Phase

implicit layer

explicit layer

planning

execution

Fig. 33 2in2-layer-model

Based on this 4 level hierarchy, an operator working onthe (implicit) task level does no longer need real roboticexpertise. With a 3D cursor (controlled by a SpaceMouse) or with a human-hand-simulator (controlled bya data-glove) he picks up any desired object in thevirtual world, releases it, moves it to a new location andfixes it there. Sequences of these kind of operations areeasily tied together as complex tasks; and before theyare executed remotely, the simulated robot engaging itspath planner demonstrates how it intends to perform thetask implying automatic collision avoidance.

Fig. 34 DLR’s universal telerobotic station MARCO(Modular A&R Controller)

Nevertheless in the explicit layer (the learning phase)the robot expert has to show and demonstrate theelementary operations including the relevant sensorypatterns and – if necessary – train the mapping betweennon-nominal sensory patterns and motion commandsthat servo into the nominal patterns later on in the realworld.

He performs these demonstrations by moving therobot’s simulated gripper or hand (preferably withoutthe arm) into the proximity of the objects to be handled(e.g. drawers, bajonet closures, doors in a labenvironment), so that all sensory patterns are simulatedcorrespondingly. The robot expert at this stage of coursemust have knowledge on position- and sensorcontrolledsubspaces (and must be able to define them, massivelysupported by MARCO functions), and he has to definehow operations (e.g. remove bajonet closure) arecomposed by elementary operations (approach, touch,grasp, turn etc.).

MARCO’s two-handed VR interface conceptThus as a general observation, on the implicit as well ason the explicit layer statement we have to move around3D-pointers or grippers / hands in the virtual labenvironment. Using classical ”immersive” cyberspace

techniques with data-glove and helmet was not adequatefor our approach, as the human arm’s reaching space isfairly small (e.g. in a lab environment) and with headmotions only very limited translational shifts of thesimulated world are feasible. As a general observationan alternative to the position control devices ”data-gloveand helmet” is the velocity control device ”SpaceMouse”, particularly if the robot system to beprogrammed has no articulated hand. Velocity controlhere means we may casily steer around an object in VRover arbitrary distances and rotations via smalldeflections (which command velocities) of an elasticsensorized cap. The second important observation(confirmed by extensive tests of car manufacturers inthe context of 3D CAD-design) is that just as in real lifetwo-handed operations when interacting with 3D-graphics are the optimum. Indeed whenever humans canmake use of both hands, they will do (e.g. when carving,modelling, cutting). In the northern hemisphere foraround 90 % of the people the right hand is the workinghand, while the left hand is the guidance andobservation hand, which holds the object to be workedon (vice versa for left-handers).

This ideal situation for a human is easily transferred tothe VR interface scenario. A right-hander preferablymoves around the whole virtual world in 6 dof with aSpace Mouse in his left hand (the guidance hand), whilewith his right hand he moves around the 3D cursor witha second Space Mouse (velocity control, Fig. 35) or asimulated hand with a data glove (position control, Fig.36).One should note that now even for the glove theproblem of limited workspace disappears, because withthe left hand the operator is always able to move thevirtual lab world around such that the objects to begrasped are very close so that even in position controlmode with a data glove only small, convenient motionsof the operator’s hand are requested to reach them.

Fig. 35 Two handed VR-interface using two Space Mice(ETS VII scenario as example)

Fig. 36 Two handed VR-interface using Space Mouse andData Glove

(space station scenario as example)

More details on MARCO’s high level user interface asare Java/VRML client techniques as indicated in Fig.37 are given in Ref. [ 3 ].

Fig. 37 Internet-Programming with VRML /3DJAVA

The sensorbased task-level-teleprogramming systemMARCO has reached meanwhile a high level ofuniversality. It was not only used as ground control

station for the ETS VII experiment, but it was used alsofor technology studies of Germany’s technology projectExperimental Service Satellite ESS, as well as forremote ground control of a new, climbing space stationrobot, and for mobile terrestrial and planetary robotprojects.

6. ConclusionSpace robots in the future will take over more and moretasks from humans. The real value of space robots liesin their remote programmability and controllability incombination with onboard autonomy, realizing theprolongation of human’s arm into space. The relevanttechnologies including lightweight arms, articulatedhands and powerful and delay compensation teleroboticsystems are available – it’s our task to convincepoliticians and decision-makers in agencies that time ismature for the robotics age in space. Indeed already atthe space station – and even for its construction – anumber of remarkable manipulator and robot systemswill be active. However all of them originally weresupposed to be exclusively operated by astronauts, andthis was one of our main concerns and disappointments.As a consequent next step we try to help in pushingforward the remote control of operational space robotsystems. In particular we cooperate with the canadianspace agency aiming at the control of the SSRMS spacestation robot via the MARCO telerobotic system [ 10 ].

7. References[ 1 ] G. Hirzinger, Sensor-based space robotics –ROTEX and its telerobotic features”, IEEE Transactionson Robotics and Automation, vol. 9, No. 5, Oct. 1993.[ 2 ] G. Hirzinger, J. Butterfaß, M. Fischer, M.Grebenstein, M. Hähnle, H. Liu, I. Schaefer, N. Sporer,“A Mechatronic Approach to the Design of Light-Weight Arms and Multifingered Hands”, IEEE Int.Conference on Robotics and Automation, SanFrancisco, April 2000[ 3 ] B. Brunner, K. Landzettel, G. Schreiber, B.M.Steinmetz, G. Hirzinger, “A universal task-level grundcontrol and programming system for space robotapplications, iSAIRAS 5th Int. Symposium on ArtificialIntelligence, Robotics and Automation in Space, 1-3June 99, ESTEC, Noordwijk[ 4 ] Dubowsky, S. and Papadopoulos, E., “TheKinematics, Dynamics and Control of Free-Flying andFree-Floating Space Robotic Systems”, IEEETransactions on robotics and Automation, 5, 1993.[ 5 ] Stieber, M.E. Hunter, D., Abramovici, A:Overview of the Mobile Servicing System for theInternational Space Station, 5th InternationalSymposium on Artificial Intelligence, Robotics andAutomation in Space, Noordwijk, Netherlands, 1999[ 6 ] Toru Kasai, Mitsushige Oda, Takashi Suzuki:Results of the ETS-7 Mission – Rendezvous Dockingand Space Robotics Experiments, Proc. Fifth

International Symposium on Artificial Intelligence,Robotics and Automation in Space[ 7 ] Hirzinger, G., Landzettel, K., Brunner, B.,Schaefer, I., Fischer, M., et al.: DLR's Robotics Lab -Recent developments in Space Robotics. 5thInternational Symposium on Artificial Intelligence,Robotics and Automation, Noordwijk, The Netherlands,ESTEC, (1999)[ 8 ] W. S. Kim, P. S. Schenker, A. K. Bejczy, S. Leake,and S. Ollendorf: An Advanced Operator InterfaceDesign with Preview/Predictive Displays for Ground-Controlled - Space Telerobotic Servicing, SPIEConference 2057, Telemanipulator Technology andSpace Telerobotics, pp. 96-107, Boston, MA, Sept.1993.[ 9 ] R. Lampariello. G. Hirzinger: Freeflying Robots –Inertial Parameters Identification and Control Strategies,ASTRA 2000, 5.-7. Dez. 2000, Noordwijk[ 10 ] K. Landzettel, B. Brunner, G. Schreiber, B.Steinmetz, E. Dupuis: MSS Ground Control Demo withMarco, iSAIRAS 2001, Montreal

Space Robotics - Towards Advanced Mechatronic Components and ...

Documents

Transcript of Space Robotics - Towards Advanced Mechatronic Components and ...