N91-240 '5 ROBOTIC CONTROL AND INSPECTION VERIFICATION Virgil Leon Davis Chief, Robotics Section, DM-MED-12 Kennedy Space Center,FL ABSTRACT This paperdiscusses three possible areas of commercialization involving Robots developed atJohn F. Kennedy Space Center (KSC), Florida. (1) A6-Degree ofFreedom (6-DOF) TargetTrackingSystem forremote umbilical operations (2) An intelligent TorqueSensing End-Effector (TSEE) foroperating hand valves inhazardous locations O) An Automatic RadiatorInspection Device (ARID), a65by 13foot robotic mechanism involving completely redundant motors, drives, and controls AsIX_ts concerning thefirst two innovations can be integratedtoenable robotsor teleoperators toperform tasks involving orientation and panel actuationoperations thatcanbe done withexisting technology ratherthan waiting fortelerobots to incorporate artificial intelligence (AI) toperform "smart" autonomous operations. These operations are applicable to Space Station work,groundaerospace launchprocessing, andhazardous petrochemical ornuclear sating operationsworldwide. Thethird robot involves theapplication of complete control hardware redundancy toenableperformance of workover andnearexpensive Shuttle hardware. The consumer marketplace maywish toexplore commercialization of similarcomponent redundancytechniques for applications when arobotwould notnormally beused because of "reliability" (when an inadvertentmove could result indamage toexpensive components or personnel). Introduction The computer hardware and software systems inthe Robotic Applications Development Laboratory (RADL) were designed tofacilitate the development and application of advanced robotic control technology. KSC not only launches spacecraft but services these spacecraft onthe ground: designing the support equipment, hunch accessories, and computer hardware/software forground spacecraft servicing. KSC has implemented an integrated system that coordinates state-of-the-art robotic subsystems. It is a sensor-based realtime robotic control system performing operations beyond the capability ofan off-the-shelf robot. The integrated system provides realtime closed-loop adaptive path control of position and orientation of all six axes of a large robot; enables theimplementation of a highly configurable, expandable testbed for sensor system development; and makes several smart distributed control subsystems (robot arm controller, process controller, graphics display, and vision tracking) appear as intelligent peripherals toa supervisory computer that coordinates the overall system. The integrated RADL system iscurrently providing an easy-to-use testbed forNASA sensor integration experiments. Advanced targettracking development isinprogress concerning the mating of umbilicals used during space vehicle hunch. Programmatic studies are underway touse laboratory capabilities toenhance the safety, productivity, and efficiency of KSC facilities for Shuttle and future ground processing operations. Projects are underway that should generate large operational cost savings through the integration of advanced technologies for ground processing operations, such asOrbiter tile and radiator damage assessment. Robotic techniques to improve "Shuttle Orbiter inspection and closeout verification" (operations involving possible human- ormechanism-induced damage) are being investigated and implemented. Nondestructive test sensors, vision systems, and various kinds of distance ranging sensor systems can be integrated with the RADL systems todevelop the prototype concepts for integrating robot parameters withlarge data-based graphics and artificial intelligence software systems. For example, the RADL robot can position a sensor with precise accuracy, report that position and orientation, provide distance sensory data, and integrate machine vision "electronic photographs" with graphics and AI software tofurnish computer printouts providing automatic sizing and highlighting of exception data. This type of system isbeing proposed for Shuttle Orbiter radiator damage inspection, Orbiter tile damage/debonding assessment, and Orbiter contour measurements. The manual methods presently employed inthese operationsare very labor intensiveand produce expensive serial-time flow constraints. 298 REMO'rUEMBILICAPLLATSDOCKING/INSERTION Realtime adaptive control isthe necessary tool for tracking a Shuttle vehicle that rocks inthe wind while stacked at the launch pad. This adaptive control isnecessary inorder todock and insert umbilicals (consisting of a ganged connection of electrical and cryogenic/hypergolic fluidfines) without damage tothe vehicle and without harardous leaks. This reduces reconnect times of 14to34hours to less than 15minutes and eliminates hazards associated with umbificals that would otherwise have tobe connected at launch. The KSC environment is demanding: the system must withstand heavy acoustical shock and see through fog caused by the dumping of thousands of gallons of water onto the flames of the Shuttle main engines (conditions that exist right after an launch when an umbilical reconnect would take place). These shock and blast conditions rule out sensitive laser tracking. The prototype system KSC isdeveloping can be upgraded quite easily for this environment. For instance, an infrared filter can be added to the CCD camera and our 5dots can be changed to infrared LED's to see through the water vapors without affecting the system architecture or changing the extensive software algorithms. Also, KSC istracking speeds atwhich the target can be blurred. KSC has already advanced the state of the art by developing algorithms and packaging off-the-shelf imaging hardware intoprocess machines that significantly reduce the blurring of moving targets to allow more precise, smoother tracking. KSC isstudying adaptive conlxol (trajectory perturbation based onrealtime sensory feedback) with heavy inertial loads. KSC isattempting toprecisely position 5,000-pound umbilicals (200 pounds initially) with pneumatic counterbalances on future heavier umbilicals. Research inindustry has been done oncompliance-aided insertion of resisters into ram-tolerance holes; however, KSC isdesigning heavy umbilicals (oriented perpendicularly with respect togravity) toinsert into ram-tolerance holes on anobject randomly moving with wind-induced perturbations. This induces high torques intoservo control motors and KSC has obtained stability under these extreme conditions. The RADL islocated ina50by I00 foot high bayfacility. Figure 1depicts the left half of the facility. It has been recently expanded toprovide space for aproduction model of the ARID robot. Several work cells (as shown in figure 1) are accessed byan ASEA IRB-90 six-axis industrial robot located on a 30-foot track. The IRB-90 has a reach of approximately 9 feet, a load capacity of 90kilograms, and a repeatability of0.005 inch. A central MicroVAX II mini-computer acts as the supervisory controller. Communication toperipheral systems (ASEA robot controller, AD/DA interface, and vision system) isestablished through custom serial connections. The system isconnected via DecNet network tothe local KSDN network and theNational NASA SPAN network. The vision system, a Data Cube MaxVideo, incorporates a pipeline design approach, a VME bus with a 68020 processor, and aprocessed throughput of 30frames per second. Additional sensors (force/torque, proximity) are also integrated in the system. Because ofassociated limitations, the original robot controller isnow being replaced by a special-purpose high-speed flexible robot controller, which will provide direct access toeach joint controller and provide 6-DOF adaptive control capabilities inrealtime (30 Hz control updates for vision and 100 Hz for force-torque tracking). The RADL prototype system consists of avision-based 6-DOF tracking system attached tothe ASEA robot and a target attached toa separate receptacle plate. Using a passive compliance end-effector, the robot isable totrack and insert an umbilical plate mockup (which incorporates fluid, electrical, and data lines) into a receptacle plate mounted on a 3-DOF device that simulates the motion of the Orbiter (see figure 2). The passive compliance device uses nontactile vision tracking and isbeing augmented byan active tactile tracking system using force/torque reactive feedback toreduce mating forces. Enhancements include the use ofa counterbalance mechanism that removes loads from the robot and enables the robot to disconnect from the umbilical after mating. The robot will be turned off and the umbilical will free float with the counterbalance removing loads from the vehicle. This reduces vehicle design weight and allows more payload toreach space at reduced costs. A floating plate isalso incorporated that reduces insertion loads, reduces connect/disconnect surges, and eliminates galling of sensitive cryogenic fluidcoupling surfaces. Improvements also have been made inalignment and capture mechanisms toreduce forces, reduce galling, and ensure positive latching. 6-DOF Tracking The major innovation isarobotic vision subsystem that measures the relative position and orientation of a specially designed target and provides realtime control toa large robotic mechanism. The subsystem uses 299 standardimage processing algorithms implemented direcdy in circuitry instead of computer programs that consume more time. This featuremakes itpossible toextract complete sets of target tracking data from successive image frames at the rate of more than 30frames per second. Asolid-state video camera views the target,which consists of five bright or reflective circles, four located at the comers of a square and the fifth located at the center of the square but offset from the plane of the square (see figure 3). Theraw image data issent toimage processing circuitry that performs a convolution difference-of- Gaussian edge-analysis filtering operation toclarify the picture dements representing the edges of the circles. On the Shuttle, the fifth circle could be implemented simply as a styrofoam thread-spool glued inthe center of four painted circles. The image isthen processed further toobtain the centroids of the five circles. The locations of these centroids relative toeach other and to the overall image frame are processed toobtain three Cartesian coordinates of the targetrelative to those of the camera. Tria gulafion calculations based onthe vector relationships among the localions ofthe five circles and the central axis of the target yield the roll, pitch, and yaw angles that describe the orientation of the target relative to the line of sight and the field of view of the camera (see figure 4). Thus, the relative position and orientation of the targetare determined inall six degrees of freedom (see figure 5). The offset of the central circle from the plane of the other four circles can be increased or decreased toincrease or decrease the sensitivity of the subsystem tothe pitch and yaw of the target and toprovide more accurate distance-to-the-target information. The outputdata may have tobe transformed into spherical orother coordinates used byanyother robot. However, this transformation can be performed easily insoftware. If the robot ischanged, itisnecessary only tochange this software. TORQUE SENSING END-EFFECTOR (TSEE) The TSEE was produced from a Small Business Innovation Research (SBIR) project todevelop an intelligent tool/gripper toopen or close valves during hazardous maintenance or emergency work. The TSEE features servo control ofjaw opening dimensions, nontactile/tactile sensors, and torque feedback todetermine and maintain optimum seat pressure settings. This feedback provides reflective force feedback to anoperator and toautomatic computer-controlled operations enabling determination of valve position and preventing damage to valve seals. A Phase I SBIR (see figure 6) produced a small working model and a Phase II SBIR (see figure 7) produced a hardened mechanism with a user-friendly data base capable of operating inhazardous NEC Class I,Division I, Group Bhypergolic/cryogenic environments. A version closer tothe Phase I unit may be more suitable for petrochemical and nuclear applications where asmaller mobile robot (see figure 8)could be used tosafe hazardous fires, or chemical or nuclear spills. The TSEE's interchangeable gripper can locate a valve position and rotate continuously toopen or close valves ranging from either 0.5 to4.0 inches or 3.5 to6.0 inches insize. Valves can be opened that have a torque range of 10to 150 inch-pounds.. The TSEE has a unique nontactile torque sensor utilizing magnetoelastic phenomena. The end-effector isused inconjunction with a computer controller that can interpret commands from an operator at a computer keyboard, from a parallel digital interface on the robot carrying the end-effector, or from a serial communications link toanother computer. This lets the system open and close valves and determine if the valve isturning asexpected from any of these operating modes. The computer that controls the end-effector stores information about arange of valves whose positions need tobe adjusted. Valves are identified tothe system by labels that may be descriptive or numeric. Information in the valve data base isused to identify valves and toprovide information onturning ranges, gripping forces, valve handle sizes, and current status orposition of the valve. Once information has been put into the system, valve operations will keep the data base up todate. ATSEE can be incorporated with the 6-DOF target tracking system toallow a remote operator (Space Station astronaut, Flight Telerobotic Servicer (FTS) remote controller, nuclear cleanup engineer, or fire fighter) to: (1) position or teleoperate the robot infront ofa panel where itcan see the target, (2) let the target tracker autonomously orient the robot atthe correct angle and distance from the panel, and (3) activate a stored program toperform pretaught routines. The TSEE locates valves (which may be inthe wrong position with stems inor ou0 through simple sensors and then performs pretaught operational sequences to "safe" an operation ina hazardous environment. This can be done with existing technology rather than waiting for telerobots to incorporate AI toperform "smart" autonomous ooerations. These two innovations (TSEE and 6-DOF 300 u'acking/orienta_on) arethekey technologies toenable more sophisticawAuseof tclerobotics, notjust telcopcrarions, sooner inSpace Stadon work,groundaerospace launchprocessing, andhazardouspetrochemical ornuclear sating operations worldwide. This productcanalso enhance roboticapplications inNASA and industry without redesigning existing valve panelfacilities forrobotics. Itwill provide safer andless serial-dine operations in hazardousenvironments. ORBITER RADIATOR DAMAGE INSPECTION KSC (NASA) isworking ona joint projectwithLockheed (Kennedy and Pale Altoorganizations) wbere Lockheed isdeveloping a sensor system toexamine and inspect the Orbiter radiatorsfor damage (delamination and meteorite dings) while NASA isdeveloping aroboticmechanism totransporttheLockheed sensor over the complex contours and 10.5 by60 foot surface of the radiators. Aprototype production system and future production models will be tested, certified, and installed inthe Orbiter Processing Facilities (OPF) at KSC. The objective of the project is todecrease the amount oftime it takesto process an Orbiter before each mission. The efficiency of the radiators todissipate radiantheatenergy isdependent ontheir surfaces being clean and damage free. Itpresently takes 16people 24 hours toinspect the Orbiter radiators todetermine damage (dings, scratches, impacts) prior tocontinuing other work in the OPF. A robotic inspection system should reduce this to two people in3 hours and provide accurate repeatable trend data and a quality inspection. Currently the operation isperformed with an XYZ mechanism over the Orbiter that moves "buckets" large enough tocarry men and equipment. The buckets are driven byhighly skilled personnel. The visual inspection isperformed using the naked eye and observations are recorded on alog sheet. This does not provide an accurate permanent record of damages and their locations. Small imperfections may be overlooked. This isa very uncomfortable, task-intensive, repetitive job. The radiator panels are normally inspected twice during allOPF flows. The fast inspection, postflight zonal inspection, isperformed todetect any damage to the radiator panels that may have occurred inflight. The second inspection, immediately prior topayload bay closeout for flight, isperformed todetect any damage that may have occurred during OPF processing. Other inspections are required if the radiators are removed from the payload bay doors and placed in storage. Operations personnel are primarily concerned with anomalies that fall intotwo categories: (1) damage tothe radiator panel surface silver Teflon tape and (2) damage totheradiator aluminum facesheet and honeycomb sandwich core. To date, inspection of the radiator panels has focused on only damage visible tothe normal unaided human eye. Once a nonconformance condition isdetected and documented, other techniques are performed toassess the damage severity. Anomalies tothe radiatorsurface include scuffs, scratches, tears, discoloration, delaminations, and bubbles in the silver Teflon tape. An automated inspection device must detect these anomalies and differentiate one from the other. The inspection tool will need tomaintain a memory (damage log) ofeach radiator panel todetermine if an anomaly was "new" orhad been previously assessed and documented. Because most damage tothe thermal control coating isnotrepaired, itisessential that the damage log provide a means forcontinual update and for referencing the cummulative surface area of minor unrepaired damage. Anomalies tothe radiator aluminum facesheet and honeycomb core include indentations (dents), scratches, pin holes, punctures, gouges, and meteorite strikes. The automated inspection device will be able todetect these anomalies and differentiate between defects thatpenetratethe facesheet from those that impact the facesheet but donot penetrate. As before, the inspection tool will maintain anautomatically operated memory (damage log) of each radiator panel todetermine Irends and toquickly locate repair areas that are tobe processed at a later date. By automatingthis operation, most of the personnel will be relieved of their inspection duties sotheir talents could be used toperform otherjobs more demanding of theirskills. Itwould also provide aquickly accessible permanent recordof radiator inspection data (damage assessment andprecise location), reduce the amount of paperworkrequiredtoget thejob done, minimize setuptime toget ready forinspection, andprovide expansion capabilities sothat otherfunctions could be performed onthe radiatorswith theautomated mechanism (e.g., cleaning). Theoverall goal istoachieve anefficient and less expensive operation. Automatic Radiator Inspection Device (ARID) The ARID mechanism was anevolutionary design culminatingina4-degree-of-freedom robot (see figure9). Lockheed originally envisioned thatthe ARID wouM bea camePdtraveling ona contoured beam shapedlikethe 301 radiatorstr,ansporteadlongtworails that lay along the edges of the Orbiter radiators (see figure 10). However, there were four major problems inimplementing this configuration: (1) Two of the major design goals were tonot impact existing operations and toinstall the ARID without major modifications to the facility. The first simple concept would have led toproducing a large mechanism (10 by60by 10by 60fee0 which would have tobe picked upby anoverhead crane and moved into place with one transport rail latched inplace tothe Orbiter Processing Facility (OPF) platforms and the other transport rail suspended over the Orbiter's radiator hinges. This hanging structure would impact other operations that require the "buckets" tobe moved. It would also have tobe removed toopen and close the radiator doors or tolower access platforms above the radiator to allow personnel access topayloads inside the Orbiter bay. Because of these facility impacts, the rail structure would have impacted serial flow, notreduced it ('2) A simple rod-like transport rail over the radiators was not feasible because of the 65-foot length required tospan the cargo bay without resting onthe Orbiter hinges. Atruss assembly would have been required toreduce bowing and sagging of the structure, making itheavier and more difficult to move. (3) Trendphotographs of an old image must be aligned with a new image for comparison. If the alignment isoff by more than 10pixels or ifthe light angles are different, then the image will look different inthe vision system processor's eyes. The bulky mechanism would have tobe fitted with alignment offsets toallow for x, y, and z, and orientation differences for each vehicle and toallow for imprecise placement of the Orbiter inthe OPF. This would have further complicated the design and introduced complex manual alignment procedures. (4) The forward radiator panel (see figure 10) has a5.7'l-degree of slope (covering an area 7 inches high by 70inches long). This eliminates a fixed track design from being able toaccess all four side panels. During the design, itbecame readily apparent that allthese problems could be minimized through a flexible robotic mechanism that could: (1) provide a quick, software-programmed "frame shift operation" eliminating parking offset adjustments, (2) be cantilevered from under access platforms located adjacent tothe radiator panels and moved out of the way when necessary soasto not impact facility operations, and (3) be repro- grammed or be expanded tosupport other future changes. This robotic flexibility will be advantageous for future update of the system toallow for cleaning of the radiators by the ARID rather than by men hanging over the sides of buckets. In the near future, as the use of the robot becomes more ofa standard operation, as people get more accustomed to using automation, and after the ARID proves itself to bereliable, a modification will be made so that itcan actually clean the radiators by "hand rubbing" the delicate surface. Figure 11depicts the first in-house mechanism concept (aPPRP robot). It consisted of a 65-foot prismatic rail to traverse the length of the four radiator panels, a prismatic cart toreach from the outer edge of the door tothe Orbiter's hinge, and another prismatic rack-and-pinion rail with a rotary joint tomove the inspection device over the contoured surface of the radiators. There were also problems with this configuration that caused itnot tobe implemented. (1) Grease drippings from the Thompson rails and the rack-and-pinion gears used in such adesign would pose problems incleanliness. For example, radiators are covered afterthey are cleaned allowing upper platforms tobe lowered (see figure 12). If they were not covered, dirt or tiny objects could fall on them as personnel walk on the platforms toaccess experiments inthe payload bay. (2) APPRP device would have created problems with the management of linesand cables as the cart moved back and forth. (3) The required inspection path would carry the vision system payload 6inches above the radiator surface (see figure 12). However, thisconfiguration required many closeup pictures and would have produced anexcessive amount ofdata. The time toprocess this data would not have significantly reduced operational timelines. To solve the first two problems, the design quickly evolved intoa PRRR mechanism (see figure 13) in which the cables could be routed within anenclosed space (bending at the joints rather than being dragged over the width of the radiators). Seals can be installed on the joints toeliminate grease drippings. 302 Inordertosolve the thirdproblem, Lockheed (the designers of the payload) requestedthatNASA KSC (therobot designers) investigate designing amechanism that would allow taking photos from 6feet away. Thisdesign would requirethatthe upperaccess platforms beraisedand another revolutejoint beinstalled to lift the nose of thedevice (refer topointA of figure 13). Also, itwould have made itdifficult todesign the robot toallow for later modification so itcan reach the radiator surface for cleaning. However, a6-foot inspection provedtobe too faraway toobtainreliable photos even with zoom lens. Also, since the designers didnotwant the device to hit and damage the radiators ifitfell, it was determined that a24-inch inspection distance would be the required sensing position above the surface of the radiators. The linklengths of the robot were then designed sothe robot could transport the sensor over the work envelope (described infigure 13) without hitting the radiators at its extreme lower position. Amechanical stopkeeps the robot withinthis work envelope. The link lengths were designed sothe stops could beremoved later and still be long enough toreach the surface of the radiators for any future tactile tasks. During the development of the ARID byNASA, Lockheed was performing prototypesensor developmental tests. These tests revealed that a small vision sensor could be built (but ithad tobe aligned with the radiator panels to pick upall defects). Therefore, touse itonthe 5.7-degree sloped surface ofpanel one, the smaller sensor required that the robot be designed with afifth rotation axis (PRRRR configuration). At this time, the design of the robot was 90 percent complete, and this modification would have added too much weight and control complexity. A larger sensor could be built that did not require precise orientation, but the larger vision sensor would not fitinto the operational envelope required totuck the robot into a stowed position. The finalsolution was to build a "crook" or offset inthe robot toenable use of the larger sensor (refer to figure 9). The tolerances aresotight in the OPFthat the robot arm must be moved inordertoopen and close the doors of the Orbiter. It was first thought that ahinge could be added inthe area indicated by point Bof figure 13. This, together with a revolute joint atpoint A, would allow the arm tobe moved to a horizontal plane and then, using the hinge, be moved out of the way and stowed inparallel with the track itself. However, the payload bay doors arenot strong enough tobe opened on the ground inzerogravity conditions; therefore, when opening and closing the radiator doors, a strong-back device isattached tothe outside surface to provide structural support. CAD drawings of the facility revealed that this strong-back device would pass at two places through the only area in which the prismatic track could be mounted (see figure 9). It would also pass through the robot arm if itwere stowed in front of the track. Twosolutions were needed before the robot could be installed without modifying structureswithin the facility. First, an area (or "cubbyhole") at one end of the payload bay doors was located tostore a folded-up arm (refer tothe crosshatches infigure 13). If the arm was not foldable, itwould have tobe removed toopen the doors. There was just enough space available inthe cubbyhole toredesign the arm and joints to fitand still enable reach capability fora 24-inch nontactile inspection path and a6-inch cleaning path. Second, inorder toeliminate all facility impacts, the track was sectioned sothat two shortremovable pieces could be quickly unbolted and slid back. These two modifications allowed the strongback passage through the track's operational envelope and resolved the trmalproblems during the design evolution of the inspection robot. RedundancyRequirements Since this isthe t-h-Strobot that will be installed next toflight hardware (andespecially since ithangs over the radiator doors), the reliability of the robot isextremely important. Recently, several suspended loads at various NASA facilities have had mechanical failures that caused them todrop onto a Spacecraft or flight hardware. If thisoccurs atthe launch pad, notonly could an$80 million Spacecraft be damaged, but itwould cause an aborted launch. Thisproduces expensive consequences resulting inlarge amounts of serial-time/money toreplace and repair. The existing inspection "buckets" have even Coll_ded-withtheradiator doors and bent them upward. The robot has been designed tonot hit any of the radiator panels even in worst-case parking conditions. However, the robot was also designed (for future use) where itcould actually touch the entire surface of the radiators. When, and if, the stops and mechanical constraints are removed, the robot control system must be so reliable that there can be no inadvertant moves to cause even "dings" or "scratches" to the sensitive surface. To prevent this from occurring, the ARID has been designed so that itcontains complete control system component redundancy and additional mechanical constraint redundancy. Eleclromechanieal design of each joint includes redundant chive shafts, bearings, harmonic drives, brakes, transmission chains, encoders, and motors (see figure 14). They are sized for the load, torques, and the space available inthe stowage position. Electrically, the control system includes redundant motor servocontrollers, redundant indexers, andtwo separate control computers. Mechanically, redundancy exists by addinga special 303 cable/pulley configuration (not shown) atjoints two andthreethatprevents thearm fromcolliding withthe radiators even ifboth redundantbrakesormotordrive chains fail. Two motors inparallel could possibly be expected to "fight" each other, but there isa400-to-1 transmission gear ratio and a small torsional compliance in the harmonic drives and development inthe laboratory has resulted in synchronous control of redundant motors. Also, drive components are balanced relative tothe load. Therefore, anynoticeable motor torque imbalance caused bycomponent failure will be detected and used toshut down the system. The internal motor resolvers are compared against duplicate incremental encoders toprovide redundant sensor feedback toenhance troubleshooting bythe two computers. Once the fault isidentified, the operator can shut down one side of the drive, release itsbrake, and use the remaining drive to fold the arm back into its tucked position and stow it inits "cubbyhole." In the case of a failure that requires repair of the robot, this will allow the doors tobe closed without having todismantle the robot arm and will enable the operational worldlow to continue. A master computer and the slave computer each compare itsown calculated kinetic positions, cross-check positions with each other, and check sensor feedback before allowing parallel-commanded moves to continue. The redundant system will be used ina "fail-safe"mode. If one computer fails, the other computer is switched inas a lone controller allowing the operation tocontinue toa safe conclusion. Additional computer interfaces include digital/discrete control of brakes, emergency stop, digital sensors, limit switches, a manual control pendant, and a Lockheed-developed vision system computer data bank. The teach pendant allows manual override and speed control, and provides individual control of the position ofall joints of the robot. Simplication isthe key to the design. Links 2, 3,and 4 of the robot will basically operate ina plane. The arm will be positioned toa location, the joints will be locked, and then the arm will travel lengthwise down a panel (using only the motors in link l's prismatic axis). Next, the revolute joints will be unlocked and the arm will be lowered toanother position that extends outward from the Orbiter (refer tofigure 13). Then the arm willbe moved lengthwise again. Photographs willbe taken without stopping as the system moves at4 inches per second. After an automatic scan, the operator may wish tolook again at an anomoly or take a "still" photograph. He then can either program that position for anautomatic move or manually drive the robot toa point that requires more resolution. CONCLUSION The first two innovations discussed inthis paper (6-DOF Target Tracking System and TSEE) were designed and developed byconwactors (Adaptive Automation Incorporated, and Automated Dynamics Corporation, respectively) working from specifications and with close guidance from NASA. These innovations can be integrated toperform tasks involving orientation and panel operations by using existing technology toperform "smart" autonomous operations. The thirdinnovation (an ARID robo0 was designed byNASA engineers. Contractor support isbeing used tofabricate and install the robot, todetail the teach pendant, and todevelop the vision sensor. NASA engineers and co-op students performed the detail design ofthe joints, links, controls, and kinematics; and they performed kinematic, static, and dynamic analyses of thisnewly developed robotic mechanism as an "in-house" NASA developmental project. It issignificant tonote that the ARID robot involves the application ofcomplete control hardware redundancy toenable performance of work above and near expensive critical Shuttle hardware (see figure 15). There are applications inwhich a robot would not normally be used because of "reliability" or in situations where an inadvertent move could result indamage toexpensive components oreven more expensive personnel. In such applications, the consumer marketplace may wish to explore commercialization of these component redundancy techniques. The three innovations discussed inthis paper are applicable to Space Station work, ground aerospace launch processing, industrial safety, and hazardous petrochemical or nuclear saf'mgoperations worldwide. 304 PAGE IS OF POOR OUAUI"Y Fipr¢ 2. Ream= UmbilicaMlains Figure 4. 6-DOP TmIIe¢DLta'imiMtioa wlm _,= StmutmarTarse¢ inthe RADL The FIVeBright Okeles of the target are positioned In such I way that the video Images ofthem canbe processed Into data onthe position _ orterttatlonof the target relative tothe camera. Figure3.Tech lldd Fiip=re5. 6-DOF Rnlxl Tsclpt Tmdd_ 305 ORIGINAL PAGE BLACK A_D WHITE PHOTOGRAPH ORIGINAL PAGE BLACK AND WHITE PHOTOGRAPH ORIGINAL PAGE IS OF POORqUJ_'Y Figure 6. SmallPhase ITSEE Figure8.RemoteOpec_en ofTSEE Figure 7. Lm_ Ptuu_nTSEE WORK ENVEL PiIm _. IImIir IlIiI_ i_x_ 306 I/ "FlXrn RADIkT01LS DrPLOY/d_L[IUtD,IATOR$ Figure 10. ARID Work Envelope SPACE SHUTTLE PAYLOAD BAY DOOR OPEN AT tSO DEGREES POSITION Figure 13. 24-InchInspection Work Space Figure 11. Early Concept Figure 14. RedundantDrive Components ORIGINAL PAGE IS OF POOR QLIALITY pA¥1.OAO SAY DOOR POS|TIONI I_|N][ I=H I|r_l G(G I_.A1'¥_ 1$ ¢0_vlII_,JI_TION* UIP CtlDmkNCE _ROM rkll_l I_MI, 6 Im CLE:_CI[ FJION ¢M¢ $11HI" e,I_.T$'g 4 IN _10 SP_IM6, 4 IX $¢_.E, ,1_$ IN • I IN _1 I*t.A; i$ Hid tIME -J PE''/_- ""' ,=.|,,= WORK ENVELO Figure12.Conceptfor@InchInspection Figure 15. Final ARID Configuration 307