This segment of the program is dedicated to the development of component technologies which have been determined to be of potential benefit in addressing multiple needs of the known robotics requirements. These elements of the program are typically long lead-time items, which may take many years to fully develop and bring to an appropriate level of readiness. However, if successfully completed, these elements typically have the potential of significantly improving or even revolutionizing the state-of-the-art in space telerobotics technology. This portion of the current program includes such elements as fundamentally new robotic joint designs, exoskeleton systems, microtelerobots, and widely-applicable proximity sensor technology. The long term goal of this effort is to develop a series of component technologies which can then be incorporated into larger robot assemblies and full application systems. This effort is phased such that technology components "spin-off" from the component development level to the next level on a regular basis. It is anticipated that this area will continue throughout the life of the program, producing an increasingly-beneficial series of fundamental technologies.
Telepresence Control At Planetary Distances
Exploration of Small Bodies
Nano-rover Technology Development
Passive Proximity Sensing
Low Mass, Compact, Muscle Actuators
Multiple Interacting Robots
Perception for Advanced Robotic Interfaces
Redundant Robot Systems
Telepresence
Interfaces for Planetary ScienceThe objective of this task is to develop and evaluate haptic recording and playback techniques, specifically for application to remote planetary science exploration. We will incorporate haptic data that captures surface texture and material properties. We will develop novel interfaces that capitalize on this sensory modality to provide operators with some "feel" of the remote environment. Throughout the program, we will conduct field trials to evaluate and validate the tools and robotic interfaces.
Geologists performing field work make extensive use of the full range of the human sensory spectrum. The primary sensory interface is, of course, stereo color vision. Often overlooked, however, is their dependence on touch and smell. The human sense of smell is basically a chemical analysis, and can be reproduced to some extent with spectral or composition analysis instruments.
The human sense of touch, however, is not normally well reproduced for field work. High-bandwidth, low time-delay force reflecting systems can reproduce contact forces for a human teleoperator, but current systems are not designed for use with significant time delay or low-bandwidth communications links to the remote location. In addition, these systems cannot easily reproduce subtle textures humans can sense by passing their fingers over a rough surface. Geologists gain a tremendous amount of information from touching, squeezing, and breaking rock and soil samples.
This task will address this problem by developing techniques for recording and playing back object surface textures and material characteristics by a combination of force sensing and surface roughness sensing, with an emphasis on techniques that will work well in the presence of lengthy time delays.
Approach:
We will divide the problem into three tasks: (i) haptic data acquisition, analysis, and compression, (ii) haptic data recovery and playback, and (iii) integration of haptic playback into a virtual environment remote science interface and science field test evaluation of the system. We will use the respective expertise of the three participating institutions to accomplish these tasks. CMU will concentrate of the first task, MIT/ARC will concentrate of the second task, and ARC will concentrate on the third task. CMU and ARC will jointly perform the science field test evaluations.
Haptic Data Acquisition
Acquiring force-torque data from a sensor attached to a manipulator is relatively simple. However, directing a remote system to accomplish these measurements without direct human-in-the-loop control is more challenging. We will develop techniques to allow a remote manipulator system to autonomously acquire such data.
The determination of the roughness of the material will require the use of compliant manipulators and vibration sensors. These sensors must be designed to sample the roughness to the point where a human operator replaying the texture can identify the sample.
We will develop probing motions to acquire the haptic data. We will assume we have an arm with a force/torque sensor, and a controller providing position and force control. To acquire surface texture, we will use the arm to drag the sensor along the object surface, while maintaining a small normal force. To measure compliance, we will use the arm to exert a force and record load versus sinkage. To measure shear strength, we will use the arm to exert a normal force and then start pulling out laterally; at the instant when the arm pulls out, we record the joint angles, forces, and torques which determine the angle of internal friction and the cohesion of the material. To measure compressive strength, we will exert direct compressive force on the sample. For plasticity, we will identify the point at which the object begins to flow.
Feature Extraction and Compression
Once the haptic data are acquired, we will analyze the data to determine surface textures and material properties. Surface texture can be determined by using the spectrogram of surface elevation data. The spectrogram gives the power at every frequency as a function of time. The material properties (compliance, strength, plasticity) can be determined from critical points in stress/strain curves.
A large compression factor will be obtained by transmitting only the coefficients of the various material properties. For surface textures, we plan to transmit non-zero elements of the spectrogram.
Haptic Data Playback
An additional task for the texture replay capability is to design and test various texture output devices. Prototypes of these include tunable vibrators in a force-reflecting hand controller and electronic braille output devices. The MIT Phantom device is an excellent candidate as a very capable haptic replay device, and will be evaluated as such for this task.
Science User Interface Integration
Once the basic techniques of haptic playback are well enough understood, we will integrate the haptic playback system into a remote science operator's interface, derived from the Virtual Environment Vehicle Interface. The task is to integrate haptic data with other science data transmitted from the remote planetary work site. Other science data includes image data, spectral data, composition data, and terrain elevation data. Haptic data in isolation will not be of much use to the remote geologists, so the integration of the parallel data sets is important.
Once the initial integration is accomplished, we will begin science field testing of the system during terrestrail analogue rover missions. Candidate missions are simulated lunar traverses (CMU), Rocky/Pathfinder tests (JPL), and Marsokhod tests (ARC). The system will be evaluated in actual field use by planetary geologists, and the feedback will be used to modify and improve the system.
Focus and Directions:
FY 1996 Develop haptic data acquisition, analysis, compression, and playback techniques for surface texture properties. Evaluate end-to-end system with test object textures ranging from rough to smooth.
FY 1997 Develop haptic data acquisition, analysis, compression, and playback techniques for material compliance properties. Evaluate end-to-end system with test objects including rock (granite, marble, pumice, etc), granular material (sand, gravel, silt, loam, etc), ice, and snow. Field test prototype system using a science team to evaluate the system effectiveness.
FY 1998 Develop haptic data acquisition, analysis, compression, and playback techniques for material strength and plasticity properties. Evaluate end-to-end system with test object textures ranging from rough to smooth. Field test prototype system using a science team to evaluate the system effectiveness.
Points of Contact:
Butler Hine
(415) 604-4379
hine@ptolemy.arc.nasa.gov
Eric Krotkov
(412) 268-3058
epk@cs.cmu.edu
Exploration of Small BodiesFuture exploration of interplanetary small bodies, such as comets and asteroids, requires technology development in a variety of areas. Landing and operation in the low gravity environment of small bodies, ranging from 10-4 to 10-2 meter per second squared, is an extremely challenging problem. The Exploration of Small Bodies task will focus on developing the enabling technologies to accomplish in situ scientific studies of these interplanetary objects. The primary objectives of this task are to develop the mechanisms and autonomous control algorithms to perform landing, anchoring, surface/subsurface sampling and sample manipulation for a complement of science instruments.
A critical enabling technology required for a majority of the scientific investigations of small bodies is the process in which the sample acquisition, manipulation and transport to the science payload is accomplished. A key technical innovation in this effort will be to perform subsurface sampling and transportation to the science instruments under the expected extreme resource constraints of Discovery class missions. Planned in this regard will be a subsurface drill mechanism with an integrated sample acquisition system. Transport mechanization to a complement of science instruments will be attempted to be embedded in the design.
Another critical enabling technology required is the landing and attachment to small bodies.
A lander could use crushable material on the underside of a base plate design to absorb almost all of the landing kinetic energy. An anchoring, or attachment system, would then be used to secure the lander and react forces and moments generated by the sample acquisition mechanisms. Impact absorption materials will be configured and characterized, as will the systems used for anchoring.
Cryogenic testing in cometary simulation materials will demonstrate realistic performance of the anchoring systems, and full-up system level demonstra=tions of a lander test bed will occur.
Focus and Directions:
FY 1996:
Design, fabricate and develop a "pre-prototype" drill and sample acquisition mechanism. Incorporate force and torque sensors and develop preliminary control algorithms. Perform preliminary impact absorption component testing. Perform preliminary anchoring design, fabrication and development. Test ambient environment drill mechanisms and anchoring components in cryogenic cometary materials (Level 1 Milestone).
FY 1997:
Integrate mechanisms, impact absorption system and anchoring system into a lander test bed. Perform "end-to-end" system level test demonstrating impact absorption, anchoring, drilling and sample manipulation (Level 1 Milestone).
FY 1998:
Design and analyze legged lander system. Develop control concepts for landing and hopping including "cat like" maneuvers for stability. Study anchoring for multiple landing, investigating removable anchors, multiple ejectable/frangible anchors and grasping anchors. Develop "pre-prototype" test leg concept. Demonstrate landing and impact absorption with a directly driven legged lander test bed (Level 1 Milestone).
FY 1999:
Utilize the lander test bed to implement advanced sample handling and mobility hardware. Integrate advanced mobility systems onto lander test bed and perform end-to-end system level testing, demonstrating hopping and advanced anchoring system
Point of Contact:
Donald R Sevilla
818-354-2136
Donald.R.Sevilla@ccmail.jpl.nasa.gov
NanoRover Technology DevelopmentRecent advances in microtechnology and mobile robotics have made it feasible to create extremely small automated or remote-controlled vehicles which open new application frontiers. One of these possible applications is the use of nanorovers (robotic vehicles of the order of 10-50 grams) in planetary exploration. Such vehicles could be used, for example, to survey areas around a lander, or even to be distributed along the lander descent trajectory, and to look for a particular substance such as water ice or microfossils. The objective of this task is twofold: to create a useful nanorover system using current-generation technology including mobility, computation, power, and communication in a 10-50 gram package, and also to advance selected technologies which offer breakthroughs in size reduction.
NASA Code-S planetary missions have been under increasing pressure to reduce their launch mass requirements so that less expensive launch vehicles can be used. For example, the Delta launch vehicle is less than one-fifth the cost of the Titan, and the Taurus is less than half the cost of the Delta. In order to launch on these inexpensive vehicles, significant reductions in mass must be achieved. For example, the Science package on post-Pathfinder landers is expected to be 20 kg or less. To achieve those aspects of scientific exploration requiring mobility, any rover component of the science payload must compete effectively in terms of mass against other payload options. Microrovers (<10kg) were first proposed in 1987 partly in response to this, and soon after nanorovers (10-50 gm) were proposed for the same reasons. Nanorover technology would allow some mobility-based science surveys, such as the search for water ice or other volatiles, microfossils, or other entities at or very near the surface with a small, perhaps negligible fraction of the science payload. This latter point makes it conceivable that nanorovers could fly on most landers using whatever mass margin is left over at launch time.
Focus and Directions:
This activity focusses in the near term on creating a vehicle of 10-50 gram mass. The vehicle has two tracks, like a miniature tank, driven by small gearmotors which are current commercial technology similar to that used for autofocussing in camcorders. The main activity of the task is to create an integrated sensing, processing, and control circuit, with a "fisheye" wide-angle lens which focusses light onto the chip for imaging. This circuit module combines the image sensing with all the processing and control functions to gather the desired information and to provide mobility control based on limited autonomy and operator commands. It also has several photovoltaic cells which provide power, and a primary lithium battery for periods of operation in the shade.
FY 1996: Creation of a tethered 10-50 gram nanorover, with sensing and actuation functions on the vehicle but computation performed off-board for convenient algorithm development.
FY 1997: Creation of the integrated sensing and control circuit module, incorporating the algorithms developed in '96.
FY 1998: Integration of the sensing and control module into the ~10 gram mobility chassis; field testing and performance characterization.
FY 1999: Integration of sensing and control module into ~1 gram mobility chassis.
Point of Contact:
Brian Wilcox
(818) 354-4626
brian.h.wilcox@jpl.nasa.gov
Passive Proximity SensingThe objective of the Passive Proximity Sensing task is to develop and demonstrate a focus-based proximity sensor that measures object distances between 50 and 500 cm with a radian-scale field of view. The sensor will operate passively (emit no electromagnetic energy); operate in ambient illumination (from dark shadows to worst-case sunlight); consume mass, volume, and power comparable to a CCD camera on small satellites; consist of components with a plausible path to flight-qualification; and require modest computation (no matching or search). The proposed sensor is particularly well-suited for two proximity sensing applications: lunar rovers, and Orbital Replacement Unit (ORU) change-out.
Approach:
The technology development will focus on the following six topics.
1. Image Formation. To eliminate moving parts, we will configure the sensor to gather light with a single lens, and split the beam into two (or more) images formed with different camera parameters. We will investigate image formation with different (fixed) apertures, and different (fixed) focusing distances. We will not use different focal lengths, since this requires additional lensing and introduces the need for significant image rescaling.
2. Image Detection. We will consider two image detection technologies: charge-coupled devices (CCDs), and position-sensitive detectors (PSDs). We will provide control of the detector integration time, which will enable operation over a wide range of illumination conditions, and if necessary can normalize the response between different detectors.
3. Image Processing, We will consider four options: a digital signal processing chip with video processing; an MPEG chip; simple bandpass filters; and optical computing.
4. Sensor Design. Once we select the sensor configuration (image formation, image detection, and image processing), we will then perform a detailed sensor design. In contrast to existing implementations of the range-from-defocus approach, we will develop a sensor with state-of-the-art components in a carefully engineered system package.
5. Calibration. The measurements will be read from the sensor in non-metric units. We will develop a calibration procedure that converts the readings into meters and degrees.
6. Performance Characterization. With the calibrated sensor, we will undertake a systematic program of performance characterization, measuring key sensor characteristics, and evaluating the range measurement accuracy and precision as a function of critical independent variables.
Focus and Directions:
FY 1996 Design and fabricate laboratory model
FY 1996 Characterize laboratory model speed and accuracy under a wide variety of target and illumination conditions
FY 1997 Fabricate, calibrate, and characterize field unit.
FY 1997 Demonstrate field unit on rover with accuracy better than 5% under Air Mass 0 conditions (Level One milestone)
Point of Contact:
Eric Krotkov
(412) 268-3058
epk@cs.cmu.edu
Artificial Muscles from Electrostrictive PolymersFY 1996 Program Plan This research area is dedicated to the development of actuation technology based on electrostrictive polymers (ESP). Muscle actuators will be developed to demonstrate the unique and superior actuation capabilities of these materials compared to electrostrictive ceramics. ESPs will be developed and fabricated from polymers bearing highly polar pendant groups to produce a strong electrostriction effect. The application of ESP is involved with manageable technical challenges which will be addressed in this proposal in order to develop breadboard muscle-actuators for space applications. The performance of ESP-muscles will be demonstrated in a series of rigorous terrestrial tests including simulated space environment
Focus and Directions:
FY 1996 - Effective ESP material will be developed using a backbone polymer with a high free-volume to allow easy alignment of the pendant groups in response to an activation electric field. Also, investigate muscle configurations using i-PMMA which was reported to have a high electrostriction coefficient.
FY 1997 - Muscle-actuated manipulators will be designed to comply with specifications that are determined in the initiation phase of this program. The manipulators will be fabricated and tested to determine the performance envelop and define requirements for ESPs..
FY 1998 - The performance of ESP muscle-actuators will be demonstrate by designing and fabricating an arm, which will be driven by ESP materials. This demonstrator arm will represent a mechanism that is determined to be relevant to a flight experiment.
Point of Contact:
Yoseph Bar-Cohen
818-354-2610
Yoseph.Bar-Cohen@jpl.nasa.gov
Multiple Interacting RobotsThe Aerospace Robotics Laboratory (ARL) at Stanford University is pursuing two fields of research as discussed below:
Control of Free-Flying Space Robot Systems
This research area is dedicated to the development of Free-Flying Space Robots that can be directed at an intuitive, task level to perform typical space operations. There are three prototype space robot vehicles employed for experimentation, and there are two major focus areas of these experiments. The first area is concerned with upgrading the FFSR's toward operational systems. This includes incremental modification our present FFSR design to incorporate various 3D capabilities and development of human-interface/robot-control techniques for greater functionality in unstructured environments. The second area of research is concerned with expanding the basic set of tasks that the FFSR's can perform (such as "capture", "move", "dock", etc.), and to increase the reliability of the current set of task capabilities through the development of new control techniques.
Focus and Directions:
FY 1994 Demonstrate neural network-based thruster map system, capable of identifying thruster misalignment or failure and adaptively adjusting controller to compensate.
FY 1995 Perform autonomous robot-satellite rendezvous using the Global Position System (GPS) as the only navigation sensor. The capability demonstrated by this experiment can 1) be directly applied to future on-orbit free-flying robot servicers as a high-precision navigation and rendezvous control system and 2) establish the basic infrastructure for precision formation flying of clusters of small satellites. This completed research project has demonstrated that GPS sensing ALONE is sufficient to perform precise intercept and capture of a free-floating target vehicle by a free-flying space robot. The robot was required to grasp inside 5cm diameter ports on the target vehicle. This represents better than two orders of magnitude improvement over conventional GPS techniques, which are limited to accuracies on the order of 10 meters.
FY 1995 Assemble structures using a team of Free-Flying Space Robots. This experiment will be a test of new robotic control theory for complex dynamic systems, including OBTLC of intimately cooperating teams of free-flying two-armed robots. It will enable multi-DOF robot-manipulator systems to perform precise position and force control autonomously for the first time in highly dynamic environments, such as space.
FY 1996 Identify, track, and manipulate objects about which the robot has no prior knowledge, using the human operator to assist in cogent object modeling through simple visual interaction. This capability will enable robots to effectively operate and manipulate objects in unstructured environments, while the human directs, at a high level, tasks to be performed with the object. The performance of the new experimental system operating in an environment where object information is deficient will be compared directly to that of the same hardware platform performing identical tasks with objects whose characteristics are known exactly and precisely a priori. Such metrics as object capture success rate, and absolute positioning accuracy will be quantified for operation in the more challenging environment where no prior knowledge of object characteristics is present. See metrics table Ic.
FY 1996 Maneuver a Free-Flying robot quickly and efficiently through a dynamically changing obstacle field. In particular, intercept and capture a target object that is moving through the obstacle field along a path that is not known beforehand. This experiment will demonstrate ongoing real-time task planning: the capability to perform sophisticated optimal trajectory generation in real-time on a robotic system moving through a changing environment. The Free-Flying robot will be able to react to unforeseen changes in the world around it, including impulsive disturbances to the moving target itself, while always flying in a time- and fuel-efficient manner. See metrics Table Id.
High Performance Control of Flexible Manipulators
This area of the research pursues general theoretical advances in automatic control theory that can be applied to greatly improve the control of a variety of robotic and other very flexible dynamic systems such as the RMS and its successors. In order to refine and transfer these theories to operational systems, all of the results are experimentally verified. This research is primarily conducted on three hardware platforms: a Single-Link Flexible Manipulator with complex dynamic payloads, a Two-Link Flexible-Drive-Train Manipulator with a mini-manipulator at the end effector, and a Two-Link Flexible Manipulator which carries a fast Mini-manipulator. There are three major thrusts of this research: Adaptive Control of Manipulators with Complex Dynamic Payloads, Optical Tracking of Objects using an Arm-Mounted Camera, and End-Point Impedance Control for a multiple-link Flexible Manipulator carrying a Fast Mini-manipulator.
Focus and Directions:
FY 93 Developed and experimentally verified optimal control techniques in conjunction with flexible mode modeling to control a two-link flexible manipulator equipped with a mini-manipulator. These techniques showed order-of-magnitude increase in speed for positioning the end-effector of a flexible robot arm.
FY 1994 Demonstrated control techniques for cooperative robotic assembly of flexible parts. This new capability expands the range of objects that autonomous robots can reliably assemble.
FY 1994 Track an object or follow a line in real time using a macro-mini manipulator equipped with a wrist-mounted camera. This capability can be applied to automate the family of surface inspection routines.
FY 1995 Demonstrate adaptive control for manipulation of payloads that possess significant internal dynamics. An example application is the shuttle RMS manipulating a satellite that contains fuel or has flexible appendages.
FY 1996 Demonstrate ability to achieve high-performance control of flexible structures with unknown multiple D.O.F. dynamic payloads. An example application is the shuttle RMS manipulating a satellite that contains fuel or has flexible appendages.
FY 1996 To obtain precise end-point control of flexible structures, an accurate dynamic model of the total system must be available. However, an unknown dynamic payload results in an uncertain total system model, which degrades the achievable closed-loop performance.
FY 1996 The result of this research will be an on-line adaptation that identifies the total system model and modifies the control system accordingly. Specifically, this adaptation is expected to lead to a closed-loop bandwidth which is a factor of three greater than that achievable through conventional control methods.
FY 1996 Use a flexible macro-mini robot arm to demonstrate stable contact with an object in the environment and regulation of the contact force between the tip of the manipulator and the object. This work is intended to resolve several fundamental control issues that are of concern to attached servicer robots currently under development.
This research will demonstrate force control bandwidth at least a factor of 4 greater than current collocated control methods and will demonstrate for the first time ever impedance control at the tip of a flexible manipulator carrying a fast mini-manipulator.
Control of Redundnt, 7-DOF Manipulators
This focus area of research concentrates on extending object-based task-level control to redundant, 3D robots. This effort is a collaboration between ARL researchers and researchers from the Intelligent Mechanisms Group (IMG) at NASA Ames Research Center. Many different algorithms exist to resolve motion redundancy based on kinematic constraints such as object avoidance, joint limits and joint-singularity avoidance. There are several 'dynamic' algorithms that minimize kinetic energy, joint forces, and acceleration profiles. However, these approaches disregard the task when constraining the internal motion of the robot. What would be desired is to decouple the task-control specification and the internal robot control into two independent problems. Thus, the end user can begin by specifying things like control bandwidth of the task, compliant and non-compliant directions of motion, etc. The robot will then configure itself and its motion to meet these dynamic constraints which were developed independent of the robot. The results of this project are applicable to many programs, specifically to control of the manipulator mounted on the Marskahod robot and the Ranger robotic system. Another area to which this work is directly applicable is of course flexible automatic assembly.
Focus and Directions:
FY 1996 Demonstrate improved task performance as a result of the new redundancy management scheme. This will be demonstrated by comparing current state-of-the-art solutions with the new algorithms. Specific comparisons that will be made: task bandwidth in the entire workspace, time to complete given tasks, and stability of system as manipulator moves through singularity. All of the performance measure will be made on the experimental system consisting of the Robotics Research Arm.
FY 1996 Build multi-use end-effector for the Robotics Research manipulator. This will be used for force contact tasks that will help to quantify performance of new algorithms.
Point of Contact:
Robert Cannon
(415) 723-3602
cannon@sun-valley.stanford.edu
Point of Contact:
Butler Hine
(415) 604-4379
hine@ptolemy.arc.nasa.gov
This effort performs research to advance fault tolerance in reconfigurable manipulator systems. For space operations, adaptability as well as reliability are required to protect space assets to ensure that robots remain capable of physical task performance over long duration missions. Specific objectives of this program are to develop commonality in fault tolerant reconfigurable robotic architectures and to accommodate adaptive control strategies for robotic manipulator systems which can dynamically respond to drastic failures in the robot manipulator mechanisms, sensors or control systems, and yet maintain stable end- point trajectory control with minimum disturbance. Desired systems level of fault tolerance is provided by selective incorporation and intelligent utilization of excess redundancy available in manipulator systems. This effort develops a roadmap. Various kinematic designs using redundant, modular, reconfigurable joint actuators and kinematic redundancies are being pursued at a fundamental level. Standardization of functional requirements (capabilities) and guidelines are being developed to provide commonality for future space based robot joint mechanisms. The commonality allows the incorporation of advanced adaptive and reconfigurable controls to provide the desired level of performance and fault tolerance in manipulator systems.
This effort enhances three areas in fault tolerant robotics: 1) develops a standardized fault tolerant robotic joint architecture which demonstrates the functional requirements of redundancy and independence in sensing and control to provide for fail-safe operations. The architecture is extensible to incorporate increased levels of fail-operational capabilities, 2) develops a system level control system architecture that incorporates robotic control, fault detection and isolation, and manages the reconfigurability of available redundant resources, while meeting the safety requirements for human space flight. Implementation of distributed servo controllers to support future embedded electronics within the manipulator are being evaluated and refined to meet the NASA safety requirements and to enhance fail-operational capabilities, and 3) develops enhanced robot mechanism component technologies for space applications, such as brakes, joint torque sensors, clutches and gearboxes, which are compact and scaleable in size. These elements will be environmentally tested and evaluated at JSC, and functionally included in internal designs.
This research of failure tolerance in manipulator system design is being conducted and demonstrated by the development of mechanism testbeds with duality embedded within the joint actuators and by the use of manipulators with excess redundant degrees of freedom. The fault tolerance concepts are validated in integrated hardware testbeds for application to space mechanism designs. This work is being conducted at The University of Texas at Austin under the sponsorship of the Johnson Space Center.
Approach:
Robot joint actuators which incorporate internal dual actuation provide the first line of defense against manipulator faults which terminate the operational use of a robot system. This project's development of a dual servo actuator which can provide cost effective commonality is the first of its kind making fault tolerance feasible for all actuated systems in space for the first time. The first module prototype showed major improvement over standard industrial practice (volume down 3x, weight down 2.5x, stiffness up 17.5x, horsepower up 1.5x, gear train reduction down 3x). Similar benefits for space are being pursued at this time by projects in development at JSC and GSFC.
No planned robot manipulator in space provides fault tolerance except in a few component technologies (communications, software, electronics), but none in the mechanical components (brakes, gear trains, clutches, etc.) or in the operational controls system software. This project makes reconfiguration of redundant resources available for space robots for the first time, which allows the robotic system to tolerate the effects of a fault and continue operations. This reconfiguration action includes a general process for fault detection and isolation (under development for 10 independent faults) and for operational criteria (joint limits, dexterity, load capacity, stiffness, tracking error, etc.) and associated real time software (less than 10 milli-sec.) for 10 or more operational criteria. Best present practice for a future space robot is 3 criteria. This program is laying the foundation to include up to 100 criteria, far beyond anything in existence anywhere in the world.
To obtain fault tolerance means that special component technologies must be developed. A clutch to decouple an incapacitated actuator drive element is under development (none now meets stringent space requirements). A fault tolerant brake has been designed which is dramatically better than standard industrial practice (torque/weight up 3x, volume down 2x, faster response up 2x, power consumption down by 700x, and redundancy for fault tolerance at 4x). The significant reduction in sustaining power consumption to hold off the brake during robot operations is already a key improvement. An actuator torque sensor of very high resolution (10x better than at present) and compactness (3x better than at present) is now being developed.
Modularity includes both hardware and software. Modularity makes rapid assembly and repair feasible which is essential for future long duration missions such as lunar base. This approach means that a very large population of robots (from 3 to 10 DOF) can be assembled from a very small collection of modules on demand. It has been argued that 10 carefully chosen modules will be sufficient to make any robot system needed for space. Development cost should come down (at least by 20x) and maintainability and up-dating capacity (logistics operations) should improve by a factor of 10x. These are fundamental to the success of all future robot missions in space and can now become the basis of mission planning (thus far achieved in terms of very expensive, monolithic, one- off designs). Also, the modular operating software (up to 1000 modules) can be selected on demand to make a universal control system essential to the above logistics argument for reduced spares, assembling systems on demand, and rapid tech mods. This type of universal software does not exist anywhere else in the field of robotics.
Focus and Directions:
FY 1996: Provide the modular electro-magnetic brake element to JSC for environmental testing and evaluation. Complete the design and initial prototype development and testing of the joint torque/position sensor for space requirements.
Goal: Determine how temperature, vacuum, etc., affect the ability of the brake to stop a load, how fast, and how frequently before operation deteriorates. Parameter characteristics and lessons learned will be used to improve the brake design for space deployment and to recommend a flight experiment.
FY 1996: Redesign the distributed controller architecture used in the 2 DOF knuckle module testbed to meet NASA safety requirements for fail-safe operations, targeting the RRC joint redundancy proposed for ISSA. Develop and demonstrate limited fail-operational capability using this level of duality.
Goal: Operate the dual motor windings, sensors and brakes from either one or both actuator controllers to show that fault tolerance at the electronic controller level is also feasible. This effort is believed to be a first in the field of robotics. The 2 DOF knuckle module has clutches, but these will be disabled to simulate the RRC motor module.
FY 1996: Extend the 2 DOF knuckle actuators beyond coupled dual motor windings with the inclusion of clutches to improve fail-operational capabilities. Complete the development and preliminary testing of the fault tolerance performance of the clutched servo drive mechanisms, with the incorporation of enhanced fault tolerance control of this configuration. Demonstrate the improved fail-operational capability from the management of redundancy independence, which reduces the single point failure modes remaining within the joint mechanism.
Goal: The final form of the 2 DOF knuckle test-bed will be wired to allow the testing for 10 distinct failures (communications, current, voltage, sensor, brake, clutch, motor, etc.) at each actuator module. Preliminary FDI software will be implemented to identify the faults and to initialize a reconfiguration of the system to best accommodate the fault. This level of complexity has never been tried before. Questions on how well and quickly a fault can be identified and how fast the system can respond to absorb and therefore tolerate the fault will be studied.
FY 1996: Provide selection criteria for providing robot fault tolerance capabilities in future robotic systems. Detail for ISSA dexterous robotics and External Work Systems.
Goal: UT-Austin is committed to work with the JSC staff as consultants on the External Work Systems robotics effort and to provide design and selection criteria for the development of a very small precision manipulator and associated end effectors. Several meetings at JSC and UT are expected to be supported during the year. This effort will be under the direction of Charles Price at JSC.
FY 1997: Provide the joint torque/position sensor element and the dual actuator clutched servo mechanism to JSC for environmental testing and evaluation.
Goal: A prototype actuator torque/position sensor will be fabricated to improve the state of the art, as found in current RRC systems, by 5x and reduce cost by 10x. Best practice today is a 22 bit resolver costing $2500 which is sensitive to environmental effects (temperature, vibration, etc.). These items remain cumbersome in their need for delicate assembly and large physical volume (10 cu. in.).
FY 1997: Develop and simulate advanced criteria based control of new dual actuator architectures based on up to 10 standardized separate operational sensors. These extend to temperature and vibration sensors for decision assessment.
Goal: Best present practice for actuator sensors for space would be 2 sensors (position and torque). This effort will show how to use 10 sensors combined to maximize performance (100x better than standard practice, as represented by RRC systems) for an actuator and to prevent the likelihood of saturation (the ultimate actuator failure). The 10 sensors then provide the foundation for a truly versatile FDI and fault tolerance at the actuator level which does not exist today. Culminated improvement will be demonstrated using 10 independent performance measures selected by NASA/JSC.
FY 1997: Further develop algorithms and provide several demonstrations of fault detection and isolation at the actuator level. Demonstrate the failure tolerant control capability by management of the reconfigurability of the redundancies incorporated at the individual joint level in the 2 DOF knuckle module.
Goal: Improve accuracy of FDI technique demonstrated in FY96 by 5x (and to reduce false alarms) and the decision making software speed by 3x for the 2 DOF knuckle module test-bed.
FY 1997: Further develop the science of criteria fusion for overall performance indices of redundant manipulator systems, establish finite number of simulations for criteria fusion, demonstrate some of these results in the RRC 17 DOF system.
Goal: Expand the number of criteria being fused from 10 to 25. Establish 5 useful overall performance indices, provide 10 unique simulations for criteria fusion, and demonstrate in the 10 DOF serial configuration and the 14 DOF dual arm configuration for the 17 DOF Robotics Research Dual Arm System in UT's laboratory.
FY 1998: Redesign the joint torque/position sensor based on JSC environmental tests and integrate in architectural design of actuator module. Further test the space related clutch design at JSC for fault tolerant actuators. Refine the design of the fault tolerant brake for space actuators and integrate in architectural design of actuator module.
Goal: Concentration on actuator component technologies will be refined to meet space requirements based on tests at JSC. Objective is to be able to transition this technology to the Robotics and Automation Division at JSC or other groups in NASA and their contractors. A workshop will be held at JSC or UT for that purpose.
FY 1998: Further develop criteria fusion for performance, reconfigurability, and fault tolerance of redundant manipulator systems which incorporate both joint and kinematic redundancies. Finalize object oriented software for real time decision and control including fault tolerance and demonstrate in the RRC 17 DOF testbed.
Goal: Expand operational criteria from 25 to 75. Expand overall performance indices from 5 to 10. Demonstrate in 20 different task regimes on the 17 DOF Dual Arm System. Show how human intervention can provide guidance to improve overall task performance (ultimate goal is to reduce burden to the astronaut, yet skill him to be knowledgeable about using this versatile and aggressive technology). Hold workshop to show the NASA community how this works and can be used in present and future robot systems in space.
FY 1998: Finalize FDI formulation and demonstrate against 10 distinct actuator faults for integration in the fault tolerant control of the next generation of space manipulator systems.
Goal: Refine the accuracy of the FDI software to simulate the integration of signals from 10 distinct actuator sensors against an array of 10 arbitrary and unplanned faults in the actuator. Formulate a structure to integrate this technology into future actuator designs for space.
FY 1998: Finalize and demonstrate first level criteria based control of full architecture of the dual actuator servo module based on 10 standardized operational sensors.
Goal: Establish and simulate a 20 performance criteria basis for the model reference control and fault avoidance of a dual actuator module suitable for future deployment in space. Use 10 sensors to obtain a full sensor model description of the system. The difference of the sensor and model references becomes the basis for control. Provide 10 different threshold indices to enhance performance or to avoid faults at the actuator level. Determine the relative accuracy of the model parameters and sensor signals required in order to make the use of the threshold indices as a reliable basis for control (i.e., do they have to be 5x more accurate than the performance indices themselves). Demonstrate the system in real time (less than 10 milli-sec.) on standard PC systems.
Point of Contact:
Del Tesar
(512) 471-3039
tesar@mex.cc.utexas.edu
with any comments, criticisms or corrections for this
page.