Farokh B. Bastani, Ph.D., Department of Computer Science
Work focused on robust planning for coordinating multiple mechanisms in achieving some goal state. The state space of the system is mapped onto a massively parallel cellular automaton, reducing the planning problem to that of computing the shortest path between the initial state and the goal state. An algorithm was developed to compute the shortest path (in the more difficult Euclidean rather than the Manhattan sense), using simple, fully decentralized actions at each cell in the cellular automaton. ISSO investigators enhanced the planning method in several ways to make both the planning process and the plan itself tolerant of the failure of cells or mechanisms. This methodology includes the use of expanded source and destination points, block paths to reduce communication requirements, and different mapping schemes (uniform grid, non-uniform grid, and quadrilateral with various degrees of redundancy).
The cellular automaton planning method is being extended to multi-agent systems that utilize multiple robots in a manufacturing plant and robots designed to explore an unknown terrain. The specific problem being addressed is technology for moving robots simultaneously while ensuring that they do not collide with each other. Systems with various degrees of sophistication of individual agents have been simulated. Results show that a few simple rules will enable the collection of agents with rudimentary intelligence.
ISSO investigators are considering two generic structures for cooperating multi-agent systems. The first is the bucket brigade model where all agents work on different portions of the same task. Each task is split into a number of categories; the failure of an agent triggers its neighbors to take over its responsibility. The goal is to achieve this level of fault-tolerance without adversely affecting the performance of the system during failure-free periods.
In the second structure, termed the parallel brigade model, each agent completes a task in its entirety, with a number of tasks to be performed. The failure of an agent, in this structure, should be covered by the remaining agents who must eventually complete all tasks. As in the first case, the goal is to ensure the normal performance of the system.
Return to Table of Contents
Guanrong (Ron) Chen, Ph.D., Department of Electrical Engineering
Intelligent control theory and techniques are being developed with application to the adaptive control of tele-robots working in an uncertain environment where sudden changes of conditions and/or unpredictable faults occur. Researchers seek to develop projects in conjunction with the Robotic Development Section of NASA-JSC. ISSO researchers seek to develop a realizable framework for the proposed intelligent control process. Strategy under study includes key components in the design of a control system in such a way that the adaptive controller can perform the following functions:
To detect impending breakdowns and diagnose the potential problems, researchers are attempting to design a realistic fuzzy detection and diagnosis subsystem for the control process.
Most of the diagnosed sources in electrical or mechanical diagnosis of robots (or, more
generally, industrial systems) are fuzzy to some degree. Some basic issues like
problem-sources or trouble-causes can be defined as fuzzy sets, basic phenomena like the
robot's "symptoms"; its electrical or mechanical reactions can be described by
fuzzy functions. Usually, diagnostic analysis employs the traditional two-valued logic
(e.g., diagnosing if there is a trouble-cause or not). In practice, however, problems and
troubles are usually multi-level, multi-degree, and/or multi-characteristic. For instance,
to report the occurrence of a problem (trouble), the final result (decision of the
engineer) is needed to grade the activity as "very severe," "relatively
severe," "not so severe," or "no problem at all," etc., in order
to suggest an accurate resolution of the problem. Thus, multi-level fuzzy system modeling
and multi-valued fuzzy logic decision-making are desirable for the diagnosis.
Therefore, efforts are focusing upon a few key components in the design of the control
system:
After diagnosis has been completed, researchers can employ the conventional technologies for adaptive controller design to recommend and execute corrective action and continue the originally designed control process, justifying decisions based upon the new output with appropriate feedback. The combination of the fuzzy detection and diagnosis subsystem and their use by the conventional adaptive controller constitute a very efficient intelligent control system design for tele-robots, an evolving technology clearly applicable to other industrial systems. The new system can work in any uncertain environment where sudden changes of conditions or unpredictable faults may occur from time to time during the entire objective-oriented control process.
Under partial support from this ISSO grant researchers expect to complete Parts (1)-(4), and to improve some of these results by the end of this year. The on going research will combine fuzzy logic principles with neural network architecture to strengthen the real-time capability of the developed fuzzy PID controller.
Partial results obtained from Parts (1)-(2) constitute the M.S. thesis of Mr. Huaidong Li, and those from Parts (3)-(4) are covered by the M.S. thesis of Mr. Weiming Tang.
The principal investigator currently has a team of graduate students working on this project.
 |
Helen Zhang (left) adjusts a robotic arm in the Robotics and Control Laboratory. Ms. Zhang, who earned her baccalaureate at East China Normal University, Shanghai, is pursuing a master's degree in electrical engineering. |
| In the Robotics and Control Laboratory (right), Darren Zheng (standing) utilizes robotic equipment intended to test the fuzzy controller. Zheng earned his baccalaureate degree in Zhejiang University, Hangzhou, China, and his master's degree from Northwestern Polytechnical University in Xian, China. He is currently pursuing his doctoral degree in electrical engineering. Zheng observes Mohan Upadhyaya studying the remote manipulation of robotics, the task made more difficult because tof the time lapse between issuance of commands on earth and the response on the moon. Although a few seconds may lapse between the command and response, the difference poses difficulties in the performance tasks. Upadhyaya earned his B.S. in electrical engineering from the University of Mangalore, India, and is pursuing his master's degree in electrical engineering. |  |
Return to Table of Contents
Thomas L. Harman, Ph.D., School of Natural and Sciences, University of Houston Clear Lake; Cathy Chilton; research assistant; James Dabney, Intermetrics, specialist in control systems and mathematical modeling; Luon An Nguyen, Ph.D., Lockheed, specialist in robotics control systems; Hatem Nasr, Honeywell, specialist in robotics vision; Saganti Premkumar, research associate, RICIS, specialist in image processing
The Robotics and Computer Control Systems Laboratory is dedicated to research in the areas of robotics, control systems, digital signal processing, and computer systems design. The laboratory is generously supported by the High Technology Laboratory, the Institute for Space Systems Operations, Motorola, Inc., and the Computer Engineering Group at the University of Houston Clear Lake.
The laboratory acquired an Eshed Robotec Robot in 1993. The 6-degree-of-freedom Scorbot ER-VII is a robotic arm and manipulator controlled by a vision system. The unit is being used for research in the areas of control and vision processing. A controller is being designed to allow direct control of the robot using a high-level command language. Preliminary work utilizing the ̉CÓ programming language has proven successful. Consequently, a simulation program for the robot kinematics is being developed to test advanced control strategies. On a second front, research is being conducted on medical applications of the robot.
The laboratory continues to build its resources with state-of-the-art equipment and software. A second CCD camera has been purchased for research in 3D vision. The new CCD camera, attached to the robot gripper, is being used for high resolution differentiation between objects. A 486 workstation with windows software was purchased for analysis and simulation. MATLAB, SIMULINK, and Matcad software packages are available at the workstation.
The laboratory continues to sponsor papers at national conferences and invites experts in the field to address basic issues on the UHCL campus.
Return to Table of Contents
J. R. Jagannatha Rao, Ph.D., and Yi-Chao Chen, Ph.D., Department of Mechanical Engineering
Efficient strategies for the Simulation and Optimization of Rocket Trajectories (SORT) are being sought complementary to those under examination by colleagues in the context of Taguchi Methods. In particular, studies are focusing upon a mathematical model that forms the basis of SORT and offers ways buy which the simulation count can be reduced for a given confidence level. Based on preliminary study, findings afford the following reasoning: If the underlying mathematical model is well behaved, then it must be possible to approximate the extreme or the outlier set for a given dispersion of parameters in a simulation. Often, these approximations of undesirable outcomes (of the simulation) can be obtained at low numerical cost. Then, the only remaining tasks are (1) to examine the quality of the approximations and (2) to establish confidence measures for such approximate characterization of "badness."
Upon this premise, the following strategies are feasible:
Researchers can deliver the following:
Return to Table of Contents
Jeong-A Lee Ph.D.Department of Electrical Engineering
Tele-robot kinematics computation is an important field in NASA and many other fields of aerospace engineering. Tele-robots working in space applications generate many inverse kinematics problems to be solved in real time; consequently the so-called Coordinate Rotation Digital Computer (CORDIC)-based engine has been proposed as an additional device to be attached to a conventional computing system to meet the real-time requirement. With the support of ISSO, researchers investigated architectural alternatives based on CORDIC for inverse kinematics computations including the pipelined architecture which utilizes fully the regularity of CORDIC with simple local communication by unfolding CORDIC recurrence equations and their implementations in a Field Programmable Gate Array (FPGA). FPGA technology was chosen for its fast turn-around time and design flexibility. The study demonstrated the feasibility of the CORDIC processor, especially for t he Stanford Manipulators. A 486 PC-based system and upgrade FPGA design tools have been purchased through this ISSO grant. Pipelined implementation requires more module area but produces faster execution time. Currently ISSO researchers are investigating ways to reduce the modules The results obtained from this study will provide insights applicable to real-time and large-scale computations of inverse kinematics of tele-robots and can contribute to many ongoing NASA projects related to robotics. ISSO has funded this research for two years.
Return to Table of Contents
Contents
ISSO -- Institute for Space Systems Operations
1992-1993 Annual Report
|
|
