The Application of Sensory Information and Multifunction Learning to Autonomous Manipulator Control
This paper describes a computer system which can learn to control a remote manipulator, including the ability to search for goals and subtasks and to carry out fine handling motions. The system is an expanded version of the Autonomous Control Subsystem described in an earlier publication. Like the earlier concept, the new system begins by observing the control actions of a human operator, and subsequently takes over control responsibility. Autonomous control functions are provided by a set of interconnected learning networks which relate the sensory experiences of the machine to the generation of arm and hand motions. For remote applications the computer can be pretrained to operate autonomously in a simulated environment and will adapt to unexpected changes in the actual working environment. The paper presents the mathematical basis of system design, and concludes with a discussion of advantages.
KeywordsDecision Space Task Goal Input Event Trajectory Control Decision Network
Unable to display preview. Download preview PDF.
- Ivancevic, Nebojsa J. and Luigi Cordella, “Artificial Tactile Perception with Computer Processing” Proceedings of the Fourth International Symposium on External Control of Human Extremities, Dubrovnik, Yugo., Aug. 28 - Sept. 2, 1972.Google Scholar
- Goto, Tatsuo, Kiyoo Takeyasu, Tadao Inuyama, Raiji Shimomura“Compact Packaging by Robot with Tactile Sensors„Proceedings of the Second International Symposium on Industrial Robots, Chicago, Illinois, 149–159, 1972.Google Scholar
- Whitney, D.E. “State Space Models of Remote Manipulation Tasks„ Ph.D. dissertation, M.I.T., 1968.Google Scholar
- Freedy, A., F.C. Hull, L.F. Lucaccini, and J. Lyman, “A computer-based Learning System for Remote Manipulator Control„ IEEE Transactions on Systems, Man and Cybernetics, SMC - 1:356–363, October, 1971.Google Scholar