AUTONOMOUS SPACE ROBOT CONCEPT WITH LEARNING CAPABILITIES - LABORATORY DEMONSTRATION OF PEG-IN-HOLE BY BI-ARM

To realize some parts of human superiority and to overcome some difficulties of the present autonomous space robot control system design philosophy, a new concept of autonomous "learning-type" space robot control system design philosophy is presented. In this concept the system is capable of learning necessary information for the given task through numerous input-output data in either on-line or off-line mode. To check the concept a classical peg-in-hole by bi-arm is chosen for laboratory experiments which indicate the concept is realizable. Experimental results show that the system can learn the operation of "peg-in-hole" and also tolerate uncertainties and unknowns in the system.