Coping with complexity in the navigation of an autonomous mobile robot
David R. Dodds · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1991
This paper discusses the integrated use of dynamic systems theory neural networks and evolutionary programming methods as a means of coping with complexity in automatic plan generation. Plan elements representing actions are mapped into phase-space and are examined for stability by searching for and identifying any " attractors. " A knowledge-based system for doing this is described. As a means of coping with unexpected environments modifications of plans are made by the planning system using an evolutionary programming method coupled with a neural network approach.