A Framework for Unknown Environment Manipulator Motion Planning via Model Based Realtime Rehearsal

摘要

In this paper, we propose a novel framework for an unknown environment path planning of manipulator type robots. Unknown environment motion planning, by its nature, requires a sensor based planning approach. The problem domain of unknown environment planning is notoriously hard, especially for difficult cases. The framework we propose herein is a sensor based planner composed of a sequence of multiple MBPs (Model Based Planners) in the notion of cognitive planning using realtime rehearsal. That is, by the proposed framework, one can use a combination of model based planners as tactical tools to resolve location specific problems in overall planning endeavor. The enabling technology for the realtime rehearsal is a sensitive skin type sensor introduced in the paper. We describe the developed sensor and demonstrate the feasibility of solving a difficult unknown environment problem using the introduced sensor based planning framework up to 3 DOF linked manipulator.