Sensor Based Randomized Diffusion Planner for Higher Order Manipulators in Unknown Environments

Abstract

Unknown environment motion planning with no world model is a daunting task, especially for higher order manipulators. Sensor based planning is a dominant trend for planners in unknown environments. However, unknown environment planning, while important, still falls short of practical solutions for higher order manipulators. An amelioration proposed herein is to make use of a ramification of model-based approaches with mobile sensation. In this paper, we demonstrate how randomized planning techniques developed for model-based planners can be adopted to deal with planning problems of higher order manipulators in the midst of unknown obstacles. No other study has reported useful results on unknown environment planning utilizing model-based theories. The proposed planner is rendered to handle cases where non-sequential random sampling or randomized road map generation is infeasible due to the absence of a world model. For simplicity, the classical lattice planner, or incremental grid sampling, is considered with mobile sensation for probabilistically biased searches. For mobile sensation, we introduce a novel collision detection sensor, namely infrared proximity array (IPA), that is designed to enable samplings in an unknown configuration space. The proposed planner together with the IPA demonstrated some useful results on unknown environment planning problems utilizing a model-based sampling approach. As a performance measure of the planner, resolution completeness of the proposed planner is investigated from the topological standpoint as well