Augmented Motion Plans for Planning in Uncertain Terrains

Abstract

In exploration of unknown planets, ground vehicles such as Mars rovers may not know exactly what terrain they will run into, causing great difficulty in meeting their goals. This paper presents a two-stage approach for motion planning in uncertain terrains. In the first stage, we utilize a specialized planner to generate motion plans to meet some arrival requirements. In the second stage, we augment the motion plans with sensing information and combine them to form a full- fledged controller in order to cope with uncertainty in the environment. This separation of planning and uncertainty management can simplify the development of planners for complicated goals. Our preliminary experiments showed that a vehicle can meet the arrival requirements with a high probability in small random graphs.