Rearrangement Planning of Multiple Movable Objects by a Mobile Robot

Abstract

A rearrangement planning methodology of multiple movable objects by a mobile robot is proposed for a certain rearrangement problem, i.e., the LP ε k (k > 2) problem. Several objects are to be moved from an initial configuration to separate destinations as a result of a robot handling operation. Proposed in this paper is a planning algorithm consisting of a global motion planner and a local motion planner. The global planner uses a precedence graph of a movable object, while the local planner uses a real-time search methodology called LRTA* (Learning Real-Time A*). Two important factors in the local planner are the generation of intermediate configurations and the design of heuristic functions. Three methods in each factor are compared in the rearrangement simulation of four, seven and 10 movable objects. The proposed method demonstrates the effectiveness of the proposed algorithm from the viewpoint of task completion time. The calculation costs for both planners are polynomial with the number of movable objects for one time activation. Simulation results indicate that the proposed algorithm can be practically applied, although not complete, to rearrangement problems with, at most, about 10 movable objects.