Graph-based optimal reconfiguration planning for self-reconfigurable robots

The goal of optimal reconfiguration planning (ORP) is to find a shortest reconfiguration sequence to transform a modular and reconfigurable robot from an arbitrary configuration into another. This paper investigates this challenging problem for chain-typed robots based on graph representations and presents a series of theoretical results: (1) A formal proof that this is a NP-complete problem, (2) A reconfiguration planning algorithm called MDCOP which generates the optimal graph-based reconfiguration plan, (3) Another algorithm called GreedyCM which can find near-optimal solution in polynomial time. Experimental and statistical results demonstrate that the solutions found by GreedyCM are indeed near optimal and the approach is computationally feasible for large-scale robots. Keywords—Modular Robots, Optimal Reconfiguration Planning, Computational Complexity