Mobile manipulator configuration optimization using evolutionary programming

Abstract

Multi-degree-of-freedom manipulators are becoming commonplace on mobile platforms. Full autonomy of mobile manipulator robotic systems will depend on the ability to resolve the inherent kinematic redundancy in task commutation. This work investigates the application of an evolutionary search strategy for determining near-optimal mobile manipulator configurations. Joint torques, obstacle avoidance and manipulability are incorporated in a multi-criteria optimization formulation. A variety of aspects of the evolutionary programming paradigm are addressed via empirical studies on a two degree-of-freedom (DOF) manipulator. These studies investigate full configuration vector versus partial configuration vector mutation as well as mutation strategies which incorporate cost and iteration number. The results of this study are then applied to a planar three DOF manipulator mounted on a single DOF mobile base. Experiments indicate that the configuration optimization problem is amenable to a variety of mutation strategies.