Collision-free motion planning of multiarm robots using evolutionary algorithms

Abstract This paper presents an evolutionary algorithm for the collision-free path planning of multiarm robots. A global path planning technique is used where the paths are represented by a string of via-points that the robots have to pass through, connected together by cubic spline polynomials. Since the entire paths of the robots are considered for optimization, the problem of deadlock between the arms and the static obstacles does not occur. Repeated path modification is done through evolutionary techniques to find an optimized path with respect to length and collision among the robots and the obstacles and the robots themselves. The proposed algorithm departs from simple genetic algorithms in that floating point vector strings represent the chromosomes and customized operators are used to improve upon the performance of the search. Moreover, a local search is carried out on each individual in addition to the global population based search. The result is a highly efficient path-planning algorithm that can deal with complex problems easily. Simulation results are presented for collision-free paths planned for two planar arms and then for two 3 degree-of-freedom (DOF) PUMA®-like arms moving in three-dimensional operational space.