Mobile robot path planning with η3-splines using spatial-fitness-sharing variable-length genetic algorithm

This paper presents a novel evolutionary G³-continuous (continuous-differentiable curvature) path planner for nonholonomic wheeled mobile robots. The evolutionary path planner generates intermediate configurations connected via η³-splines that yields a collision-free G³-continuous path, which each point on the path has a closed-form expression. The path planner is implemented as architecture of island parallel genetic algorithm (IPGA) running a variable-length genetic algorithm in each island. The techniques of spatial fitness-sharing in search space and a crowded measure of generated paths are integrated in the planner to implement a high-diversity evolutionary optimizer of paths that could self-adjust the spacing and number of intermediate configurations. The experimental result demonstrates the robustness and self-adjusting capability of evolutionary path planner in discovering shorter and smoother composite η³-splines paths in complex environments.