A deterministic sampling-based approach to global footstep planning for humanoid robots

Global humanoid footstep planning in complex domestic environments is different from mobile robot path planning because humanoid robots do not always circumvent obstacles considering their unique capabilities of stepping over or upon certain obstacles. This paper presents a deterministic sampling-based approach to global footstep planning for humanoid robots, which not only considers this unique locomotion capability, but also the legged kinodynamics and stability constraint. A dynamic footstep transition model is employed in the proposed footstep planner which improves the planning efficiency as well as the feasibility in certain environments. A frame of posture transition trajectory planning, which includes ¿offline generation - online calling¿ and ¿offline generation - online modification¿ modes, is also described. Simulation results verify the feasibilities and improved performance of the proposed footstep planning approach and its affiliated trajectory planning methods.