Optimal Trajectory Generation for Bipedal Robots

This paper proposes a new method of trajectory planning for biped robots walking on level ground. In this approach, the hip and foot trajectories are designed in Cartesian space using polynomial interpolation. The key parameters which define the trajectories are searched by genetic algorithm. The objective is to obtain the best trajectory that has large stability margin and low energy consumption. ZMP is used as the criterion to ensure physically realizable walking motion. The effectiveness of our method is verified by simulations of a humanoid robot named NUSBIP-II.