Planning walking patterns of a biped robot with uDEAS optimization

In this paper, the recent optimization method uDEAS (univariate dynamic encoding algorithm for searches) is used for generation of biped walking patterns of a humanoid robot. The walking patterns made of optimal blending polynomials are simulated for a simple six degree-of-freedom biped model. uDEAS plays a key role to optimize polynomial coefficients in trajectories assigned to each joint motor. To improve search efficiency of uDEAS, search order of each parameter is carefully determined based on kinematic and dynamic effects. Simulation result of initial stepping shows that the proposed approach is feasible for near real time generation.