Improving the Versatility of Humanoid Walking Machines

This thesis presents strategies in trajectory planning and control to improve the versatility and robustness of humanoid walking machines. By treating the task of generating reference motions for the robot as an optimization problem, the proposed trajectory planning system enables rapid changes in walking direction. The control system is enhanced by introducing feedback paths inspired from neurobiology. The reactive component utilizing the additional sensory information significantly increases the robustness of walking on rough terrain.