BipedWalking control using offline and online optimization

The paper is aimed at an energetically efficient control method for biped walking. The walking cycle is composed of successive single-support phases and passive impacts. A parametric trajectory optimization method is implemented that finds symmetric, periodic, low torque walking gaits for a planar 5-link biped robot offline. The robot's configuration is then regulated by RHC (receding horizon control) with the same optimization criterion online. Only the geometric evolution of the robot's configuration is controlled, but not the temporal evolution. The effectiveness of the proposed method is evaluated using simulated walking control. The results show lower torques and more robustness from the proposed controller compared to a hand-tuned PD servo based walking controller.