Gait Generation for Passive Running via Iterative Learning Control

This paper proposes a novel framework to generate optimal passive gait trajectories for a planar one-legged hopping robot via iterative learning control. The proposed method utilizes variational symmetry of the plant model in executing the steepest decent method in the learning algorithm. This allows one to obtain solutions of a class of optimal control problems without using precise knowledge of the plant model. Furthermore, its application to a hopping robot produces a passive running gait trajectory with zero input. Some numerical examples demonstrate the effectiveness of the proposed method