Global optimum human-like gaits for an articulated one-legged hopper

Achieving energy efficiency remains an important goal in legged robot running. This paper presents a novel offline method to generate globally efficient gaits for a point-foot articulated hopping robot. Previous gait generators found in the literature do not reach a global optimum, since takeoff and touch-down (transition) states are pre-specified. In this work, an optimization procedure produces the joint positions and velocities at transitions in addition to flight and stance phase trajectories. The gait achieves either a desired step length, gait period, or average forward velocity. The simulated optimal gait resembles a human hopping gait.