Experimental results for 3D bipedal robot walking based on systematic optimization of virtual constraints

Feedback control laws which create asymptotically stable periodic orbits for hybrid systems are an effective means for realizing dynamic legged locomotion in bipedal robots. To address the challenge of designing such control laws, we recently introduced a method to systematically select a stabilizing feedback control law from a parameterized family of feedback laws by solving an offline optimization problem. The method has been used elsewhere to design a stable gait based on virtual constraints, and its potential effectiveness was illustrated via simulation results. In this paper, we present the first experimental demonstration of a controller designed using this new offline optimization method. The new controller is compared with a nominal controller in experiments on MARLO, a 3D point-foot bipedal robot. Compared to the nominal controller, the optimized controller leads to improved lateral control and longer sustained walking.