Slip Detection and Recovery for Quadruped Robots

What should a legged robot do when it slips? When traction is lost, the locomotion can be irreversibly hampered. Being able to detect slippage at the very beginning and promptly recover the traction is crucial for body stability and can make the difference in a situation where falling is not an option. Indeed, the majority of locomotion controllers and state estimation algorithms rely on the assumption that the stance feet are not slipping. The following work presents a methodology for slip detection and estimation of the friction parameters, plus a recovery strategy which exploits the capabilities of a whole body controller, implemented for locomotion, which optimizes for the ground reaction forces (GRFs). The estimation makes use only of proprioceptive sensors (no vision). Even though the essence of the approach is quite general, the implementation is specialized for the quadruped robot HyQ. Simulation results demonstrate the effectiveness of the proposed approach while walking on challenging terrains (a slippery ramp or an ice slab).