Biologically motivated push recovery strategies for a 3D bipedal robot walking in complex environments

Though balancing is a fundamental part of human walking, it has been a challenging topic for bipedal robot. Compared to the versatile strategies of handling disturbances of human, current bipedal robots possess limited skills of managing external disturbances. Among them the capabilities of push recovery and maintaining balance are obviously of prior importance for a bipedal robot to walk in an unknown environment. Existing solutions, such as, capture point, walking phase modification, foot placement estimator and etc, are mostly based on simplifying a bipedal robot as a linear inverted pendulum, which result in omitting the effects of knee and ankle joints, natural dynamics, ground reaction forces and interaction between joints in sagittal and frontal planes in a 3D bipedal robot. To overcome those drawbacks, this paper presents a work that it classifies push types, presents a push detector, and proposes control strategies for managing continuous strong pushes occurring during walking. Based on biological and biomechanical research, more efficient strategies, e. g., hip and ankle strategy, a bent-knee strategy, can be transferred to control methods for bipeds. Through generating required torques and/or regulating joint positions in order to stabilize bipedal robot, a push recovery controller is proposed in this paper. Simulation results demonstrate that a 3D physically simulated anthropomorphic biped with 21 degrees of freedom can succeed in recovering from a strong push of up to 200 N during walking in different walking scenarios.