The Bio-inspired Behavior-Based Bipedal Locomotion Control (B4LC) system consists of control units encapsulating feed-forward and feedback mechanisms, namely motor patterns and reflexes. To optimize the performance of motor patterns and reflexes in terms of stable locomotion on both even and uneven terrains, we present a learning scheme embedded in the B4LC system. By combining the Particle Swarm Optimization (PSO) method and the Expectation-maximization based Reinforcement Learning (EM-RL) method, a learning unit is comprised of an optimization module and a learning module embedded in the hierarchical control structure. The optimization module optimizes the motor patterns at hip and ankle joints with respect to energy consumption, stability and velocity control. The learning module generates compensating torques against disturbances at the ankle joints by combining the basis function derived from state information and the policy parameters. The optimization and learning procedures are conducted on a simulated robot with 21 DoFs. The simulation results show that the robot with optimized motor patterns and learned reflexes performs a more robust and stable locomotion on even and uneven terrains.

Human shows remarkable skills in reactive balancing control on unknown disturbances while standing and walking. Though current bipedal robots can walk, run and step obstacles, they normally perform in a well-controlled environment. Unexpected perturbations can cause the tumbling of bipedal robots when they possess limited capability of rejecting disturbances. Studies upon neurology and psychophysics of human stance attempt to trace how the human deals with external disturbances. This paper introduces a methodology for disturbances estimation and compensation (DEC) for bipedal robot during standing. Previous psychophysical studies of human self-motion perception indicate that humans estimate and compensate disturbances as follows: firstly, multi-sensory inputs are fused to provide explicit measures and then the estimations of the external disturbances are performed based on them. Then, the estimations are fed into a local feedback control loop, compensating the disturbances. Thus, an approach of disturbance estimation and compensation is developed according to the psychophysical aspects of human. Various experiments, for instance, standing on a rotating plate with varying frequency and amplitudes and continuous external contact forces upon the torso of a bipedal robot, are implemented. Through analyzing and verifying the experimental results on a simulated biped, one can state that the DEC approach successfully serves the bipedal robot during perturbed stance.