Design of Footstep Planning Controller for Humanoid Robot in Dynamic Environment

In the problem of humanoid robot stepping over dynamic obstacle in the sagittal plane, the duration, the length of the step, and the maximum step height of the biped robot are key factors which influence successful obstacle avoidance. In this paper, an approach of footstep planning for biped robot has been presented. Fuzzy Q-Learning algorithm has been developed to learn these crucial parameters. Cubic spline interpolation has been applied in order to interpolate the trajectories of the swing foot between the designated points such as starting point, landing point and maximum step height. Inverse kinematics has been adopted to compute the joint trajectories of the swing leg. The simulation results show that the robot can step over dynamic obstacle successfully.