Task generation for humanoid robot walking using human motion by phase sequence

The motion division, called "phase", for generation "task" of humanoid robot is effective method. Task could be realized by method for phase transition, defined as "phase sequence", which can refer a skilled actual human motion without learning. In the method, guarantee for connection between the phase transitions is important problem. This study aim to generate human mimicking task for a humanoid robot based on an analysis of the human motion strategy with phase sequence method. In the approach, human motions were can be classified into several phases based on the patterns of joint angle and floor reaction force. Injecting human motion data into a humanoid robot, we convert the joint angles into desired trajectories, which were transformed by the third-order Bezier curve and constraint function of configuration from actual motion. To realize, these methods, we did motion extraction experiment from human walking with capturing system. Furthermore, phase transition method was applied by a given set of constraint conditions. As a result, task was generated by simulator successfully. Then, walking task for humanoid robot could be verified by proposed algorithm.