Humanoid motion planning using multi-level DOF exploitation based on randomized method

This paper addresses a multi-level exploitation of degree of freedom (DOF), for humanoid motion planning based on a randomized method. The improvement of autonomy and mobility is required so that humanoid robot can perform tasks in various environments. Although a humanoid robot has many DOFs, all of them do not have to be controlled depending on the situation and the required tasks. Utilizing rapidly-exploring random trees (RRTs) as an efficient planning tool, a method of multi-level DOF exploitation is developed that adjusts the controlled DOFs according to the detected environmental situation. That allows the planner to deal with only the necessary sets of DOFs instead of exploring the search space of all the DOFs, which leads to efficient humanoid motion planning. The proposed method is verified through simulations using software platform OpenHRP and HRP-2 humanoid robot model.