Closed loop trajectory optimization based on reverse time tree

This paper addresses the general methods for creating the approximately optimal closed loop stabilization controllers that depend on given rigid body systems. The optimal stabilization controllers calculate the optimal force (torque) inputs that depend on the current states of the systems. In this paper, a creation method for approximately optimal controllers named closed loop optimizer based on reverse time tree (CLO-RTT) is proposed. In the open loop optimization phase, this method creates approximately optimal open loop solutions using rapid semi-optimal motion planning (RASMO). In the closed loop optimization phase, this method selects a solution from the RTT according to the measured current state of a system. The proposed method was validated in the time optimal stabilization problem of a double inverted pendulum model. The proposed method successfully stabilized the model quickly. When the resolution of RASMO was higher, the motion time was shorter.