118 非ホロノミックロボットの誤差学習による終端状態制御(初期入力の決定方法)

概要

論文の詳細を見る
This paper considers decision of initial inputs for final-state control with error learning of a two-link cat robot. Since the final-state control method we use can not avoid the state that the denominator of nonholonomic constraint condition becomes zero, to obtain the feedforward input depends on input decision by trial and error. The simulation results we have obtained shows that the direction of waist rotation in early time of twisting motion is most important. From this knowledge we clarify that learning started from the initial inputs that give the right direction to the cat motion realizes the cat twisting motion. Furthermore, we design a two-degree-of-freedom control system to follow the trajectory obtained by the feedforward inputs against variation of the damping coefficient around the waist.
社団法人日本機械学会の論文
2000-09-01