Toshihiko Shimizu Web

Dynamic motions, such as walking, running, jumping and etc, have been one of most difficult tasks in the fields of robotic control and machine learning. I have been interested in a method which achieves dynamic tasks from reusing the experiences of successful motions.
One of the ideas to reuse the experiences is to figure out few critical states and actions to achieve the dynamic task. Especially, we inspired from the transitions of motion states, and proposed the novel motion feature named changing order. So far, we have proposed a method which searches the transitions from Human-Robot Interaction and reuse it in novel scenario.