로봇 프로그래밍을 위한 인간 시연 동작 경로의 중요도 측정 방법

To represent plans of a robot using motion signals captured from a few demonstrations of a human, we propose a framework to possess following three abilities: (i) to divide motion trajectories of a whole task into meaningful subtasks, (ii) to measure degree of motion significance at every time index, (iii) to represent such plans using PDDL (Planning domain definition language). To validate our proposed method, this framework is evaluated using a robot arm that performs a manipulation task: the task of inserting a ball into a ring using a cup.