Trajectory planning for reaching an object on a table using accumulated action data

This paper describes a robot trajectory planning method for grasping an object placed on a table. In conventional methods, the more obstacles are on the table, the more calculation costs are needed to find a course for avoiding the obstacles; to overcome this problem we use Accumulated Action Data from previous situations. The Action Data is composed of obstacles arrangement and manipulation trajectory and the manipulator acts on the information that is decided by comparing data arrangement with the present arrangement. For evaluating this method, we performed experiments in different conditions on a simulator.