Motion generation for human-robot collaborative pick and place based on non-obstructing strategy

This paper presents a model and motion generation method for the task of collaborative pick and place between a robot and a human. It often happens that the performance of a human is higher than that of a robot, in the case where the robot needs to recognize the environment as well as human behavior and move in a safe manner. This fact requires that the robot needs to make motions considering the timing of human motions and properly choose targets of manipulation beyond just considering one stroke of the human motion to enhance the total performance of the human-robot system. Accordingly, in this paper, we first express the robot and human using a state transition model. Then we present a method to generate a motion of the robot to increase future task executability of the human. This approach consists of two steps. In the first step, the system of the robot and human is simulated using 3D model till the near future. In the second step, the utility of the possible object-place layout is evaluated based on possible transportation hypotheses. We also evaluate the effectiveness of the proposed approach in simulation.