Research on Optimal Motion Decision and Trajectory Generation for Table Tennis Robot

The success for robot's playing table tennis often hinges on quickly and accurately deciding when, where and how to hit the ball. In this article, we formulate this task as a non-linear optimization problem which aims to minimize a cost function related to input aggressiveness. Besides, we proposed a data-driven method for robot's motion decision based on Gaussian process regression. Furthermore, considering the robot's kinematic and dynamic constraints, we introduce a 2-stage trajectory generator with low computational cost. Eventually, we evaluate their performance in simulation with Kuka LBR iiwa 7 R800 and the results indicate that the presented methods are highly efficient in robot's hitting table tennis.