Learning control method for throwing an object more accurately with one degree of freedom robot

摘要

The present paper proposes the throwing manipulation method which cannot only control the position and orientation of the rigid object at the final goal, but also make the object track the desired trajectory which minimizes the impact at landing to avoid damages to the object as much as possible. Due to a model uncertainty, it is generally difficult to obtain the exact model and find appropriate control inputs to realize the throwing manipulation by experiments. Instead of raising the accuracy of the throwing model itself, we overcome the problems of the approximate model with errors by proposing the position-based iterative learning control method. We show experimentally that the proposed control method can gain the high positioning performance for throwing a rigid object with one degree of freedom robot.