Dynamics simulation on object handling tasks by using a myoelectric upper limb prosthesis with bimanual coordination

We have proposed the upper limb prosthesis control system which helps an amputee perform a two-handed task with the healthy arm and the prosthesis in our previous work. Our first focus was on motion generation of the prosthesis. However, ideal motion of the prosthesis does not guarantee success of a two-hand task because of interaction between the prosthesis and the handled object. In this paper, we demonstrated dynamics simulation of lifting-up-a-box tasks with the healthy arm and the prosthesis in order to evaluate effect of interaction between objects such as contact and friction on task performance. We prepared the five boxes with different size in order to consider training error and generalization error of the neural network based motion generation. We showed that lifting-up-a-box tasks with different boxes can be successfully completed on the dynamics simulation considering contact and friction by using our proposed prosthesis control system.