Synchronous dressing support based on working-status prediction

This paper describes a method for helping a person dress with the help of a robot. We place strong emphasis on the idea of assisting someone who is acting proactively while considering their work capacity. The present study aims to develop a technique for dressing support that fits such a policy. Therefore, we propose a method to support dressing according to a person’s movement by predicting their working status. We define the working status as the positions of the human shoulder and hand and external forces applied to the robot’s wrist and present the neural network structure that predicts them. We also use the prediction results to show a model predictive control (MPC) method. This method allows the robot to determine a hand trajectory that reduces undesirable events such as collisions while proceeding with dressing support in accordance with the person’s movements. In the evaluation, experiments were conducted to support the motion of inserting an arm into a sleeve. We confirmed that the proposed technique can generate smooth end-effector trajectories with little burden while respecting the user’s proactive movement.