Long Short-Term Human Motion Prediction in Human-Robot Co-Carrying

Abstract

In this paper, a long short-term human motion prediction method based on long short-term memory (LSTM) is proposed in human-robot co-carrying task. Long-term human motion prediction is utilized to predict the human motion target in the task. If robot is aware of the target, robot will lead the task to reduce human's control efforts. If robot does not know the human motion target, human will lead the task towards the target. Short-term human motion prediction is involved in hybrid visual-force controller in dynamical level to reduce the motion delay in the process. Finally, the method is verified by experiment of human-robot co-carrying task.