A Short-term Motion Prediction Approach for Guaranteed Collision-Free Planning

Abstract

To enable safe and efficient human-robot collaboration in shared workspaces, it is important for the robot to take possible future movements into account and predict the reachable occupancy of human when performing a task. While human motion is fast and changeable, predicting human motion for tasks unknown a priori is very challenging. However, the existing methods lack the ability to adapt to time-varying of human behaviours. Moreover, many of them do not quantify uncertainties in the prediction. This paper proposes a simple and fast approach calculating the reachable occupancy of human arms in Cartesian space. We use a second order kinematic model which is based on the constraints of human motion such as the maximum velocity and acceleration constraints collected from the demonstrations of different people. The constraint prediction model is conservative and can accommodate the time-varying behaviours of human. Finally, this model has been used to predict the motion of different people. The experiment results show that the proposed method is computationally efficient and robust for all relevant movement.