Interacting Multiple Model-Based Adaptive Trajectory Prediction for Anticipative Human Following of Mobile Industrial Robot

摘要

In smart manufacturing, the introduction of mobile industrial robot has facilitated efficient and flexible production. Mobile industrial robot has the capability of following human operators to coordinate with them to complete complicated operations. In this context, following target person robustly is a significant prerequisite for mobile robot offering assistance. However, the mobile robot with limited field of view may lose the dynamic target during following, which easily leads to following failure. In this paper, therefore, an anticipative human following approach is adopted. This method is appropriate for omnidirectional mobile industrial robot equipped with a visual sensor with limited perception range. We propose a novel Interacting Multiple Model-based adaptive trajectory prediction algorithm. The algorithm integrates two physics-based motion models and adaptively adjusts model parameters for increasing accuracy of prediction. Based on the prediction, mobile robot plans its path and configuration during following in advance to avoid obstacles and achieve robust following. Comparison results demonstrate that the proposed approach can predict human trajectory more accurately, and reduce the deviation between the human and the center of view of the robot.