Robust Human Motion Estimation for Human Robot Collaboration

Abstract

Accurate prediction of human motion trajectory is a key factor to ensure the safety of human-machine collaboration. To improve the safety and effectiveness of human-robot collaboration, the robot must make accurate intention prediction in the early stage of human motion and realize active robot behavior. This paper proposes a robust human upper limb end trajectory prediction algorithm, which can effectively Improve the accuracy of trajectory prediction. The basic framework can be described as follows:1) According to the complex characteristics of human upper limb motion patterns, a robust Gaussian mixture model (RGMM) is proposed to model the motion trajectory. It can consider more comprehensive parameter information in the optimization process, and reduce the impact of traditional algorithms that are sensitive to initial values, easy to fall into local convergence and need a priori clustering value. 2) Gaussian mixture regression (GMR) is used to predict cluster trajectories and obtain statistical values of future trajectories. The experimental results show that the RGMP algorithm proposed in this paper can accurately describe human movement and make predictions, which not only ensures the safety of the human body but also has important significance for improving production efficiency.