Data Driven Models for Human Motion Prediction in Human-Robot Collaboration
Qinghua Li,Zhao Zhang,Yue You,Yaqi Mu,Chao Feng
DOI: https://doi.org/10.1109/access.2020.3045994
IF: 3.9
2020-01-01
IEEE Access
Abstract:To improve the safety and effectiveness of human-robot collaboration (HRC), the robot must plan a safe trajectory before the human movement is finished. Therefore, it is necessary to enable proactive robot behavior by making accurate intention prediction decisions early in a human motion. Furthermore, it is desirable to not only provide the long-term trajectory prediction of human motion but also characterize the uncertainty around it. In this paper, we present a human motion prediction framework to predict the motion trajectory of human arm in a reaching task. The proposed framework combines partial trajectory classification and human motion regression. By leveraging on the partial trajectory classification, our framework makes it possible to recognize the human action and to provide a trajectory prediction before the human movement is finished. The human motion regression can compensate the low accuracy of the representative trajectory through the fusion strategy. The proposed framework consists of two phases: online phase and offline phase. The offline phase aims to learn a regression model with optimized hyperparameters and a fusion strategy combining different prediction algorithms. In the online phase, based on the partial motion classification, the future reaching trajectory in a given time step is predicted by using a multi-step Gaussian process regression and representative trajectory. Experimental results show that our proposed framework achieved significant performance.
computer science, information systems,telecommunications,engineering, electrical & electronic