Understanding Dynamic Human Intentions to Enhance Collaboration Performance for Human-Robot Partnerships

Human-robot collaboration is being implemented into manufacturing processes at a higher rate than ever before. However, many areas within human-robot collaboration still need development in order for robots to understand and work with humans in a human-human collaborative manner. Further investigation will allow for increased safety and comfortability for human workers as well as higher quality for complex, varying tasks. In this study, we propose a dynamic human intention understanding model based on the optical flow algorithm for human-robot teams to improve their collaboration performance. Our approach allows the robot to evaluate and follow its human partner's operation intentions dynamically during collaborative tasks. The proposed model is experimentally implemented by different human participants in real-world human-robot collaborative contexts with accuracy and stability. Future work for alleviating the limitations of the developed approach is also discussed.