Safety evaluation method of physical human-robot collaboration based on human upper limb model

In view of the special security risks faced by physical human-machine collaboration in heavy-duty assembly scenarios, this paper introduces human model simulation into the study of physical human-robot collaboration, and proposes a method for human motion recognition and safety evaluation which is suitable for the special application scene. First, we collected upper limb motion data, processed the data by human model simulation analysis, and set up a mapping model between hand/shoulder relative position and upper limb joint angle through neural network. Then, we analyzed and calculated the weight value of each joint degree of freedom, and established a safety evaluation method through fuzzy comprehensive evaluation. In this research, the weights assigned to shoulder flexion/extension, abduction/adduction, externalrotation/ internal-rotation and elbow flexion/extension were 0.1103, 0.3579, 0.3229 and 0.2089. The experiment shows that, shoulder abduction/adduction and external-rotation/internal-rotation joints at high risk and above the safety risk level of the time accounted for 70.33% and 86.81% of the total safety risk time, elbow flexion/extension only entered the low-risk state, shoulder joint flexion/extension remained safe throughout the whole process, the weight value of this research is set reasonably, the method can more accurately recognize upper limb posture and estimate the safety risk in physical human robot collaboration.