Research On Human Motion Intention Prediction For Industrial Human-Robot Collaboration

With the development of human-robot collaboration,industrial human-robot collaboration is applied in various complex manufacturing environments.human-robot collaborative manufacturing combines the advantages of both industrial robots and workers,making manufacturing develop towards high speed,high precision,and high efficiency.The ideal human-robot collaborative manufacturing needs to ensure the safety of workers and improve the efficiency of human-robot collaboration,that is,industrial robots need to accurately obtain the content of workers 'actions in advance to ensure the efficiency of human-robot collaboration,and at the same time need to obtain the position information of workers' joints to ensure their safety.At present,human motion prediction research mainly focuses on single content information prediction or location information prediction,which cannot guarantee the efficiency and safety of human-robot collaboration at the same time.Therefore,industrial robots need to make more comprehensive predictions of human motion information.This paper studies the human motion prediction problem in the industrial human-robot collaboration scenario,and proposes a human motion intention and location information prediction method suitable for the industrial human-robot collaboration manufacturing environment.The main research contents of this article are as follows:(1)Prediction of human motion position information in industrial human-robot collaboration based on joint point repair.Use the 3D sensor Kinect to directly obtain joint point information and analyze human motion characteristics during industrial human-robot collaboration.A Kalman filter algorithm is proposed to eliminate the jitter phenomenon of obtaining joint points,and the partial point delimitation method is used to solve the joint point offset phenomenon,so as to achieve accurate prediction of human joint points.On this basis,the hybrid prediction algorithm is introduced to predict the position of human motion during the manufacturing process of human-robot collaboration,improving the accuracy of predicting human motion trajectories,reducing prediction errors,and providing a reliable basis for position prediction in human-robot collaboration systems.And lay a good foundation for industrial robots to assist workers in manufacturing tasks.(2)Research on prediction of human motion intention in industrial human-robot collaboration based on spatial attention.By analyzing the impact of environmental factors such as dismantling items and disassembling tools on human motion intent estimation in human-robot collaboration scenarios,video data containing scene information is used to model human motion content information prediction problems.A feature extraction method based on the spatial attention mechanism is introduced to extract key areas of human motion in video data to improve the accuracy of human motion recognition.On this basis,combined with the improved loss function and intention discrimination of the adjustment factor,the optimal early discrimination result is solved to realize the prediction of human motion intention information.(3)Design and development of human motion prediction prototype system for industrial human-robot collaboration.In order to verify the effectiveness of the content-based early prediction algorithm based on spatial attention and the trajectory prediction algorithm based on joint point repair.Based on industrial robot systems,vision sensors and other hardware equipment,a human motion prediction system for industrial human-robot collaborative manufacturing is designed and developed.The sensor is used to collect picture data and skeletal flow data,and the data is processed by the proposed method,and an experimental system is established to verify the algorithm performance.The experimental results show that the proposed algorithm can achieve good prediction of motion intention information and location information in human-robot collaborative manufacturing environment.