Research On Human Motion Tracking And Predicting For Human-robot Collaboration In Manufacturing

Human-robot collaboration combines the advantages of industrial robots and human workers effectively which reduces production costs and improves product quality,make manufacturing more flexible and personalized and develop to high speed,high precision and high efficiency,in manufacturing.It is important for human-robot collaboration in manufacturing to ensure the safety of workers.In other words,when industrial robots cooperate with workers in the same environment,they will not harm workers directly or indirectly.Therefore,it is an urgent problem to ensure that industrial robots can accurately obtain the current and future positions of workers to ensure workers' safety and understand workers' intentions.In order to solve the occlusion problem of human motion tracking and the accuracy of human motion predicting for human-robot collaboration in manufacturing,this paper researches the method of human motion tracking and predicting method for human-robot collaboration in manufacturing and the main contents of this paper are as follows:(1)Research on human skeleton information fusion for human-robot collaboration in manufacturing.According to the characteristics of human joints,a suitable human posture description model is established.The specific human-robot collaborative manufacturing environment is analyzed to determine the number and location of visual sensors in the collaborative environment.Based on the evaluation of the data quality collected by different visual sensors in the same environment,the measurement confidence of each vision sensor is obtained.A data fusion model is established to solve the problem that information data acquired is incomplete or instable due to the interference between multiple vision sensors and occlusions.(2)Research on human motion tracking and predicting for human-robot collaboration in manufacturing.In order to predict the human motion for human-robot collaborative in manufacturing,a new method is proposed to track and predict the human motion obtained by depth vision sensor.A filtering algorithm is utilized to preprocess the raw data of human motion to reduce jitter caused by sudden changes in human motion.In addition,to handle the unstable prediction results caused by different movement speeds of human joints,an interpolation method based on the human joint constraints and human motion velocities is used to adjust the parameters of the prediction algorithm adaptively.The two proposed algorithms are combined to improve the accuracy of the predicting results and reduce the predicting error.As a result,it provides a reliable basis for the predicting of human intentions in the human-motion collaboration,and lays a good foundation for industrial robots to assist workers in carrying out manufacturing tasks.(3)Design and implementation of human motion tracking and predicting system for human-robot collaboration in manufacturing.A human motion tracking and predicting system in human-robot collaborative manufacturing is designed and developed.The multi-visual sensor system is used to collect the human skeleton information,and the proposed method can realize data fusion,motion tracking and predicting accurately.An experimental system is established to adjust the parameters of the algorithm and verify the performance of the algorithm.Results demonstrate the effectiveness and feasibility of the algorithm.