Research On Human-robot Collaboration Based On SEMG And IMU

With the continuous development of artificial intelligence technology,"human-robot collaboration" came into being.Human-robot collaboration is not a simple division of labor between humans and robots,but to fully develop the complementary capabilities of humans and robots to maximize team performance.This article is oriented to the scene of human-robot collaborative handling,and researches on how to achieve efficient collaboration between humans and robotic arms in this scenario.The main contents of this article are as follows:(1)According to the research needs,the design plan of the platform is determined,and the man-robot verification platform based on sEMG and IMU is built.The secondary development of the EMG sensor and the attitude sensor is carried out,and the upper computer software based on Py Qt is built.The robot arm,the host computer and the sensor are connected to the unified local area network through the router,and the communication is realized through the TCP/IP protocol.The data of the sensors and the posture information of the robotic arm are uploaded to the host computer on the server through the network for data visualization,and the robotic arm can also be controlled through the host computer.(2)Research on one-way adaptation of human-robot collaboration methods,aiming at the strategy of how robots adapt to humans in collaborative scenarios,research on methods of perception and prediction of upper limb movement intention based on sEMG and IMU.First,the recognition method based on the surface EMG signal was analyzed,and the data of 10 subjects were verified through random forest and SVM.Secondly,the shoulder and elbow joint angles of the subject's upper limbs in the sagittal plane were calculated by the azimuth angle of the IMU.Aiming at the problem of measurement lag in IMU data,Bi LSTM was used to predict the joint angle data.Finally,the features of sEMG and IMU are combined.After combining the two features,the accuracy of intention recognition is significantly improved.(3)Aiming at the problem that the one-way adaptation of human-robot collaboration cannot fully consider human adaptability,the research is based on the Q-learning algorithm for human-robot mutual adaptation.The experimental analysis of 5 subjects under different collaboration methods,the statistical analysis of the muscle contribution of the subjects through the non-negative decomposition matrix,and the comparison of the cumulative inclination of completing the same task under different strategies.The result shows that the human-machine collaboration method based on the Q-leanring algorithm effectively reduces the accumulated tilt and improves the efficiency of collaboration.In summary,this article takes human-robot collaborative handling as the starting point to study the collaboration strategy of human and robot.Aiming at the problems of the existing one-way human-robot collaborative methods,the Q-learning algorithm realizes human-robot mutual adaptive collaboration.Way to obtain more efficient collaboration efficiency.