Research On Teaching Technology Of Cooperative Robot Based On Multi-sensor Fusion

With the development of intelligent manufacturing,industrial robots play a more and more important role in industrial production.Among them,a large number of cooperative robots that can interact with people and work together appear in the market.However,due to the low efficiency and high professional threshold of the traditional robot teaching technology,it is more and more difficult to satisfy the requirements of the work and production at present.Although many teaching techniques with good interaction have been developed by researchers in domestic and overseas,due to the limitations of their respective sensors,they have not been able to realize the teaching control of cooperative robots very well.Therefore,this paper proposed a cooperative robot teaching technology which combines Leap Motion,accelerometer,magnetometer and gyroscope to obtain the hand motion of the operator,so as to realize the intuitive control of the robot.First of all,the overall teaching control scheme was put forward.The somatosensory equipment Leap Motion and three kinds of sensors were preliminarily developed,and the data of the sensors were read and analyzed by writing computer programs,and then their own inherent coordinate system was obtained.The attitude of the carrier was obtained by integrating the data of accelerometer,magnetometer and gyroscope.The kinematic model of the cooperative robot used in this paper was established,and the mapping relationship among human hands,sensors and robots was constructed.Secondly,in order to solve the problems of low accuracy,limited measurement range and environmental noise in the detection of hand motion by single sensor,the method of sensor data fusion was used to make up for the defects of each sensor.In the summary of the existing main sensor data fusion algorithms,combined with the specific requirements of this topic,an algorithm fusion model based on back propagation neural network(BP neural network)was constructed.Through the adjustment of the weights of each neuron and the gradient drop method,the neural network was finally optimized.Finally,a perfect robot teaching system was constructed.Aiming at the noise problem and the drift problem of accelerometer,magnetometer and gyroscope in the process of somatosensory equipment Leap Motion detection,Kalman filter was used to eliminate the noise problem and the drift problem of accelerometer,magnetometer and gyroscope respectively.Based on the theory of admittance model in the field of force control,an online variable scale factor algorithm based on acceleration information observation was constructed to adjust the velocity mapping relationship in the process of teaching control.The teaching software and human-computer interaction interface were programmed with C # language,and all the above functions were integrated so that they have the functions of online real-time control and offline speed reproduction.Finally,the feasibility and practicability of the system were verified by experiments.