Research And Application Of Collision Detection Algorithm For Sensorless Cooperative Robots

With the "Made in China 2025" and "the 14th Five-Year" robot industry development Plan "proposed,China’s cooperative robot industry has been rapidly developing,and human-machine cooperation environment human and robot will inevitably collision,so safety has become the primary key issue.In this paper,by studying the collision detection mechanism of robots,a method based on machine learning is proposed to compensate the modeling errors of dynamic models,so as to improve the detection accuracy of traditional collision detection schemes and ensure the safety of human-machine collaboration.In this paper,the Denavit-Hartenberg(D-H)parameter is used to establish the model of the cooperative robot,the Newton-Euler method is used to build the dynamic model of the cooperative robot,and the viscous friction and coulomb friction are used to establish the friction model on the basis of considering the current offset of the motor.Aiming at the problem of parameter identification of the dynamic model,this paper linearized the dynamic model,used the linear dynamic equation to construct an incentive trajectory for the cooperative robot,and adopted the pattern search algorithm to optimize the incentive trajectory under specified constraints.Kinematic data,including joint Angle,angular velocity and joint torque,were collected during the execution of the incentive trajectory on the robot experimental platform.Finally,the least square method was used to identify the kinetic parameters.Due to the inevitable errors in the process of production,assembly and model parameter identification of cooperative robots,the establishment of dynamic model and friction model is inaccurate.Therefore,this paper proposes a dynamic error compensation model based on Long Short-Term Memory Network(LSTM)network.This model has multiple sets of inputs and multiple sets of outputs.The network uses the mot ion data of several groups of cooperative robots to train and predict the difference between the predicted torque and the actual torque of the robot dynamics model,so as to improve the prediction accuracy of the dynamics model and the collision detection accuracy.Finally,two ways of offline and online experiments were used to verify.As for the offline experimental scheme,the end of the cooperative robot is collided during its point-to-point movement.The collision detection experiment is conducted using the data of the fifth joint of the cooperative robot.The experimental results show that the proposed collision detection scheme can effectively improve the collision detection accuracy of the cooperative robot compared with the uncompensated scheme.For the online experimental scheme,the end of the cooperative robot is still collided during its point-to-point operation in this paper.The test results show that the cooperative robot can effectively detect the collision through the proposed collision detection algorithm and stop the operation.