Research On Sensorless Collision Detection Method For Modular Robot Based On Observer

Modularization is an important feature of the current development of industrial robots.Modular robots have an irreplaceable position in traditional industrial robots due to their reconfigurable characteristics.In order to improve production efficiency,robots need to cooperate with workers to complete production tasks,so robots need to have collision detection function to ensure the safety of man-machine cooperation.In this paper,the main work of robot collision detection without sensors is as follows :Firstly,the Newton-Euler method is used to establish the dynamic model of the robot,and the minimum model parameter set is determined by simplifying the linearized dynamic model.In order to obtain high precision dynamic model parameters,the excitation trajectory is designed and optimized based on finite Fourier series.Based on the optimized excitation trajectory,the least square method is used to identify the minimum parameter set of the dynamic model,and the accuracy of parameter identification is verified by multiple sets of trajectories.Secondly,this paper takes the modular joint driven by harmonic reducer as the research object,and proposes to use the torque estimation algorithm to obtain the joint torque information,avoiding the installation of torque sensors in the joint.The proposed torque estimation algorithm describes the relationship between the deformation and torque of harmonic reducer with structural damping model,and identifies the parameters of nonlinear structural damping model by grey wolf algorithm.The torque estimation experiment is carried out by using the identified model parameters,and the experimental results show that the accuracy of torque estimation is significantly improved.Then,in order to improve the real-time and robustness of the robot to perceive the external torque,a high-order disturbance observer is proposed based on the first-order disturbance observer.The simulation program is written based on the identified robot dynamic model,and the observation results of the high-order disturbance observer and the first-order disturbance observer are compared.It is determined that the high-order disturbance observer has better real-time performance and robustness.In order to distinguish between unexpected collision and expected contact in human-computer cooperation,a combined threshold detection scheme for collision detection is proposed.The proposed combined threshold detection scheme is verified by applying simulated expected contact and unexpected collision in the simulation system.Finally,a collision detection experimental platform is built,and the collision detection algorithm based on high-order disturbance observer is verified on this experimental platform.In order to determine the actual effect of different collision response strategies,multiple sets of experiments were designed to compare the effects of four collision response strategies,namely,stop,reverse avoidance,zero force control and reflection control.The comparison results show that the reflection control has the best protection effect.