Kinematics Parameter Identification And Compensation Of An Industrial Robot

Industrial robots have gradually been applied in various fields.Traditionally,industrial robots are programmed by teaching programming method.With the development of intelligent manufacturing technology,offline programming has become an important development direction of industrial robot technology.The offline programming method of industrial robot requires the robot to have high absolute positioning accuracy,which is much lower than its repeatability.Kinematic calibration is an important means to improve the absolute positioning accuracy of industrial robots.As a key part of calibration,parameter identification aims to identify the kinematic parameters of industrial robots and provide a basis for error compensation.In this paper,a hybrid optimal parameter identification method based on BPNN-PSO is proposed.The results of simulation and experiments demonstrate that the method can greatly improve the efficiency of parameter identification of industrial robots.Taking the ER20-C10 robot as the research object,the kinematics model of the robot is established using the MDH method.The kinematics forward and inverse solutions of the robot are discussed.The kinematics model of the robot is verified using Matlab Robotic Toolbox.The error model and the pre-identification error model of the robot are established,which provided theoretical support for the parameter compensation of simulation and experiment.The principle of the traditional least squares parameter estimation is studied.To solve the problem of non-convex optimization of the least squares method,a parameter identification method of industrial robots based on PSO is proposed.However,PSO has the problems of large numbers of iterations and slow convergence in high dimension space.Then,a method of parameter identification for industrial robots based on BPNN-PSO is proposed.To verify the effects of the above three methods,the kinematics parameter errors of the industrial robot are identified by simulations,respectively.The results of the simulations show that the parameter identification based on the least squares method can reach the best accuracy among the three algorithms,but the influence of redundant parameters should be considered and the memory is greatly occupied.Although the standard PSO can also effectively identify the parameter error,but it has large numbers of iterations and slow convergence.Comparatively,the BPNN-PSO can effectively reduce the number of iterations and improve convergence speed.The three parameter identification methods are studied experimentally.The data acquisition and error compensation platform for industrial robots is set up.Joint angle data and position data of the industrial robots are acquired.The results of parameter identification experiments show that the BPNN-PSO can identify the error of the industrial robot's parameters rapidly.Compared to the other two methods,the method based on BPNN-PSO has the ability of strong fault tolerance,which is suitable for processing practical data with noise.Error compensation experiments are carried out with the results of the identifications based on the three algorithms.The experimental results reveal that the parameter identification based on BPNN-PSO not only is insensitive to outside noise,but also has a stronger global search capability than the other two algorithms.After calibration,the positioning error of the industrial robot is reduced from the 0.643 mm to 0.327 mm,which means that the absolute positioning accuracy is enhanced by 49%.