Robot calibrate error compensation method based on fuzzy nerve network

Abstract

The invention discloses a robot calibrate error compensation method based on a fuzzy nerve network; the method comprises the following steps: compiling a homogeneous transformation matrix between adjacent link rods; calculating a kinetics equation and an error equation general formula of a robot end performer according to the transformation matrix; generating a kinetics equation and an error equation of an angle [theta i] according to a geometry parameter nominal value, the kinetics equation and the error equation general formula; compensating a first error compensation for the robot according to the kinetics equation and error equation; applying a fuzzy nerve network model to carry out a second error compensation for the robot; the method can enable the robot error compensation model to be faster, and more accurate with strong robustness.

Description

technical field [0001] The invention belongs to the field of mechanical processing and manufacturing, and mainly relates to a robot calibration error compensation method based on a fuzzy neural network. Background technique [0002] Nowadays, the robots produced by robot manufacturers have relatively high repeat positioning accuracy, but low absolute positioning accuracy. As the off-line programming technology of robots becomes more and more widespread, improving the absolute positioning accuracy of robots has become one of the key technical issues. The calibration of industrial robots is generally divided into four steps: modeling, measurement, parameter identification and compensation, among which compensation is the last step of calibration. Due to the positioning error, installation error and working error (transmission, deformation and other errors) of the robot in the manufacturing and installation process, the absolute positioning accuracy of the working robot is low...

AI Technical Summary

Problems solved by technology

Due to the positioning error, installation error and working error (transmission, deformation and other errors) of the robot in the manufacturing and installation process, the absolute positioning accuracy of the working robot is low, and the error needs to be compensated. Usually, the inverse kinematics problem is used to solve the problem. The joint angle is compensated, and the Newton-Raphson method is generally used to solve the inverse compensation problem, but this method has the following two disadvantages: the geometric error must be small, otherwise the algorithm convergence cannot be guaranteed; the calculation amount is large, and it is not easy to compensate online in real time

The application number is 201110113246.6, which discloses a method for compensating the accuracy of spatial three-dimensional grids for industrial robots. This method utilizes the characteristics of industrial robots with high repeatability positioning accuracy, and uses laser trackers to establish the distance between theoretical coordinates and actual positioning coordinates. For any point in a cubic grid divided in the envelope space, the theoretical coordinates of the robot are corrected by using the space interpolation method to complete the absolute positioning accuracy compensation of the robot at this point. The disadvantage of this method The reason is that the coordinates of most of the points in the process of robot movement are obtained by interpolation of a few accurate point coordinates obtained through measurement. Due to the nonlinear characteristics of the robot error, the interpolation points cannot meet the accuracy of the offline programming of the robot.

The document titled "BP Neural Network Compensation for Parallel Robot Positioning Error" ("Optical Precision Engineering", 2008, Vol. 16, No. 5, pp. 878-883), uses neural network technology to calculate the terminal pose of a 6-DOF precision parallel robot The error is compensated, but the neural network has no classification ability, and the newly added samples will affect the successfully learned network, so the number of features required to characterize each input sample must also be the same

Images