Method for compensating positioning errors of robot based on deep neural network

Abstract

The invention discloses a method for compensating positioning errors of a robot based on a deep neural network. The method comprises the following steps: taking theoretical coordinate data of each sampling point and corresponding actual positioning errors as one group of sample data; constructing a robot positioning error prediction model, and training and testing the robot positioning error prediction model by adopting the sample data, wherein the robot positioning error prediction model is used for predicting positioning errors of an inputted target point to obtain a positioning error prediction value; and inputting theoretical pose data to the robot positioning error prediction model to obtain prediction errors and compensating and correcting target point coordinates. According to the method disclosed by the invention, the influence of the position and the pose of a positioning point on the positioning errors is comprehensively considered; and a complex mapping relation between theoretical pose and actual positioning errors of a robot is expressed by using the deep neural network to predict the positioning errors of the target point and compensate the errors. By use of the method disclosed by the invention, absolute positioning precision of the robot can be remarkably improved.

Description

technical field [0001] The invention relates to the technical field of robot positioning error compensation, in particular to a robot positioning error compensation method based on a deep neural network. Background technique [0002] In recent years, with the development and deepening of robot technology, robots have begun to be used in high-precision manufacturing fields such as aircraft assembly, flexible grinding, and laser cutting. However, under normal circumstances, the repetitive positioning accuracy of the robot is high, which can reach ±0.1mm, while the absolute positioning accuracy is only ±1-2mm, which is difficult to meet the technical requirements of high precision and high efficiency in fields such as aircraft assembly. Therefore, it is urgent to study an accuracy compensation method to improve the absolute positioning accuracy of the robot. [0003] The accuracy compensation method, that is, the error compensation method, uses artificially generated errors to...

AI Technical Summary

Problems solved by technology

The patent of publication number CN104535027A realizes the absolute positioning accuracy compensation of the robot at the pose point by identifying the actual kinematic parameter errors in different spaces of the robot, and then performing kinematic inverse solution to solve the actual pose of the robot. This method requires the establishment of complex motion The geometric error model is used, and only the geometric error is considered, which limits the accuracy compensation effect

The patent of Publication No. CN102230783A divides the grid in the robot space, performs spatial interpolation on the actual positioning error of the eight vertices of the cube containing the target point to obtain the actual positioning error of the target point, and realizes positioning error compensation after correcting the target point. The compensation of this method The effect is affected by the size of the meshing step size, and does not take into account the influence of the attitude on the error

The existing robot error compensation methods generally have the following problems: (1) The complex kinematic error model needs to be established, and the calculation process is cumbersome; (2) The model only considers the position factor, and does not consider the impact of positioning attitude on positioning error. influence etc.

Images