Compensation of Robot-Manipulator Positioning Errors in Work Space of Technological Equipment

Abstract

Industrial robots are widely used in modern enterprises. Although high repeatability of industrial robots can satisfy most process needs, their low absolute positioning accuracy in the workspace cannot meet the requirements of some high-precision tasks. An example, is that the action of a robot relies on real-time sample data rather than simulated data. The low absolute robot’s accuracy to positioning leads to the fact that work productivity is significantly different from what is expected. Therefore, the purpose of this paper is to analyze the developed program to increase the accuracy of robot-manipulator’s positioning in real time under the compilation with the programs of integration automation projects.
 The paper considers the deformation of the links, which is one of the robot manipulator’s production errors. The effect of applied forces on the connections and links can affect the deformation of the robot posture in different ways, and these changes are difficult to account for using linear algebra, so recently the demand for "offline programming" has increased, which will allow you to simulate the work of a robot manipulator with a change in time and will have a reasonable price. Therefore, the application of artificial neural networks based on optimization algorithms to compensate for absolute error was considered. After analyzing the existing technical solutions, a genetic algorithm was chosen that satisfies the conditions of ease of implementation, clarity, lack of redundant calculations, flexibility to the data types used. The paper also presents a general scheme of algorithms for optimizing the parameters of artificial neural networks and an iterative process of finding optimal values of hyperparameters using a genetic algorithm.
 After choosing the modelling method and optimization algorithm, the developed algorithm on an open data set was tested. The obtained results showed an increase in accuracy from 22 to 77% (positioning accuracy without the use of compensation methods). Therefore, this article describes a method of increasing the absolute accuracy of positioning under the condition of compilation with programs of integration automation projects. The result is a program based on the above method.