IDENTIFICATION OF THE DYNAMICS OF THE CONTROL OBJECT THE SYMBOLIC REGRESSION METHOD

Abstract

Background. This paper addresses a pressing issue in autonomous robotics. It focuses on solving a navigation problem using a highly accurate internal mathematical model. This model can be constructed using modern computational methods and installed on the robot's onboard processor during full-scale testing. Materials and methods. The paper considers the problem of a robot determining its location in the absence of global positioning systems. A new method for describing symbolic regression is proposed. This method, like artificial neural networks, allows for the approximation of any complex function. However, unlike neural networks, symbolic regression provides the researcher with a mathematical expression for the desired function. An example of identifying a quadcopter model based on experimental data obtained during the object's movement along two different complex closed trajectories, one circular, the other a figure-eight, is presented. Results. Using the internal model, it is possible to both assess the accuracy of incoming information and refine the model itself. Conclusions. Quadcopter trajectories along circular and octal trajectories are considered. The results obtained are tested on a real object.