Neural Network Synthesis Algorithms of Adaptive Position-Trajectory Control Systems of Moving Objects (In the Case of Multi-Link Manipulators)

This article is devoted to neural network synthesis algorithms for adaptive position-trajectory control systems of moving objects. The dynamics and tracking error of a multi-link industrial robot manipulator and the boundary conditions of its dynamic characteristics are presented, and the mathematical basis for determining the tracking error and filtered tracking errors of the control system of a multi-link industrial robot manipulator is proposed. Based on these mathematical models, a structural scheme of the adjuster with an adaptive control system based on the approximation of the differences between the current and specified coordinates of the manipulator handle is presented, which serves to develop neural network models taking into account the boundary conditions of the specified coordinates of the manipulator handle. Neural network models are minimized by periodically observing the movement of the robot links and synthesizing their current state deviations and their velocity errors between set states. In the control according to boundary conditions, the distributed neural network provides reliability and accuracy of movement by adjusting the velocities and accelerations at the specified coordinate trajectory and position of the manipulator based on synthesis models, which allows to increase the stability of the system and the efficiency of monitoring. Also, by modeling the control system using computer technology, the relationship between the dynamics and links of the manipulator with the kinetic and potential energy along the specified coordinate trajectory and position, and the graphical characteristics of its determination and synthesis as a holistic system are presented.