Data-Driven Digital Direct Position Servo Control by Neural Network With Implicit Optimal Control Law Learned From Discrete Optimal Position Tracking Data

Abstract

To get better control performance in motor control, more and more researches tend to apply non-linear control laws in the field of motor control. However, most conventional non-linear control theory is based on explicit model of controlled object and often resulting in complexity. Besides, the control parameters tuning is mainly aiming at stability of the system. No valid direct performance-oriented nonlinear control theory has been proposed. Facing the limitations, this paper presents a direct motor position control in an implicit data-driven manner. Unlike conventional non-linear motor controls that are based on explicit models and with stability-based parameters tuning, this study gives performance-oriented non-linear control by mastering non-linear discrete optimal control law in an implicit data-learning manner. Firstly, optimal data of position tracking problem is obtained by solving optimization problem. Secondly, the implicit discrete optimal control law hidden in data is learned by a BP neural network. Finally, the learned control law is implemented in real-time control to reproduce optimal control performance. Simulation and experiment results validated the feasibility of the data-driven controller, which could be helpful for performance-oriented non-linear control designs. The merits and further improvements are also discussed.INDEX TERMS Position control, implicit discrete optimal control, artificial neural network, motor, data learning.

Neural Network Predictive Control for Improved Reliability of Grid-Tied DFIG-Based Wind Energy System under the Three-Phase Fault Condition

“…The output layer of neurons executes related controls, and in this example, the kth output ok is as follows [66]: In this example, we use a backpropagation algorithm to showcase the feedforward ANN. Figure 4 shows the BP ANN's basic working schematic diagram [65]. The BP ANN configuration comprises input, hidden, and output layers.…”

A Physics-Guided Data-Driven Feedforward Tracking Controller for Systems With Unmodeled Dynamics—Applied to 3D Printing

“…To handle both nonrepeating trajectories and the unmodeled dynamics with unknown or complex structure, neural networks (NN) have also been incorporated into FF or other controllers, due to their loose structure and ability to approximate any mathematical function [24], [25], [26], [27], [28], [29], [30]. However, their high dimensional and nonlinear nature typically make them difficult to be trained online or their stability to be rigorously analyzed to ensure safe learning and control.…”

“…However, their high dimensional and nonlinear nature typically make them difficult to be trained online or their stability to be rigorously analyzed to ensure safe learning and control. As a result, most existing works on NN-based controllers either ignore the stability of the NN portion [24], [25], [26], [27], [28] or analyze its stability under very strict conditions that are not validated experimentally [29], [30].…”

Comparative Analysis of Optimal Control Strategies: LQR, PID, and Sliding Mode Control for DC Motor Position Performance

“…This approach is widely acknowledged for systematically designing controllers to optimize performance according to a specified index. In the literature, there are studies in the field of optimal control of DC motors using various artificial neural networks (Khomenko et al, 2013;Wang et al, 2019) or metaheuristic algorithms (Mamta & Singh, 2020;Rasheed, 2020). However, despite yielding successful results, these algorithms come with a high computational burden.…”