Reinforcement Learning-Based Adaptive Speed Control of Electric Drives for Autonomous Industrial Automation

Abstract

Current industrial automation, robotics and smart manufacturing systems heavily rely on electric drive systems because they must have a higher level of control over speed to operate reliably. The most common forms of control in traditional methods include the proportional -integral-derivative (PID) control and the vector control which are successful in controlling motor speed, but fail when subjected to nonlinear processes and variations in the characteristics of the system parameters, and the load conditions of the system which have varying elements and parameters. To overcome these constraints, this paper will suggest a reinforcement learning-based adaptive speed control system on the basis of a Deep Q-Network (DQN) algorithm on intelligent electric drive systems. The suggested solution combines a learning agent based on DQN and the control architecture of the electric drive to empower the independent decision-making process and ongoing optimization of control activities depending on the system states, including the speed error, the motor current, and the load changes. An electric drive system simulation model is created in order to assess the efficiency of the proposed controller in various conditions of operation. The critical performance indicators such as accuracy of speed tracking, set-tling time, admission, as well as dynamic response are measured and compared with the traditional methods of control. The findings of the simulation indicate that the proposed DQN -based controller can greatly enhance the speed tracking performance, and minimise overshoot and improve the system response under load disturbances. The results identify the promise of reinforcement-based learning control measures to enable smart, adaptive, and autonomous electric drive systems in innovative automation industrial systems of the future.