Reinforcement Learning-based Target Tracking Control for a Vehicle with Input Delay

In this article, a reinforcement learning-based controller is proposed to address the target tracking problem for vehicles with inherent delays in the drivetrain. To enhance the performance of this controller, we incorporate a compensatory signal in addition to the actual control inputs to better handle delays. We introduce a method for designing a reward function that enables the reinforcement learning-based controller to effectively tackle this problem. The controller is trained using the Soft Actor Critic algorithm by combining the developed simulation model. Finally, its performance and robustness are evaluated in simulation. Our controller demonstrates the capability to navigate from random initial points to target points and robustness to variations in delay parameters.