Kinematic Model-Based Lateral Trajectory Tracking Control Design and Experimental Validation for Unmanned Autonomous Cars

As intelligent transportation systems evolve, the Unmanned Autonomous Cars (UAC) has become a crucial focal point in research aimed at achieving efficient and safe adaptive driving within complex traffic environments. To ensure the precision and effectiveness of control, this paper presents an adaptive trajectory tracking control strategy based on the Deep Deterministic Policy Gradient (DDPG) algorithm. Initially, a kinematic model for the monorail UAC is established, providing the foundation for motion control. By carefully designing the network structure of the control framework and introducing self-attenuating Ornstein-Uhlenbeck (OU) noise, the controller's initial exploration and subsequent stability are enhanced. Using the Python Gym library and PyTorch library, three simulation scenarios of varying complexities are created. A reward function is thoroughly analyzed and designed to determine the optimal discount rate that balances immediate and long-term returns, thereby verifying the effectiveness of the DDPG-controlled lateral trajectory tracking for UACs. Finally, through the Robot Operating System (ROS)-based UAC experimental platform, the accuracy, robustness, and effectiveness of the designed controller in the real-world environment are demonstrated.