Font Size: 

This study presents the replication and extension of an Adaptive Cruise Control strategy originally developed using a classical PID architecture within a CarSim-Simulink cosimulation environment. The reconstructed control system was validated against multiple driving scenarios to ensure robust longitudinal vehicle behavior under real world traffic dynamics. Building upon this classical framework, a novel reinforcement learning-based approach using the Twin Delayed Deep Deterministic Policy Gradient algorithm was developed to adaptively tune controller gains in real time. The learned policy successfully replaced the PID controller and demonstrated generalization across nonlinear scenarios, including gear shifts, aerodynamic drag, and sudden lead-vehicle maneuvers. The integration of learning-based models reflects a shift toward more intelligent and adaptable autonomous driving systems, highlighting the potential of neural controllers to replace conventional rule-based logic in safety-critical applications.