Self-Driving Car Simulation Using Reinforcement Learning for Driving on Roads with Potholes

We proposed the application of reinforcement learning (RL) techniques to simulate autonomous driving on pothole-laden roads. Our goal is to develop an RL-based framework that enables autonomous vehicles to learn robust driving behaviors, effectively perceive road conditions, and make decisions to avoid potholes while maintaining safe, efficient, and comfortable driving. The results of the experiments show that the agent could adjust its speed in a human-like manner and attempt to avoid potholes as much as possible by selecting routes with smoother surfaces efficiently. In future research, we can apply the results of these experiments to other types of roads and enhance the reliability of autonomous driving by using accurate road data in training and testing.