A custom reinforcement learning environment in Webots features an autonomous vehicle on a fixed track with randomized obstacles. SAC and PPO are used for training, with Uniform Domain Randomization to enhance generalization. Results show improved stability, efficiency, and robustness, supporting sim-to-real transfer.