An Analysis of the Effects of Hyperparameters on the Performance of Simulated Autonomous Vehicles

Reinforcement learning (RL) is emerging as an effective technique to study autonomous vehicles (AVs) that are capable of navigating their surroundings safely and accurately. This is due to the fact that with RL, an agent can evaluate its surroundings and make appropriate decisions to maximize rewards without the need for human intervention. RL offers an alternative solution to complement Supervised learning solutions in the AV field and it offers some additional flexibility that makes it ideal to study the subject and test-focused solutions. To apply RL to AVs, we train the algorithm on a simulator, called AWS’s DeepRacer, first. The scope of this paper focuses on the hyperparameters of the algorithm and studies the performance of the model and how the hyperparameters affect it. As the need for autonomous vehicles increases to reduce traffic congestion and car crashes, it becomes significantly important to study the performance of AVs based on the hyperparameters of reinforcement learning.