Evaluating the Robustness of Deep Reinforcement Learning for Autonomous and Adversarial Policies in a Multi-agent Urban Driving Environment

Deep reinforcement learning is actively used for training autonomous and adversarial car policies in a simulated driving environment. Due to the large availability of various reinforcement learning algorithms and the lack of their systematic comparison across different driving scenarios, we are unsure of which ones are more effective for training and testing autonomous car software in single-agent as well as multi-agent driving environments. A benchmarking framework for the comparison of deep reinforcement learning in a vision-based autonomous driving will open up the possibilities for training better autonomous car driving policies. Furthermore, autonomous cars trained on deep reinforcement learning-based algorithms are known for being vulnerable to adversarial attacks. To guard against adversarial attacks, we can train autonomous cars on adversarial driving policies. However, we lack the knowledge of which deep reinforcement learning algorithms would act as good adversarial agents able to effectively test autonomous cars. To address these challenges, we provide an open and reusable benchmarking framework for systematic evaluation and comparative analysis of deep reinforcement learning algorithms for autonomous and adversarial driving in a single- and multi-agent environment. Using the framework, we perform a comparative study of five discrete and two continuous action space deep reinforcement learning algorithms. We run the experiments in a vision-only high fidelity urban driving simulated environments. The results indicate that only some of the deep reinforcement learning algorithms perform consistently better across single and multi-agent scenarios when trained in a multi-agent-only setting.