Navigating Autonomous Vehicle at the Road Intersection Simulator with Reinforcement Learning

In this paper, we consider the problem of controlling an agent that simulates the behavior of an self-driving car when passing a road intersection together with other vehicles. We consider the case of using smart city systems, which allow the agent to get full information about what is happening at the intersection in the form of video frames from surveillance cameras. The paper proposes the implementation of a control system based on a trainable behavior generation module. The agent’s model is implemented using reinforcement learning (RL) methods. In our work, we analyze various RL methods (PPO, Rainbow, TD3), and variants of the computer vision subsystem of the agent. Also, we present our results of the best implementation of the agent when driving together with other participants in compliance with traffic rules.