Road Traffic: Deep Q-learning Agent Control Traffic lights in the intersection

In recent decades, road traffic has increased in line with the attractiveness of cities. As a result, motorists are increasingly faced with traffic jams, which have many consequences. To find solutions to this problem, it is necessary to understand the origin of congestion. this is why this document proposes a strategy for managing intersections by controlling traffic signals at an intersection aimed at reducing the rate of congestion where we present a deep reinforcement learning (Deep RL) model that is a development and implementation work of the deep Q-learning algorithm that manages an agent in a simulated traffic environment using the SUMO (Simulation of Urban Mobility) traffic Road Simulator.

I. Introduction

The development and evaluation of reinforcement learning techniques in real-world problems are far from trivial. One such task is to simulate the dynamics of an environment and the behavioral interactions of an agent [1]. Nevertheless, reinforcement learning has been applied successfully and has given promising results in several areas, such as traffic, networks, intelligence, robotics, and games, among others [2]. The issue of traffic is especially critical. Even in small communities, it is well recognized as one of the difficulties we encounter daily. As a result, the artificial intelligence community has paid special attention to traffic. Because of its distributed and autonomous nature, traffic has proven to be an intriguing testbed for reinforcement learning systems [3].