Investigation of DQN Algorithm for Driving Control for Smart Cities and Traffic Safety

Abstract

With the rapid increase in the world population day by day, the migration of people from rural areas to cities is also increasing rapidly. This rapid population growth in cities brings with it many problems such as transportation, accommodation and education that need to be solved. Traffic and transportation problems in cities are at the forefront of these problems. Due to the increasing traffic day by day, the time people spend in traffic is increasing, and depending on the traffic density, there can be great variability in the acceleration and deceleration of the drivers. In addition, studies have shown that accidents due to excessive acceleration and deceleration are more common at times when traffic density is low. The fact that vehicle drivers can drive comfortably and safely in traffic can bring many advantages. Many parameters, from driving safety to fuel consumption, from environmental pollution to comfort, can be provided by driving control. The main goal of this work is to create an action strategy that maximizes the overall reward in the long run. In this study, an optimum speed has been tried to be suggested for vehicles in urban traffic by using deep reinforcement learning. © 2023 IEEE.