Towards Mitigating Probable Road Mishaps through DQN Based Deep Reinforcement Learning

Road accident is a widespread problem in many urban areas nowadays. Reckless Driving (Over-Speed) is one of the attributes leading to the enormous destruction of pedestrian life. Autonomous vehicle speed control is an important research area in the Intelligent Transportation System (ITS) to mitigate road accidents in densely populated areas. Many existed works utilized Reinforcement Learning (RL) to control the vehicle speed intelligently. However, controlling the over-speed of the vehicle using RL under the speed limit of the different zones (i.e., Town, Rural Main Road, Highways, and so forth) is also another demanding task because of the speed limit variation. Therefore, we propose a Deep-Q-Network (DQN) based Deep Reinforcement Learning (Deep RL) method to overcome the over-speed problem in various speed limit areas. In this paper, the proposed system will reduce the vehicle speed intelligently if the driver passes over the speed limit in a particular zone. The simulation results show that our proposed solution can successfully reduce the vehicle speed under various zone-based speed limits.