Safe Navigation Based on Deep Q-Network Algorithm Using an Improved Control Architecture

In the last decades, the application of deep reinforcement learning algorithms in autonomous vehicles (AVs) has become a popular research topic due to their effectiveness of controlling the cars based on current states of the environment. In this paper we trained autonomous vehicle to make decisions regarding navigation safely in the traffic using deep reinforcement learning technique and Carla simulator. the study presents the integration of depth, segmentation camera and collision sensor to allow the vehicle discover the environment and reach its destination without collide with the road users. The model is tested on 3 different paths in present of other road users in Carla simulator. Our results show that the proposed technique achieves better performance in terms of the estimation of distance and heading angel over time, and success rate.