International Journal of Science and Technology (IJST)

Enhanced Deep Q-Learning for 2D Self-Driving Cars: Implementation and Evaluation on a Custom Track Environment

This research presents the implementation of a Deep Q-Learning Network (DQN) for a self-driving car on a 2-dimensional (2D) track, aiming to enhance the performance of the DQN network. It covers the development of custom driving environment with pygame on the track around the University of Memphis map and design and implementation of the DQN model. The algorithm utilizes data from 7 sensors collected by sensors installed in the car, based on the distance between the car and the track. These sensors are positioned in front of the vehicle, spaced 20 degrees apart, enabling them to sense a wide area ahead. We successfully implemented DQN and also modified the DQN with priority-based action selection mechanism and referred to it as modified DQN. The model is trained on 1000 episodes and the average reward received by agent is found to be around 40, which is around 60% higher than the original DQN and around 50% higher than the vanilla neural network.