Reinforcement learning for autonomous vehicle navigation in Urban environments

Praggnya Kanungo *

Student, Computer Science, University of Virginia, USA.
 
Research Article
World Journal of Advanced Engineering Technology and Sciences, 2024, 11(01), 457-466.
Article DOI: 10.30574/wjaets.2024.11.1.0019
Publication history: 
Received on 06 December 2023; revised on 21 January 2024; accepted on 24 January 2024
 
Abstract: 
This paper presents a comprehensive study on the application of reinforcement learning (RL) techniques for autonomous vehicle navigation in complex urban environments. We propose a novel deep RL framework that combines state-of-the-art algorithms with realistic urban traffic simulations to train robust navigation policies. Our approach leverages a hierarchical learning structure to decompose the challenging urban driving task into more manageable sub-tasks. Extensive experiments in simulated urban scenarios demonstrate that our method significantly outperforms baseline approaches in terms of safety, efficiency, and adaptability to diverse traffic conditions. We also conduct real-world validation tests to verify the transferability of learned policies to actual autonomous vehicles. Our results highlight the potential of RL-based techniques to enable safe and efficient autonomous navigation in dynamic city environments.
 
Keywords: 
Reinforcement Learning; Autonomous Vehicles; Urban Navigation; Hierarchical Learning; Traffic Simulation; Deep Learning; Self-Driving Cars; Policy Optimization; Urban Environments; Safety Systems
 
Full text article in PDF: