The advancement of autonomous driving technology has conditioned the need for robust and efficient automated systems to ensure safe navigation in different traffic situations. Despite the existence of models that incorporate a variety of risks, there is still a possibility of conditions that cause discrepancies, which could, in turn, produce hazardous situations. This research paper proposes an approach with multi-objective optimization problem that integrates a critical safety metric into a reinforcement learning (RL) algorithm to improve autonomous vehicle (AV) safety. The aim is to develop an adaptive AV decision-making framework that includes a control strategy enabling the agent to improve its driving performance through interaction with simulated urban environments. The primary focus is on the Advanced Longitudinal Control system, particularly in scenarios involving critical vehicle maneuvers, such as cut-in in dense traffic flow. The learning algorithm is implemented using MATLAB-Simulink® software, providing a robust simulation environment. This approach allows for an in-depth understanding of urban traffic challenges using high-fidelity simulations for the RL model's training.
Adaptive Decision-Making Framework for Autonomous Vehicles: A Reinforcement Learning Approach to Urban Traffic Safety
Proceedings in Automotive Engineering
International Congress of Automotive and Transport Engineering ; 2024 ; Brasov, Romania November 06, 2024 - November 08, 2024
CONAT 2024 International Congress of Automotive and Transport Engineering ; Kapitel : 12 ; 136-147
20.11.2024
12 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
ArXiv | 2025
|Heuristics‐oriented overtaking decision making for autonomous vehicles using reinforcement learning
Wiley | 2020
|