Multi-Objective Deep Reinforcement Learning for Variable Speed Limit Control

• DOI: 10.1145/3651671.3651719
• Publication Date: 2024-06-07T22:55:50Z
• Authors: Asmae Rhanizar; Zineb El Akkaoui
• Keywords: Speed limit

This study proposes a novel approach to address the multi-objective challenge of traffic control on congested freeways using Deep Reinforcement Learning (DRL). The involves developing Learning-based Variable Speed Limit (VSL) agent specifically designed optimize efficiency while enhancing road safety. employs reward function, striking delicate balance between safety and mobility by optimizing speed limits minimize collision risks maximize flow simultaneously. To illustrate architecture Framework, Q-Networks (DQN) topology is presented. Through comprehensive simulation section A1 freeway in Morocco, our DRL-VSL Framework showcased significant improvements, with mean increasing 2.55% Time Collision values demonstrating 57.33% enhancement. These results highlight Framework's ability simultaneously improve real-world scenarios, contributing advancement intelligent systems through DRL.