Icon Legend

This session is not in your schedule.

This session is in your schedule. Click again to remove it.

Back

12 Views

ICTD 2025

Artificial Intelligence (AI) in Transportation

POSTER RECEPTION (View list of posters here)

Multi-agent reinforcement learning for integrating traffic signal control and vehicle routing

Tuesday, June 10, 2025

4:45 PM – 6:00 PM MT

Primary Author(s)

Yiwei Wang, n/a

Ph.D Candidate
University of Hawaii at Manoa
HONOLULU, Hawaii, United States

Presenter(s)

Yiwei Wang, n/a

Ph.D Candidate
University of Hawaii at Manoa
HONOLULU, Hawaii, United States

Co-Author(s)

Hanyi Yang, Ph.D

Postdoctoral Research Assistant
University of Hawaii at Manoa
Honolulu, Hawaii, United States
CZ

Changjian Zhang, Ph.D

Postdoctoral Resesarch Associate
University of Hawaii
Honolulu, Hawaii, United States
Guohui Zhang, PhD

Professor
University of Hawaiʻi at Mānoa
Honolulu, Hawaii, United States

Submitter(s)

Yiwei Wang, n/a

Ph.D Candidate
University of Hawaii at Manoa
HONOLULU, Hawaii, United States

Abstract: Integrating traffic signal control with vehicle routing represents a promising strategy for mitigating urban traffic congestion. Within the realm of intelligent transportation systems, the role of autonomous vehicles (AVs) has emerged as a significant area of research. This paper explores the potential of multi-agent deep reinforcement learning to coordinate traffic signal control and vehicle routing, specifically in a fully autonomous vehicle environment, where the routing decisions of AVs can be effectively managed. Our framework utilizes dedicated agents: signal control agents that fine-tune traffic light timings at intersections and vehicle routing agents that adaptively select optimal routes for AVs in response to real-time traffic conditions. By enhancing the interaction and cooperation among these agents, we create a collaborative learning environment that boosts overall network efficiency. A decentralized reinforcement learning scheme is leveraged to effectively manage the complexities of this multi-agent system. We evaluate our approach by incorporating the SUMO-Gym framework, which integrates traffic environments from the SUMO (Simulation of Urban MObility) simulator into the Petting-Zoo platform. This integration allows us to assess our method across various scenarios, comprehensively analyzing its performance in different traffic conditions. The comprehensive experiments demonstrate that our method reduces congestion and significantly outperforms conventional methods that treat traffic signal control and vehicle routing separately in practice.

Learning Objectives:

Attendees can expect to learn the following from this session:

study traffic management knowledge
develop a multi-agent reinforcement learning model
propose a joint optimation method for integrating traffic signal control and vehicle routing