Zhou, Jingyuan (National University of Singapore), WANG, ZHICHENG (National University of Singapore), Yang, Kaidi (National University of Singapore)

Mixed-Autonomy Platoon Control As Sequence Modeling: A Fairness-Aware Multi-Agent Transformer Approach

Scheduled for presentation during the Invited Session "S23c-Trustworthy AI for Traffic Sensing and Control" (TH-LA-T23), Thursday, November 20, 2025, 16:20−16:40, Coolangata 2

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on October 18, 2025

Abstract

It is widely recognized that the control of mixed-autonomy platoons comprising connected and automated vehicles (CAVs) and human-driven vehicles (HDVs) can enhance traffic flow. Among existing approaches, Multi-Agent Reinforcement Learning (MARL) emerges as a promising control strategy due to its ability to handle complex scenarios in real time. However, current research on MARL-based mixed-autonomy platoon control faces two key limitations. First, existing methods do not explicitly account for the problem structure of the platoons, which may impair cooperation among agents. Second, current approaches typically optimize a global reward without considering fairness among individual vehicles. To address these gaps, we propose a Fairness-Aware Multi-Agent Transformer (FMAT) for controlling mixed-autonomy platoons. FMAT formulates platoon control as a sequence modeling problem, modeling the platoon from head to tail in alignment with the natural direction of disturbance propagation. Furthermore, we incorporate fairness into reward shaping by penalizing reward disparities among agents, thereby promoting equitable control performance while maintaining overall efficiency. Simulation results demonstrate that the proposed strategy significantly enhances CAV cooperation and improves fairness among CAVs within the platoon.