Collaborative Multi-Agent LLMs for Mixed-Autonomy Traffic - CoMAL Framework

Watch a 25-minute research presentation exploring the CoMAL (Collaborative Multi-Agent LLMs) framework and its innovative approach to managing mixed-autonomy traffic systems. Dive into how autonomous and human-driven vehicles can coexist through Large Language Models (LLMs) that enable high-level reasoning, collaboration, and real-time decision-making. Learn about the framework's sophisticated architecture where autonomous vehicles communicate, assign dynamic roles, and coordinate traffic strategies through modules handling perception, memory, collaboration, reasoning, and execution. Understand how LLMs generate human-readable traffic scenario descriptions for natural, interpretable decision-making while combining with rule-based models like the Intelligent Driver Model (IDM) for precise low-level control. Explore the advantages and limitations of this LLM-centric approach compared to reinforcement learning (RL) alternatives, and discover potential future directions for integrating multiple AI approaches to optimize collaborative traffic systems.