Ego-centric Learning of Communicative World Models for Autonomous Driving

2025-06-09

AI Summary

Multi-agent reinforcement learning (MARL) for autonomous driving faces challenges of partial observability and environmental non-stationarity; existing information-sharing approaches suffer from high communication overhead and poor scalability. Method: This paper proposes CALL, a lightweight communication framework grounded in a generative world model. Its core innovation lies in leveraging the world model’s latent-space representation to encode and share low-dimensional ego-vehicle states and intentions—enabling vehicle-centric, decentralized learning. Contribution/Results: We theoretically characterize the predictive accuracy gain from intention sharing and its impact on narrowing the policy performance gap. Evaluated on local trajectory planning in CARLA, CALL significantly improves prediction accuracy and planning robustness while reducing communication overhead by over 70%. It scales to collaborative scenarios involving hundreds of vehicles, demonstrating both efficiency and scalability.

Technology Category

• Multiagent Systems: Agent Communication
• Planning, Routing, and Scheduling: Planning with Language Models
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

We study multi-agent reinforcement learning (MARL) for tasks in complex high-dimensional environments, such as autonomous driving. MARL is known to suffer from the extit{partial observability} and extit{non-stationarity} issues. To tackle these challenges, information sharing is often employed, which however faces major hurdles in practice, including overwhelming communication overhead and scalability concerns. By making use of generative AI embodied in world model together with its latent representation, we develop {it CALL}, underline{C}ommunicunderline{a}tive Worunderline{l}d Modeunderline{l}, for MARL, where 1) each agent first learns its world model that encodes its state and intention into low-dimensional latent representation with smaller memory footprint, which can be shared with other agents of interest via lightweight communication; and 2) each agent carries out ego-centric learning while exploiting lightweight information sharing to enrich her world model, and then exploits its generalization capacity to improve prediction for better planning. We characterize the gain on the prediction accuracy from the information sharing and its impact on performance gap. Extensive experiments are carried out on the challenging local trajectory planning tasks in the CARLA platform to demonstrate the performance gains of using extit{CALL}.

Problem

• MARL suffers partial observability non-stationarity autonomous driving
• Information sharing faces communication overhead scalability issues
• CALL enables lightweight communication ego-centric learning world models

Innovation

• Uses generative AI for world model learning
• Encodes states into low-dimensional latent representations
• Lightweight communication for information sharing