PAGNet: Pluggable Adaptive Generative Networks for Information Completion in Multi-Agent Communication

2025-02-06

AI Summary

In partially observable multi-agent cooperative tasks, inefficient policy learning arises from communication mechanisms that lack explicit information weighting and multi-message aggregation capabilities. To address this, we propose a plug-and-play adaptive generative communication framework that decouples communication modeling from policy learning. Our approach introduces, for the first time, a generative global state representation coupled with an explicit information-weighting mechanism, realized through a lightweight communication model integrating a variational autoencoder (VAE) and attention-based weighting. The framework employs a separately trained policy network and a dedicated communication adapter. Evaluated on standard benchmarks—including MPE and StarCraft II—our method achieves an average 18.7% improvement in win rate and reduces communication overhead by 42%, while significantly enhancing global state reconstruction fidelity and inter-agent cooperation consistency.

Technology Category

• Multiagent Systems: Agent Communication
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

For partially observable cooperative tasks, multi-agent systems must develop effective communication and understand the interplay among agents in order to achieve cooperative goals. However, existing multi-agent reinforcement learning (MARL) with communication methods lack evaluation metrics for information weights and information-level communication modeling. This causes agents to neglect the aggregation of multiple messages, thereby significantly reducing policy learning efficiency. In this paper, we propose pluggable adaptive generative networks (PAGNet), a novel framework that integrates generative models into MARL to enhance communication and decision-making. PAGNet enables agents to synthesize global states representations from weighted local observations and use these representations alongside learned communication weights for coordinated decision-making. This pluggable approach reduces the computational demands typically associated with the joint training of communication and policy networks. Extensive experimental evaluations across diverse benchmarks and communication scenarios demonstrate the significant performance improvements achieved by PAGNet. Furthermore, we analyze the emergent communication patterns and the quality of generated global states, providing insights into operational mechanisms.

Problem

• Enhance multi-agent communication efficiency
• Model information-level communication effectively
• Reduce computational demands in MARL

Innovation

• Pluggable adaptive generative networks
• Weighted local observations synthesis
• Reduced computational demands training