Integrated adaptive communication in multi-agent systems: Dynamic topology, frequency, and content optimization for efficient collaboration

Jianrui Wang, Yi Li, Yitian Hong, Yang Tang

Abstract: In multi-agent systems (MAS), effective communication is essential for coordination and achieving common goals, especially in complex environments. However, existing communication methods face significant challenges, such as high resource consumption and limited adaptability to dynamic environments. To address these, we propose the integrated adaptive communication network (IACN) built on multi-agent proximal policy optimization (MAPPO), which enhances communication efficiency and adaptability in MAS. IACN dynamically adjusts communication topology using a learnable graph and optimizes content based on task relevance. Additionally, it incorporates an adaptive frequency adjustment mechanism to balance communication demands based on task urgency and environmental changes. Experiments in multi-agent particle environments demonstrate that IACN significantly outperforms existing methods in terms of overall performance, coordination effectiveness, and adaptability.