Problem
Research questions and friction points this paper is trying to address.
Multi-Robot Systems
Adaptive Collaboration
Unseen Partners
Heterogeneous Multi-agent Zero-Shot Coordination by Coevolution
🤖 AI Summary
This work addresses heterogeneous multi-agent zero-shot coordination (ZSC): enabling agents to collaboratively accomplish diverse tasks with previously unseen partner types—without any prior cooperative experience. We formally define this problem for the first time and propose a general framework based on dual-population coevolution, comprising three stages—pairing, updating, and dynamic selection—that integrate coevolutionary algorithms, multi-agent reinforcement learning, and a heterogeneity-aware partner sampling strategy. Empirical evaluation across multiple heterogeneous task domains demonstrates substantial improvements over existing homogeneous ZSC methods, validating the necessity of explicit heterogeneity modeling. The proposed framework exhibits strong generalization across unseen tasks and partners, high task adaptability, and robustness to partner diversity—establishing a novel paradigm for open-world multi-robot collaboration.
📝 Abstract
Generating agents that can achieve zero-shot coordination (ZSC) with unseen partners is a new challenge in cooperative multi-agent reinforcement learning (MARL). Recently, some studies have made progress in ZSC by exposing the agents to diverse partners during the training process. They usually involve self-play when training the partners, implicitly assuming that the tasks are homogeneous. However, many real-world tasks are heterogeneous, and hence previous methods may be inefficient. In this paper, we study the heterogeneous ZSC problem for the first time and propose a general method based on coevolution, which coevolves two populations of agents and partners through three sub-processes: pairing, updating and selection. Experimental results on various heterogeneous tasks highlight the necessity of considering the heterogeneous setting and demonstrate that our proposed method is a promising solution for heterogeneous ZSC tasks.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Co-evolutionary Method
Multi-robot Collaboration
Adaptability in Dynamic Environments
🔎 Similar Papers
No similar papers found.
