Multi-party Agent Relation Sampling for Multi-party Ad Hoc Teamwork

TL;DR

This paper introduces MARs, a relational modeling approach for multi-party ad hoc teamwork, enabling controlled agents to coordinate with multiple unfamiliar groups more effectively than existing methods.

Contribution

The paper proposes MARs, a novel relational modeling technique that captures cross-group dynamics in multi-party ad hoc teamwork scenarios, improving coordination and convergence speed.

Findings

MARs outperforms MARL and AHT baselines in experiments.

MARs converges faster than existing methods.

MARs effectively captures cross-group dynamics.

Abstract

Multi-agent reinforcement learning (MARl) has achieved strong results in cooperative tasks but typically assumes fixed, fully controlled teams. Ad hoc teamwork (AHT) relaxes this by allowing collaboration with unknown partners, yet existing variants still presume shared conventions. We introduce Multil-party Ad Hoc Teamwork (MAHT), where controlled agents must coordinate with multiple mutually unfamiliar groups of uncontrolled teammates. To address this, we propose MARs, which builds a sparse skeleton graph and applies relational modeling to capture cross-group dvnamics. Experiments on MPE and starCralt ll show that MARs outperforms MARL and AHT baselines while converging faster.