← Back ICRA 2026

Event-Driven MARL for Collaborative Swarm Confrontation in Asynchronous Environments

Qizhen Wu, Lei Chen, Kexin Liu, Jinhu Lv

PDF

AI summary

Key figure (auto-extracted from paper)

Integrating event-driven task termination with Transformer-enhanced QMIX and a synchronized experience selection scheme enables robust collaborative strategy learning in asynchronous swarm confrontations.

Event-driven MARL Swarm confrontation Asynchronous learning Transformer networks Collaborative control Multi-agent reinforcement learning

Problem

Event-driven multi-agent reinforcement learning reduces decision jitter but disrupts agent coordination due to misaligned information sharing and inconsistent strategy updates across diverse timescales. This hinders effective collaboration in large-scale, dynamic swarm confrontation scenarios.

Approach

The method introduces an experience selection scheme to filter and synchronize joint information across asynchronous agents, paired with Transformer-based actor-critic networks to process complex historical confrontation data for coordinated decision-making.

Key results

Event-driven termination eliminates decision jitter
Experience selection synchronizes training across timescales
Transformer QMIX captures temporal correlations for coordination
Outperforms conventional MARL in large-scale confrontations

Why it matters

Provides a scalable coordination framework for robotic swarms operating in dynamic, asynchronous environments, advancing autonomous multi-robot systems.

Abstract

Multi-agent reinforcement learning (MARL) pro- vides a flexible solution for tackling task and motion planning challenges, particularly in swarm confrontation scenarios. By customizing termination conditions for diverse tasks, event- driven MARL reduces decision jitter caused by frequent task switching. However, it hinders robots from updating strategies on a consistent timescale, leading to misaligned information sharing that disrupts agent coordination. To address this, we propose a novel event-driven MARL approach that fa- cilitates collaborative strategy learning under asynchronous conditions. The approach introduces an experience selection scheme tailored to diverse timescales, ensuring efficient training through synchronized information sharing among robots. By incorporating Transformers, our method enables robots to infer others’ behaviors from historical data, optimizing collaborative strategies. Extensive experiments validate the effectiveness of our proposed approach.

Index terms

Task and Motion Planning Reinforcement Learning Swarm Robotics