IDEA: Insensitive to Dynamics Mismatch via Effect Alignment for Sim-to-Real Transfer in Multi-Agent Control

2026-06-24

AI Summary

This work addresses the high sensitivity of existing learning-based multi-agent control methods to dynamics mismatch during sim-to-real transfer, which hinders stable deployment in complex real-world environments. To overcome this limitation, the authors propose a semantics-level policy learning framework that is robust to dynamics discrepancies by abstracting actions into semantic commands through effect alignment. The approach integrates closed-loop control, stochastic environmental structures, and a novel action synchronization mechanism to effectively mitigate temporal inconsistencies among agents. Notably, this is the first method to jointly leverage effect alignment and action synchronization for multi-agent sim-to-real transfer. Evaluated across four navigation tasks, the proposed framework significantly outperforms state-of-the-art baselines, achieving higher training efficiency, substantially improved task success rates, and enhanced system stability in real-world deployments.

Abstract

Complex multi-agent control tasks remain challenging for traditional rule-based and model-based approaches, motivating the adoption of learning-based methods. However, learning-based methods often struggle with sim-to-real transfer because they rely on accurate dynamics modeling or system identification and learn policies in low-level control spaces that are highly sensitive to dynamics mismatch, making them costly and fragile in complex environments. To address this issue, we propose a sim-to-real method for multi-agent control, which is insensitive to dynamics mismatch via effect alignment. Our method combines random environmental structure with discrete semantic actions through closed-loop control, elevating policy learning to a semantic abstraction level. Additionally, we develop an action synchronization mechanism that mitigates inter-agent action timing mismatches, thereby enhancing the temporal consistency of the system. Experiments on four multi-agent navigation tasks demonstrate that our method substantially improves training efficiency over mainstream transfer methods and achieves higher success rates in real-world scenarios, thereby improving the robustness and deployment stability of multi-agent systems under dynamics mismatch.

Problem

• sim-to-real transfer
• dynamics mismatch
• multi-agent control
• policy learning
• robustness

Innovation

• effect alignment
• sim-to-real transfer
• semantic abstraction
• multi-agent control
• action synchronization