# Meta-path guided policy distillation for resilient coordination in autonomous unmanned swarm

## TL;DR

This paper introduces a new framework for training resilient coordination policies in autonomous drone swarms using meta-path guided reinforcement learning.

## Contribution

The novel framework combines path-specific graph attention with policy distillation to better model complex dependencies in swarm coordination.

## Key findings

- MPGPD-RC outperforms existing methods in handling structured disruptions in autonomous swarms.
- The framework effectively models high-order dependencies through meta-path guided embeddings and contrastive learning.

## Abstract

Enhancing the resilience of Autonomous Unmanned Swarms (AUS) requires policies that remain effective under severe, structured disruptions while respecting the heterogeneous semantics of inter–subsystem interactions. Existing reinforcement learning (RL) approaches typically aggregate first–order neighborhoods in a path–agnostic manner, thereby blurring typed, ordered, and directed multi–hop dependencies encoded by domain meta–paths. We propose MPGPD-RC, a Meta- Path Guided Policy Distillation framework for Resilient Coordination that couples: (i) meta-path–guided embeddings learned by path-specific graph attention with contrastive reconstruction and attention fusion, and (ii) a teacher–student scheme in which a PPO teacher trained with a relaxed meta-path mask provides trajectories, and a student aligns both action distributions (KL) and trajectory-level structural codes via path-aware contrastive learning. Empirical evaluations validate that MPGPD-RC consistently surpasses state-of-the-art baselines across diverse perturbation scenarios by modeling complex, high-order dependencies that underpin resilient coordination.