AeroSecure: Blockchain-federated Reinforcement Learning Framework for Autonomous Drone Swarm Coordination

Abstract

Autonomous aerial swarms are revolutionizing sectors such as defense, agriculture, and disaster response by enabling distributed and intelligent coordination. However, ensuring secure collaboration, real-time adaptability, and explainability among autonomous drones remains a significant challenge. This paper introduces AeroSecure, a novel Blockchain-Federated Reinforcement Learning (BFRL) framework for trustworthy drone swarm coordination. The proposed model employs Multi-Agent Reinforcement Learning (MARL) to optimize collective drone behaviors while preserving local autonomy. A permissioned blockchain ledger records mission data, model updates, and trust metrics, ensuring integrity and accountability. To enhance human trust and operational transparency, Explainable Reinforcement Learning (XRL) modules using SHAP and policy visualization provide interpretable insights into swarm decision logic. Simulation results on benchmark drone coordination tasks demonstrate that AeroSecure achieves superior performance in navigation efficiency, security, and interpretability compared to traditional centralized control systems. The framework paves the way for transparent, secure, and cooperative autonomous drone ecosystems.