Dynamic Reconfigurable Aerial Governance and Optimization Network for UAV–RIS-Empowered 6G IoT Communications

This study is motivated by the growing need for resilient and energy-aware Unmanned Aerial Vehicles-Reconfigurable Intelligent Surfaces (UAV-RIS) systems in 6G IoT environments where dynamic obstacles and mobility hinder connectivity. The framework uniquely integrates hierarchical deep reinforcement learning (DRL), MetaBeam for adaptive beamforming, and federated DRL for decentralized optimization. This paper presents Dynamic Reconfigurable Aerial Governance and Optimization Network (DRAGON), a novel 6G framework that integrates Unmanned Aerial Vehicles (UAVs) and Reconfigurable Intelligent Surfaces (RIS) for enhanced IoT connectivity. Using a hierarchical DRL approach, DRAGON dynamically optimizes UAV trajectories and RIS configurations to adapt to changing network conditions. Simulations show 37% higher energy efficiency, 52% improved data rates, and 28% more IoT device admissions compared to Proximal Policy Optimization - Probabilistic (P-PPO), Centralized Deep Deterministic Policy Gradient (C-DDPG), and Multi-Policy Training (MPT). DRAGON supports RIS densities of 64-256, UAV fleets of 4 to 16 drones, and sub-second deployment latency, proving its scalability and effectiveness in dense, dynamic 6G environments.