Intelligent Multi-UAV Navigation in ITNTNs: A Hierarchical LLM Approach

2026-07-20

AI Summary

This work addresses the challenge of coordinating multiple high-speed unmanned aerial vehicles (UAVs) in highly dynamic integrated terrestrial–non-terrestrial networks (ITNTNs), where existing approaches struggle to simultaneously ensure real-time control and strategic adaptability amid complex flight dynamics and multi-tier network handovers. To bridge this gap, the authors propose a cloud–edge–device collaborative hierarchical agent architecture: a cloud-based large language model (LLM) deployed on high-altitude platforms performs global semantic-level task planning and load balancing; an onboard lightweight edge LLM translates local environmental perception into tactical subgoals; and a low-level deep reinforcement learning (DRL) controller generates collision-avoidant, seamless-handover-supporting flight trajectories. This architecture uniquely integrates the strategic reasoning of LLMs with the real-time control capabilities of DRL, enabling efficient coupling between semantic planning and physical execution. Simulations demonstrate that the proposed system significantly reduces collision rates and improves network throughput compared to state-of-the-art baselines.

Abstract

The deployment of high-speed Uncrewed Aerial Vehicles (UAVs) in 3D aerial highways necessitates robust coordination of physical flight kinematics and multi-tier network handovers. While Deep Reinforcement Learning (DRL) offers rapid tactical control, it lacks the zero-shot strategic reasoning required to quickly adapt to dynamic Integrated Terrestrial and Non-Terrestrial Networks (ITNTNs). Conversely, Large Language Models (LLMs) excel at semantic reasoning but suffer from high inference latency, rendering them unsuitable for real-time aerodynamic control. To bridge this gap, we propose a novel Hierarchical LLM-driven control framework. A massive cloud-based LLM deployed on a High-Altitude Platform Station (HAPS) manages slow-timescale global load balancing, while lightweight edge-LLMs on individual UAVs translate local observations into tactical sub-goals. These sub-goals guide a fast-timescale physical DRL controller to execute collision-free, handover-aware trajectories. Simulation results demonstrate that our agentic architecture significantly reduces collision rates and improves aggregate system throughput compared to existing baselines.

Problem

• Intelligent Multi-UAV Navigation
• Integrated Terrestrial and Non-Terrestrial Networks
• 3D aerial highways
• network handovers
• real-time aerodynamic control

Innovation

• Hierarchical LLM
• Multi-UAV Navigation
• Integrated Terrestrial and Non-Terrestrial Networks (ITNTNs)
• Deep Reinforcement Learning (DRL)
• Edge Intelligence