Hierarchical LLM-Driven Control for HAPS-Assisted UAV Networks: Joint Optimization of Flight and Connectivity

2026-05-12

AI Summary

This work addresses the joint optimization of flight control and communication connectivity for multi-UAV systems operating in dynamic, partially observable space-air-ground integrated networks. To tackle this challenge, the authors propose a hierarchical multi-rate control framework grounded in large language models (LLMs). The framework formulates a hierarchical multi-objective partially observable Markov decision process, wherein high-altitude platform stations (HAPS) perform global long-term planning while UAVs leverage LLM-based spatial reasoning fused with reinforcement learning for rapid local responses. This study pioneers the integration of LLMs into hierarchical control architectures for space-air-ground UAV networks, enabling zero-shot generalization in high-dimensional spaces and coupled optimization of communication and control objectives. Experimental results demonstrate significant improvements over existing methods: a 14% increase in transportation efficiency, 25% higher communication throughput, and a 23% reduction in collision rate, collectively enhancing dynamic adaptability and handover stability.

Abstract

Uncrewed aerial vehicles (UAVs) are increasingly deployed in complex networked environments, yet the joint optimization of multi-UAV motion control and connectivity remains a fundamental challenge. In this paper, we study a multi-UAV system operating in an integrated terrestrial and non-terrestrial network (ITNTN) comprising terrestrial base stations and high-altitude platform stations (HAPS). We consider a three-dimensional (3D) aerial highway scenario where UAVs must adapt their motion to ensure collision avoidance, efficient traffic flow, and reliable communication under dynamic and partially observable conditions. We first model the problem as a hierarchical multi-objective partially observable Markov decision process (H-MO-POMDP), capturing the coupling between control and communication objectives. Based on this formulation, we propose a large language model (LLM)-driven hierarchical multi-rate control framework. At the global level, an LLM-based controller on the HAPS performs long-term planning for load balancing and handover decisions. At the local level, each UAV employs a hybrid controller that integrates a slow-timescale LLM for high-level spatial reasoning with a reinforcement learning agent for faster UAV-to-infrastructure (U2I) communication and motion control. We further develop a high-fidelity 3D simulation platform by integrating the gym-pybullet-drones environment with 3GPP-compliant RF/THz channel models. Numerical results demonstrate that the proposed framework significantly outperforms state-of-the-art baselines, achieving a 14% increase in transportation efficiency and a 25% improvement in telecommunication throughput. Additionally, it achieves a 23% reduction in physical collision rates, demonstrating strong handover stability and zero-shot generalization in dynamic scenarios.

Problem

• UAV networks
• joint optimization
• connectivity
• flight control
• HAPS

Innovation

• LLM-driven control
• HAPS-assisted UAV networks
• H-MO-POMDP
• multi-rate hierarchical framework
• 3D aerial highway