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

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. 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 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 model the problem as a hierarchical multi-objective partially observable Markov decision process (H-MO-POMDP) and propose an LLM-driven hierarchical multi-rate control framework: a HAPS-level LLM for long-term planning and handover/load balancing, and local controllers combining slow-timescale LLM reasoning with faster RL-based UAV control. We develop a high-fidelity 3D simulation integrating gym-pybullet-drones with 3GPP RF/THz channel models; results show notable gains in transportation efficiency, throughput, and collision reduction compared to baselines.