LLM-Driven Control for UAV Networks with HAPS

A new paper on arXiv proposes a hierarchical LLM-driven control framework for multi-UAV systems in integrated terrestrial and non-terrestrial networks (ITNTN) with HAPS. The study models the joint optimization of flight and connectivity as a hierarchical multi-objective partially observable Markov decision process (H-MO-POMDP). It addresses a 3D aerial highway scenario where UAVs must balance collision avoidance, traffic flow, and communication under dynamic conditions. The framework operates at a global level using large language models to coordinate decisions. The work aims to solve fundamental challenges in multi-UAV motion control and connectivity.