Optimal base station selection for GNSS-denied UAV localization

Abstract Accurate and robust localization of UAV swarms in GNSS-denied urban environments remains a major challenge because signal blockage, multipath, and poor anchor geometry can severely degrade positioning performance. This paper extends our previous work on degree-centrality-based clustering for UAV swarm relative localization by introducing a new optimization layer for ground base-station deployment. While the earlier framework organized the swarm into non-overlapping clusters and used fixed base stations to support hierarchical localization through beacon drones, the present work focuses on direct optimization of the number and placement of pre-deployed base stations using a genetic algorithm (GA). The proposed formulation evaluates candidate base-station subsets through a multi-objective fitness function that combines geometric dilution of precision, line-of-sight (LoS) awareness, infrastructure cost, and vertical diversity constraints in a realistic urban canyon model. The GA is validated on a representative 50-UAV cluster in a GNSS-denied urban environment and is compared against greedy, random, and robust static baselines. Simulation results demonstrate that the proposed approach achieves a mean localization error of 1.08 m with a standard deviation of 0.18 m, outperforming all baselines and remaining reliable even under severe non-line-of-sight (NLoS) conditions. Furthermore, the theoretical error predictions match well with the Monte Carlo results, which validate the proposed formulation. The results show that the adaptive selection of base stations can significantly improve localization accuracy, robustness, and infrastructure efficiency for UAV swarms in complex urban scenarios.