Robust and Adaptive UAVs-Based Localization Without Predefined NLoS Error Models

In UAV-based localization systems utilizing Time of Arrival (ToA) measurements, Non-Line-of-Sight (NLoS) conditions present a persistent challenge by introducing significant errors that degrade localization accuracy. Traditional techniques rely heavily on prior knowledge of NLoS error statistics or measurement noise characteristics. These dependencies make such methods computationally intensive and less adaptable to dynamic or large-scale scenarios. This paper presents a low-complexity localization algorithm that overcomes these limitations by eliminating the need for prior NLoS error statistics or path status information. The proposed approach dynamically identifies and excludes ToA measurements affected by severe NLoS errors while refining localization accuracy through iterative updates. A two-stage Robust Regression Algorithm (RRA) is employed, combined with an adaptive UAV selection strategy, ensuring both computational efficiency and precise positioning. Theoretical convergence analysis verifies the algorithm’s robustness in selecting reliable UAVs and estimating the accurate position of the target. Simulation results show the algorithm’s superior performance compared to state-of-the-art methods, achieving higher accuracy and efficiency even under severe NLoS conditions. The proposed method’s adaptability, scalability, and robustness make it a valuable solution for accurate localization in complex and dynamic environments, including 5G ultra-dense networks and UAV-based deployments.