Bayesian Cooperative LOS/NLOS Classification With Domain Insights and Model Refinement for UAV Communication

Effective classification of Line-of-Sight (LOS) and Non-Line-of-Sight (NLOS) conditions is essential for optimizing communication performance in UAV-assisted networks, where signal quality, transmission reliability, and network efficiency are directly influenced by propagation conditions. This paper presents a novel Bayesian Cooperative Classification Model that integrates domain-specific insights and Bayesian optimization to achieve real-time LOS/NLOS differentiation with a classification accuracy of 94%. Leveraging Channel Quality Indicators (CQI), Downlink Modulation and Coding Scheme (DL MCS), and bit rates, our approach combines DKA with MBA to enhance feature representation and refine loss functions. By embedding domain-critical dependencies and introducing Bayesian-driven loss regularization, our model offers a robust and adaptive solution for varying UAV communication environments. Extensive experiments demonstrate the superiority of our method over traditional approaches, highlighting its potential to enhance UAV communication reliability and operational adaptability across diverse and dynamic conditions. The findings underscore the strategic importance of LOS/NLOS classification in ensuring optimal performance for next-generation UAV-assisted communication networks.