Оптимізація початкових параметрів фільтра калмана для підвищення точності інерціальної навігації бпла

Abstract

This study addresses autonomous UAV navigation under GNSS denial caused by electronic warfare. In such environments, navigation relies on MEMS sensors and the Extended Kalman Filter (EKF), whose accuracy depends critically on correctly initialized noise covariance matrices. To overcome the inefficiencies of manual tuning, this research applies global stochastic optimization methods: Genetic Algorithms, Particle Swarm, and Bayesian Optimization. By minimizing RMSE or NIS, this automated multi-parameter tuning approach significantly reduces navigation errors, enhancing UAV autonomy and survivability