Optimization of classification thresholds for states of transionospheric radio links described by the normal distribution for ensuring the accuracy of uav positioning

Abstract

Nowadays the development of satellite navigation systems is largely integrated into modern society. The GPS (USA) and GLONASS (Russia) navigation systems are most commonly used. However, all the satellite navigation systems are negatively affected by the artificial irregularities arising in the ionosphere, which introduce the greatest error in the accuracy of an unmanned aerial vehicle (UAV) positioning. In such cases, radio link perturbations are defined by one of the random distributions: normal distribution, Rayleigh distribution, Rice distribution, Nakagami distribution. Existing methods of counteracting the perturbations cannot adequately suit the emerging practical needs since they are based on the predictive models. One of the promising ways to solve this problem is using of the methods of automated monitoring and control of satellite links state. The control and monitoring of such systems is determined by solving a problem of recognition and classification of an object, but such approach leads to inevitable occurrence of type I (false alarm) and type II (anomaly undetection) errors. In order to minimize the anomaly undetection errors when crossing classes, it is necessary to determine the classification threshold. In this case, it is possible to solve the problem mathematically using the available values of the system parameters. The solution to the problem is expressed with simultaneous equations which are put in a convenient form for solving by three quasi-Newtonian methods that allow to simplify the algorithm (Powell, Broyden, and Krylov-Newton). #CSOC1120