Analysis of Doppler Collision Prediction using Supervised Machine Learning

Abstract

Navigation with Indian constellation (NavIC) has been developed by Indian Space Research Organization (ISRO). NavIC satellite constellation contains four geosynchronous satellites and three geostationary satellites. There are many factors which affect the navigation system. Doppler Collision is one such factor which leads to tracking errors due geostationary satellites. It occurs between geostationary satellite combination IRNSS 1C-1G, 1C-1F, 1F-1G. When the relative Doppler frequency of satellites is less than the code tracking loop bandwidth Doppler Collision (DC) period is observed. The positional accuracy of NavIC system is affected by DC. Due to DC, the more effected geostationary satellite pair is 1C-1G. To mitigate the DC, the prediction of DC using machine learning algorithms will be very much useful for improving positional accuracy. The parameters considered for prediction are relative Doppler, satellite position, satellite velocity, duration of occurrence and relative Doppler. Three supervised machine learning algorithms such as Linear regression, Random Forest regressor and K-Nearest Neighbors (KNN) regressor are used for prediction. Among these three algorithms, random forest regressor predicted the Doppler Collision accurately.