Unsupervised learning-based satellite selection algorithm for GPS–NavIC multi-constellation receivers

The lower bound of the GDOP is an important parameter to benchmark satellite selection algorithms. Existing GDOP lower bound formulations do not consider the satellite azimuth and elevation angle constraints with respect to the user for Geostationary Earth Orbit (GEO) and Inclined Geosynchronous Orbit (IGSO)-based regional navigation constellations. A GDOP lower bound formulation considering the azimuth and elevation angle constraints is formulated for GEO- and IGSO-based navigation constellations. Using numerical simulation, it is demonstrated that the GDOP lower bound for using the Navigation Indian Constellation (NavIC) is significantly higher than the Global Positioning System (GPS), whereas the existing GDOP lower bound formulation provides comparable GDOP lower bound for the GPS and NavIC. It also indicates that one or more navigation constellations should be used with the NavIC to achieve better position accuracy. In this context, an unsupervised learning-based satellite selection (ULiSeS) algorithm is also proposed and the effectiveness of the algorithm is demonstrated through numerical simulation for the GPS and the NavIC constellations. A meta-cognitive component is also introduced to enable the ULiSeS algorithm to decide when to learn and when to use the available model. The ULiSeS algorithm selects a better set of satellites than the Quasi-optimal selection algorithm and requires 89.12% less processing time than the fast satellite selection algorithm.