Study Of GNSS Satellite Selection Algorithm Based On Support Vector Machine

If all visible satellites are involved in solving process of satellite positioning,the difficulty of the receiver design will increased,but the positioning results will not significantly improve.With the gradual improvement of the various GNSS systems,especially the BDS is about to run a global network,the number of visible satellites will be greatly increased during the same epoch,so selecting the appropriate visible satellites in position resolution has certain practical significance.The main contents are as follows:(1)GNSS satellite selection methods are outlined from theoretical basis,development status and evaluation methods,and considered for future GNSS multi-system integration applications,exploring an efficient and robust good intelligence satellite selection algorithm is very necessary;(2)GDOP are derived from the basic principles of satellite positioning,and different forms of GDOP values in multi-GNSS positioning system combination are given,and the relationship between terrestrial coordinate system and ECEF are given;(3)for the largest polyhedral volume satellite selection method,a method based on Hamilton route to solving the inner sphere polyhedral volume method is given,the stimulation results proved that the largest polyhedral volume satellite selection method is necessary in satellite selection process;for clustering satellite selection method,the stimulation results give the appear frequency of satellite which has extremum elevation angle in minimum GDOP satellite set and the significant of grouping the visible satellite based on elevation angle;(4)introduced the SVM's application in the field of GNSS satellite selection algorithm,the coverage of city in mainland are analyzed when GNSS system is complete in STK scenario,GNSS satellite selection algorithm based on SVM is analyzed with stimulation test of the GNSS satellite orbit data broadcasted by NORAD.Simulation results show that the method presented ensuring the small change of positioning accuracy,and real-time,robustness has been greatly improved compared with the conventional machine learning algorithms.