Machine Learning Algorithms for FCB Estimation in PPP-AR Technique

Abstract

Precise Point Positioning (PPP) technique have shown continuous improvement regarding positioning. The recent developments in PPP-AR (Ambiguity Resolution) have facilitated the resolution of integer ambiguity. Thereby, the observation period needed for the convergence time, which is considered as a disadvantage of PPP, has been shortened. Furthermore, the integer ambiguity is not independent source of error but influenced by the Fractional Cycle Bias (FCB) products. This study aims to investigate the effects of the FCBs in PPP-AR technique, which reduces the convergence time. For FCB estimation, machine learning was applied by modifying the functional model of the Single Difference Between Satellite (SDBS) technique. The incorporation of these algorithms enables estimation of FCB values, even for relatively small values. It can be asserted that the support vector machine performs than both the random forest and the SDBS model regarding success. For the integer ambiguity solution PPP-AR demonstrates superior performance compared to PPP.