A Fuzzy Weighted Kalman Filter for GPS Positioning Precision Enhancement

Abstract

Kalman Filter (KF) is the predominant implemented method where continuity in state space of latent variables is presented. The filter applications are chiefly in the fields of control, navigation, signal processing and robotic motion planning. This work proposes a novel fast weighted KF based on fuzzy classifier. This approach improves the functionality of this digital filter in general. The presented extension of KF is analyzed and validated for accurate positioning of moving objects which is one of the main issues that KF is expected to address. Accordingly, raw Global Positioning System (GPS) data, as the predominant means of obtaining position, is utilized as input observations for this filter. In weighting phase, the fuzzy classifier segregates observations based on different features of their signal. Therefore, it alleviates effects of deteriorating factors such as power and quality loss and multipath error. The suggested approach has been vetted upon three distinctly different motion scenarios. Evaluating the positioning results indicated that the new method has raised precision up to 35 percent, compared to Recursive Least Square (RLS) and classic KF.