A Fault-tolerant Filter for Robust GNSS Positioning Using a GM-Estimation Approach.

Effective global navigation satellite system (GNSS)-based positioning is an enabling issue to support many location-based applications of the Internet of things (IOT). In some specific scenarios, the fault tolerant capability of the positioning method is of great significance. In this paper, we present a robust nonlinear filtering algorithm for GNSS positioning with probabilistic measurement faults. In the proposed approach, the conventional measurement update process of a nonlinear cubature filter is modified to be a standard linear regression problem, and the generalized M-estimation technique is utilized to achieve fault-tolerant estimation with an online adaptive measurement covariance. Cook's distance is involved to determine the generalized equivalent weight, which is capable of revealing the negative effects on the estimation performance from the probabilistic sensor faults. Both the nonlinearity approximation ability and robustness to the faults of the proposed filter are noted by means of an adaptive strategy for the constraint coefficient. The advantages of this method lie in its simplicity and its insensitivity to fault conditions. The results from simulations demonstrate the improved performance of the proposed filter over the conventional cubature Kalman filter and the Huber M-estimation-based solutions.