The Effect of Linear Approximation and Gaussian Noise Assumption in Multi-Sensor Positioning through Experimental Evaluation

Abstract

Assumptions of Gaussianity in describing the errors of ranging data and linearization of the measurement models are well-accepted techniques for wireless tracking multi-sensor fusion. The main contribution of this paper is the empirical study on the effect of these assumptions on positioning accuracy. A local positioning system (LPS) was set up and raw data were collected using both the global satellite navigation system (GNSS) and the LPS. These data were fused to estimate position using both an extended Kalman filter (EKF) and a particle filter (PF). For these data, it was shown that the PF had an improvement in accuracy over the EKF of 67 cm (72%) with achieved accuracy of 26 cm. This imp..