Low-cost MEMS-INS/GNSS integration using quaternion-based nonlinear filtering methods for USV

Abstract

This paper addresses the problem of integration of Inertial Navigation System (INS) and Global Navigation Satellite System (GNSS) in the field of developing low-cost, robust and highly accurate navigation systems. With the implementation of the navigation system of an unmanned surface vehicle (USV) as background, a tightly-coupled integration filter based on quaternion has been proposed. The proposed filter avoids the singularity problems inherited in Euler angle-based approaches, and the quaternion vector involves less elements than DCM-based approaches. Due to the nonlinearity of process and observation models of the system, as well as the non-Gaussian distribution of stochastic measurement errors of MEMS (micro-electromechanical systems) sensors, a modified unscented particle filter (UPF) that combines the particle filter (PF) and unscented Kalman filter (UKF) has been presented. For nonlinear and non-Gaussian estimation problem, PF is extremely theoretically attractive to be used. However, PF is seldom to be seen in practical engineering application for its large computation burden. Considering the purpose of releasing the computation burden, meanwhile, guarantying the estimation accuracy, quaternion based UPF is a good choice to implement INS/GNSS integration for USV navigation system.