Integrating GPS and INS: comparing the Kalman estimator and particle estimator

Abstract

This report describes how a relatively new type of state estimator, called the particle estimator (PE), has been evaluated and compared with the well established method based on a linear Kalman estimator (KE). The estimators have been applied to the problem of integrating information from a Global Satellite Navigation System (GNSS) and an Inertial Navigation System (INS) expressed in the navigation frame. No attitude is included in this system, i.e. the vehicle is considered as a point moving in space. Therefore the INS consists only accelerometer signals. The two estimators are compared, essentially regarding their robustness against different types of unmodelled errors in the three acceleration measurements. The errors consist of different combinations of white noise components and constant components (biases). KE uses a continuous linear error model. The task for the KE is to estimate the errors of an Inertial Navigation System (INS) by using the difference between the GPS solution in velocity and position and the integrated INS velocity and position. PE uses a nonlinear full state discrete model. The GPS receiver position and velocity are used as measurements to the estimator. The PE does neither require white Gaussian noise distribution nor linear equations. The computational load for the PE increases with the complexity of the problem, e.g. the number of states.