MEMS INS/GPS Data Fusion using Particle Filter

摘要

Targeting low cost solution for land vehicles, Micro-Electro-Mechanical Systems (MEMS) based inertial sensors are used. These sensors suffer from complex stochastic error characteristics that are difficult to model. Kalman filter (KF) has limited capabilities in providing accurate estimation of such system parameters, because KF is restricted to use only Gaussian linear models for these sensors' stochastic errors. EKF makes linearization of non-linear model, and after solve problem optimally by KF. This first order linearization in EKF introduces additional errors and difficulty in estimating process [1,2]. It is becoming important to include elements of nonlinearity in order to model accurately the underlying dynamics of inertial system. To solve the problem of nonlinear filtering, the particle filter (PF) was proposed. It was first introduced by Gordon et al. (1993). PF exploits numerical representation techniques for approximating the filtering probability density function (PDF) of inherently nonlinear non-Gaussian systems. Using these methods, the obtained estimates can be set arbitrarily close to the optimal solution (in the Bayesian sense) at the expense of computational complexity [2, 3].