ENABLING ACCURATE LOW COST POSITIONING IN DENIED GPS ENVIRONMENTS WITH NONLINEAR ERROR MODELS OF INERTIAL SYSTEMS

The complementary characteristics of GPS and inertial sensors motivate their integration for more reliable positioning information in challenging GPS environments. Last decade has witnessed an increasing trend of utilizing MEMS-grade inertial sensors in the integration due to their low cost. For this research, only one MEMS-grade gyroscope and the vehicle built-in odometer are used together with two accelerometers to build a reduced inertial sensor system (RISS). This system is integrated with GPS to provide a low cost 3D positioning solution. As a linear estimation technique, Kalman Filter (KF) is adequate for the data fusion of GPS and high-end inertial sensors. However, MEMS-grade inertial sensors suffer from severe sensor errors including non-stationary stochastic drifts and nonlinear inertial errors, which undermine the effectiveness of KF. To overcome the problem, Fast Orthogonal Search (FOS) algorithm is employed in this research to identify the higher order RISS errors. With a tolerance of arbitrary stochastic noise, FOS is able to build an accurate nonlinear model that predicts RISS errors. KF can then be augmented by FOS to estimate and reduce both linear and nonlinear inertial errors, thus enhancing the navigation performance. The road test trajectory is conducted to examine the proposed method. The results demonstrate the advantages of the proposed method over the stand alone KF approach.