High-Precision Integrated Navigation System for Intelligent Vehicles Using an Extended Kalman Filter

Abstract

Maintaining high-precision continuous positioning for intelligent vehicles in GNSS-limited environments (such as urban canyons and forest areas) has become a core challenge. A single Global Navigation Satellite System (GNSS) or Inertial Navigation System (INS) can hardly meet decimeter-level navigation requirements, making multi-source information fusion critically important. This paper reviews the optimal design of GNSS/INS integrated navigation systems based on the Extended Kalman Filter (EKF). First, it reviews the basic mechanism of the conventional EKF following the “ INS prediction, GNSS update ” architecture; then, it elaborates on two improved approaches: the maximum Correntropy Extended Kalman Filter (RMCEKF), for adaptive estimation of noise covariance under dynamic external noise, and the Lie group-based theory (LIEKFN), for reconstructing the state manifold space and overcoming linearization error under large misalignment angles. Finally, the conclusion points out that deep integration of adaptive robustness and Lie group manifold geometry is key to overcoming the bottleneck of a single algorithm, providing a theoretical reference for designing high-precision, high- robustness navigation. This research not only provides a new theoretical framework for high-precision integrated navigation algorithm design in complex environments, but also offers an important reference for engineering implementation of highly reliable intelligent-vehicle navigation systems.