Adaptive Extended Kalman Filter in SINS/GPS Tightly Coupled Navigation System

Abstract

In the practical application, the SINS/GPS tightly integrated navigation system is faced with the situation that state error model is inaccurate or noise statistical characteristics are inconsistent with the reality. However, traditional Extended Kalman Filter (EKF) method can not effectively solve it, causing large filter errors in this system. This paper presents a method of GDOP estimation and adaptive Extended Kalman Filter, which combines satellite positions of ephemeris with the output of corrected SINS position instead of GPS positioning results to calculate GDOP; On this basis, using the GDOP value and innovation realizes the online real-time estimation of measurement noise variance matrix Rk while taking advantage of innovation realizes the online estimation of state noise variance matrix Qk to achieve the adaptive effect and improve navigation accuracy in tightly integrated navigation system. Introduction Strap-down Inertial Navigation System (SINS) does not depend on any external information, autonomy is strong and data update speed is fast, but the error of SINS will diverge over time, so SINS is not suitable for long time navigation; Global Positioning System (GPS) can work with high precision and long term, its error does not diverge over time, but there are also some shortcomings such as multipath effects and easily-affected radio interference, resulting in low stability of navigation. Therefore, GPS can’t perform well as a standalone system. In the high accuracy navigation field, GPS and SINS are mostly integrated together to overcome their own shortcomings, which makes the navigation performance better than the performance of two separate systems. SINS/GPS tightly integrated navigation system uses more raw data from GPS receiver pseudo-ranges and pseudo-range rates and has obvious advantages. Therefore, this paper focuses on the study of SINS/GPS tightly coupled system in order to achieve better navigation performance. In practical applications, due to the error of inertial device itself and inaccurate external environment information, there is a system error model inaccuracy and noise uncertainty in the SINS/GPS nonlinear systems. As for the uncertainty of system noise and statistical properties of measurement noise, people proposed various adaptive filtering algorithms. Literature [2-4] proposed a method of real-time estimating the system noise and measurement noise based on the innovation. The method does not increase the system dimensionality, only needs limited innovation to memory, and the calculation is simple and reliable, but cannot effectively solve the question that the system noise and measurement noise are inaccurate at the same time. This paper presents an adaptive filtering method for online estimation of the measurement noise covariance matrix Rk and the system noise covariance matrix Qk based on estimated GDOP value and innovation in EKF, and then applying this method to the SINS/GPS tightly coupled navigation system and verifying the validity of the adaptive algorithm. SINS/GPS tightly integrated system model State equation of integrated system is as follows :