Model Predictive Based Unscented Kalman Filter for Hypersonic Vehicle Navigation With INS/GNSS Integration

Gaoge Hu,Wei Wang,Xinhe Zhu,Bingbing Gao,Longqiang Ni,Yongmin Zhong

doi:10.1109/access.2019.2962832

Abstract

The INS/GNSS integration is the commonly used technique for hypersonic vehicle navigation. However, owing to the complicated flight dynamics with high maneuverability and large flight envelope, the dynamic model of INS/GNSS integration inevitably exists errors which degrades the navigation performance of a hypersonic vehicle seriously. In this paper, a new model predictive based unscented Kalman filter (MP-UKF) is proposed to address this problem. The MP-UKF employs the concept of model predictive filter for the establishment of a dynamic model error estimator, and it subsequently compensate the model error estimation to UKF for nonlinear state estimation. Since the MP-UKF could predict the dynamic model error persistently and correct the filtering procedure of UKF online, it improves the UKF adaptiveness and is promising for the performance enhancement of INS/GNSS integration for hypersonic vehicle navigation. Simulation results and comparison analysis have been conducted to demonstrate the effectiveness of the proposed method.

Highlights

Hypersonic vehicle, which is a kind of vehicles at the speed of Mach 5 or above, provides a cost-effective way to access space by reducing the flight time
This paper proposes a new model predictive based unscented Kalman filter (MP-unscented Kalman filter (UKF)) for hypersonic vehicle navigation with INS/GNSS integration to address the performance degradation due to the dynamic model errors involved
The MP-UKF employs the concept of model predictive filter (MPF) to improve the UKF adaptiveness and furtherly resist the effect of dynamic model error on navigation solution

Summary

INTRODUCTION

Hypersonic vehicle, which is a kind of vehicles at the speed of Mach 5 or above, provides a cost-effective way to access space by reducing the flight time. In most practical applications especially for the hypersonic vehicle, as it is difficult to satisfy these assumptions due to the high maneuverability and large flight envelope, the nonlinear system model should be employed by INS/GNSS integration to describe the complete propagation process of system error and reflect the real system characteristics. Cho and Choi developed a sigmapoint based receding horizon Kalman filter (SPRHKF) to improve the UKF adaptiveness against dynamic model error and temporarily unknown sensor bias [21]. Due to the real-time performance in state estimation and the correction of dynamic model, the MPF is capable to achieve superior filtering performance in presence of dynamic model error in comparison with EKF and UKF [26]. This paper presents a novel model predictive based unscented Kalman filter (MP-UKF) for hypersonic vehicles navigation with INS/GNSS integration. Where f (·) is a discretized nonlinear function describing the dynamics of system state; and w(k) is the discrete-time process noise

MEASUREMENT MODEL

MP-UKF ALGORITHM

SIMULATION ANALYSIS AND DISCUSSION

Findings

CONCLUSION