Low-cost integrated INS/GNSS using adaptive H∞ Cubature Kalman Filter

Abstract

We proposed an adaptive H-infinity Cubature Kalman Filter (AH∞CKF) to improve the navigation accuracy of a highly manoeuvrable unmanned aerial vehicle (UAV). AH∞CKF fuses the Inertial Navigation System (INS) and Global Navigation Satellite System (GNSS) measurements. Traditional state estimation filters like extended Kalman filters (EKF) and cubature Kalman filters (CKF) assume Gaussian noises. However, their performance degrades for non-Gaussian noises and system uncertainties encountered in real-world applications. Thus, designing filters robust to noise and distribution is crucial. AH∞CKF combines H∞CKF design with an added adaptive factor to adjust the state estimation covariance matrix according to measurements by exploiting the square root method to yield more numerically stable results (SrAH∞CKF). We conducted multiple dynamically rich flight tests to validate our claims using a UAV equipped with a commercially well-known GNSS solution. Results show that the SrAH∞CKF state estimation outperforms EKF and CKF methods on average by 90% in various statistical measures.

A Novel Adaptive Robust Cubature Kalman Filter for Maneuvering Target Tracking with Model Uncertainty and Abnormal Measurement Noises

“…It is a particular form of Kalman filter [ 15 ]. For the nonlinear discrete system proposed in the previous section, the H-infinity filter design idea is that when , , and reach the upper limit, there is a specific cost function to minimize [ 24 ]: where represents the posterior covariance matrix and and are the initial covariance matrix and the system’s initial state, respectively. The standard symbol operations used in the formula are as follows: .…”

“…It is a particular form of Kalman filter [15]. For the nonlinear discrete system proposed in the previous section, the H-infinity filter design idea is that when P k , Q k−1 , and R k reach the upper limit, there is a specific cost function to minimize [24]:…”

A novel Gaussian sum quaternion constrained cubature Kalman filter algorithm for GNSS/SINS integrated attitude determination and positioning system

“…That is to say, GNSS/SINS-IADPS is a nonlinear non-Gaussian system. In such case, if the random model is mismatched, it may affect the accuracy of filtering estimation and even lead to accuracy divergence in severe cases ( Duong and Chiang, 2012 ; Wang et al, 2020 ; Taghizadeh and Safabakhsh, 2023 ). Therefore, if the QCKF algorithm can be extended to take into account both the effect of non-Gaussian noise and normalization constraints, it will show better estimation performance.…”