Adaptive “Cubature and Sigma Points” Kalman Filtering Applied to MEMS IMU/GNSS Data Fusion during Measurement Outlier
Adaptive “Cubature and Sigma Points” Kalman Filtering Applied to MEMS IMU/GNSS Data Fusion during Measurement Outlier作者机构:International Institute for Advanced Aerospace Technologies of Saint-Petersburg State University of Aerospace Instrumentation 67 Bolshaya Morskaya Saint Petersburg Russia SET Laboratory (Systèmes Electriques et Télecommande) of Electronic Department of Saad Dahlab University of Blida Soumaa Algeria
出 版 物:《Journal of Sensor Technology》 (传感技术(英文))
年 卷 期:2013年第3卷第4期
页 面:115-125页
学科分类:1002[医学-临床医学] 100214[医学-肿瘤学] 10[医学]
主 题:IMU MEMS GPS GNSS Kalman Filtering Cubature Rule Sigma Points Unscented Kalman Filter
摘 要:In this paper, adaptive sensor fusion INS/GNSS is proposed to solve specific problem of non linear time variant state space estimation with measurement outliers, different algorithms are used to solve this specific problem generally occurs in intentional and non-intentional interferences caused by other radio navigation sources, or by the GNSS receiver’s deterioration. Non linear approximation techniques such as Extended Kalman filter EKF, Sigma Point Kalman Filters such as UKF and CDKF are computed to estimate the navigation states for UAV flight control. Several comparisons are conduced and analyzed in order to compare the accuracy and the convergence of different approaches usually applied in navigation data fusion purposes. The last non linear filter algorithm developed is the Cubature Kalman Filter CKF which provides more accurate estimation with more stability in Tracking data fusion application. In this work, CKF is compared with SPKF and EKF in ideal conditions and during GNSS outliers supposed to occur during specific interval of time, innovation based adaptive approach is selected and used to modify the covariance calculation of the non linear filters performed in this paper. Interesting results are observed, discussed with real perspectives in navigation data fusion for real time applications. Three parallel modified algorithms are simulated and compared to non-adaptive forms according to Root Mean Square Error (RMSE) criteria.