Simultaneous Adaptation of the Process and Measurement Noise Covariances for the UKF Applied to Nanosatellite Attitude Estimation

Abstract

A common technique for improving the estimation performance of the Kalman filter and making the filter robust against any kind of faults is to adapt its process and measurement noise covariance matrices. Although there are numerous approaches for the adaptation such as full estimation or scaling, simultaneous adaptation of these two matrices is an ongoing discussion. In this paper, firstly, two common problems for the attitude estimation of a nanosatellite are solved by adapting the process and noise covariance matrices. Then these two adaptation methods are integrated with an easy to apply scheme and the matrices are simultaneously adapted. The newly proposed filtering algorithm, which is named Robust Adaptive Unscented Kalman Filter, considerably increases the estimation performance and is fault tolerant against the sensor malfunctions.