TRIAD aided adaptive Kalman filter for fault tolerant attitude estimation of a nanosatellite

Abstract

In this study, a fault tolerant attitude estimation system is proposed for a nanosatellite, combining the TRIAD algorithm, a vector-based attitude determination method, with an adaptive extended Kalman filter. As the first step of the estimation system, the TRIAD algorithm produces an initial coarse attitude estimation using three-axis magnetometer and sun sensor measurements. Then, this coarse estimation is filtered via an adaptive extended Kalman filter to obtain the final estimation. Besides filtering and providing more accurate estimations, the adaptive filter also makes the system fault tolerant by maintaining the estimation accuracy in cases where sensor faults exist and measurements become unreliable. This is made possible by tracking the filter's innovation history and re-adjusting the Kalman gain using appropriately calculated fading factors. In order to verify the performance of the proposed system, two different simulations are performed where the attitude sensor outputs are disturbed by additional measurement noise.