A Study on Reasonably Reliable Attitude Determination Algorithms for Microsatellites

Abstract

This study investigates adaptive based reasonably reliable attitude determination algorithms for microsatellites which use simple attitude sensors and on-board computer with limited performance. The advance reasonably reliable attitude determination algorithms should have high cost-effective on consumption power per estimated qualities or calculation effort per estimated qualities. Within this study, three methods are proposed for biased sensor systems, for efficient and flexible power consumption estimation, and recovery systems, alternately. All of proposed methods are carefully analyzed and compared with conventional, flight confirmation methods in term of cost-effective, convergence speed, estimated accuracy, and robustness characteristics based on the Monte Carlo numerical simulations. The Monte Carlo numerical results show that the proposed methods provided advance tools to control the payment cost for archiving a reasonably estimated quality. Finally, these proposed methods are implemented and tested as flight software of attitude determination system in a specific microsatellite. The first contribution of this thesis is the development of a real-time tuning separate-bias extended Kalman filter (RTSEKF) for robust spacecraft attitude estimation in the presence of measurement biases. The adaptive mechanism applied during calculation of Kalman gain in the bias filter could help the attitude determination system have a better response to the larger initial estimated error and unpredicted sensor bias models. The RTSEKF has higher accuracy and faster convergence speed than the conventional methods like separate-bias extended Kalman filter, extended Kalman filter , and event unscented Kalman filter. The computational cost of RTSEKF could be controlled by choosing the accuracy of the optimal process or using the filter reduction form. The secondary contribution of this thesis is the development of an automatic and real-time tuning procedure for the process (Q) and measurement (R) noise covariance matrices of the UKF. The proposed method is applied for efficient and flexible power consumption of attitude determination. It uses attitude data from a higher-accuracy estimator as a truth reference data to compare with estimated attitude data of a lower-accuracy estimator. Based on the difference between the two attitudes above, the cost function is generated then minimized by a numerical optimization process using the downhill simplex algorithm. Through the tuning process, the convergence speed and estimated accuracy of the lower-accuracy estimator are improved. Based on that, to customize the system power consumption, the proposed filter is used with the suitable duration and frequent repeat of turn-on time of the higher-accuracy estimator data. The third contribution of the thesis is the development of a residual-based adaptiveunscented Kalman filter (AUKF) for attitude estimation and fault detection and diagnosis (FDD). In the proposed AUKF, the measurement noise covariance matrix is real-time updated in case the FDD system gives out the warning signal. The FDD includes two filters, residual generators, statistical tests and isolation process. The proposed algorithms in this chapter are light and fast due to the use of simple algorithms during residual generation parts. Therefore not only the requirement of the onboard computer is reduced but also the delay in isolation/adaptive processes are cut down. Through proposed methods, users can easier to control the payment cost, such as computational cost or power consumption cost, of attitude determination system for archiving a reasonably reliable estimated result. This content makes attitude determination systems are more reliable in practical applications. Moreover, with the adaptive mechanisms applying in general part of filters, these proposed methods also can be expanded for other applications in the field optimal estimation of dynamic systems.