Nanosatellite attitude estimation via triad-aided kalman filters

Abstract

Increasing demand for the space operations, space industry turns its face to cost effective solutions. Small satellites, due to their size and cost, are receiving interest from many organizations. In 2018, NASA sent two MarCO cubesats to Mars. Their mission was to relay the landing vehicle data back to Earth. Restricted size comes with its own challenges. The amount of attitude determination and control equipment that can be placed in small satellites are considerably lower than a the regular size satellite. In this work, using common sensors, couple of filters are design to overcome to attitute determination problem. Two of most common sensors that are being used in nanosatellites are magnetometers and sun sensors. Magnetic dipole model is selected for magnetic field model. VSOP87 theory is used for sun direction vector. Using these two models, sensor measurement models have been established. For attitude representation of the spacecraft euler angles are selected. Using these angles, equations of motion of the spacecraft are obtained. One of the earliest attitude determination method is algebraic method. Using sun sensor as the first triad, additional two triads have been constructed. Constructed three vectors form a direction cosine matrix. Body angles are obtained from this matrix. In order to increase the accuracy of the satellite motion parameters, three different methods have been analyzed. Extended Kalman filter, unscented Kalman filter and adaptive fading Kalman filters are derived and designed for the system. Comparison of these three filters are studied. Designing the filters, a new sensor type, gyroscopes are used. Measurement model of the gyroscope is obtained. Body angles that are produced by algebraic method are used as linear measurements to the Kalman filters. Hence, two methods are integrated for achieving better accuracy for body angles and angular velocities. Analytic jacobian matrices for EKF have been calculated. Magnetometer and gyroscope biases are estimated. Instead of using magnetometer measurements, magnetometer is corrected with estimated biases. Same process has been applied to the gyroscope measurements. Sensor measurements can be corrupted for many reasons. Aadaptive filters are applied to both EKF and UKF to increases robustness of the both filters. All of the algorithms are designed in MATLAB. Simulation are also conducted at this package program.