Magnetic attitude control of a nanosatellite

Abstract

In this thesis, a PD and LQR controller is used to control attitude of a nanosatellite with using only magnetorquers. Three different approaches are used for that. Firstly, a deterministic approach is used with the assumption of exact knowledge of data. Secondly, magnetometer, Sun sensor, horizon sensor, and gyroscope are modeled as noisy devices to make the simulations more realistic. Measurements of vector measurement devices are used in TRIAD method to determine attitude. Thirdly, these measurements are filtered with a basic linear Kalman Filter. All these three test cases are used for both controllers with two different satellite models. One of the model is a gravity gradient stable satellite and used as a verification model. The other model is a basic 3U CubeSat orbiting in Sun-synchronous orbit. Tuning the gains of PD controller and Q matrix of LQR is also considered during this thesis. Gain matrices of PD controller is assumed diagonal and tried to be optimized for all the test cases and models. Similarly, Q matrix of LQR controller is considered as diagonal matrix to optimize it. The aim of this optimization process is to obtain a performance increase on attitude control. Also, particle swarm optimization is used for this process. Lastly, magnetorquers, that are used in 3U CubeSat model, are designed in the thesis. One torquerod for two axes and one air core torquer for one axis are designed. The design space of magnetic torquers is created with respect to a literature review for magnetic torquers. Then, the optimized design is selected from this design space with the help of particle swarm optimization. Results of the thesis showed that, achieving 1-2 degree pointing accuracy for each Euler angle with optimized parameters is possible when deterministic case is used. On the other hand, adding noise to measurements and using TRIAD for attitude determination decreased the performance of optimization. It is really hard to work on randomly changed solution space with basic optimization algorithms. While the optimization is running, the run is repeated 10 times to get more stable and high performance results. TRIAD cannot reach high accuracy like deterministic case, but the strategy that is followed for optimization is work for next case. Then, applying Kalman Filter to results of TRIAD allows to an important improvement on attitude control. Kalman Filter almost neglects the effects of noises. Also, in this case the optimized parameters provide similar accuracy as the deterministic approach for some cases. More importantly, it is possible to obtain optimized parameters that are gives stable and accurate results.