Controlling a system with constraints is a challenging problem. In addition, if robustness against uncertainties in the system is taken into consideration, it becomes even more challenging. In control theory, MPC can be used to overcome this. Model Predictive Control (MPC) is widely used today in many application areas such as plants as process control or autonomous vehicle control. The main idea behind MPC is building a model and solving optimization problems online to find an optimum input to a dynamic system. MPC takes account of finite time-horizon, thus, is able to estimate the future state of the system. In theory and practice, MPC is one of the successful approaches to control a constrained system. In this thesis, robustness is incorporated into MPC and a spaceship is chosen to study as the dynamic system that is controlled.
We first test existing models and study how they fail when sensors fail in different ways. Then, a Robust MPC model is proposed, and the fault-tolerance of the model is studied. The robustness comparison procedure is implemented via simulating different failure levels of sensors which are plausible in real-world applications. Finally, the spaceship system is simulated in a space flight simulation game called Kerbal Space Program that provides realistic scenarios for space flight travel.