Non-linear Model Predictive Controller for the autonomous descent and landing of a parafoil system in multi-planetary applications.

The objective of this thesis is to evaluate the applicability of a non-linear Model Predictive Controller in multi planetary environments for control of an autonomous parafoil system. The work stems from many examples of the use of parafoils and parachute as a landing system on planets with suitable atmospheric conditions. As in terrestrial applications, from the thrill of paragliding to the need of deploying resources to the battlefield, or in space exploration where landing critical instruments on a planet leaves no room for error, parafoils have always been explored as a possible solution. The inherit aerodynamic characteristic of this type of system, requires specific atmospheric condition to be considered as a suitable candidate solution to the descent and landing phases of a mission. But it provides considerable benefits from the perspective of energy expenditure and complexity. This work focuses on implementing a simulation environment able to correctly represent the dynamics of the system at hand and the different environments considered. The data generated is used in the system identification of the reduced order models employed in the controller. The controller parameters are subsequentially tuned using a weight scheduling technique, to allow for the different required behaviour during the descent of the system and optimized for each environment. Finally, the performance is analysed in different atmospheric conditions, obtaining information on the critical aspects of the application. Scenarios for the simulations differ for each planetary application, as the atmospheric modellization and the parameters that are relevant for the analysis. For Earth, a wind disturbance parametrized as a fraction of the planar velocity of the system is considered, as to generate a significant disturbance which is still manageable for the controller. On Mars a constant wind disturbance is considered, aligning this work to the other studies present in literature. On Titan, an exponential parametrization of the wind is employed, stemming from the research carried out by NASA with the Huygens spacecraft. The obtained performance is analysed critically and empirically, fitting the landing point obtained from the Montecarlo simulations to a multivariate Gaussian distribution. The covariance error ellipses obtained this way allows for quantification of the performance in a compact numerical form. The obtained results are comparable to other techniques for the application on Earth and on Mars, while the performance on Titan is adequately better than studies previously carried out in this field.