Nonparametric System Identification of an Autonomous Helicopter

The unmanned helicopter Skeldar relies on model based control to per form its tasks. System identification is an integral component in the process of deriving models of helicopter dynamics. This thesis aims to investigate how nonparametric methods of system identification can support the current modelling and system identification practices at Saab.Nonparametric system identification does not require a pre-defined model structure. Models estimated with this methodology may be used to validate parametric models, which are necessary for the implemen- tation of the model based control system. This thesis examines several nonparametric methods of system identification in both the frequency and time domains. The theory of these methods is presented and their performance is analyzed on data from flight tests as well as from simu- lated systems.Analysis of the results shows that models are highly dependent on the choice of input signal spectrum. To best take advantage of nonpara- metric system identification in this application, experiments should be designed with special regard to the system properties sought to be mod- elled. Nonparametric system identification can then be used to provide a good understanding of the system properties in the excited frequency region.In the specific case of helicopter dynamics, of which the principles are very well understood at Saab, it can be concluded that the exist- ing system identification process is sufficient to provide well performing models. However, a nonparametric model could be estimated and used as a tool for comparison and validation in the process of identifying a paramteric model.