Data-driven modelling of LTI systems using symbolic regression

The aim of this project is to automate the task of data-driven identification of dynamical systems. The underlying goal is to develop an identification tool that models a physical system without distinguishing between classes of systems such as linear, nonlinear or possibly even hybrid systems. Such an identification tool would be able to mine data generated by the system to infer such structural knowledge, without relying on the expertise of a skilled user. This will allow researchers to shift their focus back from the modelling task to the actual utilization of the model. Such a research objective requires the identification technique to employ tools that are not targeted towards nuanced modelling tasks, but remain applicable for a very broad range of systems. Hence, we seek to develop a new framework for system identification that uses generic tools.