MED 2017 Plenary lectures and tutorial: Modelling, estimation and identification of spatio-temporal and multiscale systems

Advances in sensing and data acquisition systems in a multitude of domains, combined with information networks, have expanded the scope of monitoring and characterising dynamic behaviour of systems in many applications. The data and signals that are acquired from the system are typically multichannel, multivariate and in some cases, spatially organized. The focus of the talk is on systems from which spatio-temporal data are obtained. The talk will begin with a background to estimation and system identification giving details of the methods used. This is followed by a selection of problems that are characterised by spatio-temporal processes that set the scene for the classes of problems being considered. Specifically, it highlights three case studies which serve to illustrate the three different spatio-temporal models used in estimation and identification. The first case study is from the engineering domain with application to wind turbines. The problem of estimating the wind velocity and pressure fields from LIDAR-type sensor measurements is considered. The second case study is from the healthcare domain with application to neuroscience. The problem of developing a patient-specific model from intracranial EEG array signal that can characterize changes during epilepsy is considered. Finally, the third case study is from the social science domain with application to conflict modelling. The problem of developing a predictive model of the dynamics of conflicts from event data is addressed. The talk concludes with the challenges in spatio-temporal modeling.