Modelling is an essential tool to understand and design complex systems, even more in the current era in which computational devices are embedded into and interact with physical world. Whether these cyber-physical systems have the task to sense the world or are actively interacting with it, their design is highly complex, and modelling is needed to take into account aspects like openness, uncertainty in the environment, spatial distribution and mobility. 

Models themselves are going to be complex and computationally costly to simulate, particularly when uncertainty is taken into account, adding stochastic ingredients into models. 

In this context, it is of fundamental importance to have techniques, based on simulations, capable of analysing efficiently such models, of identifying and verifying  behavioural emergent properties, of synthesising controllers. As simulation is a typical computational bottleneck, approximations and machine learning techniques play a central role in designing algorithms to perform these tasks.  

A key player in modelling is data, which can be available abundantly nowadays. Data-based modelling beyond parameter identification, e.g. black box predictive models, and combining them with mathematical ones, is becoming more and more relevant. 

Specific research projects regarding this area include: efficient verification of properties and counterexample generation under high-dimensional uncertainty using machine learning, efficient design and control synthesis of cyber-physical systems (e.g. IoT), runtime monitoring of spatio-temporal properties, using machine learning for model abstraction and simulation acceleration, combining classical and data-based models to improve accuracy and reliability of predictions.  

Application of these techniques go beyond fluid mechanics and applications in earth sciences, touching areas like robotics, internet of things, biology, model-based engineering.