Simulation and control in data-driven and interactive contexts hold out the promise of completely new opportunities in scientific fields in which the systems studied can be described only partially or at a very high cost using classical modeling approaches. However, to fully exploit this huge potential, new data-integrated approaches face the challenge of having to work in dynamically changing and heterogeneous environments.
This concerns, first, the computational setup during the active simulation and, second, the use of simulations for active real-time control. Interestingly, in both cases, the same aspects need to be considered regarding dynamic changes if we are to achieve our long-term Visionary Examples. These include
- changes in the availability and type of computational resources due to shared usage, hardware faults, or volatile connectivity of mobile networks;
- changed user requirements, such as maximal response time, required accuracy, or configuration parameters, in interactive settings;
- variations in the characteristics of the considered system such as aging or surgical changes of a simulated human organ, or outer disturbances acting on a control loop; and
- real-time availability/generation of data, e.g., eye-tracking information in pervasive simulations. Moreover, we are confronted with a broad heterogeneity of models (classical and data-based components, multi-X modeling) and data, e.g., structured versus unstructured, time-dependent versus static, and precise versus uncertain.
Tackling the challenges defined by this dynamical and heterogeneous context requires an interdisciplinary effort both for automatic control as well as for computational tasks in simulation.