Our research focuses on interdisciplinary basic research aimed at computing complex real-world problems efficiently. The problems considered include floods, movement mechanisms in human muscles, and storage and optimum use of renewable energies.

Important predictions can be obtained from simulations of these phenomena. Our research concentrates on ensuring that the properties of those predictions can be quantified as accurately as possible.

The kinds of questions we explore:

• How credible are simulations? How accurate or inaccurate are the predictions they provide?
• How can we guarantee that we can finish a simulation “in good time” and with maximum usability of the results even if we only have a small amount of:
  
  computing power (as with a smartphone),
  computing time (predictions have to be supplied as quickly as possible in the event of a disaster),
  or, indeed, “knowledge” about the underlying phenomena (e.g., complicated, difficult-to-predict weather phenomena)?

You try!

Our special sandbox exhibit shows how flood scenarios can be simulated for islands. You can move the sand around to create a new landscape. It will then be scanned and processed and then a wave-induced flood will be simulated in real time and projected onto your landscape.

Using this method and virtual reality, you can see directly whether a wave would flood a Maldive island, for example, or whether a quickly constructed dam could prevent flooding.

One Step Ahead of Reality

Our research focuses on solving complex real-world problems, like floods and muscle movements, using advanced simulations. We aim for accurate predictions and tackle questions about the credibility of simulations and ensuring timely results with limited resources. Join us in our sandbox exhibit to explore flood simulations in real-time and see if you can save an island from being flooded!

Often, it’s not possible to calculate complex real-life problems precisely because, for example, the underlying differential equations that describe the problem cannot be easily solved. So simulations typically use approximations of the actual problem by discretizing space and time, for instance, or simplifying the underlying model.

Usually, the closer you get to the correct solution, the longer the calculation takes or the larger the computer needed. Conversely, the quicker the calculation, the bigger the errors in the simulation. 

Time and Computing Power of the Essence

In many real-life scenarios, e.g., providing disaster warnings, time is of the essence. Another point is that lots of places (hospitals, for instance) do not have powerful computers. So our research concentrates on developing efficient methods that deliver fast yet precise results and provide a good balance between computing time and simulation error rates.

Apart from the systematic errors described above, simulations also have to take uncertainties into account. They can occur when input data isn’t measured properly, for instance, or not all of the necessary information is available.

Randomizing for More Accuracy

A familiar example is weather forecasts, which cannot predict with certainty whether and how much it will rain. That uncertainty can be quantified more accurately by running the simulations several times in random conditions.