Arctis

High Performance Computing Center Stuttgart (HLRS), Universität Stuttgart
Prof. Dr.-Ing. Michael M. Resch, Uwe Wössner

..in collaboration with VISENSO

COVISE is an interactive rapid prototyping platform which allows collaborative analysis of ongoing simulations to optimize products or carry out experiments. Our approach is to integrate the simulation, analysis and physical prototype or experiment into one consistent environment to not only analyse a problem but also modify and thus optimize the product we are simulating.

By using parallel high performance computers, we can obtain high quality simulation results with interactive response times. We are using advanced visualization techniques like direct volume rendering, Virtual- and Augmented Reality to visualize the simulation results. Physical objects provide easy to use user interfaces to interact with both the simulation and the analysis of the simulation results. Virtual environments provide high quality visualization at scale 1:1 and allow for collaborative analysis of the results.

We demonstrate the applicability of our approach with two applications, an interactive air-conditioning optimization as well as air-flow simulation in urban planning. The multidisciplinary nature of most of the problems currently solved with numerical simulations typically involves a number of working groups or departments. In international research projects and today’s globally operating companies, those teams are not necessarily collocated but often spread around countries or the whole globe. Thus collaborative analysis of simulation results can help reducing development times by increasing the speed and quality of the communication. Therefore we also support collaborative working in this environment.

The first application we present is a 3D CFD air-conditioning simulation of our booth at SC05 conference. A model at scale 1:20 is used as user interface. The user can move around model furniture to try out different configurations and after pressing a button, a computational grid is generated which represents the new positions of all objects. This mesh is split up into pieces by a domain decomposition module. Those pieces are then transferred to the supercomputer which carries out the simulation. After a few seconds the first results are sent on to the visualization machine which performs the post-processing, e.g. calculates particle traces or isosurfaces. The results are displayed through an Augmented Reality HMD, overlaid over the model or the real booth, in an immersive virtual environment or on the desktop. A similar model of buildings provides an intuitive interface for the second application as well.

In addition to geometry definition, the physical model can also be used to interact with the visualization. Markers can be moved around in the model to position cuttingplanes or start particle-traces. In a collaborative analysis session, all partners have a common view on the dataset, no matter whether they are joining with an immersive virtual environment, an augmented reality setup or just from their desktop. Furthermore it is possible to interact with the visualization, simulation and grid generation from each of the participating sites. In this specific application, this means that one could change geometry parameters like the position of the booth equipment, or its shape, boundary conditions of the simulation, like the velocity at the inlet, or grid generation parameters like the resolution of the mesh and the number of partitions, numerous parameters of the simulation or visualization, such as position of cutting planes or particle traces or iso values.