by David Boyd and Julian Gallop, Rutherford Appleton Laboratory, and Jeremy Walton, NAG Ltd.

Users of IRIS Explorer have been long familiar with the way in which it can be used to create applications to visualise their data in a 3D representation which is viewed in a 2D window on a monitor screen. This data-centered interaction mode, in which a camera is manipulated around the 3D representation, can be a powerful way of interacting with some forms of data but another paradigm might, under some circumstances, offer greater flexibility and appear more intuitive. For example, the user might find it better to be freely navigating through the 3D space whilst looking around and interacting with the scene. This user-centered approach is the one which is often employed within virtual environments.

Recently we have applied this user-centred approach to data visualization by using a virtual environment (VE) to construct a new interface-called VIVRE-for a data visualization system such as IRIS Explorer. The objective of this work is to allow the user to 'steer' the visualization task via the VE to more easily gain the insight they are seeking from their data, whilst retaining the functionality and extensibility of the visualization system. Unlike previous work in this area, which has often been concentrated on building the components of the system from scratch, we have used commercial systems in order to leverage existing user experience. The project has also been driven by user requirements, and VIVRE is being applied to their real commercial and industrial problems.

Four groups of end-users drove the technical directions for the VIVRE system. Their applications covered a wide range of domain and scale, including the visualization of combustion and gas flow in industrial processes, potential leakage paths surrounding underground storage caverns, the dynamic operation of a flexible fluid control valve, and complex natural microstructures and their properties.

The users required a range of visualization actions (such as moving cutting planes and changing isosurface thresholds) to be accessible directly from the VE. Other actions such as substituting a different visualization function in the network required the user to interact directly with the visualization system. The users’ requirements were amalgamated into a single prioritized list which guided the development work and provided the basis for user evaluation of the VIVRE system.

The VE selected for the VIVRE system was dVISE, a mature commercial product which provides the required extension mechanisms, thus facilitating construction of an interface to the visualization system. In addition, it supports a variety of user input/output devices, ranging from the 2D window plus mouse, through a six degree of freedom input device to stereoscopic projection and immersive head-mounted display. We extended IRIS Explorer by supplementing its module suite (which already contains standard visualization modules which produce geometry-a cutting plane, for example-from numerical data) with new modules which handle data coming from the VE and for 'tagging' geometry with an object identifier.

In the development of VIVRE, we had to take account of a number of technical issues, including the way in which the scene in the VE is updated as a result of changing visualization parameters in the IRIS Explorer map, the ability to distinguish between multiple objects (e.g. isosurfaces, cutting planes, enclosing geometries) in the VE scene, the efficient transfer of geometry between IRIS Explorer and the VE and the way in which IRIS Explorer is controlled from within the VE.

Experiences from the VIVRE users (none of whom had previously worked with a VE) were largely positive. Although no difference in performance between IRIS Explorer and the VE could be detected for small datasets, there was some consensus that the VE allowed a comparatively smooth navigation of significantly large datasets, to an extent that user actions were translated into navigation in a direct way not normally associated with general purpose visualization systems. Similarly, users reported being able to see advantages of the user-centered approach for visualizations which resulted in a large collection of 3D objects. In addition, the direct interaction with cutting planes in the 3D scene was said to be helpful. The user was able to manipulate a plane’s position and orientation in the VE; upon releasing the plane, its new location was passed to IRIS Explorer which recalculated the image to be displayed on the plane.

Not all of the VIVRE users worked with an immersive system. Of those that tried it, all reported getting better insight into their data, particularly for data where spatial relationships were important, and-again-where there was a large number of 3D objects.

Future work here includes the extension of the system to include other visualization techniques such as streamlines, the incorporation of time-dependent data and the improvement of color control and the interrogation of intersections between objects. Finally, we note that NAG has said that it intends to incorporate the VIVRE system into a future release of IRIS Explorer.

More information on VIVRE, which was a project in the HPCN sector of the EC 4th Framework Programme can be obtained from http://www.tessella.co.uk/literature/articles/tessarchive/vivre.htm.