Abstract
Computer vision, with its ability to efficiently generate geometric or photo-metric models of "things", has become a key resource in the process of virtualizing actual scenes or artifacts. Because of the key role of vision as a sensory modality, we often associate Virtual Reality with visualization, an association which has progressively been enriched by additional modalities such as haptics or touch, as well as by the inclusion of (usually) simple behaviors necessary to provide a certain sense of "presence".In this presentation, we discuss a class of Virtual Environments which target support for critical interventions -applicative areas where operational mistakes may have severe human / economic consequences and must be avoided. The level of strategic and tactical support - ranging from scenario optimization to direct supervisory control - which is required in such cases implies much more than an a believable sensory experience, but rather an infrastructure where detailed, accurate (within specified bounds), coherent and robust physical representations and behaviors are maintained. While interactive visualization remains a key component of such a system, it needs to shifted away from its core and be integrated as a satellite to a simulation engine capable of high performance computation consistent with hard real-time requirements. This is especially true during the intervention phase, in an AugmentedReality mode, when actual tools operate upon actual objects. This raises a number of design and implementation challenges which we will consider in the context of industrial and medical applications which are currently being pursued in our Laboratory.