This report is also available as an Acrobat file.
The Design of Virtual Environments with particular reference to VRML
Design into production
Preceding sections of this report have deliberately addressed issues which are at several removes from the immediate business of making a Virtual Environment. Here we approach more closely the transition from concept to project.
Some metrics of reality
An unspoken assumption in some discussion of VR is that its technological development is well advanced and that the path before it is unproblematic. We might almost think that VR is within sight of the goal of perfect replication. By contrast, we have emphasised that a VR world is a representation, not a reconstruction of reality. The only representation that would theoretically approach reality would be boundless in size and of infinitely small granularity, modelling the behaviour of sub-atomic particles! This is a practical problem, not a philosophical one, and an important one for education. Not only must those who are building models recognise the limits of the representations they build, but so must the academics and learners making use of them.
Items from a reality checklist
It may be useful to look at a few of the aspects of reality which could in principle be represented using computer models, and notice how only some are possible in VR or even simulation systems. We have tried where possible to describe these metrics in non-computer terms.
- Extent and scaling
- Height, depth, breadth
- Position and movement
- Location in three-dimensional space; translation, rotation, including non-linear velocities
- Passive visual qualities
- Colour: hue, saturation, value; transparency, translucency; reflectivity (matt, lustrous, shiny, gloss, part-mirrored); texture (surface texture, solid texture)
- Active visual qualities
- Light source type eg light bulbs, phosphor, lightning; colour of light; distance, direction and spread of light sources
- Dynamics and dependencies
- Free movement independent of all other objects; rigid fixing to other objects; able to penetrate some objects, not others; internal object constraints (degrees of freedom); dependent object constraints (within, outside, hinge, slide); hierarchical object constraints (eg. arm moves hand moves finger, and reverse: finger moves hand moves arm)
- Physical qualities - passive
- Mass, hardness, brittleness, flexibility, crystallinity eg. splintering on breaking
- Isotropy and directionality
- Insulating/conducting properties for heat, sound, electrical current etc.
- Physical qualities - active
- Emitting heat and other forms of radiation
- Textural qualities
- Fluffiness, cloudiness
- Tendencies to fall, continue along a trajectory, change colour, decompose, grow, subside, disperse, cool, lose or gain moisture, expand or contract
- Animal behaviours and propensities
- Human behaviours and propensities
- Atmosphere: colour, opacity, humidity, air currents
- Responsive sounds
- When moved eg. scraping; when struck eg. thud; when squeezed eg. escaping air
- Autonomous sounds
- eg boiling water, sounds of animate creatures
Even this brief list indicates many aspects of reality which are currently difficult to model in VR systems (together with others which are widely available). Importantly, VR systems have tended - by contrast with various kinds of simulation - have tended to model passive visual qualities, rather than those of activity, behaviour and process.
Before embarking on VR...
Up-to-date technical tutorials can provide advice on the practical steps necessary in order to undertake a VRML project. Here we offer broader questions which should be asked when embarking on a project.
Is the project enhanced by 3D?
Clearly a fundamental issue, this is not always easy to answer. Some help may be provided by the following questions:
- If the project replaces other materials, say student handouts of worksheets, do those existing materials use 3D graphics? If not, why not?
- If the project is one of visualisation, are you proposing more dimensions in the output than in the data? For example it is all too common in business graphics and news media to see two-dimensional data 'enhanced' by a redundant third dimension - thickened bar charts, pie charts with meaningless heights applied to the segments, and so forth. Similarly, as an aid to navigation a map may do as well as a 3D model, or better.
- Are any graphics needed at all (let alone 3D)? Well designed tables can be as effective as graphics (though the literacy and numeracy of the audience must as always be taken into account).
- Is 3D too much? Perhaps what is proposed is a grand interface to rather trivial content. A number of Virtual Gallery projects have suffered this failing, providing a lavish means of obtaining sparse and finally unrewarding information.
Can you justify the following requirements in relation to any expected gains?
- Development time: 3D modelling involves significant labour, some of it uninteresting repetitive production work (though this decreases if many similar models are built, since libraries of components can be built up). Currently, output from 3D modellers is not always in perfect VRML form. Debugging of VRML files may be necessary, as may optimisation (for example to reduce polygon count).
- Effort: in addition to perhaps needing to master the relevant software packages for creating 3D models and successfully testing and debugging VRML files, it will be important to maintain and develop the models once built. New versions of browser software may reveal hitherto hidden failings in the model, and user expectation may demand enhancements as new versions of VRML emerge and browsers become more sophisticated.
- In summary, there may be a need for a commitment of time and effort at three stages: 1 pre-development, including mastering software packages and techniques; 2 development of models, animations and VRML code; 3 maintenance and enhancement
- Loss of audience: is there a danger that users will be sacrificed who do not have access to suitable bandwidth, hardware and software? For example, developing on-line courses dependent on VRML might be counter-productive, if what is needed is greater access for home-based students.
- Direct costs: in addition to the demands on bandwidth, hardware and software for users, there will be direct costs associated with the machines needed for development.
Perception and cognition
In this report we have commented on the possible dangers of replacing abstract and symbolic forms of knowledge with those which are more direct and concrete.
Designing the world - making the most of 3D
- Are the possibilities of 3D worlds being fully exploited? Are the respective strengths of spatiality and virtuality being fully explored?
- What use can be made of the space as a place for collaboration? What benefits arise from combining the strengths of visualised worlds with those of collaborative exploration, research and projects?
Designing the world - practical constraints
Many of the constraints on VRML models at any given time are contingent on particular versions of software - things which are problematic one month are easily solved the next. However, there are some general guidelines which are likely to remain valid:
Some simple steps which may be taken include:
- Keep models small: even though bandwidth will increase and computers become more powerful (and optimised for 3D), the need to keep models to manageable size will continue as network traffic also increases, consuming the increasing network capability. There will never be a situation in which models move too fluidly in the browser, or the user's actions are responded to too quickly.
- Use in-line instances where possible: to avoid unnecessary downloading, build in a modular way where possible, reusing predefined components, so that each is downloaded only once and used repeatedly.
- Avoid texture maps if possible: texture maps are data-hungry compared with the geometry of objects, and the computation of the mapping slows the rendering process.
- Test the use of LODs (levels of detail) with various browsers. With LODs, navigation may be stop-go, travelling smoothly while a model is in RAM and then lurching when the next LOD is loaded. However, without LODs, navigation may be slow all the time.
- Though an increasing number of modellers can produce VRML, test models and effects (such as transparency) at an early stage, to avoid wasted effort. Put all models through economisers and parsers to reduce their size and check their validity.
Designing the world - managing the process
The design of a VR world can be, and should be, almost as difficult as real architectural design. Indeed, as modelling and modellers become more sophisticated, their construction will increasingly resemble architecture (though not necessarily visually). For example, as third-party libraries become increasingly available, there will be problems with consistent quality and maintenance. Likewise management issues will become important as model-building is farmed out to large teams, some of them sub-contracted. And issues of control and negotiation will arise as worlds are connected to each other and virtual worlds come to be seen as public spaces.
Virtual Environments Visualisation