The most long-standing virtual humans have been built as accompaniments to computer games. The similarities between these humanoids and the ones built to inhabit virtual worlds end skin deep. In the case, of computer games, the virtual humans need only convey a few traits and have little functionality. However, today's virtual beings have to be able to communicate with users in a convincing fashion. The problem is that human communication remains an immensely complicated function to model computationally.

"Playing chess is easy, but making breakfast is enormously complicated.
This complexity stares us in the face every morning, yet it is invisible."
-- Phil Agre, 1985

RESEARCH FACT FILE ON UCL: The Virtual Environments and Computer Graphics group at University College London (UCL) have been undertaking engrossing research in the design of virtual humanoids, in conjunction with our psychology departments and other universities around the globe. Most of the research at UCL focuses on making our virtual collaborators more realistic and believable. Virtual humans at UCL can be used as an interface to represent us while visiting virtual environments (VEs) or they can be built as a self-sufficient sentient to co-inhabit with users in VEs. This involves not only designing and building physically believable virtual humanoids but also embodying them with behaviour traits to enable them to interact with other users in the virtual environments.


Fig 1: A VE user interacting with an avatar in a shared IVE in the ReaCTor at UCL.


Fig 2: A user in a HMD interacting in the same-shared IVE as Fig 1.

Some works hypothesise that behaviour realism is more important to aid believability. However, recent works at UCL into eye-gaze behaviour in avatars suggest that there is a correlation between physical and behavioural realism. In fact it was found that embodying a photo-simplistic with complex inferred behaviour was detrimental to the levels of believability and face-to-face effectiveness experienced in the shared virtual environment. Users of VEs expect a behaviour model consistent the physical appearance of the virtual human in order to believe that the virtual human is real and suspend their sense of disbelief.

AGENTS EMPLOYED AT UCL: Existing virtual humans at UCL vary in physical appearance. They have been built over recent years to be incorporated into various projects and are available for use in future projects.

The avatar shown in Fig 4 is a photo-simplistic avatar with added features. The avatar was designed with pronounced eyes and with less emphasis laid on the other facial features. Experiments are carried out at UCL to investigate if people in a VE react to the existing agents as they would in real life to positive, neutral or negative experiences. The results up to date are very promising and showed that their experiences were very similar to real life situations.


Fig 3: The most simplistic avatar available at UCL. It was used to assess Internet2 and the theory of emotional conveyance through a network. This work was carried out in conjunction with the University of North Carolina, Chapel Hill and Massachusetts Institute of Technology.


Fig 4: A photo-simplistic avatar with realistic eye-gaze behaviour and minimal limb animations. This was one of the avatars used in our most recent work on behaviour realism pertaining to eye-gaze.

Fig 5: Relatively more photo-realistic and gendered virtual humans available at UCL. These were used in the experiments dealing with behaviour realism, fear of public speaking, virtual acting, and social paranoia.

Research into "better looking" virtual humans is ongoing. With the current rate of advancements in the rendering of virtual objects, better avatars and agents are constantly emerging. Needs of a virtual humanoid other than good looks include:
* Consistent photo realistic appearance
* Corresponding behaviour realism
* Crowd behaviour
* Emotional and Personality Identikits
* Skinning and Clothing
* Usability across various types of virtual worlds ...

UCL undertakes work in all these areas of research.

For more information on the work done with avatars at UCL, go to our group website "VR at UCL" or try the papers section at our equator site [Papers are in PDF].