Optimising Viewpoint Control in Virtual Environments

Paul Milgram and Wenbi Wang (*)
Dept. Mechanical and Industrial Engineering
University of Toronto
(* also Defence Research and Development Canada, Toronto)


One of the well-known problems associated with navigating through essentially 
any environment, be it real or virtual, is trading off the need for a 
"ground-level" or "out-the-window" viewpoint with the need for a global 
"bird's eye view" of the world being traversed.  Whereas the former is known 
to be optimal for control of low-level locomotion, the latter is necessary for
 acquiring and maintaining optimal global awareness of the environment.  
In real-world situations, such as remote control of manipulators or vehicles,
 or endoscopic surgery, one is constrained by the ability to place and control
 the remote sensors, typically video cameras. With virtual environments, on 
the other hand, one frequently has the ability to manipulate the operator's / 
participant's viewpoint quite flexibly. 

In our talk we shall deal with the challenge of optimising control of the 
viewpoint of a generalised avatar within a (large-scale) virtual environment.
 The discussion will centre on the concept of tethering the virtual camera 
viewpoint to the avatar being controlled. The tether metaphor is very useful, 
as it allows us to incorporate a number of key manual control principles, 
including compensatory versus pursuit tracking and ego-centric versus 
exo-centric viewpoints. In our treatment of the topic, we shall extend the 
tether concept to include dynamic tethering, a modification which is 
completely analogous to the concept of frequency separated displays.

By adding dynamics, using a mass-spring-damper metaphor, between the virtual 
camera viewpoint of the operator / participant and the movements of the 
remotely controlled avatar, we should, in theory, be able to optimise the 
dynamic, task specific parameters of the tether, for all six possible degrees 
of tether freedom (that is, including both translational and rotational 
control), and thereby maximise both local control performance and global 
situational awareness, with respect to such task related parameters as 
forcing function bandwidth. In the talk we shall present our experimental 
results, which illustrate many of the benefits of dynamic tethering and 
confirm the existence of optimal levels of tether damping, rigidity, and 
length, for a generalised navigational control task.