|
The field of virtual human autonomous behaviour has become of great interest in
recent years. Games, real time simulation systems, animated films and cinema all
benefit by incorporating thousands of virtual humans, which traverse their environment
automatically. Imagine an animation system that has thousands of virtual agents,
which are moving and interacting with their environment. It is impossible for an
animator to deal with all of these virtual agents in a reasonable amount of time.
Consequently some techniques are required for automatically controlling the behaviour
of virtual humans in a realistic manner. The field of study governing the autonomous
behaviour of virtual humans is partitioned into two major parts: real-time and offline
simulation. Offline systems are frequently adopted by the film industry, where each
frame of the movie can take a considerable amount of time to generate. In contrast,
real-time systems are required to create and display 30 frames every second, for
continuous motion to be perceived. Consequently this places stringent constraints
on the amount of time and processor power required.
|
Figure 1: Real Time Crowd Behaviour |
Usually realistic crowd simulations have to re-plan the behaviour of the virtual
agents for each frame. Planning virtual agent behaviour frequently does not consist
of simple algorithms. Indeed advanced algorithms have to be executed in every animation
step for a realistic simulation. In addition to the behaviour, virtual agents have
to be rendered to the screen.
The quality of a virtual environment simulation depends heavily on rendering
the environment and virtual agents in three dimensions.
Due to the high demand on quality rendering and realistic behaviour algorithms,
real time virtual crowd simulation has become a modern challenge and a field of
study.
The aim of this project is to obtain automatically, realistic and varied behaving
crowd members while keeping the simulation efficiency in real-time limits with as
many virtual agents as possible.
SYS Consulting Ltd.
