Behavioral Animation - Computer Animation: Algorithms and Techniques - page 342

in direct opposition to the collision avoidance urge. InReynolds' flockmodel, a boid stays part of the flock
by a flock centering force [ 34 ] . But as he points out, a global flock centering force does not work well in
practice because it prohibits flock splitting, suchas that oftenobservedwhena flockpasses aroundanobsta-
cle. Flock centering should be a localized tendency so that members in the middle of a flock will stay that
way and members on the border of a flock will have a tendency to veer toward their neighbors on one side.
Localizing the flocking behavior also reduces the order of complexity of the controlling procedure.
Another force that is useful in controlling the member's reaction to movements of the flock is veloc-
ity matching , whereby a flock member has an urge to match its own velocity with that of its immediate
neighbors. By keeping its motion relative to nearby members small, velocity matching helps a flock
member avoid collision with other members while keeping it close to other members of the flock.
Local control
Controlling the behavior of a flock member with strictly local behavior rules is not only computation-
ally desirable, it also seems to be intuitively the way that flocks operate in the real world. Thus, the
objective is for control to be as local as possible, with no reference to global conditions of the flock
or environment. There are three processes that might be modeled: physics , perception , and reasoning
and reaction . The physics modeled is similar to that described in particle systems: gravity, collision
detection, and collision response. Perception concerns the information about the environment to which
the flock member has access. Reasoning and reaction are incorporated into the module that negotiates
among the various demands produced as a result of the perception.
Perception
The main distinction between flocking and particle systems is the modeling of perception and the sub-
sequent use of the resulting information to drive the reaction and reasoning processes. When one local-
izes the control of the members, a localized area of perception is modeled. Usually the “sight” of the
member is restricted to just those members around it, or further to just those members generally in front
of it. The position of the member can be made more precise to generate better defined arrangements of
the flock members. For example, if a flock member always stays at a 45 angle, to the side and behind,
to adjacent members, then a tendency to form a diamond pattern results. If a flock member always stays
behind and to the side of one member with no members on the other side, then a V pattern can be
formed. Speed can also affect perception; the field of view (fov) can extend forward and be slimmer
in width as speed increases. To affect a localized fov, a boid should do the following:
Be aware of itself and two or three of its neighbors
Be aware of what is in front of it and have a limited fov
Have a distance-limited fov
Be influenced by objects within the line of sight
Be influenced by objects based on distance and size (angle subtended in the fov)
Be affected by things using an inverse distance-squared or distance-cubed weighting function
Have a general migratory urge but no global objective
Not follow a designated leader
Not have knowledge about a global flock center