The intention generator of a prey fish is a specialization of the
generic intention generator of Fig. 
as
shown in Fig. . The two characteristic behaviors
of prey fish are schooling and evading predators. We briefly present
the implementation of the two behaviors.
Figure: Portion of intention generator of prey.
Schooling is a complex behavior where all the fishes swim in
generally the same direction (see Fig. ). Each
fish constantly adjusts its speed and direction to match those of
other members of the school. They maintain a certain distance from one
another, roughly one body length from neighbors on average
[Wilson and Wilson1985]. Each member of a school of artificial fish acts
autonomously, and the schooling behavior is achieved through sensory
perception and locomotion [Reynolds1987]. An inceptive school is
formed when a few fish swim towards a lead fish (see
Fig. ). Once a fish is in some proximity to some
other schooling fish, the `schooling' behavior routine outlined in
Fig. is invoked.
Figure: Schooling behavior routine.
The intention generator ensures that the fish do not get too close
together, because the avoid collision intention has highest
precedence. To create more compact schools, the collision sensitivity
region of a schooling fish is decreased once the fish gets into
formation. When a large school encounters an obstacle, the autonomous
obstacle avoidance behavior of individual fishes may cause the school
to split into two groups (obstacle avoidance behavior has higher
priority than the schooling behavior). Once the obstacle is cleared,
the behavior memory of individual fishes ensures that the
schooling behavior routine regains control and hence the school
rejoins (Fig. (a)-(d)).
Similar to the flocking of boids [Reynolds1987], the schooling
behavior of artificial fish is achieved through the interaction of
two behaviors--to stay in the neighborhood of other schooling fish
(by using chasing-target) and to avoid being too close to any of
them (by invoking avoiding-fish). Each fish acts independently
based on its perception of its neighborhood. One can adjust the
density of a school by adjusting the collision sensitivity regions of
the schooling fish (Fig. (b)). For instance,
increasing the sensitivity region can create more loosely coupled
schools. In our implementation, there is a designated `leader fish'
that swims in front of all other fish in the school. The general path
that the school takes is determined by the leader fish. Of course, the
leader fish could also be decided dynamically at run time, i.e.
whichever fish is frontmost will become the leader.
Figure: A small school of angelfish.
Figure: An inceptive school.
When a predator gets too close to any fish the escaping
behavior is triggered. This examines the relative position between the
fish and the predator against a set of rules. Each rule suggests a
proper action in a certain situation. For example, If predator is
closely behind and swims in roughly the same direction and is to the
right of fish, then turn to the left. One may consider all these
rules as the specifications of one general rule, that is, to swim away
from the predator as fast as possible. Rules are interpreted as detailed
geometrical/mathematical conditions, and their quantities determine
the values of motor control parameters such as the angle of a turn.
Currently there are eleven rules and the resulting behavior looks
natural. Fig. and show two
instances of a small school of prey scattering in terror when a
predator comes close. Note how each of the prey fish tries
independently to flee from the pursuing predator.
Figure: School scatters in terror.
Figure: Fleeing from predator.
The eleven rules based on which a prey fish flees from a predator are listed below (note ``me'' or ``mine'' indicate the fleeing prey fish): if predator is far:
if predator is close:
| Xiaoyuan Tu | January 1996 |
