Even when an intention has been decided, it remains to determine how the intention should be fulfilled through the successive selection of motor actions. The selected sequence of actions, hence the behavioral outcome, depends on the dynamics of the 3D virtual world as well as the dynamics of the multiple desires of the fish. The latter manifests its effect through the motor preferences and hence, although there is only one intention or winning behavior at any instant in time, other behaviors can nonetheless influence the creature's actions.
We can further analyze the control at this level at two stages--the target stage where the target, i.e. the object of interest given the current intention, is chosen and the motor command stage where the final, detailed motor command is calculated. The target stage control is governed by the focusser and the motor command stage control is governed by the behavior routines.
The control process at the target stage can be viewed as the
`competition' amongst the set of potential targets to become the
current target. This is particularly obvious in the cases described in
Section 
, where the fish tries to decide to which
food morsel or mate it should attend. Since the desirability of each
potential target may change over time (due to the dynamics of the
virtual world and the motor preferences), the target may change over
time. Consider the case when the fish intends to eat and there are
multiple available food sources; it is possible that a fish may
perform an interesting zigzagging behavior due to the competition
between potential targets.
Once the current target is chosen, control is then passed to the
active behavior routine. The final choice of a motor controller and
the appropriate control parameters are computed based on the current
state of the fish and the target, as well as the values of the motor
preferences (see Section for examples and
details).
| Xiaoyuan Tu | January 1996 |