next up previous contents
Next: Persistence in Targeting Up: Persistence in Behavior Previous: Behavior Memory

Inhibitory Gain and Fatigue

To further ensure persistence in a behavior that has been interrupted and then resumed, we adopt the concept of inhibitory gain from ethology [Ludlow1980]. Inhibitory gain was first proposed by Ludlow to model the control mechanism in animals that balances persistence and opportunism. It is elicited by the winning behavior and is a negative value imposed on the activation energy (strength of desire) of all the competing behaviors. This helps to keep the activation of the winning behavior relatively high and hence induces persistence in that behavior. The higher the value of the inhibitory gain, the more persistent the behavior that releases it. In our case, since behaviors are layered in a hierarchy by their intrinsic priorities, the number of competing behaviors is reduced to those in the same layer rather than to all of the other behaviors had a distributed structure been used. The inhibitory gain tex2html_wrap_inline3600 (where tex2html_wrap_inline3280 is the activation value of the active behavior) is added to the competing behaviors only when the current behavior is one resumed from the behavior memory. Let us take the same example as above. When the mating behavior is resumed from the obstacle avoidance behavior, tex2html_wrap_inline3604 is added to H to help maintain the persistence/continuity of the mating behavior by inhibiting further interruptions, say, from opportunistic feeding.

Excessive persistence in behavior can be undesirable in certain situations. For example, if we do not put food in the virtual marine world, a hungry fish would not engage in any behavior (in the same or lower priority layer) other than foraging. This can be avoided by modeling fatigue in behavior. Fatigue is also important for behavior time-sharing [McFarland1993a], in which low priority activities are given a chance to execute despite the presence of a higher priority activity. We model fatigue of a behavior by simply multiplying the corresponding activation value of that behavior with a decay function of time. A similar model of inhibitory gain and fatigue was proposed by Blumberg Blumberg94.


next up previous contents
Next: Persistence in Targeting Up: Persistence in Behavior Previous: Behavior Memory
Xiaoyuan TuJanuary 1996