Game Development Reference
In-Depth Information
To do this, we judge the weight (now being used figuratively) of new informa-
tion not only against the weight of the previous information, but against that
weight times the velocity that the action is already traveling. If we already have
committed heavily (see how these weight words work so well?) to a decision, the
momentum of the system will be fairly significant. If information of equal weight
to what we already used is placed in front of our decision, the idea of decision mo-
mentum tells us that we would continue on in the same direction. However, if we
encounter new information that has
greater
weight than what we originally had
considered, the pace of our overall decision system will slow down somewhat. The
greater the difference in the weight of the information, the greater the effect it has
on our decision. Eventually, if the new information becomes massive enough, it will
act as the bowling ball above did, and our decision will reverse in the face of this im-
posing new mass of data.
Keeping this in mind, rather than focusing on stopping our decisions from
strobing
after the fact
, we can ensure that enough of the relevant information is
available for the agent to take into account so that the problem doesn't occur in the
first place. Often, by making the agent a little more aware of information and a lit-
tle more mindful of ramifications, we are giving it more “mass� and, therefore,
more decision momentum. Rather than changing a decision on the fluctuations of
the moment, our agents can persevere with their chosen course of action.
Time Commitment
In the Flotilla of Futility example above, there was one more piece of information
that, if the AI had considered it, may have made the difference in its decision:
time
.
In Chapter 7, we spent some time pondering the utility of time as a component
to consider in decisions. We discussed time
spent
, time
saved
, time
wasted
, and
other such ways that we can view chronological matters. In one example, we set out
a scenario in which we were deciding whether we wanted to accomplish goal A be-
fore goal B or vice versa. We illustrated how the decision depended on factors such
as the relative value of the goals and the time it would take to travel to them. We
complicated the situation a little more by introducing decay in the value of the goals
over time. By doing so, we caused the time we spent traveling to the goals to be even
more influential in the final decision calculation. All of those factors are transfer-
able to the example of the aimless armada.
In the example above, the AI calculated the relative utility of attacking the two
cities (and presumably others as well). The AI would have to weigh various pros
and cons about his own fleet, the defenders, and so on. This is, of course, similar to
the Dude Assault example from Chapter 14. This is where the calculation stopped,
however. Unlike our decision regarding which Dude to fire a single shot at, there is
