Game Development Reference
In-Depth Information
it in the ways we have defined, and arrive at a single score do we come to the most
logical answer:
Use our rocket launcher to kill Evil Genius.
Because there are eight enemies and four weapon choices for each, there are 32
entries in our target list. Rather than clutter things up, we will only list the best
weapon for each of the eight targets. (It's not much of a surprise that it's usually the
rocket launcher, is it?)
Name
Weapon
Threat
Urgency
Score
Baddie 1
R/L
14.1
0.5218
27.0
Baddie 2
R/L
4.0
0.3101
13.1
Baddie 3
R/L
2.9
0.2992
9.7
Baddie 4
R/L
250.7
0.3950
634.7
Evil Genius
R/L
3.2
0.4229
7.7
Evil Knievel
M/G
8.0
0.5626
14.3
Evilmeister
R/L
3.3
0.3445
9.8
Boss Man
R/L
2.6
0.3101
8.5
As we ponder the stats, we begin to see why Evil Genius didn't immediately
attract our attention. He ranks third on threat ratio (remember, lowest is most
threatening) and third on urgency (higher is more urgent). If we had looked only
at the disconnected data such as distance to us, distance to the detonator, weapon
strength, and so on, we would not have detected the combination of information
that makes Evil Genius the highest-priority target. (Rather ingenious of him, in an
evil sort of way… don't you think?)
While it seems like it was a long road to accomplish the decision of which Dude to
assault with which weapon, by breaking it down into individual steps, we were able
to simplify the
entire
process.
One of the important decisions
we
made about constructing our
agent's
deci-
sion is what the decision was going to entail. As we theorized early in the chapter
and subsequently confirmed in our last example, we could not have separated the
two individual choices of who to attack and what weapon to use. Doing so may
have led us to less-than-optimum results. This is reminiscent of the Prisoner's
