Utility Theory in AI

Published March 09, 2010
Advertisement
What is "utility theory". It is related to intrinsic value versus perceived value. Maximization of utility theory is a merit function.

Utility theory is related to game theory and decision theory. By ascribing value to things, we can analyze them. Raw numbers (distance, ammo, health). Value allows comparisons, and by assigning value, we can compare. Utility is not a universal value. It is relative to outside factors (i.e. how much you want pizza depends on how much you've had and how long it's been).

Marginal utility - each additional unity is worth more than the one before. Lego blocks, for example. More legos means more cool stuff you can make.

For realistic behavior, you'll need more than just distance functions to determine the "attack" vs "do not attack" threshold.

Also, thresholds don't need to just be "attack" vs "do not attack".

Utility theory is intended to maximize utility factoring in several variables. This also applies to your supplies. Having low health is more concerning than having low ammo.

Time can be converted into a utility value. Time can make an attractive goal into a less attractive one.

Don't perform all calculations in every frame. Use triggered updates or update every N frames. Or split these calculations off into their own threads.

In utility-based architectures, you continuously analyze all options.
0 likes 0 comments

Comments

Nobody has left a comment. You can be the first!
You must log in to join the conversation.
Don't have a GameDev.net account? Sign up!
Advertisement