Chapter 13. Decisions Under Uncertainty—Bayesian Techniques

This chapter introduces Bayesian inference and Bayesian networks and shows how you can use these techniques in games. Specifically, we’ll show you how to use these techniques to enable nonplayer characters (NPCs) to make decisions when the states of the game world are uncertain. We’ll also show you how simple Bayesian models enable your computer-controlled characters to adapt to changing situations. We’ll make heavy use of probability, so if that subject is not fresh in your mind, you might want to read Chapter 12 first and then come back to this chapter.

Before getting into the details of Bayesian networks, let’s discuss a hypothetical example. Suppose you’re writing a role-playing game in which you enable players to store valuables in chests located around the game world. Players can use these chests to store whatever they want, but they run the risk of NPCs looting the chests. To deter looting, players can trap the chests if they have the skill and materials to set such traps. Now, as a game developer, you’re faced with the issue of how to code NPC thieves for them to decide whether to open a given chest that they discover.

One option is to have NPCs always attempt to open any given chest. Although simple to implement, this option is not so interesting and has some undesirable consequences. First, having the NPCs always open the chest defeats the purpose of players trapping chests as a deterrent. Second, ...

Get AI for Game Developers now with the O’Reilly learning platform.

O’Reilly members experience live online training, plus books, videos, and digital content from nearly 200 publishers.