Game AI Part 2: Problem Spaces
This is the second part in an series of posts on Game AI. The first post can be found here
When building AI agents that solve complex problems or play games, it helps to consider the problem in this way:
- Consider the starting conditions of the puzzle or game as an Initial State
- Naturally, your goal is to get the game in a certain state (or any such state that satisfies certain conditions). We’ll call these Winning States
- Now consider a wide open space, where every point represents a game state. (You can also see this as a maze of states). We’ll call this the Problem space
- So now solving the problem equates getting from the initial state to a winning state. Sometimes this will require you to FIND the winning state, and sometimes you’d know the exact winning state, but you’d have to figure out the path from an initial state to the winning state. This process is called the Search
...