I was looking at a problem in the early chapters of Artificial Intelligence: A Modern Approach. It’s called Vacuum World. This is a very simple agent problem consisting of a vacuum placed on a grid. The grid has only two squares, each of which is either dirty or clean. The vacuum has three actions: move left, move right, and suck.
I wrote a simple implementation of this in Java (because the course I’m taking uses Java), starting out something like this:
public class Vacuum {
private GridSquare location;
public void moveLeft() { ... };
public void moveRight() { ... };
public void suck() { ... };
}
Obvious, beginner’s OOP, right? But then I start to second-guess myself. Is location really a property of the vacuum? Or is it a property of the vacuum in its environment? That is, does the location variable belong in the Vacuum class, in the Grid class (the environment), or in some “world” object that encompasses both the vacuum and the grid? Should Vacuum even be allowed to have a reference directly to its location, or should it query the environment indirectly, as in “Is the square I am currently occupying (wherever it may be) dirty?”
The example code on the A.I.M.A. web site seems to go out of its way to be inscrutably OO, with the behavior of the vacuum divided among half a dozen abstract “agent” classes. I think the vacuum there stores its location as an integer, but I couldn’t swear it to you.