So after this introduction of what may go wrong in the world, I would like to introduce

you to an example we are going to use in the remainder of this chapter.

We call it the furniture coloring example.

So the idea is that you have furniture, a chair and a table and you want to color them

in matching colors.

And now we set the whole thing up so that it's partially observable or possibly even

non-deterministic the environment.

So we have an initial state where we have two cans of paint of unknown color.

So the color of the furniture is also unknown to us.

And only the table is in the agent's field of view, as you can see on the picture up

there.

Now, we have a couple of actions.

One is to remove the lid of the can.

You can imagine that only once you have opened the can, can you actually see the color of

the paint in that.

And once you've opened it, you can paint the furniture with the paint from an open can.

Okay, very simple.

Let's see how we can formalize this.

We're going to use PDDL as introduced above.

So we have a domain file where we describe the domain.

So the main furniture coloring has a couple of predicates.

We can say what it is to be an object and a can and whether something is in view.

We're going to model percepts later.

And we can ask ourselves is X of color Y?

So this is very simple.

Of course, the actions and, by the way, the percepts will actually come in a minute.

So the problem file describes our setting.

So we have, of course, the domain we just introduced.

We have a table and a chair and two cans, can one and can two.

And in the initial state, we have tables and chairs being objects.

C1 to C2 are cans.

And we're looking at the table.

The goal is to color the table and the chair in some joint color C.

Note that this C here is a variable.

So we don't care what the color is.

It just should be the same.

So here are two actions, these remove lid and paint.

So they both have arguments.

And the first time X, which is a can, which is the precondition and the effect is that

the can is open, obvious.

And you can paint if you have an X with the color from Y.

So therefore, for that, X must be an object.

Y must be a can.

The color in X is C and X must be open.

And the effect is that the color of X is actually C.

So it's the color of the color in the can is C. It's afterwards also C.

So we have a universal variable here in this with the color.

Again, contradiction to the existential reading in the goal.

So a sensuous plan obviously is we can open one can, paint the chair and the table in

whatever color we find in this can.