Say you're in a real world situation.

You have a car and you have a flat tire.

Okay? What do you do?

Of course, you have a couple of actions in the real world.

You have a setup which we'll call start,

which is where you have a spare tire.

That's not flat, which is good to know.

The spare tire is also intact, very important.

It's not on the wheel,

right? Because it doesn't help if it's already on one of the other wheels.

You have another tire, which is tire one, this guy here, which is also flat.

Now you can do...

You want to have a goal state, which essentially has a car that can drive,

namely every X is on its wheel, every tire is on its wheel,

and they're all not flat.

Okay?

Say you have a remove action,

which you have to have a mounted tire,

and then afterwards is unmounted,

and then afterwards you have the hub, which is clear.

Then you can put a tire onto something,

and then you can inflate a tire.

That's a planning problem.

Now that has problems.

We don't have all the knowledge,

or some of our knowledge might actually be wrong.

This is what I always think of the situation when I start my car.

Okay?

I'm pretty sure that my tire in the back is not flat.

Do I know that?

No.

I haven't...

Well, actually, I should probably change the examples

because my car doesn't even have one.

But most people never look at their spare tire,

and it's usually kind of flat, half flat.

So what do we do if things go wrong?

Right?

Another thing is that the remove action might actually fail.

I've had that.

Try to open the nuts.

They're just too tight.

Can't do it.

Right?

Or if you have aluminum wheels, then even if you have all the nuts down,

it might actually be corroded to me as well.

What do you do?

That's really what we want to look at, namely,

things can go wrong in the real world.

Right?