By the way, we're not using strips for this.

Why might that be?

You have to be a little bit more explicit about.

Well, it's hard to model this problem with strips

because it changes while the elevator is acting.

For example, some other passenger arrives

and then the plan has to be redone or something like that.

So the elevator has made some kind of plan

to satisfy all the people and then the CEO is hungry

and wants to go down.

That's not the reason.

But that's a good concern.

You can always think of this problem as a static one.

You have one world state and you plan an action.

You commit to something.

And when the CEO gets hungry, you

have to solve a new problem.

And when, I don't know, B has reached his destination,

you have a new problem.

So you can re-plan every, I don't know, second or so,

which might be confusing to customers, which is probably

what you mean by saying, well, if the new information arrives

and then the elevator says, ah, no, I'm not going to go down.

After all, I'm going to go up and fetch the CEO.

That might cause trouble with people always being in a loop

because the elevator can't make up its mind, things like that.

But really, the main problem is that strips,

which is propositional logic of a particular kind,

is just too inconvenient.

You can imagine what an action description that

takes this into account would look like.

You'll have long sets of free conditions, long sets of add

and delete lists, what happens to the requests, and so on,

things like that.

So there is an evolution of programming, planning

languages, and the kind of current contender

is called PDDL.

I want to show you a little bit of it.

It's not a very new language, but having standards there

is actually good because you can make competitions

and you can actually try out other people's planners,

say for the elevator company.

This is important because their own planner might actually

be outperformed by a new planner coming out from somewhere.

Wouldn't it be nice if we could just basically take

the planner out, replace it by a different one,

and do something?

So we have the planning domain description language,

which is really an extension of strips

with practical extensions that come from applications.