Yeah

good.

So welcome to everybody who has made it to the university

even with the news now.

We were talking about planning.

Planning is essentially logic-based world descriptions with a special trick for change.

So instead of representing time and change by having kind of an explicit representation

of change or time

we actually represent change by equipping all of our actions

the agents'

actions with a way of

in a way

down-dating the state.

Not only do we have an action say

what are the new facts

the things that I've achieved

by this action

but also which of them have I destroyed by this action.

Okay

so what we're seeing

we call that a strips task that is essentially a search problem

right?

We have a couple of states, only that it's logic-based.

We're using a logic to describe the world state.

In this particular case

we're using propositional logic or

in practical applications

PLNQ.

The actions are represented as triplets and set of things

a set of facts.

Those are the things that make up the state that are added.

A couple of facts that are deleted if they're there.

And a couple of preconditions, which is very useful.

Remember

we have always the problem that certain actions are only applicable to certain states.

Since states are sets of facts

we can just say

oh

these facts must be present.

And there are a couple of side conditions that are kind of obvious.

You don't want to add a fact that you delete at the same time, right?

Doesn't make any sense.

And then the transition model of that search problem is very simple.

Applying an action in a state S gives you a new state that is the old state

all the

facts you had before

plus the ones you add

minus the ones that you delete.

And it's only applicable if the preconditions are a subset of the state.

Otherwise undefined.