Re: Planning under Description Logic ?--an obstacle towards WSAC

On Dec 10, 2004, at 5:41 PM, Manshan Lin wrote:

>> Basically, inference-rules are like operators except that they don't
>> occur in the final plan.
> I suddently find that there is a problem when treating inference-rules
> as operators. If we teat them as operators, they are such special
> operators with the following characteristics:
> (1) As you mention, they don't occur in the final plan
> (2) They must be "executed" whenever their can be apply. It is
> not like other operators, which will never be excuted if you don't
> choose it in your plan.
> This will add unexpected complexity in planning algorithm.

See the SHOP papers. It's not that bad assuming you have the 
expressivity you need in the first place.

It's ofttimes unaesthetic (i.e., in our work we used SHOP's so called 
"bookkeeping" operators which, well, are a hack; I'd ratehr use 
inference rules as inference rules ;))

> Considerting the following situation (sorry for not being able to
> provide a pratical example):
> (1) inference rule: x:ClassA and x:ClassB -> x:Complementof(ClassC)
> (2)
> action-1:
>      Precondition: x:ClassC
>      Effect: x:ClassA
> action-2:
>      Precondition: x:ClassC
>      Effect: x:ClassB
> action-3:
>      Precondition: x:ClassC and x:ClassA
>      Effect: x:ClassD
> (3)
> initial: a: ClassC
> goal: a:ClassB and a:ClassD
> If we don't consider the characteristic (2), the generated plan would
> probably be: action-1, action-2, action-3, while the proper plan
> should be action-1, action-3, action-2. How to take these inference
> rules into consideration when planning is really a problem.

Really? First, I don't see that you've used inference rules as 
operators, but be that as it may. And I'm not sure how you plan to 
interpret the complement of, but be *that* as it may, the sensible 
difference, in this case, between the inference rules and regular 
operators is that the rules aren't *optional* if the  "preconditions" 
hold. So the first plan won't be correct.

In this example, since I don't see any deletions, action three will 
likely have a contradictiory state of the world, so failure.


Received on Saturday, 11 December 2004 13:27:52 UTC