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

On Dec 13, 2004, at 2:14 AM, Drew McDermott wrote:

>> [Manshan Lin]
>
>> If we don't consider the characteristic (2), the generated plan would
>> probably be: action-1, action-2, action-3, while the proper plan=20
>> should be action-1, action-3, action-2. How to take these inference
>> rules into consideration when planning is really a problem.
>
> The only real problem is getting the semantics of actions right.

Ah, actions.

>   Most
> planning research makes some version of the Strips assumption, that it
> is possible to enumerate all the effects of an action.  Actions are
> simulated by changing the representation of a situation, erasing some
> formulas and adding others.  Inference rules (or, more precisely,
> axioms) complicate the picture because they make it harder to
> enumerate the effects, and more difficult to spell out exactly what
> gets erased or added.  The usual solution is to classify predicates as
> "primitive" or "derived."  Only the former can be specified as effects
> of actions.
[snip]

Just to extend this a bit, consider an action that is specified to 
delete a proposition entailed by some axiom. Just deleting the asserted 
version of the proposition (assuming there *is* one) won't help, since 
the system will (should) just infer it. Suppose you had some truth 
maintence information, you could try to delete some crucial aspect of 
the support of the proposition. BUT this is likely to be highly 
non-deterministic. *Which* premise do you delete? Or do you delete one 
of the axioms (with broad effect)? What if it is an axiom that, by 
itself, entails your proposition? What will *actually* happen at 
execution time? How do you delete the support --- by planning for 
another action?

It gets ugly.

In a system with classical negation, you might try adding the negation 
of the to-be-deleted proposition *and* deleting the explicit assertion. 
If the resultant state is contradictory, you should just give up on the 
plan. The planning guys I suggest this too tend to freak a bit :) The 
objections seem to fall into two camps, 1) they don't know how to 
interpret what's going on or 2) they think that such domains are poor 
style, likely to mess thing up. I don't think I agree. 1 usually 
follows from the idea that general axioms are expressing "facts of 
nature" or some other immutable thing so that actions purporting to 
affect derived predicates are just wrong headed. I don't think that's 
right.

Cheers,
Bijan Parsia.

Received on Sunday, 12 December 2004 23:19:03 UTC