Re: Homework 4

[Subbarao Kambhampati <rao@asu.edu>]

At 11:31 PM 11/19/2001 -0700, you wrote:
>Hi Rao,
>
>I just had a couple of questions for part 1 of homework 4.  If you could
>clarify these for me (or tell me if there's a place in the book to find the
>answers) I would appreciate it.
>
>* In question 3 (making the planning graph), can action O4 be added in the A1
>stage, where its preconditions (R & S) are mutex?  In general, can an action
>be added to the planning graph in a stage where any/all of its preconditions
>are mutex?

There is no point in adding such actions...

>* For question 4, which heuristic do we use for estimating state
>costs? (set-level without mutex?)

The best heuristic you have--set-level with mutex..


>* In converting the graph plan to a CSP, should the domain of a variable
>contain "#" if it only exists due to an actual action (not a propogation)?  I
>think I may be confused about the meaning of "#" in the CSP conversion (I
>cannot find where I wrote down your explanation).  My question arises from
>your example slide of this conversion where h-A-1 contains "#" in its domain,
>but first appears in the plan graph only due to the action Pick-A-1 (not a
>propogation).  Please clarify before I confuse myself further ;).

No--all prop variables MUST have # value as one of the values--having # 
means we dont need ANY action to support that proposition (either real 
action or a persist action)


>* Could you also briefly explain action constraints in the CSP conversion
>(this ties into my confusion of "#").  I understand the mutex constraints.
>
>Also, is there a section in the book about the conversion of plan graphs to
>CSPs?  I could not find it to clarify these questions.

Unfortunately, no. But I discussed this in the class at length. I will be 
happy to answer questions in person


rao