Optimize for two costs

[sachi...@gmail.com]

Hello,

I apologize, this is not a Fast-Downward problem.

My domain has three actions pay-for-object, move-object-a, move-object-b.

I have defined them as below.

(: action pay-for-object

:parameters ( ... )

:precond (... )

:effect (..... (increase (total-cost) 1 ) )

)

(: action move-object-A

:parameters ( ... )

:precond (... )

:effect (..... (increase (total-cost) 1 ) )

)

(: action move-object-B

:parameters ( ... )

:precond (... )

:effect (..... (increase (total-cost) 1 ) )

)

I want to find solutions such that the cost of the plan is minimum with respect to the number of moves and also number of objects.  In my problem definition I use (:metric minimize (total-cost). I am using Fast-Downward with LM-cut heuristic to find minimum cost solutions.

Would it be possible to know if I have the right definition and planner configuration to do this?

Thank you very much for your time

[Florian Pommerening]

Dear sachini,

can you be more specific with what you mean by "minimum with
respect to the number of moves *and also* number of objects". For
example if one solution uses 5 moves and 2 objects, and another solution
uses 3 moves and 5 objects, which one do you want? In the simple case
where you want to minimize the sum of moves and objects, your approach
would work: give both actions a cost of 1 and run an optimal
configuration (like "astar(lmcut())"). That will minimize the cost, so
it will find a plan with a minimal number of those actions.

If you want a weighted sum (like 5*#objects + 3*#moves), you can adjust
the action costs accordingly.

If you want something more fancy that considers both cost functions
independently (like finding many pareto-optimal solutions) things get
more tricky and basic Fast Downward will probably not be enough. I'm not
too familiar with the research on this but you can probably find help on
this here: theplanningcommunity.slack.com/


Cheers
Florian

On 02.10.20 23:53, sachi...@gmail.com wrote:
> Hello,
>
> I apologize, this is not a Fast-Downward problem.
> My domain has three actions pay-for-object, move-object-a, move-object-b.
> I have defined them as below.
>
> (: action pay-for-object
> :parameters ( ... )
> :precond (... )
> :effect (..... (increase (total-cost) 1 ) )
> )
> (: action move-object-A
> :parameters ( ... )
> :precond (... )
> :effect (..... (increase (total-cost) 1 ) )
> )
> (: action move-object-B
> :parameters ( ... )
> :precond (... )
> :effect (..... (increase (total-cost) 1 ) )
> )
>
> I want to find solutions such that the cost of the plan is minimum with

> respect to the number of moves *and also* number of objects.  In my

> problem definition I use (:metric minimize (total-cost). I am using
> Fast-Downward with LM-cut heuristic to find minimum cost solutions.
>
> Would it be possible to know if I have the right definition and planner
> configuration to do this?
>
> Thank you very much for your time
>
>
>
>

> --
> You received this message because you are subscribed to the Google
> Groups "Fast Downward" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to fast-downwar...@googlegroups.com
> <mailto:fast-downwar...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/fast-downward/caae283e-802d-4c4b-96a8-db520e0982e8n%40googlegroups.com
> <https://groups.google.com/d/msgid/fast-downward/caae283e-802d-4c4b-96a8-db520e0982e8n%40googlegroups.com?utm_medium=email&utm_source=footer>.