Hi to you all dear Oz users,
I am tackling now a scheduling problem known as "flexible flow shop" where,
very similar to the "job shop" problem (jobs with tasks on known machines)
but characterized by the possibility to process every task on equivalent
(possibly not equal) machines.
So, if we espress a task record in a job-shop problem for example:
task(id:'drill' job:'Art01' res:'DrillMach1' start:S{10#200} dur:45)
the same in a flex-shop could be expressed as:
task(id:'drill' job:'Art01' res:R{3 5 11 12} start:S dur:45)
where the resource R could assume one of the 4 possible values in the group
of the (say equivalent) "Drilling Machines" .
Actually, my approximated solution is to create a "resource distributor"
which can choose among the alternatives knowing the "cost" of the resource,
in this case the "load".
After this first distribution all the resources for the tasks are known and
the script can continue with the classical distribution on S and
"deep-first" search .
But, after a while, it comes clear that this could be only a rough
approximation due to the calculation of the cost function, that,in the case
, is only global, so you can have a resource, with an heavy global load but
with local holes where tasks could accomodate (with the obvious objective to
minimize the makespan on the tardiness of the job).
After the reading of the dr.Shulte's "Programming Constraint Services" , I
would like to thank for his clear and elegant exposition of Oz insights, it
comes evident to me tha the solution could be to build a "Best-Solution"
search engine where every choice is cost-driven.
But, and here is the question probably directed to dr.Shulte, how can I
express a global cost (say the job makespan) that could be precisely known
only at the end of the search ?.
Thanks for your patience,
Adriano Volpones
-
Please send submissions to users@mozart-oz.org
and administriva mail to users-request@mozart-oz.org.
The Mozart Oz web site is at http://www.mozart-oz.org/.
Please send bug reports to bugs@mozart-oz.org.