[rules-users] Planner: What is a good example/algo to chain entities?

[Geoffrey De Smet]

Op 08-02-13 09:20, Michiel Vermandel schreef:
> Hi,
>
> Any one got an idea?
>
> Thanks
> -----------------
> http://www.codessentials.com - Your essential software, for free!
> Follow us at http://twitter.com/#!/Codessentials
> ------------------------------------------------------------------------
> *From:* Michiel Vermandel <mvermand at yahoo.com>
> *To:* Rules lists.jboss.org <rules-users at lists.jboss.org>
> *Sent:* Wednesday, February 6, 2013 8:44 AM
> *Subject:* [rules-users] Planner: What is a good example/algo to chain 
> entities?
>
> Hi,
>
> I need to keep some Planning entities (Tasks) together and I am 
> looking for some insights/ideas/examples in how to tackle this.
> Tasks are grouped by a groupId (Groups are problem facts).
> The Tasks do not have to be grouped like a pillar but more like an 
> accordion.
> One of the planning variables is "Period", a Period typically is 3 
> weeks and we have 12 Periods to plan the tasks into.
>
> The requirement is that all tasks within  a group need to be kept 
> together in such a way that:
> - the spread (nr of periods) of the tasks is minimal
> - the spread is never more than the sum of the duration of the 
> separate tasks in the group.
>
> So given a group of 3 tasks:
>
> 1)This is good:
>
> Period 12345789ABC
> T1:      xxx
> T2:         xx
> T3:           x
>
> 2)This is better:
>
> Period 12345789ABC
> T1:      xxx
> T2:       xx
> T3:         x
>
> 3)This is (one of) the best:
>
> Period 12345789ABC
> T1:      xxx
> T2:       xx
> T3:        x
>
> 4)This is bad:
>
> Period 12345789ABC
> T1:      xxx
> T2:       xx
> T3:          x
Is 4) worse than 1)?
>
> Of course 3 is favorite, but maybe not possible due to other planning 
> variables (-> availability of people to execute the tasks).
> I need to plan a few hundred of such groups of tasks which all share 
> the same Period space and list of available people.
> (A person can only do one task per period)
>
> I have got the DRL rule that keeps the tasks together (rule counts the 
> gaps) , but giving penalty for making gaps (by shifting a single task) 
> makes it hard to find a good solution.
> What would be the best approach to solve this problem?
>
> - A combination of custom mover and swapper which moves/swaps groups 
> instead of single tasks?
> - The above in combination with a group shuffle factory (swapping 
> tasks in a single group)?
Not sure what you mean with these last 2 questions,
but course grained moves are probably a good way to get out this situations.
For example: move all T's in the same period (or that share a period).
<pillarSwapMoveSelector> does this if the value is exactly the same 
(you'll need variableNameInclude since 6.0.0.Alpha7 if you have multiple 
vars...)
Otherwise, take a look at BedDesignationPillarPartSwapMoveFactory for 
inspiration (it's pretty different, but interesting to learn).
>
> Other ideas?
>
>
> Thanks a lot!
> -----------------
> http://www.codessentials.com - Your essential software, for free!
> Follow us at http://twitter.com/#!/Codessentials
>
> _______________________________________________
> rules-users mailing list
> rules-users at lists.jboss.org <mailto:rules-users at lists.jboss.org>
> https://lists.jboss.org/mailman/listinfo/rules-users
>
>
>
> _______________________________________________
> rules-users mailing list
> rules-users at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/rules-users/attachments/20130211/028f5d38/attachment-0001.html