Thanks a lot Guido/Christian, wonderful tips, i already downloaded the MPG to improve my knowledge on Gecode and in a soon future replace ILOG.
Regards, Jefferson Soares BiernastkiJSK Soluções em Tecnologia Ltda Desenvolvimento - Novas Tecnologias 41 88426214 http://www.onlaboral.com.br http://pixeljef.wordpress.com Subject: Re: [gecode-users] Problem solving simple model From: [email protected] Date: Wed, 7 Sep 2011 14:39:47 +0200 CC: [email protected] To: [email protected] On 6 Sep 2011, at 15:57, Jefferson Soares Biernastki wrote: > Hi people, > > i m working on Gecode for 2 weeks and i model a simple constraint problem > already resolved in ILOG CP. > > Problem: Two trains (X01 and X02) must arrive in final station, each > ocupation must be released, normally 1 minute for releasing. > > the model bellow tries to represent this problem. > > somebody could help me on what i m doing wrong? The reason your code can't solve the problem is the search strategy. ILOG CP uses advanced automatic search heuristics that work well for scheduling problems. The INT_VAR_SIZE_MIN search in your Gecode model simply makes bad decisions early on and can't recover from them (see the huge tree when you run with -mode gist). For your problem, a simple dynamic heuristic (INT_VAR_SIZE_AFC_MIN) together with a restart based search seems to work - at least it does find a solution and can prove optimality, although I'd expect a more specific search strategy to perform better. I attached a modified version of your model, it does the restart search when you invoke it with -search restart. There are a number of small things you can improve in your model, too: * use unary instead of nooverlap (see attached file) * do not initialize the IntVarArrays during cloning (just call update, see attached file) * not store the instance data in the Scheduling class (to avoid copying it during search, where it is not needed) Cheers, Guido -- Guido Tack, http://people.cs.kuleuven.be/~guido.tack/
_______________________________________________ Gecode users mailing list [email protected] https://www.gecode.org/mailman/listinfo/gecode-users
