Re: [eclipse-clp-users] bb_min problem

From: Stephan Schiffel <stephan.schiffel_at_googlemail.com>
Date: Mon, 4 Apr 2011 17:03:08 +0000
Hi Bogdan,

On Monday 04 April 2011 15:25:55 Bogdan Tanasa wrote:
> I have a piece of CLP which is designed to do some kind of bin packing. If
> I use bb_min with continue strategy the same solution is printed several
> times. Can someone please have a look on my code am give me some hints ?

This seems like an awful lot of uncommented code to go through. Without 
looking at it, I suspect the following:
In your input you have several signals with the same parameters, e.g.,
signal(7640, 8000, 8000, 64).
signal(7640, 8000, 8000, 64).
signal(7640, 8000, 8000, 64).

When setting up the constraints you probably treat these as different signals 
and, because they have identical parameters, they can be freely exchanged in 
the solution of your problem (whatever the problem is).

As far as I know, there are two ways to remove those symmetric solutions:

1. Add additional constraints that break the symmetry. E.g., if you know that 
signals 1 and 2 are equal, add a constraint that ensures that signal 1 is 
always used before signal 2.

2. Reformulate the problem variables and constraints such that they do not 
contain symmetries. E.g., instead of a list of signals (that may contain 
duplicates) use a set of pairs Signal-Number that specifies for each signal how 
often it occurs.

Best regards,
Stephan
Received on Mon Apr 04 2011 - 17:03:20 CEST

This archive was generated by hypermail 2.2.0 : Thu Feb 02 2012 - 02:31:58 CET