[gecode-users] Cumulative constraint propagation

[Mikael Zayenz Lagerkvist]

Hi,

On 10/4/06, Jérémie Vautard <jeremie.vautard at laposte.net> wrote:
> I saw that the cumulative constraint in Gecode only supports value
> consistency. But does it means that this constraint performs no
> propagation at all and simply checks a total assignment or is there
> still some propagation when only a few variables get instanciated ?

The simple answer is that the propagation done by the cumulatives
propagator is stronger than naive value consitency, but it is also
much weaker than what could be done. But it does reason on bounds.

The detailed answer is that the cumulatives propagator implements the
sweep-algorithm described by Beldiceanu and Carlsson in the paper that
introduced the constraint. It does not, however, implement the task
intervals propagation nor the constructive disjunction on
machine-assignments that was described in the paper.

So, in this particular case, value/bounds consitency is not really the
best classification-scheme.

Cheers,
Mikael

-- 
Mikael Zayenz Lagerkvist, http://www.ict.kth.se/~zayenz/