Operations Research Techniques in Constraint Programming
Willem-Jan van Hoeve
Abstract:
In this thesis we investigate the application of efficient operations research
techniques in constraint programming for solving NP-complete combinatorial
problems. Such problems often occur in practice, for example in logistics.
In constraint programming, combinatorial problems are being solved by
systematic search in combination with so-called propagation. Propagation
reduces the search space, based on the constraints of the problem. We have
applied operations research techniques during search as well as propagation.
In particular, we present a generic method to apply propagation in case the
problem is over-constrained. In such cases we minimize some measure of
violation for each constraint. Our method uses graph-theoretic techniques
for network flows.
We also apply linear and semidefinite relaxations of the problem to speed
up the search process. The relaxations are used to find a solution faster
and to prove optimality of such solution earlier. Experimental results
show the applicability and effectiveness of our method.
Keywords: Operations Research, Constraint Programming