From Linear Relaxations to Global Constraint Propagation

We present a method for propagating linear constraints. Our technique exploits the fact that the interior point method converges on a central point of the polytope. A variable assigned to an extreme point is therefore assigned to this extreme point in all solutions. We show how linear relaxations and the interior point method can be combined to prune variable domains. We also describe a class of constraints where our propagator enforces global arc consistency.

• Type: Conference
• Year: 2005
• Where: CP
• Authors: Claude-Guy Quimper, Alejandro López-Ortiz
• Keywords: CP 2005, Extreme Point, Interior Point Method, Linear Constraints, Programming Languages