Reducing Symmetry in a Combinatorial Design Problem

Proceedings of CP-AI-OR'01, pp. 351-359, April 2001.
Barbara M Smith

Abstract

A combinatorial design problem is considered which can be modelled as a constraint satisfaction problem in several different ways. The models all have a large number of symmetries which cause difficulties when searching for solutions. Different approaches to reducing the symmetry are discussed: remodelling the problem; adding constraints to the model at the outset; and adding constraints during search to prevent symmetric assignments being explored on backtracking. The most successful strategy for the problem of this paper employs a complex model with less inherent symmetry than the others, combined with symmetry breaking during search.