Automatic generation of constraints for partial symmetry breaking

Constraint Satisfaction Problems (CSPs) are often highly symmetric. Symmetries can give rise to redundant search, since subtrees may be explored which are symmetric to subtrees already explored. To avoid this redundant search, constraint programmers have designed methods, which try to exclude all but one in each equivalence class of solutions. One problem with many of the symmetry breaking methods that eliminate all the symmetry is that they can have a large running overhead. To counter this flaw many CP practitioners have looked for methods that only eliminate a subset of the symmetries, so called partial symmetry breaking methods, but do so in an efficient manner. Partial symmetry breaking methods often work only when the problem is of a certain type. In this paper, we introduce a new method of finding a small set of constraints which provide very efficient partial symmetry breaking. This method works with all problem classes and modelling techniques.