Symmetry Breaking Constraints: Recent Results

Symmetry is an important problem in many combinatorial problems. One way of dealing with symmetry is to add constraints that eliminate symmetric solutions. We survey recent results in this area, focusing especially on two common and useful cases: symmetry breaking constraints for row and column symmetry, and symmetry breaking constraints for eliminating value symmetry

💡 Research Summary

The paper provides a comprehensive survey of recent advances in symmetry‑breaking constraints (SBCs) for combinatorial optimization and constraint satisfaction problems. Symmetry, which manifests as interchangeable variables, rows, columns, or values, inflates the search space by creating multiple equivalent solutions, thereby degrading solver performance. The authors focus on two prevalent symmetry classes—row‑and‑column symmetry and value symmetry—and examine how modern SBC techniques balance the competing goals of completeness (eliminating all symmetric solutions) and propagation efficiency (keeping the added constraints cheap to enforce).

For row‑and‑column symmetry, the classic lex‑leader approach enforces a global lexicographic ordering over all possible permutations of rows and columns. While theoretically complete, this method suffers from factorial blow‑up in the number of constraints, making it impractical for all but the smallest instances. Recent work therefore introduces a spectrum of approximations: partial lex‑leader constraints restrict the set of permutations considered; dynamic lex‑leader constraints generate the necessary ordering constraints on‑the‑fly during search; sorted‑constraints and global ordering constraints impose a weaker, but still effective, ordering on rows or columns without enumerating every permutation. Empirical studies reported in the survey show that these approximations reduce propagation overhead by an order of magnitude while still cutting search time by 30‑70 % compared with a baseline solver lacking symmetry breaking.

Value symmetry is tackled with value‑lex constraints, which impose a lexicographic minimum over all value permutations, and with value‑ordering constraints that specify a fixed precedence among values. Direct application of value‑lex suffers from the same combinatorial explosion when the domain size is large. To mitigate this, researchers have proposed value grouping and value‑symmetry‑class techniques. Value grouping clusters interchangeable values and only permits permutations within each cluster, while symmetry‑class methods identify equivalence classes of values and select a canonical representative, dramatically shrinking the number of required constraints. When combined with domain‑filtering propagation, these methods achieve strong early pruning, often reducing memory consumption by more than 40 % and search time by 25‑55 % on benchmark problems with large value domains.

Beyond specific constraint families, the paper discusses meta‑modeling approaches that automatically detect symmetry patterns in a problem specification and generate an appropriate SBC set. It also covers methods for preserving symmetry‑breaking information when translating problems into SAT or SMT encodings, ensuring that the benefits of SBCs survive the compilation step.

The survey concludes by highlighting open challenges: achieving full completeness with linear‑size constraint sets, extending the techniques to richer symmetry groups (e.g., combined row‑column‑value symmetries), and leveraging machine learning to predict the most effective SBC configuration for a given instance. Overall, the literature indicates a clear trend toward SBC designs that sacrifice a small amount of theoretical completeness in exchange for substantial gains in practical propagation efficiency, making symmetry breaking a mature and indispensable tool for modern combinatorial solvers.