Polynomial-time separation of simple comb inequalities

Lodi, A and Letchford, A N (2002) Polynomial-time separation of simple comb inequalities. In: Integer Programming and Combinatorial Optimization : Proceedings of the 9th International IPCO Conference. Lecture Notes in Computer Science . Springer, USA, pp. 93-108. ISBN 3-540-43676-6

Abstract

The comb inequalities are a well-known class of facet-inducing inequalities for the Traveling Salesman Problem, defined in terms of certain vertex sets called the handle and the teeth. We say that a comb inequality is simple if the following holds for each tooth: either the intersection of the tooth with the handle has cardinality one, or the part of the tooth outside the handle has cardinality one, or both. The simple comb inequalities generalize the classical 2-matching inequalities of Edmonds, and also the so-called Chv´atal comb inequalities. In 1982, Padberg and Rao gave a polynomial-time algorithm for separating the 2-matching inequalities – i.e., for testing if a given fractional solution to an LP relaxation violates a 2-matching inequality. We extend this significantly by giving a polynomial-time algorithm for separating the simple comb inequalities. The key is a result due to Caprara and Fischetti.