Use of local search in the metric traveling salesman problem

The traveling salesman problem (TSP) for points in the Euclidean plane is considered. For several approximation algorithms, such as the tree algorithm and algorithms based on matching, a local search method using permutation in groups of closely placed points in a route is applied. Some modifications of these algorithms are presented. The effect of applying the local search method for partially constructed routes is examined for the modifications, in this paper, experimental results from statistical modeling with randomly equidistributed points are discussed. It allows one to conclude on the advantage of one or the other algorithms. ©1999 Kluwer Academic/Plenum Publishers.