EXCHANGE OPERATORS AND A RELAXATION METHOD FOR THE TRAVELING SALESMAN PROBLEM

We give a new approach to find acceptable low-cost minima for the Travelling Salesman Problem, in a metric space. We introduce a new relaxation method, and above all we use topological criteria to define the exchange operators. This approach leads to a new algorithm better than Lin et Kerninghan algorithm. For a 532 cities path, the average accuracy is 1% and the best result is the true optimum. Running times grow empirically as O (N1.5).