Heuristics for Large Geometric Problems

Abstract

In the previous chapters we have considered several heuristics for finding approximate solutions of traveling salesman problem instances. We have also shown how the use of candidate subgraphs can speed up computations enormously. There may be situations, however, where even these efficient implementations are too slow. If very restrictive real time constraints have to be observed, the methods derived so far may not be appropriate. In this chapter we address the question of finding traveling salesman tours for very large instances in short time. It is clear that we will have to accept some loss in quality.