The Metric Travelling Salesman Problem, henceforth metric TSP, is a fundamental problem in combinatorial optimization which consists of finding a minimum cost Hamiltonian cycle (also called a TSP tour) in a weighted complete graph in which the costs are metric. Metric TSP is known to belong to a class of problems called NP-hard even in the special case of graph-TSP, where the metric costs are based on a given graph. Thus, it is highly unlikely that efficient methods exist for solving large instances of these problems exactly.
In this thesis, we develop a new heuristic for metric TSP based on extending ideas successfully used by Mömke and Svensson for the special case of graph-TSP to the more general case of metric TSP. We demonstrate the efficiency and usefulness of our heuristic through empirical testing.
Additionally, we turn our attention to graph-TSP. For this special case of metric TSP, there has been much recent progress with regards to improvements on the cost of the solutions. We find the exact value of the ratio between the cost of the optimal TSP tour and the cost of the optimal subtour linear programming relaxation for small instances of graph-TSP, which was previously unknown. We also provide a simplified algorithm for special graph-TSP instances based on the subtour linear programming relaxation.
| Identifer | oai:union.ndltd.org:uottawa.ca/oai:ruor.uottawa.ca:10393/35613 |
| Date | January 2016 |
| Creators | Desjardins, Nicholas |
| Contributors | Boyd, Sylvia |
| Publisher | Université d'Ottawa / University of Ottawa |
| Source Sets | Université d’Ottawa |
| Language | English |
| Detected Language | English |
| Type | Thesis |
Page generated in 0.0023 seconds