Full text
| |
| Author(s): |
Marcelo Pinheiro Leite Benedito
Total Authors: 1
|
| Document type: | Master's Dissertation |
| Press: | Campinas, SP. |
| Institution: | Universidade Estadual de Campinas (UNICAMP). Instituto de Computação |
| Defense date: | 2018-07-31 |
| Examining board members: |
Lehilton Lelis Chaves Pedrosa;
Luis Augusto Angelotti Meira;
Cid Carvalho de Souza
|
| Advisor: | Lehilton Lelis Chaves Pedrosa |
| Abstract | |
In the Hub Location Problem (HLP), the input is a metric space composed of clients, locations and a set of pairs of clients; a solution is a subset of locations to open hubs and an assignment for each pair of clients to a route starting in the first client, passing through one or two hubs and ending in the second client. The objective is to find a solution that minimizes the length of all routes plus the cost of opening hubs. The currently known approximation algorithms consider only the case in which the set of hubs is given as part of the input and the problem is assigning clients to hubs; or when the space is defined on special classes of graphs. In this work, we present the first constant-factor approximation algorithms for the problem of, simultaneously, selecting hubs and allocating clients. The first part of the thesis derives approximation algorithms for several variants of the problem. The main strategy is to reduce the hub location problems to classical location problems, such as Facility Location and k-Median. The reduction transforms an instance of hub location into an instance of a corresponding location problem, which is then solved by known approximation algorithm. The algorithm¿s output induces a solution of the original problem within a constant loss in the approximation ratio. In the second part, we focus on the Single Allocation Hub Location Problem (SAHLP), that is the variant in which a client must be connected to only one hub and there is no limit on the number of open hubs. Our main contribution is a 2.48-approximation algorithm for the SAHLP, based on the rounding of a new linear programming formulation. The algorithm is composed of two phases: in the first one, we scale the fractional solution and open a subset of hub locations, and in the second one, we assign clients to open hubs. The first phase follows the standard filtering framework for location problems. The latter, however, demanded the development of new ideas and is based on a multiple criteria assignment. The main technique is assigning a client to a closest open hub only if there are near open hubs, and otherwise selecting the hub which balances multiple costs (AU) | |
| FAPESP's process: | 16/12006-1 - Approximation algorithms for hub-location problems |
| Grantee: | Marcelo Pinheiro Leite Benedito |
| Support Opportunities: | Scholarships in Brazil - Master |