Use this URL to cite or link to this record in EThOS:
Title: A genetic algorithm approach for combinatorial optimisation problems
Author: Chu, Paul C. H.
ISNI:       0000 0004 2737 3384
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 1997
Availability of Full Text:
Access from EThOS:
Access from Institution:
This thesis deals with new heuristic algorithms for solving combinatorial optimisation problems. In this thesis we consider four well-known TVf-hard problems: the set covering problem, which is the problem of finding the minimum cost cover of a given matrix; the set partitioning problem, which is closely related to the set covering problem and is the problem of finding the minimum cost partition of a given matrix; the generalised assignment problem, which is the problem of finding a minimum cost assignment of a set of jobs to a set of agents subject to resource constraints and; the multiconstraint knapsack problem, which is the problem of finding a subset of projects that maximises profit subject to resource constraints. These problems are models for many important practical applications in areas such as resource allocation, crew scheduling and capital budgeting. We propose a heuristic method based on a genetic algorithm framework to solve these problems. The theoretical development of genetic algorithms are reviewed. A survey of existing exact and heuristic algorithms will be given for each of the four problems considered. We develop a hybrid steady-state genetic algorithm, which combines the genetic search framework and a problem-specific heuristic improvement procedure for each of the problems. We demonstrate that this hybrid approach makes the genetic search strategies more effective. Althought the zero-one integer formulation of the problems makes the genetic representation of a solution straightforward, we show that alternative representation methods can be effective in some cases. A new method, not involving use of a penalty function, for dealing with constraints is also suggested. The performance of the proposed algorithms are evaluated on standard test sets as well as on additionally generated larger-sized problems. Computational results are presented to indicate that the genetic algorithm based heuristics are capable of producing very high quality solutions, in many cases superior to all known heuristic methods in the literature.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available