Use this URL to cite or link to this record in EThOS:
Title: The application of Jostle-based hybridisation techniques to genetic algorithms for circle packing
Author: Stoner, M. A.
Awarding Body: University of Wales Swansea
Current Institution: Swansea University
Date of Award: 1998
Availability of Full Text:
Access from EThOS:
This thesis investigates the application of Jostle-based hybridisation to packing-order encoded genetic algorithms. It shows that some forms of Jostle-based hybridisation are extremely effective, while others are not worthwhile. This work addresses a practical transportation application which is modelled by the packing of non-congruent circles - a problem which has been poorly covered in the literature from a practical viewpoint. It begins by showing that the Jostle heuristic can be used for circle-packing with a reasonable degree of success, and then analyses the influence of the contents of the dataset being packed upon the performance of Jostle. A series of genetic algorithm formulations are implemented using a novel variation upon the standard permutation encoding which greatly reduces the search space in many cases. The overall performance of these formulations is found to be similar to that of Jostle, and a degree of complementary performance is observed. Surveying the literature reveals that the use of hybridisation for packing-order encoded genetic algorithms has not been widespread, principally because the structure of the encoding makes the standard local search techniques which are typically used for this purpose prohibitively expensive. Using the best solutions from runs of Jostle to "initialise" packing-order encoded genetic algorithms is shown to have two major advantages over the dimensional-ordering techniques which have typically been used, and gives significant performance gains at little or no extra cost, improving the performance of the genetic formulations in 70 to 90% of cases. A modified version of Jostle is found to produce small improvements at very low cost when used for post-processing. It is shown that the hybridisation techniques described could be readily applied to a wide range of practical packing applications.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available