Date of Award

2006

Document Type

Doctoral Thesis

Degree Name

Doctor of Philosophy

Department

Cork Institute of Technology

First Advisor

Dr. Paul Walsh

Abstract

The permutation with repetition representation is a powerful representation primarily used in the GA-JSSP domain. Recently, attention has shifted from this representation for simple GA and towards JSSP resolution using messy GA. This work returns to the permutation with repetition representation for simple GA with the intention of exploring some previously unexplored aspects. A new 1-point crossover operator is introduced which is shown to be more successful than the current standard operator for this representation. A technique known as morphogenic computation is employed with this representation for the first time and a thorough analysis of its performance is presented. While achieving superior average end-of-run makespans for a number of problems, the inclusion of this technique serves to highlight a possibility for new exploration for JSSP with simple GA. Due to the lack of conclusions as to the efficacy of the non-random initialisation of populations using this representation, a successful seeding mechanism is proposed and introduced. A comparison is made with this system and the latest advances in messy GA, which shows that messy GA is not currently as adept at solving JSSP, though significant advances have been made recently. Further analysis of these contributions is provided with local search hybridisation. The successful application of these new operators and techniques, particularly morphogenic computation, indicates a further path of possibility for this representation employed for JSSP. It is shown that enhancements to simple GA are achievable without recourse to local search hybridisation and advanced parallel models by focusing on the core elements of the GA — representation and operators.

Access Level

info:eu-repo/semantics/openAccess

Share

COinS