Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.638587
Title: Examination scheduling using the Ant System
Author: Pugh, N. J.
Awarding Body: University of Wales Swansea
Current Institution: Swansea University
Date of Award: 2004
Availability of Full Text:
Access from EThOS:
Abstract:
This work is concerned with heuristic approaches to examination timetabling. It is demonstrated that a relatively new evolutionary method, the Ant System, can be the basis of a successful two-phase solution method. The first phase exploits ant feedback in order both to produce large volumes of feasible timetables and to optimise secondary objectives. The second phase acts as a repair facility where solution quality is improved further while maintaining feasibility. This is accomplished without increasing computational effort to unrealistic levels. The work builds on an existing implementation for the graph colouring problem, the natural model for examination scheduling. It is demonstrated that by adjusting the graph model to allow the accommodation of several side constraints as well incorporating enhancement techniques within the algorithm itself, the Ant System algorithm becomes very effective at producing feasible timetables. The enhancements include a diversification function, new reward functions and trail replenishment tactics. It is observed that the achievement of second-order objectives can be enhanced through a variety of means. A modified elitist strategy (ERF) significantly improves the performance of the Ant System due to the extra emphasis on second-order feedback. It is also shown that through the incorporation of the ERF, trail limits and, in particular, 19th century evolutionary theory the area of the solution space explored by the ants during the infancy of the search can be reduced. In addition, a good level of exploration is maintained as the search matures. This balance between exploration and exploitation is the main determinant of solution quality. The use of a repair facility, as is common practice with evolutionary algorithms, encourages fitter solutions. The interaction between Lamarckian evolution and searching in an extended neighbourhood through the graph theoretic concept of Kempe chains leads to better overall solutions.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.638587  DOI: Not available
Share: