David Woods

Citation

Thesis Title
Modular local search: A framework for self-adaptive mateheuristics

Read article at Massey Research Online:

Metaheuristics are advanced problem solving algorithms that attempt to find the best solution to complicated combinatorial problems such as vehicle routing, airline scheduling, and inventory control. This research developed a framework such that metaheuristics can be expressed as subsets of “modules” from a common library, with a common structure. The standardized modules and structure allow these algorithms to modify themselves during their execution, by varying parameters and modules. This ability introduces the potential for semi-intelligent algorithms that are capable of learning. Novel metaheuristic concepts were developed and tested on the Arc Subset Routing Problem, a new problem which involves scheduling a vehicle to service a subset of roads in a network, such that benefit is maximised subject to a distance constraint. Several new methods for creating test problems and modelling results were also introduced. The findings suggest promising outcomes for these new metaheuristics concepts.

Supervisors
Professor Mark Bebbington
Dr Jw Giffin