Simulation-Based Fitness Landscape Analysis and Optimisation of Complex Systems
2013
Vitālijs Boļšakovs

Defending
09.12.2013. 14:30, Datorzinātnes un informācijas tehnoloģijas fakultātē, Meža ielā 1/3, 202. auditorijā

Supervisor
Gaļina Merkurjeva

Reviewers
Zigurds Markovičs, Pēteris Grabusts, Stefan Wagner

The doctoral thesis is dedicated to the development of methods and algorithms for a simulation-based fitness landscape analysis and optimisation of complex systems. Metaheuristic methods for NP-hard optimisation problems and their practical application are investigated in the research. This is motivated by a wide spread of such problems nowadays and a relevance of improvement of their solution methods. Research object is development of methods and algorithms for the simulation-based fitness landscape analysis, which will ensure a selection and configuration of the metaheuristic algorithm for parametric simulation-based optimisation of complex systems. Application of the fitness landscape analysis methods in the engineering field and tasks of simulation-based optimisation are reviewed and analysed in this thesis. A formalised scheme of simulation-based optimisation enhanced by fitness landscape analysis is developed. Strength and weakness of the fitness landscape analysis is researched on the benchmark landscapes, both with relations between efficiency of the optimisation algorithm and structural features of the corresponding fitness landscapes. Methods and algorithms for the simulation-based fitness landscape analysis are developed based on the proposed formalised scheme. The prototype of a software tool for the fitness landscape analysis of simulation optimisation problems is developed. Approbation of the developed methods is performed in optimisation of vehicle schedule and routes in the delivery planning task. Various metaheuristic optimisation scenarios with application of the fitness landscape analysis are investigated. The work contains 135 pages, 14 tables, 73 figures and 2 appendixes.


Keywords
simulation, optimisation, fitness landscape analysis, metaheuristic methods, vehicle scheduling

Boļšakovs, Vitālijs. Simulation-Based Fitness Landscape Analysis and Optimisation of Complex Systems. PhD Thesis. Rīga: [RTU], 2013. 135 p.

Publication language
English (en)
The Scientific Library of the Riga Technical University.
E-mail: uzzinas@rtu.lv; Phone: +371 28399196