Optimization methodologies are particularly relevant nowadays due to the ever-increasing power of computers and the enhancement of mathematical models to better capture reality. These computational methods are used in many different fields and some of them, such as metaheuristics, have often been found helpful and efficient for the resolution of practical applications where finding optimal solutions is not straightforward. Many practical applications are multi-objective optimization problems: there is more than one objective to optimize and the solutions found represent trade-offs between the competing objectives. In the last couple of decades, several metaheuristics approaches have been developed and applied to practical problems and multi-objective versions of the main single-objective approaches were created. The Alliance Algorithm (AA) is a recently developed single-objective optimization algorithm based on the metaphorical idea that several tribes, with certain skills and resource needs, try to conquer an environment for their survival and try to ally together to improve the likelihood of conquest. The AA method has yielded reasonable results in several fields to which it has been applied, thus the development in this thesis of a multi-objective variant to handle a wider range of problems is a natural extension. The first challenge in the development of the Multi-objective Alliance Algorithm (MOAA) was acquiring an understanding of the modifications needed for this generalization. The initial version was followed by other versions with the aim of improving MOAA performance to enable its use in solving real-world problems: the most relevant variations, which led to the final version of the approach, have been presented. The second major contribution in this research was the development and combination of features or the appropriate modification of methodologies from the literature to fit within the MOAA and enhance its potential and performance. An analysis of the features in the final version of the algorithm was performed to better understand and verify their behavior and relevance within the algorithm. The third contribution was the testing of the algorithm on a test-bed of problems. The results were compared with those obtained using well-known baseline algorithms. Moreover, the last version of the MOAA was also applied to a number of real-world problems and the results, compared against those given by baseline approaches, are discussed. Overall, the results have shown that the MOAA is a competitive approach which can be used `out-of-the-box' on problems with different mathematical characteristics and in a wide range of applications. Finally, a summary of the objectives achieved, the current status of the research and the work that can be done in future to further improve the performance of the algorithm is provided.
Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:744344 |
Date | January 2017 |
Creators | Lattarulo, Valerio |
Contributors | Parks, Geoffrey Thomas |
Publisher | University of Cambridge |
Source Sets | Ethos UK |
Detected Language | English |
Type | Electronic Thesis or Dissertation |
Source | https://www.repository.cam.ac.uk/handle/1810/270076 |
Page generated in 0.0016 seconds