Date of Award

5-31-2017

Document Type

Open Access Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Science

First Advisor

Junchi Suzuki

Second Advisor

Dan A. Simovici

Third Advisor

Ding Wei

Abstract

Multi-objective optimization problems require more than one objective functions to be optimized simultaneously. They are widely applied in many science fields, including engineering, economics and logistics where optimal decisions need to be taken in the presence of trade-offs between two or more conicting objectives. Most of the real world multi-objective optimization problems are NP-Hard problems. It may be too computationally costly to find an exact solution but sometimes a near optimal solution is sufficient. In these cases, Multi-Objective Evolutionary Algorithms (MOEAs) provide good approximate solutions to problems that cannot be solved easily using other techniques. However Evolutionary Algorithm is not stable due to its random nature, it may produce very different results every time it runs. This dissertation proposes an Evolutionary Game Theory (EGT) framework based algorithm (EGTMOA) that provides optimality and stability at the same time. EGTMOA combines the notion of stability from EGT and optimality from MOEA to form a novel and promising algorithm to solve multi-objective optimization problems. This dissertation studies three different multi-objective optimization applications, Cloud Virtual Machine Placement, Body Sensor Networks, and Multi-Hub Molecular Communication along with their proposed EGTMOA framework based algorithms. Experiment results show that EGTMOAs outperform many well known multi-objective evolutionary algorithms in stability, performance and runtime.

Share

COinS