Date of Award
Campus Access Dissertation
Doctor of Philosophy (PhD)
Most of real world optimization problems have several conflicting objectives. The solutions for these problems is a set of trade-offs between objectives, which is called Pareto optimal set. Meanwhile, the search space is large and complex that makes them hard to be solved by optimal techniques within a reasonable time. Recently, Evolutionary Multiobjective Optimization Algorithms (EMOAs) have been found to be effective means of solving such problems. Traditional EMOAs often rank individuals (i.e. solution candidates) based on the dominance relations among them and exploit their ranks in selection operator. This process is called dominance ranking. Although dominance ranking based algorithms work well in solving many problems, they perform poorly in many-objective problems since most solutions in the solution set are non-dominated. Other research trend in design of EMOAs is to adopt the notion of indicator-based selection, which replaces dominance ranking with quality indicator. Among many quality indicators, R2-indicator possesses desirable properties to quantify the goodness of a solution or a solution set. On the other hand, R2-indicator can be computed quickly. This dissertation proposes evolutionary multiobjective optimization algorithms (EMOAs) that eliminate dominance ranking in selection and perform indicator-based selection with the R2 indicator. The proposed EMOAs, are designed to obtain a diverse set of Pareto-approximated solutions by correcting an inherent bias in the R2 indicator. (The R2 indicator has a stronger bias to the center of the Pareto front than to its edges.). In order to solve broader range of problems, constraint and noise handling mechanisms are also proposed and integrated into these algorithms. Two real world applications are presented in this dissertation. The first application is route optimization in PDP-TW-D, Pickup and Delivery Problem with Time Windows and Demands. Putting multiple optimization objectives into consideration, solving PDP-TW-D is to find a set of Pareto-optimal routes for a fleet of vehicles in order to serve given transportation requests. The second application is communication optimization in cloud-integrated body sensor networks. The optimization goal is to seek Pareto-optimal data transmission rates for sensor nodes in the networks. The proposed EMOAs are evaluated using several well-known testing problems together with these two real world applications.
Phan, Dung Huy, "R2 Indicator based Evolutionary Multiobjective Optimization Algorithms and Applications" (2015). Graduate Doctoral Dissertations. 237.