Dynamic multi-objective optimization: a two archive strategy

Chen, Renzhi (2018). Dynamic multi-objective optimization: a two archive strategy. University of Birmingham. Ph.D.

Preview

Chen18PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (4MB) | Preview

Abstract

Existing studies on dynamic multi-objective optimization mainly focus on dynamic problems with time-dependent objective functions. Few works have put efforts on dynamic problems with a changing number of objectives, or dynamic problems with time-dependent constraints. When problems have time-dependent objective functions, the shape or position of the Pareto-optimal front/set may change over time. However, when dealing with problems with a changing objective number or time-dependent constraints, the challenges are different. Changing number of objectives leads to the expansion or contraction of the dimensions of the Pareto-optimal front/set manifold, while time-dependent constraints may change the shape of feasible regions over time. The existing dynamic handling techniques can hardly handle the changing number of objectives. The state-of-arts in constraints handling techniques are incapable of tackling problems with time-dependent constraints. In this thesis, we present our attempts toward tackling 1) the dynamic multiobjective optimizing problems with a changing number of objectives and 2) multi-objective optimizing problems with time-dependent constraints. Two-archive Evolutionary Algorithms are proposed. Comprehensive experiments are conducted on various benchmark problems for both types of dynamics. Empirical results fully demonstrate the effectiveness of our proposed algorithms.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):