An Investigation of Hybrid Framework for Dynamic Multi Objective Problems

Abstract

Multi-objective evolutionary algorithms and selection hyper-heuristics are adaptive methods that can handle different types of dynamism which may occur in the environment. In this study, a hybrid framework combining these methods is presented for solving dynamic multi-objective optimization problems. In this framework, hyper-heuristics are used to select the heuristic that will generate the individuals in the population. The performance of the approach, along with the effect of different heuristic selection methods used in the selection hyper-heuristics, is experimentally examined over a set of dynamic multi-objective optimization problems. The empirical results show that the selection hyper-heuristics with learning perform well in the framework. It is also shown that the proposed approach can compete with the well-known methods from literature.

Keywords

• Dynamic multi-objective optimization problems
• Hyper-heuristics
• Multi-objective evolutionary algorithms