Variable Dependence Interaction And Multi-objective Optimisation

Interaction among decision variables is inherent to a number of real-life engineering design optimisation problems. There are two types of interaction that can exist among decision variables: inseparable function interaction and variable dependence. The aim of this paper is to propose an Evolutionary Computing (EC) technique for handling variable dependence in multi-objective optimisation problems. In spite of its immense potential for real-life problems, lack of systematic research has plagued this field for a long time. The paper attempts to fill this gap by devising a definition of variable dependence. It then uses this analysis as a background for identifying the challenges that variable dependence poses for optimisation algorithms. The paper further presents a brief review of techniques for handling variable dependence in optimisation problems. Based on this analysis, it devises a solution strategy and proposes an algorithm that is capable of handling variable dependence in multi-objective optimisation problems. The working of the proposed algorithm is demonstrated, and its performance is compared to that of two high performing evolutionary-based multi-objective optimisation algorithms, NSGA-II and GRGA, using two test problems extracted from literature. The paper concludes by giving the current limitations of the proposed algorithm and the future research directions.