Novel Objective-Space-Dividing Multi-objectives evolutionary algorithm and its convergence property

Abstract

To overcome the shortcomings of Multi-Objectives Evolutionary Algorithms (MOEAs) based on the notion of Objective-Space-Dividing (OSD) with high calculation complexity, this paper proposes an improved algorithm called OSD-MOEA. The proposed algorithm supports the following features: 1) transforming the Pareto relationship among individuals to the ranking relationship of the total value of indexes in divided space; 2) simple and efficient environment choosing method based on index ranking; 3) an individual crowding algorithm which rapidly chooses the nearest individual to the origin. Convergence analysis shows the convergence property of the proposed algorithm. Simulation results of the proposed algorithm OSD-MOEA are compared with NSGAII and PSFGA and high efficiency, low time complexity and good convergence are noticed.