Improving NSGA-II using a Dynamic Average Distance Selection Strategy

Read full text Summarise with VIRAT
Authors: Jiezhen Yang, Yan-Zuo Chang, Qi-Hong Tang, Guan-Hong Xie, Yongqing Wang, Zheng-Kuan Deng, Zirui He, Kai-Ming Chen, Yu-Xuan Chen, Hong-Rui Yang, Wen-Min Wen
Cited by: 0
References: 6

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The traditional Non-dominated Sorting Genetic Algorithm II (NSGA-II) struggles to maintain a uniformly distributed solution set across the entire Pareto front when dealing with non-uniform, non-convex, or discontinuous Pareto fronts. This limitation arises because its crowding distance metric relies solely on local linear spacing, making it prone to issues such as the loss of boundary solutions or local redundant clustering. To address this problem, this paper proposes an improved NSGA-II algorithm, whose core mechanism is the introduction of a dynamic average distance selection strategy into the original framework. Instead of using the traditional local crowding distance metric, the proposed algorithm constructs an "influence rectangle" for each individual using a dynamic scaling factor. This transforms the occupancy relationship of individuals in the objective space into the degree of geometric overlap between these rectangles, enabling the identification and elimination of redundant individuals. Experiments are conducted using ZDT series test functions, and the Spacing (SP) indicator is employed to evaluate the distribution uniformity of the obtained solution sets. Simulation results demonstrate that, while maintaining good convergence, the SP indicator values of the improved algorithm on the ZDT1, ZDT2, and ZDT3 test functions are significantly reduced, with a decrease ranging from 56.10% to 59.10%. This fully verifies the effectiveness of the dynamic average distance strategy in enhancing the distribution uniformity of the solution set. When addressing problems with discontinuous and concave fronts, the algorithm exhibits excellent robustness and uniform distribution capability. By incorporating an adaptive geometric evaluation criterion, the improved NSGA-II algorithm provides more reliable and stable decision support for complex multi-objective optimization problems.

Fields of Study & Subjects

Advanced Multi-Objective Optimization Algorithms Metaheuristic Optimization Algorithms Research Evolutionary Algorithms and Applications

Publication details

Journal / source
International Journal of Advanced Engineering Research and Science
Volume
13
Issue
3
Pages
22–32
Publication date
1 January 2026
Language
EN