Page 73 - 2024F

66 UEC Int’l Mini-Conference No.53

An Adaptive Evolutionary Algorithm for dynamic Multi-
objective Optimization

Tasin Mobin Mohtadi and Hiroyuki Sato
UEC Exchange Study Program (JUSST Program)
Department of Informatics
The University of Electro-Communication Tokyo, Japan

Introduction
o Dynamic multi-objective optimization problems (DMOPs) pose
unique challenges due to changing objective functions over time.
o Adaptive DNSGA-II outperforms the Conventional DNSGA-II,
does not run on fixed reinitialization ratio (ζ) unlike DNSGA-II
o The reinitialization ratio (ζ) is critical to performance, and fixed
values may not be optimal for all scenarios.
Problem Statement
Figure 3:Sorting Example of initial Solutions
o A fixed reinitialization ratio (ζ) value is not optimal for the
problem output.
o It may not suit all problems requiring time-consuming parameter
tuning

Figure 4:Conventional DNSGA-II

Figure 2: comparative example graph between conventional and dynamic DNSGA-II
Objective
o Automatically determine the reinitialization ratio (ζ) value
comparing the distance of initial solutions while search
Methodology
Figure 5:Proposed Dynamic DNSGA-II
o Metrics used: Hypervolume, IGD, Spacing. Analysis
o Calculate solution distance between before and after the
problem change. o Pareto front analysis shows that Adaptive DNSGA-II
outperforms Conventional DNSGA-II in complex cases
o The greater the distance, the larger the .. o Adaptive DNSGA-II provides better spacing, ensuring diverse
o Implementation tools: MATLAB with PlatEMO framework. solutions.
o The performance changes due to the variation of the
reinitialization ratio (ζ)
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 2000 4000 6000 8000 10000 12000
Figure 1: Work Process of NSGA-II z = 0.2 z =0.3 z = 0.4 z =0.6 z =0.8 expected
Expected Outcome Figure 6:Comparison Graph of different zeta and expected result
Conclusion
o Ensure Adaptive DNSGA-II generates a consistent,
adaptable sequence of solutions for dynamic problems. o Adaptive DNSGA-II is ideal for complex, dynamic environments
o Demonstrate Adaptive DNSGA-II's capability for real-time, where flexibility is key.
dynamic applications. o Conventional DNSGA-II is suitable for simpler, static problems.
o Identify the exact reinitialization ratio (ζ) ratio of a particular o Choosing the right approach depends on the nature of the
problem that enhances convergence and diversity. optimization problem.
K. Deb, A. Pratap, S. Agarwal, and T. Meyarivan || K. Deb, U. Bhaskara Rao N., and S. Karthik, Dynamic multi-objective