PWDCMO

MATLAB code for the paper:

D. Gong, M. Rong, N. Hu, Y. Wang, W. Pedrycz, and S. Yang, "A Prediction and Weak Coevolution-Based Dynamic Constrained Multiobjective Optimization," IEEE Transactions on Evolutionary Computation, 2025, 29(4): 1328-1342. DOI: 10.1109/TEVC.2024.3418470.

Repository: https://github.com/MiaoRong-Lab/PWDCMO

This repository contains two entry points:

• Standalone/: standalone reproduction code derived from the original experiment scripts.
• scripts/run_paper_experiment.m: standalone reproduction script based on the original experiment code.
• PlatEMO/: a PlatEMO-style adapter containing the PWDCMO algorithm and PWDCMO_DCMOP dynamic constrained test problem.

Requirements

• MATLAB R2018a or later.
• For the PlatEMO adapter: PlatEMO v4.x or later.

The standalone path no longer requires normrnd or pdist2; the refactor replaced those calls with base MATLAB equivalents. The hypervolume helper still contains a Monte Carlo branch for four or more objectives, but the provided DCMOP instances are bi-objective.

Repository Layout

PWDCMO/
  Standalone/   Original-style code for reproducing the paper experiments
  PlatEMO/      Files that can be copied into or added alongside PlatEMO
  scripts/      Small entry scripts for smoke tests and experiments
  README.md
  CITATION.cff

Standalone Usage

Run a quick check:

run('scripts/run_smoke_test.m')

Run the full standalone experiment:

run('scripts/run_paper_experiment.m')

The full script defaults to DCMOP1, N = 200, pps = 100, 20 independent runs, and time points 0:0.25:20. Results are saved under results/.

PlatEMO Usage

Copy or add the repository's PlatEMO folder to the MATLAB path together with PlatEMO:

addpath(genpath('path/to/PlatEMO/PlatEMO'));
addpath(genpath('path/to/PWDCMO/PlatEMO'));

Problem   = PWDCMO_DCMOP('N',100,'D',10,'maxFE',20000,'parameter',{1,20,4});
Algorithm = PWDCMO('save',1,'metName',{'IGD','HV'});
Algorithm.Solve(Problem);

PWDCMO_DCMOP parameters are:

• instance: DCMOP instance number from 1 to 8.
• taut: number of generations for one environment.
• nt: number of time steps per unit time. The default nt = 4 gives time increments of 0.25.

Files

• Standalone/main.m: standalone experiment loop.
• Standalone/NSGA2_1.m: static optimizer with constrained and diversity archives.
• Standalone/dynamic_response_1.m: prediction-based response after an environmental change.
• Standalone/variable_move.m: infeasible solution movement toward feasible nondominated guidance solutions.
• PlatEMO/Algorithms/Multi-objective optimization/PWDCMO/PWDCMO.m: PlatEMO-compatible algorithm class.
• PlatEMO/Problems/Multi-objective optimization/PWDCMO-DCMOP/PWDCMO_DCMOP.m: PlatEMO-compatible dynamic constrained problem class.

License

This repository is released under the research-use license in LICENSE, following the research and educational use spirit of PlatEMO. Commercial use requires separate permission from the copyright holders.

Notes Before Public Release

• Validate the PlatEMO adapter against the final paper tables before submitting it upstream.
• If this code or the PlatEMO adapter is used with PlatEMO, acknowledge and cite PlatEMO as required by the PlatEMO project.

Citation

@article{Gong2025PWDCMO,
  title   = {A Prediction and Weak Coevolution-Based Dynamic Constrained Multiobjective Optimization},
  author  = {Gong, Dunwei and Rong, Miao and Hu, Na and Wang, Yan and Pedrycz, Witold and Yang, Shengxiang},
  journal = {IEEE Transactions on Evolutionary Computation},
  volume  = {29},
  number  = {4},
  pages   = {1328--1342},
  year    = {2025},
  doi     = {10.1109/TEVC.2024.3418470}
}

When using PlatEMO, also cite:

@article{PlatEMO,
  title   = {{PlatEMO}: A {MATLAB} platform for evolutionary multi-objective optimization},
  author  = {Tian, Ye and Cheng, Ran and Zhang, Xingyi and Jin, Yaochu},
  journal = {IEEE Computational Intelligence Magazine},
  volume  = {12},
  number  = {4},
  pages   = {73--87},
  year    = {2017}
}