Dynamic multi-objective evolutionary algorithm with objective space prediction strategy

To solve dynamic multi-objective optimization problems, a dynamic multi-objective evolutionary algorithm (DMOEA) must be able to deal with the dynamics of the environment, and such modifications can lead to new optimal solutions over time. Various algorithms have been proposed that modify the way a...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Applied soft computing Jg. 107; S. 107258
Hauptverfasser: Guerrero-Peña, Elaine, Araújo, Aluizio F.R.
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier B.V 01.08.2021
Schlagworte:
Dynamic multi-objective optimization
Evolutionary computation
Forecast
Prediction-based methods
Probabilistic modeling
Prediction-based methods
Probabilistic modeling
Dynamic multi-objective optimization
Evolutionary computation
Forecast
ISSN:1568-4946, 1872-9681
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:To solve dynamic multi-objective optimization problems, a dynamic multi-objective evolutionary algorithm (DMOEA) must be able to deal with the dynamics of the environment, and such modifications can lead to new optimal solutions over time. Various algorithms have been proposed that modify the way a change is handled. Among them, prediction-based methods are promising for solving this kind of problem. They provide guided direction for population evolution through a prediction mechanism that assists the DMOEA to respond quickly to new changes. Based on these strategies, we propose a dynamic non-dominated sorting differential evolution improvement with prediction in the objective space (DOSP-NSDE). The proposal uses the objective space prediction (OSP) strategy for both the static evolutionary process (between changes) and the change reaction mechanism to predict the new optimal front location. Experiments were performed on a real-world problem and four sets of test problems: FDA, dMOP, UDF, and DF. Comparison of DOSP-NSDE with several algorithms in the literature, considering three metrics, is presented, showing that the proposal is competitive with most problems. •A new algorithm to deal with dynamic multi-objective optimization problems.•Probabilistic models used both in the evolutionary process and environment changes.•Statistical information extracted from both the decision and objective space.•Both the population distribution and local information of each individual considered.•A dynamic real-world multi-objective optimization problem solved.
ISSN:1568-4946
1872-9681
DOI:10.1016/j.asoc.2021.107258