Study of multi-objective optimization and its implementation using NSGA-II
Rahul Kumar Das, Chittaranjan Samal, Suryakant Mallick
Abstract
Rahul Kumar Das, Chittaranjan Samal, Suryakant Mallick
Abstract
This project investigates the Multi-objective optimization strategies and their solutions using Multi-objective evolutionary algorithms (MOEAs). Multi-objective evolutionary algorithms (MOEAs) that use non-dominated sorting and sharing are criticized mainly for their; a) computational complexity, b) lack of elitism, c) need for specifying sharing parameter. In this paper the Non- Dominated Sorting Genetic Algorithm (NSGA) is studied and NSGA-II as proposed by Deb et. al. has been implemented, which alleviates the above three difficulties. In this study different objectives have been considered with different variables and constraints. The algorithm yielded satisfactory simulation results in all the different cases. The effect of the genetic parameters on the Pareto-Optimal front in all the cases has been studied. The results show that NSGA-II find much better spread of solutions and better convergence near the true pareto optimal front compared to other elitist MOEAs.
OpenAlex reports 1 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
This project investigates the Multi-objective optimization strategies and their solutions using Multi-objective evolutionary algorithms (MOEAs). Multi-objective evolutionary algorithms (MOEAs) that use non-dominated sorting and sharing are criticized mainly for their; a) computational complexity, b) lack of elitism, c) need for specifying sharing parameter. In this paper the Non- Dominated Sorting Genetic Algorithm (NSGA) is studied and NSGA-II as proposed by Deb et. al. has been implemented, which alleviates the above three difficulties. In this study different objectives have been considered with different variables and constraints. The algorithm yielded satisfactory simulation results in all the different cases. The effect of the genetic parameters on the Pareto-Optimal front in all the cases has been studied. The results show that NSGA-II find much better spread of solutions and better convergence near the true pareto optimal front compared to other elitist MOEAs.
Key concepts: Sorting, Multi-objective optimization, Mathematical optimization, Evolutionary algorithm, Convergence (economics), Genetic algorithm, Computer science, Pareto optimal