Improving System Architectures Using Multi-Objective Particle Swarm Optimization

Architectural definition of complex systems requires analyzing competing objectives and constraints in high-dimensionality problems due to the multitude of design choices. There are many optimization approaches which could be used for this process. Common approaches in the systems engineering literature are genetic algorithms (GA) and a multi-objective extension, non-dominated sorting genetic algorithm II (NSGA-II). Simpler alternatives exist, potentially without sacrificing optimization performance. In this research, simpler metaheuristic algorithms, specifically particle swarm optimization (PSO) and its multi-objective extension MOPSO, were evaluated and compared with the common GA methods across single and multi-objective optimization (SOO, MOO) problems. The GA approach was compared with PSO for SOO, and NSGA-II with MOPSO for MOO. Results show that performance (lowest minima) for both algorithms in the SOO case was similar (??0 rejected on 3 of 5 cases), suggesting that the simpler PSO approach may offer implementation and computational resource advantages. The simpler MOPSO algorithm showed similar performance to the NSGA-II algorithm when comparing search space coverage via hypervolume (??0 rejected on 2 of 5 cases) and via ??+-indication (??0 rejected on 3 of 5 cases). An example of MOPSO and NSGA-II applied to architecture definition is provided as a case study from the research literature. These findings suggest that PSO and MOPSO may have advantages relative to the GA and NSGA-II approaches in terms of simplicity with similar performance. MOPSO could lead to improved systems engineering via simpler optimization, more complete architectural analysis, and improved understanding of the viable design options during architecture definition.

Author

• Hampshire, Kenneth

Language

• en

Keyword

• multi-objective optimization
• genetic algorithms
• particle swarm optimization
• architecture definition
• architecture evaluation
• MOPSO

Date created

• 2026

Type of Work

• Dissertation

License

• All rights reserved

Rights statement

• In Copyright

GW Unit

• Systems Engineering

Degree

• Ph.D.

Advisor

• Grenn, Michael

Committee Member(s)

• Mazzuchi, Thomas A.
• Grenn, Michael
• Fossaceca, John
• Etemadi, Amir
• Sarkani, Shahram

Persistent URL

•  https://scholarspace.library.gwu.edu/etd/8p58pd98k