Enhancing the diversity of solutions in system design optimization
Abstract
During the design and optimization of complex systems, decision makers may be interested in obtaining a set of high-performance solutions that present a diversity of characteristics of interest. Quality-Diversity (QD) optimization algorithms model the latter as feature functions which are discretized. The aim is to find a single performing solution with respect to the objective function inside each combination of feature function intervals (called a niche). For example, in sounding rocket design, one may want to find the lightest architecture for a set of orbit and payload specifications. However, several system architectures embedding different technologies (e.g., number of rocket stages, type of rocket propulsion) can grant equivalent performances in a given niche. Current QD approaches only allow to identify the most performing one. This work proposes a novel QD framework where different solutions can occupy the same niche to provide more knowledge to the decision makers when deciding which solution to adopt for future design phases. In this paper, a mathematical formulation of the problem is developed. Four methods are proposed and their associated algorithms are detailed. Numerical simulations are done on an analytical test case and a sounding rocket design problem to assess their performances. It is shown that one of them is able to find the very best solutions in each niche and two others grant the discovery of a large batch of solutions. The last one allows to find a tradeoff between finding performing solutions and enhancing a great diversity of designs.
Related
- AutoEnergyDesign: An LLM-Driven Agent for Automatic Energy System Design Optimization
- Enhancing Interpretability of Stochastic Programming Solutions: A Multiparametric Approach
- Global solutions for the sensors placement problem via weakly convex optimization
- A selection criterion of solutions to a system of eikonal equations
- Differentiating Minimal-Norm Solutions to Parametric Optimization Problems
BibTeX
@article{baraton2026enhancing,
title = {{Enhancing the diversity of solutions in system design optimization}},
author = {Lucas Baraton and Annafederica Urbano and Loïc Brevault and Mathieu Balesdent},
journal = {Aerospace Science and Technology},
year = {2026},
month = aug,
doi = {10.1016/j.ast.2026.113578},
url = {https://hal.science/hal-05761196v1},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Enhancing the diversity of solutions in system design optimization
AU - Lucas Baraton
AU - Annafederica Urbano
AU - Loïc Brevault
AU - Mathieu Balesdent
PY - 2026
DA - 2026/08/22
JO - Aerospace Science and Technology
DO - 10.1016/j.ast.2026.113578
UR - https://hal.science/hal-05761196v1
AB - During the design and optimization of complex systems, decision makers may be interested in obtaining a set of high-performance solutions that present a diversity of characteristics of interest. Quality-Diversity (QD) optimization algorithms model the latter as feature functions which are discretized. The aim is to find a single performing solution with respect to the objective function inside each combination of feature function intervals (called a niche). For example, in sounding rocket design, one may want to find the lightest architecture for a set of orbit and payload specifications. However, several system architectures embedding different technologies (e.g., number of rocket stages, type of rocket propulsion) can grant equivalent performances in a given niche. Current QD approaches only allow to identify the most performing one. This work proposes a novel QD framework where different solutions can occupy the same niche to provide more knowledge to the decision makers when deciding which solution to adopt for future design phases. In this paper, a mathematical formulation of the problem is developed. Four methods are proposed and their associated algorithms are detailed. Numerical simulations are done on an analytical test case and a sounding rocket design problem to assess their performances. It is shown that one of them is able to find the very best solutions in each niche and two others grant the discovery of a large batch of solutions. The last one allows to find a tradeoff between finding performing solutions and enhancing a great diversity of designs.
SN - 1270-9638
SN - 1626-3219
ER -