Variable Size Design Space Formulations Applied to Optimal Satellite Orbits for Regional Lunar Coverage
Year: 2026
Type: conference-paper
Abstract: Designing lunar positioning, navigation, timing, and communications (PNT+C) constellations presents unique optimization challenges due to variable constellation sizes and complex orbital dynamics. Existing approaches restrict designs to set patterns, limiting geometric diversity, or use sequential optimization, which carries forward dominated solutions inefficiently. Focusing on the most used technique in constellation design literature, genetic algorithms, this work evaluates three variable-size formulations for multi-objective optimization: Hidden Gene Genetic Algorithms with binary tags, dimensional variable encoding, and variable-length chromosomes. Through theoretical analysis of chromosome redundancy, empirical testing on the Schwefel 2.26 benchmark function, and application to a lunar coverage scenario, the dimensional variable formulation is shown to achieve superior performance with minimal implementation complexity. On the benchmark, the dimensional variable method consistently achieved low residuals with a standard deviation over a factor of ten less than the variable length form. For the lunar application, optimizing coverage, orbital stability, and constellation size, the dimensional variable formulation showed improved evaluation times and superior hypervolume performance, partially due to an improved ability to maintain small constellations. These findings show dimensional variable formulations as a useful alternative that directly handles the challenges of the lunar PNT+C design problem with minimal implementation barriers for orbital designers.
Institution: Georgia Institute of Technology
Language: English
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7202395943
