A Layered NSGA-II Method for LEO Target Census Constellation Design

To address the pressure of space target census posed by low-orbit mega-constellations such as Starlink, this paper proposes an LEO target census constellation design method based on a layered NSGA-II algorithm. Six decision variables-orbital altitude, inclination, number of orbital planes, number of satellites per plane, sensor field of view, and focal length-are considered. Coverage rate, revisit period, and constellation cost are taken as optimization objectives to establish a multi-objective optimization model. To overcome the issues of slow convergence and susceptibility to local optima in high-dimensional decision spaces faced by classical multi-objective optimization algorithms, the decision variables are divided into the orbit layer, configuration layer, and sensor layer. NSGA-II evolution is performed layer by layer, with elite retention and global archive passing of high-quality solutions. Using high-precision 24-h ephemeris of 496 Starlink satellites generated by STK as the simulation object, the proposed layered NSGA-II is compared with classical NSGA-II, MOPSO, MOEA/D, and SPEA2. Results show that the layered NSGA-II achieves a hypervolume (HV) of 13.16, outperforming the other algorithms. Under the condition of maintaining 99.6% coverage, the recommended constellation solution achieves a revisit period as low as 7.6 h and a constellation cost of 0.429, demonstrating significantly better comprehensive performance than other algorithms. Convergence speed is improved by approximately 36% compared to classical NSGA-II. This method provides an efficient and engineering-applicable optimization approach for LEO target census constellation design.