Abstract: In response to the problem of trade-offs between coverage performance and constellation cost in the design of regional remote sensing satellite constellations, a constellation model with a short average revisit time and the minimum number of satellites as the objective function was established by utilizing the characteristics of regional coverage satellite constellations and the Walker constellation. The optimization variables involved include satellite orbit altitude, inclination, number of orbital planes, and number of satellites per orbital plane. Aiming at the shortcomings of traditional algorithms for solving the satellite constellation optimization problem, such as slow convergence speed and easy fall into local optimization, an improved crayfish optimization algorithm IMOCOA (Improved Multi-objective Crayfish Optimization Algorithm) is proposed, and the IMOCOA algorithm incorporates the refractive inverse learning strategy into The IMOCOA algorithm incorporates the refractive inverse learning strategy into the population initialization process to achieve a more uniform distribution of individuals in the solution, as well as the strategy of introducing a nonlinear convergence factor in the position update to enhance the global optimization search capability. The evaluation results of the ZDT (Zitzler-Deb-Thiele) series of multi-objective test functions in terms of both convergence and diversity show that the IMOCOA algorithm outperforms the NSGA-2, MOPSO, and MSSA algorithms. Its optimality metrics on IGD, GD, and HV, as well as SP, are improved over the other three algorithms by 54.4%, 78.7%, respectively, 3.6% and 27.3%, verifying the advantages of IMOCOA algorithm over the other three algorithms in convergence speed, convergence stability and diversity. The IMOCOA algorithm is used in the optimization design of a remote sensing satellite constellation in the Beijing-Tianjin-Hebei region to solve the problem of the minimum number of satellites under the requirement of revisit time. The simulation results show that the selection of three satellites can achieve an average revisit coverage of the Beijing-Tianjin-Hebei region of 1.54h, which further verifies the validity of the IMOCOA algorithm in solving this kind of problem. Given the problem of long simulation time, the introduction of reinforcement learning can be considered in future work to train and predict the model, to improve the optimization efficiency, and at the same time, the regional constellation design method based on the improved crayfish optimization algorithm can be used as a reference for the future construction of low-orbit constellations.

Key words: remote sensing satellite, multi-objective optimization, constellation optimization, crayfish optimization algorithm； regional remote sensing