近年来，商用超密集低地球轨道卫星星座已开始部署，旨在提供无缝覆盖的全球互联网服务。为提升卫星网络传输效率，并为未来第六代用户提供鲁棒的广域计算服务，基于低轨卫星的计算网络即地面用户可向其卸载计算任务日益受到关注。然而，如何在考虑不同区域计算需求差异的前提下，设计适用于计算网络的低轨卫星星座，仍是一个有待解决的问题。本文研究一种超密集低轨卫星计算网络，地面用户终端可将部分计算任务卸载至卫星进行在轨计算。超密集星座设计问题被构建为一个多目标优化问题来最大化平均覆盖率、传输容量与计算能力，同时最小化卫星数量。为描述算力卫星网络的连通特性，提出了一种星地连通性模型，以确定不同地面区域的覆盖率。设计了一种优先级自适应算法来求解该多目标优化问题得到最优的倾斜轨道星座参数。仿真结果验证了所提连通性理论模型的准确性，并展示了在给定服务质量要求下的最优星座部署方案。在部署相同数量低轨卫星的情况下，所提出的星座性能优于现有方案，尤其在平均覆盖率方面实现了25%–45%的性能提升。

Commercial ultra-dense low-Earth-orbit (LEO) satellite constellations have recently been deployed to provide seamless global Internet services. To improve the satellite network transmission efficiency and provide robust wide-coverage computing services for future sixth-generation (6G) users, growing attention has been focused on LEO-satellite-based computing networks, to which ground users can offload computation tasks. However, how to design a LEO satellite constellation for computing networks, while considering discrepancies in the computing requirements of different regions, remains an open question. In this paper, we investigate an ultra-dense LEO-satellite-based computing network to which ground user terminals (UTs) offload part of their computing tasks to satellites. We formulate the ultra-dense constellation design problem as a multi-objective optimization problem (MOOP) to maximize the average coverage rate, transmission capacity, and computational capability, while minimizing the number of satellites. In order to depict the connectivity characteristics of satellite-based computing networks, we propose a terrestrial–satellite connectivity model to determine the coverage rate in different regions. We design a priority-adaptive algorithm to design the optimal inclined-orbit constellation by solving this MOOP. Simulation results verify the accuracy of our theoretical connectivity model and show the optimal constellation deployment, given quality-of-service (QoS) requirements. For the same number of deployed LEO satellites, the proposed constellation outperforms its existing counterparts; in particular, it achieves 25%–45% performance improvements in the average coverage rate.