Limitations, Discussion and Future Work¶

With over 9,000 operational satellites in orbit as of December 2025 and the continuous deployment of new satellites, the density of Starlink satellites across various orbital shells is steadily increasing. Consequently, with the limited spatial resolution of 123x123 pixels in obstruction maps, it is becoming more challenging to distinguish adjacent satellites with similar azimuth and elevation Demystifying Starlink Network Performance under Vehicular Mobility with Dynamic Beam Switching based on their projected pixel locations. This could affect the accuracy of identifying the exact communicating satellite IDs for UTs. However, considering the satellite trajectories and their relative elevations to UTs, it may be feasible to cluster regions of adjacent satellites with similar performance characteristics, as they often follow similar paths and might connect to nearby ground stations. Unlike other LEO network operators such as OneWeb, without Starlink exposing the exact connected satellite IDs, obstruction map-based methods remain the only feasible option for identifying communicating satellites for Starlink UTs. We acknowledge Starlink’s efforts to increase transparency by updating comprehensive event logs 6 . We hope that future firmware updates will restore useful diagnostic information, such as connected satellite IDs, connected ground stations, or physical layer metrics such as SNR or RSSI, to facilitate the development of LEO-aware transport protocols and applications, further improving the overall user experience.

Limitations¶

卫星密度与分辨率的冲突:
- 随着 Starlink 卫星数量不断增加(截至 2025 年 12 月已超过 9,000 颗), 轨道壳层上的卫星密度稳步提升
- 受限于遮挡图像素的低空间分辨率, 区分具有相似方位角和仰角的相邻卫星变得愈发困难, 这可能会影响识别确切通信卫星 ID 的准确性
聚类分析的可行性:
- 尽管精确识别存在挑战, 但考虑到卫星轨迹及其相对于 UT 的仰角, 将相邻卫星聚类为具有相似性能特征的区域仍然是可行的, 因为它们通常遵循相似的路径并可能连接到附近的地面站
方法的必要性:
- 与 OneWeb 等其他 LEO 运营商不同, Starlink 不公开连接的卫星 ID, 因此基于遮挡图的方法仍然是目前识别 Starlink UT 通信卫星的唯一可行方案
对未来的期望:
- 作者承认 Starlink 在更新详细事件日志方面提高透明度的努力, 但希望未来的固件更新能恢复有用的诊断信息(如连接的卫星 ID, 地面站信息, SNR 或 RSSI 等物理层指标), 以促进 LEO 感知传输协议的开发并提升用户体验

Discussion and Future Work¶

网络性能改进 (Network Performance Improvements):
- 对于具有固定路线的移动场景(如公共交通或铁路), 利用历史网络测量数据来学习性能模式是可行的, 从而克服车辆运动不可预测带来的挑战
- 通过优化上层协议和应用(例如: 在同时配备 Starlink 和蜂窝服务的车辆上进行多路径数据包调度), 可以隐藏瞬时遮挡对传输层协议及应用造成的负面影响
  - 上述 Starlink + Cellular 多路径调度的想法, 建议参考 Starlink vs. 5G
推广至其他 LEO 卫星网络 (Generalization to Other LEO Satellite Networks):
- 除了 Starlink, OneWeb 等其他 LEO 卫星网络也支持移动和海事应用场景
- 鉴于 OneWeb 具有不同的星座拓扑(含约 650 颗卫星的 "Walker-Star" 结构)以及不同的波束分配与切换策略, 未来值得深入研究 OneWeb 终端在瞬时遮挡和视场快速变化下的移动性能表现