跳转至

Background

Starlink UT’s built-in gRPC Remote Procedure Calls (gRPC) interface provides various diagnostic information to users, such as the UT’s orientation, GPS location, and network performance statistics. It does not reveal raw antenna metrics such as the received signal strength indicator (RSSI), SNR, or communicating satellite IDs. Internally, the UT uses the RSSI and SNR to assess the signal quality of the communicating satellites at specific beam steering angles [17, 19]. If the SNR drops below a certain threshold, the UT identifies obstructions in the designated direction. The UT projects the azimuth and elevation of communicating satellites on a two-dimensional (2D) obstruction map, which illustrates the accumulated trajectories of connected satellites over time. Give the dense constellation of Starlink satellites and the compact 123x123 pixel resolution of the obstruction map, changes in the communicating satellite trajectories are visible in pixel coordinates, but with limited temporal resolution [1, 10, 16, 22].

Previously, Starlink UTs clear and reconstruct obstruction maps upon boot. Early researchers have to frequently reboot the UT to clear the obstruction map to avoid overlapping trajectories [10, 22]. Starlink now adds the capability to allow users to manually reset the obstruction map and initiate the reconstruction process via gRPC commands or through the Starlink mobile application, for instance, when the physical orientation of the UT has significantly changed or the UT is relocated to a new location. It typically takes a few hours for the obstruction map to fully converge, at which point the entire FOV is covered by the trajectories of communicating satellites, as shown in Fig. 1a and Fig. 1b. The UT in Fig. 1a is mounted on a building rooftop, with a clear LoS of the sky, resulting in an obstruction-free map. In Fig. 1b, the UT is situated in a residential neighborhood surrounded by trees, resulting in an FOV obstruction rate of 17.9%, as reported by the UT’s gRPC interface.

When the UT is mounted on a vehicle, unlike cameras, the obstruction map available through the Starlink mobile application or directly obtained from the gRPC interface does not provide a real-time reflection of obstructions across the entire FOV. It is updated only with the trajectories of communicating satellite between successive frames. For instance, Fig. 1c shows an intermediate obstruction map captured during driving, not long after the obstruction map was initially cleared. Due to the constraints of a 2D representation without temporal details, over time, the cumulative obstruction map gradually fills up as shown in Fig. 1d. It is unable to convey meaningful information about transient obstructions at specific times along different sections of the route. The accumulated satellite trajectories in obstruction maps are originally intended to help users identify structural obstructions in the LoS for fixed-site installations. In this paper, we integrate the obstruction map with UT mobility and utilize temporal information to develop a mobility-aware satellite identification method for mobile Starlink UTs.

Starlink UT 遮挡图 (Starlink UT Obstruction Maps)

  • 诊断数据与生成原理:
    • Starlink UT 通过 gRPC 接口提供诊断信息, 但不直接公开原始天线指标(如 RSSI, SNR 或通信卫星 ID)
    • UT 内部利用 SNR 评估信号质量, 若低于特定阈值则判定为遮挡, 并将通信卫星的方位角和仰角投射到二维(像素)遮挡图上, 显示随时间累积的卫星轨迹
  • 重置与收敛:
    • 与早期必须重启设备不同, 现在用户可以通过 gRPC 或 App 手动重置并重建遮挡图(例如: 在设备位置或朝向改变时)
    • 遮挡图通常需要数小时才能完全收敛以覆盖整个视场(FOV)
  • 移动场景的局限性:
    • 在车载移动场景中, 遮挡图无法实时反映整个 FOV 的遮挡情况, 仅更新帧与帧之间的卫星轨迹
    • 由于缺乏时间分辨率, 累积的遮挡图难以有效传达如桥梁等瞬时遮挡的信息
  • 研究应用: 本文通过整合遮挡图、UT 移动性以及时间信息, 开发了一种适用于移动 Starlink UT 的卫星识别方法

alt text

2.2 Reference Frames in Obstruction Maps

Starlink UT utilizes phased array antennas to steer beams and maintain connectivity with communicating satellites [17]. To accurately track communicating satellites through beam steering, the UT requires a reference frame to determine its orientation. Since firmware update in September 2024, the 2D obstruction maps obtained through the gRPC interface are represented in two different reference frames, depending on the user’s subscription plan for the UT, as illustrated in Fig. 2a to Fig. 2c. Such information is indicated by the mapReferenceFrame field in the gRPC response of the dish_get_obstruction_map method call.

For UTs subscribed to the residential and other fixed-site plans, the reference frame (FRAME_EARTH) is based on the Earth-Centered, Earth-Fixed (ECEF) reference system. Inactive UTs, even without active subscriptions, can still access the satellite link and reach certain Internet resources, including Starlink’s website and the billing system [18]. Both inactive UTs and those subscribed to the Mobile and Roam plans use the local reference frame (FRAME_UT) based on the UT’s local coordinate system. Both the Mobile and Roam plans support mobility use cases. Without a fixed reference system, the UT in motion must rely on its local alignment to determine beam steering angles.

In Fig. 2, we illustrate different obstruction maps from two UTs (UT A and UT B) of the same model (rev3_proto2) placed in distinct obstructed environments. Note that for the same UT hardware, the reference frame changes when the user switches to different subscription plans. Fig. 2b and Fig. 2c were retrieved from the UT’s gRPC interface when UT A was associated with the regular Residential plan and the Standby mode 1 , respectively. The structural obstruction pattern indicates the edge of a nearby building’s rooftop. UT A has a tilt angle of 26.5 ◦ , and it is facing North with a boresight azimuth of about 3.2 ◦ . In FRAME_EARTH obstruction maps (Fig. 2b), the top-center pixel aligns with North, consistent with the top-down view from the Starlink mobile application (Fig. 2a). In contrast, in FRAME_UT obstruction maps (Fig. 2c), the bottom-center pixel always aligns with the UT’s boresight direction, displaying the FOV from the UT’s perspective. For UT B in Fig. 2, it has a tilt angle of 7.7 ◦ , and it also has the FRAME_UT obstruction map reference frame type. Metal shields were intentionally placed to the North of the UT, while the UT is facing North. Consequently, the obstruction map in Fig. 2f indicates obstructions in the bottom-center region, aligning with the UT’s boresight direction. Additionally, as indicated by Fig. 2c and Fig. 2f, the reference point [1], i.e., the center pixel of the ellipse FOV in the FRAME_UT obstruction map, along with the major and minor axes, changes in response to the tilt adjustments of the UT.

遮挡图中的参考系 (Reference Frames in Obstruction Maps)

  • 两种参考系: 自 2024 年 9 月固件更新后, 根据用户的订阅计划, gRPC 获取的遮挡图使用两种不同的参考系.
  • FRAME_EARTH (固定用户): 适用于住宅(Residential)等固定位置套餐. 基于地心地固(ECEF)坐标系, 地图顶部中心对准正北方向, 与 App 中的俯视图一致.
  • FRAME_UT (移动用户): 适用于移动(Mobile/Roam)套餐及未激活设备. 基于 UT 的本地坐标系, 地图底部中心始终对准 UT 的视轴(boresight)方向(即设备前方). 移动中的 UT 必须依赖此本地对准来确定波束转向角度.
  • 硬件与订阅的关系: 即使是同一硬件(如 rev3_proto2), 切换订阅计划后参考系也会随之改变. 在 FRAME_UT 参考系下, 视场椭圆的中心及轴线会随 UT 的倾斜角度(tilt)调整而变化.