跳转至

Experiments

A. Experimental Setup

Satellite Constellation: First, experiments are carried out driven by the Starlink and Kuiper constellation traces that are obtained from [27]. And we can also get the information of ground stations from [28]. Finally, we have built a platform for simulating real-trace dynamics of LEO satellite constellation using skyfield [29] based on these constellation and ground stations data.

System Prototype: Driven by the above platform, we have built a prototype by combining UERANSIM [30] with Open5GS [31] for 5G and beyond non-terrestrial network, which achieve handover in mobile satellite network. UERANSIM is a widely utilized simulator for both UE and SgNBs implementation in 5G network, while Open5GS is employed for implementing the 5G core network. Following the 5G standard signaling flow, we have made modifications to UERANSIM to support Xn-based handover, as shown in Fig. 7. The built prototype operates on a commodity laptop with 2.5 GHz CPU core and 16 GB RAM.

Comparison Schemes: We compare our proposed handover scheme with the following three handover schemes, in order to demonstrate the higher efficiency of the proposed handover strategy.

• NTN refers to the standard handover process specified in 5G NTN, which is described in II-B. We compare the proposed handover scheme with the NTN handover scheme to show the performance improvement brought by the modified handover.

• NTN-GS refers to the handover procedure assisted by nearby ground stations, which implies that ground stations close to the LEO satellite are leveraged to record handover information, inspired by the handover strategies designed in IP satellite network [32], [33]. Thus, the controlling signaling for handover does not need to be transmitted to core network and then the handover latency can be reduced.

• NTN-SMN refers to the handover procedure assisted by nearby space network (SMN) [34]. Similar with the NTNGS handover scheme, satellites act as the access and core network simultaneously in this strategy. In this way, the handover latency can also be reduced.

不说了，直接看论文吧

B. Experimental Results

该研究的实验结果表明，其提出的新型切换方案性能卓越且具备可行性，主要结论如下：

  1. 切换延迟显著降低：

    • 新方案的平均切换延迟仅为 21 ms，相比标准5G NTN方案的250 ms，性能提升超过10倍，且远优于其他两种优化方案（NTN-GS和NTN-SMN）
    • 对接入卫星进行同向选择的优化至关重要，能将延迟再降低约6倍
    • 该方案在星链（Starlink）和柯伊伯（Kuiper） 星座上均表现稳定，证明了其普适性

  2. 用户体验显著改善：

    • 新方案能将用户的通用网络中断时间减少89%，TCP应用的中断时间减少33%，直接提升了用户体验
    • TCP性能提升受限的主要瓶颈在于卫星间固有的长传播延迟，而非切换方案本身

  3. 预测算法计算可行：

    • 核心的预测算法在普通笔记本上处理一个地面站服务1万名用户的计算任务耗时仅约2秒，满足系统5秒更新一次的要求
    • 未来用户量增加带来的计算压力可通过升级硬件解决

  4. 方案鲁棒性高，异常情况可控：

    • 对于正常移动的用户（包括飞机），预测不准导致的 “异常切换”概率极低（如 10−6 量级），几乎可以忽略
    • 仅在用户高速移动且长时间（如10分钟）无网络活动的极端情况下，异常切换概率会升高（7%-18%）。但在这种情况下，系统会自动回退到标准5G NTN切换流程，保证了连接的可靠性