[论文解读] Improving the Performance of OTDOA based Positioning in NB-IoT Systems

Sha Hu, Axel Berg+2 more|arXiv (Cornell University)|Apr 18, 2017

Power Line Communications and Noise参考文献 4被引用 4

一句话总结

本文提出一种用于窄带物联网（NB-IoT）系统中基于OTDOA的定位的EM-SIC算法，该算法在低采样率（例如1.92 MHz）下联合估计剩余频偏、多径衰落和到达时间（ToA）。通过使用迭代干扰消除和基于插值的ToA精化方法，该方法在信噪比（SNR）增益和定位精度方面表现显著，尤其在小区间干扰和剩余频偏条件下优于传统检测器，在AWGN和瑞利衰落（ETU-3Hz）信道中均表现更优。

ABSTRACT

In this paper, we consider positioning with observed-time-difference-of-arrival (OTDOA) for a device deployed in long-term-evolution (LTE) based narrow-band Internet-of-things (NB-IoT) systems. We propose an iterative expectation-maximization based successive interference cancellation (EM-SIC) algorithm to jointly consider estimations of residual frequency-offset (FO), fading-channel taps and time-of-arrival (ToA) of the first arrival-path for each of the detected cells. In order to design a low complexity ToA detector and also due to the limits of low-cost analog circuits, we assume an NB-IoT device working at a low-sampling rate such as 1.92 MHz or lower. The proposed EM-SIC algorithm comprises two stages to detect ToA, based on which OTDOA can be calculated. In a first stage, after running the EM-SIC block a predefined number of iterations, a coarse ToA is estimated for each of the detected cells. Then in a second stage, to improve the ToA resolution, a low-pass filter is utilized to interpolate the correlations of time-domain PRS signal evaluated at a low sampling-rate to a high sampling-rate such as 30.72 MHz. To keep low-complexity, only the correlations inside a small search window centered at the coarse ToA estimates are upsampled. Then, the refined ToAs are estimated based on upsampled correlations. If at least three cells are detected, with OTDOA and the locations of detected cell sites, the position of the NB-IoT device can be estimated. We show through numerical simulations that, the proposed EM-SIC based ToA detector is robust against impairments introduced by inter-cell interference, fading-channel and residual FO. Thus significant signal-to-noise (SNR) gains are obtained over traditional ToA detectors that do not consider these impairments when positioning a device.

研究动机与目标

• 为解决NB-IoT设备在低采样率和低成本射频硬件限制下实现低复杂度、高精度定位的挑战。
• 缓解基于OTDOA的定位中因小区间干扰、剩余频偏（FO）和多径衰落导致的性能下降问题。
• 设计一种实用的ToA检测方法，在保持低计算复杂度的同时，通过互相关插值实现精细时间分辨率。
• 在密集蜂窝网络部署中实现鲁棒定位，其中由于延迟增加，PRS信道抑制（muted PRS）不可行。

提出的方法

• EM-SIC算法通过低采样率（例如1.92 MHz）的时域PRS信号，迭代估计剩余频偏、多径信道抽头以及首次到达路径的ToA。
• 该方法采用两阶段ToA检测：首先，在固定迭代次数后，通过EM-SIC获得粗略的ToA估计。
• 其次，仅将每个粗略ToA中心的小范围搜索窗口内的相关值，使用低通插值方法上采样至更高采样率（例如30.72 MHz），以提高ToA分辨率。
• 迭代多径检测（MPD）算法利用PRS信号的自相关函数（ACF）特性，提升首次到达路径ToA检测的准确性。
• 该算法结合连续干扰消除（SIC）技术，以抑制来自强邻小区的小区间干扰。
• 首先应用基于阈值的检测步骤（使用公式12）以确认PRS的存在，随后进行迭代优化。

实验结果

研究问题

• RQ1 如何使基于OTDOA的定位在典型低功耗NB-IoT设备常见的低采样率约束下保持鲁棒性？
• RQ2 剩余频偏和小区间干扰对NB-IoT定位中ToA估计精度有何影响？
• RQ3 通过EM-SIC联合估计衰落信道、剩余频偏和ToA，是否能相比传统检测器提升定位性能？
• RQ4 在不增加计算负载的前提下，低速率相关性的插值能在多大程度上提升ToA分辨率？
• RQ5 与理想AWGN条件相比，该方法在真实衰落信道（如ETU-3Hz）下的表现如何？

主要发现

• 所提出的EM-SIC检测器相比未考虑剩余频偏、小区间干扰或衰落信道的传统ToA检测器，实现了显著的SNR增益。
• 在AWGN信道下，FOC与上采样结合可显著提升ToA估计精度，降低估计误差方差。
• 在ETU-3Hz衰落信道中，尽管上采样带来的增益有限（因ToA估计本身较差），但FOC过程仍显著提升了定位成功率。
• 与非IC检测器相比，EM-SIC检测器的定位比率（误差小于500米的成功率）明显更高，尤其在干扰受限场景中表现更优。
• 该方法通过仅在粗略ToA估计附近的小范围搜索窗口内进行插值，避免全带宽上采样，从而保持了低复杂度。