[论文解读] The Tale of Two Localization Technologies: Enabling Accurate Low-Overhead WiFi-based Localization for Low-end Phones

Ahmed Shokry, Moustafa Elhamshary+1 more|arXiv (Cornell University)|Jun 25, 2021

Indoor and Outdoor Localization Technologies参考文献 34被引用 25

一句话总结

混合定位 HybridLoc 借助具 BLE 的高端手机 crowdsourcing WiFi 指纹识别，以在仅有 WiFi 的低端设备上实现高准确性、低开销的室内定位，在动态条件下达到房间级到米级的精度。

ABSTRACT

WiFi fingerprinting is one of the mainstream technologies for indoor localization. However, it requires an initial calibration phase during which the fingerprint database is built manually. This process is labour intensive and needs to be repeated with any change in the environment. While a number of systems have been introduced to reduce the calibration effort through RF propagation models or crowdsourcing, these still have some limitations. Other approaches use the recently developed iBeacon technology as an alternative to WiFi for indoor localization. However, these beacon-based solutions are limited to a small subset of high-end phones. In this paper, we present HybridLoc: an accurate low-overhead indoor localization system. The basic idea HybridLoc builds on is to leverage the sensors of high-end phones to enable localization of lower-end phones. Specifically, the WiFi fingerprint is crowdsourced by opportunistically collecting WiFi-scans labeled with location data obtained from BLE-enabled high-end smart phones. These scans are used to automatically construct the WiFi-fingerprint, that is used later to localize any lower-end cell phone with the ubiquitous WiFi technology. HybridLoc also has provisions for handling the inherent error in the estimated BLE locations used in constructing the fingerprint as well as to handle practical deployment issues including the noisy wireless environment, heterogeneous devices, among others. Evaluation of HybridLoc using Android phones shows that it can provide accurate localization in the same range as manual fingerprinting techniques under the same conditions. Moreover, the localization accuracy on low-end phones supporting only WiFi is comparable to that achieved with high-end phones supporting BLE. This accuracy is achieved with no training overhead, is robust to the different user devices, and is consistent under environment changes.

研究动机与目标

• 说明在基于 WiFi 的室内定位中存在的高校准成本和设备异质性挑战。
• 提出一个众包感知框架，利用配备 BLE 的高端手机来自动构建 WiFi 指纹。
• 在低端设备上实现无需显式校准的准确 WiFi 定位。
• 在指纹构建中解决地面实况 BLE 定位误差、RSS 值缺失和设备异质性问题。
• 在真实环境中评估性能并与手工指纹技术进行比较。

提出的方法

• 通过将 WiFi RSS 与来自高端手机的 BLE 估算位置相关联，以众包方式构建 WiFi 指纹。
• 使用 IncVoronoi BLE 定位为离线指纹构建生成带置信度的位置标签。
• 基于网格（单元格）的指纹构建以减少校准开销，并使移动中的用户能够贡献数据。
• 对每个 AP 的 RSS 进行高斯建模以处理缺失值和噪声，从而实现概率指纹。
• 通过一个共同的 RSS 偏置 lambda 进行偏移校正（设备异质性），在无需对每个设备单独校准的情况下对齐设备。
• 通过贝叶斯规则进行离散位置估计，使用 P(s|g) 作为各 AP 高斯分布的乘积；通过质心和时间平均实现连续跟踪。

实验结果

研究问题

• RQ1 是否可以通过来自高端设备的基于 BLE 的地面真实定位可靠地为低端设备构建准确的 WiFi 指纹？
• RQ2 在设备异质性和嘈杂的无线信道条件下，基于网格的概率指纹识别性能如何？
• RQ3 地面实况定位精度对仅 WiFi 定位性能的影响有多大？
• RQ4 基于置信度的分配策略与仅基于位置的分配策略在准确性和开销方面的对比如何？

主要发现

• HybridLoc 的中位数精度在相同部署条件下可与手动指纹相当。
• 在 HybridLoc 下，低端仅 WiFi 定位的准确性与支持 BLE 的高端设备相当。
• 基于置信度的分配方法优于仅位置分配，其中加权置信度提供最佳或接近最佳的结果。
• 网格单元表示（质心）比几何中心略有更好地准确性。
• 在不同设备上进行训练和测试时，对设备异质性的处理使准确性提高约 14%。