This study evaluates the effectiveness of using detailed cellular network signalling data for travel time estimation and route classification. Here, the authors propose a processing pipeline for estimating travel times and route classification based on Cell ID and received signal strength (RSS) measurements from a cellular network. The pipeline combines cellular fingerprinting, particle filtering, integrity monitoring, and map matching based on a hidden Markov model (HMM). The method is evaluated using a dataset of 11,000 cellular RSS measurements with corresponding GPS locations for the city of Norrköping, Sweden. The basic fingerprinting method has a CEP-67 location accuracy of 111 m and both particle filtering and integrity monitoring improved the results: 79 and 38 m for particle filtering and particle filtering with integrity monitoring, respectively. The route classification method resulted in a precision of 0.83 and a recall of 0.92, which are clear improvements compared to basic map matching of fingerprinting estimates. This new type of noise-adaptive travel time sampling in combination with an HMM-based route classification shows promising results and can potentially support large-scale estimates of both route choice and travel times using detailed cellular network signalling data in urban areas.

中文翻译：

---

基于蜂窝网络信令数据的行程时间估计的集成跟踪和路线分类

这项研究评估了使用详细的蜂窝网络信令数据进行旅行时间估计和路线分类的有效性。在这里，作者提出了一种处理管道，用于基于蜂窝网络的ID和接收信号强度（RSS）测量来估计旅行时间和路线分类。该管道结合了基于隐马尔可夫模型（HMM）的蜂窝指纹，颗粒过滤，完整性监控和地图匹配。使用瑞典Norrköping市的11,000个蜂窝RSS测量数据集以及相应的GPS位置来评估该方法。基本的指纹识别方法的CEP-67定位精度为111 m，并且粒子过滤和完整性监控均改善了结果：粒子过滤和粒子过滤以及完整性监控的结果分别为79和38 m，分别。路线分类方法的精度为0.83，召回率为0.92，与指纹估计的基本地图匹配相比，明显改善。这种新型的噪声自适应行进时间采样与基于HMM的路线分类相结合，显示出令人鼓舞的结果，并可能使用市区内详细的蜂窝网络信令数据来支持路线选择和行进时间的大规模估计。