Dataset: Ultra-Wideband Ranging Measurements Acquired With Three Different Platforms (Qorvo, TDSR, 3db Access)

This dataset contains distance measurements acquired with three different ultra-wideband (UWB) platforms developed by Qorvo (DW3000), TDSR (P452A), and 3db Access (3DB6380C) at the same locations. The dataset accompanies the paper: "Challenges in Platform-Independent UWB Ranging and Localization Systems" by Laura Flueratoru, Elena Simona Lohan, Dragoș Niculescu, published in the 16th ACM Workshop on Wireless Network Testbeds, Experimental evaluation and Characterization (WiNTECH) 2022. If you find this dataset useful, please consider citing our paper. The dataset (<strong>uwb_multiple_platforms.zip</strong>) contains the following directories: <strong>parallel_measurements</strong> -- The actual dataset, containing all the measurements acquired with the three UWB platforms at the same locations. This directory contains three subdirectories, one for each device. The structure of the subdirectory of each platform is the following: <strong>[location_name] </strong> <strong>[LOSi/NLOSi]</strong> -- where i is the index of the recording and LOS/NLOS indicates whether that recording was acquired in LOS or NLOS <strong>info.csv</strong> -- CSV file which contains information about the recording, such as: the device it was acquired with, the LOS/NLOS condition, the type of obstruction (if any), etc. <strong>unaligned_processed_data.csv</strong> -- CSV file which contains the data. Each row has the following fields: timestamp, true distance, measured distance, time of arrival index, channel impulse response (stored as a list), and the LOS/NLOS label. <strong>split_train_test_val</strong> -- Datasets that were used to train and test the models from Section 4 from the paper. The datasets contain the same information as the directory <strong>parallel_measurements</strong>, only aligned to the TOA and randomized according to the procedure described in the paper. We include the generated sets to ensure the repeatability of our results. <strong>trained_models_error_prediction</strong> -- Models trained for error prediction that were used to obtain the results from Section 4 from the paper. We also provide code examples for reading the data, training and testing the models, and analyzing the data at the following repository: https://github.com/lauraflu/uwb-multiple-platforms The accompanying code is subject to change in the case of bugs/errors. For more information about how the measurements were acquired, please refer to the file <strong>documentation_dataset.pdf</strong>, which includes detailed information about each of the rooms, the device setup, the structure of the directories, etc. For any questions, do not hesitate to contact the authors of the paper. <strong>Note</strong>: The dataset (in the <strong>parallel_measurements</strong> directory) contains measurements acquired with the devices at fixed locations and also "free movement" measurements, during which one of the devices was moved freely around a certain area. Therefore, free-movement recordings with the same name but from different devices were <em>not</em> acquired at exactly the same locations, only in the same rooms. The free-movement recordings were not used in the paper (because they do not contain ground truth distances), but we nevertheless include them in this dataset, as they can be useful to test future algorithms.本数据集包含由Qorvo（DW3000）、TDSR（P452A）及3db Access（3DB6380C）开发的三款不同超宽带（ultra-wideband, UWB）平台在同一测点采集的距离测量数据。本数据集配套论文《面向平台无关的UWB测距与定位系统的挑战》（*Challenges in Platform-Independent UWB Ranging and Localization Systems*），作者为Laura Flueratoru、Elena Simona Lohan、Dragoș Niculescu，发表于2022年第16届ACM无线网络测试床、实验评估与特性化研讨会（WiNTECH 2022）。若本数据集对您的研究有所助益，请引用该论文。 数据集文件为**uwb_multiple_platforms.zip**，包含以下目录： 1. **parallel_measurements**：核心数据集，存储三款UWB平台在同一测点采集的全部测量数据。该目录下设三个子目录，分别对应每一款设备。各平台子目录的结构如下： - **[location_name] [LOSi/NLOSi]**：其中`i`为采集记录的索引，LOS（视距，Line-of-Sight）/NLOS（非视距，Non-Line-of-Sight）表示该次采集所处的传输环境。 - **info.csv**：记录采集信息的逗号分隔值（Comma-Separated Values, CSV）文件，包含采集所用设备、LOS/NLOS环境、遮挡类型（若存在）等字段。 - **unaligned_processed_data.csv**：存储原始处理数据的CSV文件，每行包含以下字段：时间戳、真实距离、测量距离、到达时间（Time of Arrival, TOA）索引、以列表形式存储的信道冲激响应，以及LOS/NLOS标签。 2. **split_train_test_val**：用于训练与测试论文第4节所述模型的数据集。该数据集与**parallel_measurements**目录中的数据内容一致，仅针对TOA进行了对齐，并按照论文所述流程进行了随机划分。我们提供该划分后的数据集以确保研究结果的可复现性。 3. **trained_models_error_prediction**：用于误差预测的训练模型，对应论文第4节的实验结果。 我们在以下开源仓库中提供了数据读取、模型训练与测试、以及数据分析的示例代码：https://github.com/lauraflu/uwb-multiple-platforms。配套代码若存在漏洞或错误，可能会进行更新。 有关测量采集的更多信息，请参阅**documentation_dataset.pdf**文件，该文件包含各测试房间、设备部署、目录结构等详细说明。如有任何疑问，请联系论文作者。 **注意**：**parallel_measurements**目录中的数据集包含两部分测量数据：固定测点采集的数据，以及“自由移动”采集的数据——后者指其中一台设备在特定区域内自由移动时采集的数据。因此，不同设备同名的自由移动记录并非在完全相同的测点采集，仅处于同一房间内。本论文未使用自由移动采集的数据（因其未包含真实距离标签），但我们仍将其纳入本数据集，以供未来算法测试使用。