Supplementary Materials for "Influence of Measured Radio Environment Map Interpolation on Indoor Positioning Algorithms"
收藏资源简介：
This dataset was created as suplementary material for research article: Influence of Measured Radio Environment Map Interpolation on Indoor Positioning Algorithms This package contains packet capture files of 802.11 probe requests captured at Geotec office at University Jaume I, Spain by 5 ESP32 microcontrollers. The packet capture files are in the standardized *.pcap binary format and can be opened with any packet analysis tool such as Wireshark or scapy (Python packet analysis and manipulation package). The data are split between radio map data captured at all accessible reference positions in our office spread in 1m grid and evaluation data gathered alligned to 0.5m grid, as well as in hard to access locations. The location the data were collected are available in the office. The dataset has 4 parts, and all subsets of the dataset can be generated from the captured pcap files: Data This folder contains pcap files from all 5 ESP32 stations representing the whole radio environment map. The folder name stands for each of the 5 ESP32 sniffer stations and the name of the file points to a reference location the data were captured in. Example of the coordinates matching the reference location grid names are in following table: Data Point Coordinates X Y X Y ... A1 0.85 0.1 B1 1.85 0.1 ... A2 0.85 1.1 B2 1.85 1.1 ... A3 0.85 2.1 B3 1.85 2.1 ... ... ... ... ... ... ... ... A11 0.85 10.1 B11 1.85 10.1 ... Data_Eval This folder contains pcap files from all 5 ESP32 stations with data captured at 31 locations not found in the original reference location grid. The naming corresponds to the X and Y location in which the data were collected. Processed_Data Additionally, there are 3 folders with processed CSV files. One folder that combines all radio map values, second folder contains combined evaluation values and third is with linearly interpolated radio map values. The CSV files are in a format: X, Y, RSSI_1, RSSI_2, RSSI_3, RSSI_4, RSSI_5 Data_Scenarios This folder for the ease of use, contains data for exact reproducibility of our results in the paper. There 14 scenarios described in the following table: Scenario Descriptions Data Name Scenario Description GPR00 Only measured data, 50 samples per reference position GPR01 Measured data with empty spots filled using Linear interpolation, 50 samples per reference position GPR02 Gaussian Regression trained only on measured data - 1m output grid, 50 samples per reference position GPR03 Gaussian Regression trained only on measured data - 0.5m output grid, 50 samples per reference position GPR04 Gaussian Regression trained on linearly interpolated data - 1m output grid, 50 samples per reference position GPR05 Gaussian Regression trained on linearly interpolated data - 0.5m output grid, 50 samples per reference position GPR06 Gaussian Regression trained selection of linearly interpolated data - 1m output grid, 50 samples per reference position GPR07 Gaussian Regression trained selection of linearly interpolated data - 0.5m output grid, 50 samples per reference position GPR08 Gaussian Regression trained only on measured data - 1m output grid, 1 sample per reference position GPR09 Gaussian Regression trained only on measured data - 0.5m output grid, 1 sample per reference position GPR10 Gaussian Regression trained on linearly interpolated data - 1m output grid, 1 sample per reference position GPR11 Gaussian Regression trained on linearly interpolated data - 0.5m output grid, 1 sample per reference position GPR12 Gaussian Regression trained selection of linearly interpolated data - 1m output grid, 1 sample per reference position GPR13 Gaussian Regression trained selection of linearly interpolated data - 0.5m output grid, 1 sample per reference position The folder contains 4 files for each scenario. The Beginning of the filename corresponds to the data name, with suffix describing what data are in the file. The descriptions of used suffixes are in the following table: File Suffix Descriptions Suffix Suffix Description _trncrd Training Labels _trnrss Training RSSI Values _tstcrd Evaluation Labels _tstrss Evaluation RSSI Values These data are in format compatible with systems that apart from X and Y coordinates also detect, building, floor etc. The RSSI data are in format: RSSI_1, RSSI_2, RSSI_3, RSSI_4, RSSI_5 The Labels are in format: (Since we only use positioning in 1 office, apart X and Y coordinates are set to 0) X, Y, 0, 0, 0
本数据集作为研究论文《实测无线电环境地图（Radio Environment Map）插值对室内定位算法的影响》的补充材料构建。本套件包含由5个ESP32微控制器在西班牙胡米利亚大学（University Jaume I）Geotec办公室采集的IEEE 802.11探测请求数据包捕获文件。数据包捕获文件采用标准化的*.pcap二进制格式，可通过Wireshark、scapy（Python数据包分析与操作库）等任意数据包分析工具打开。 本数据集涵盖两类采集数据：一是在办公室内以1米网格布设的所有可访问参考位置采集的无线电地图数据，二是以0.5米网格布设的评估数据，同时包含难以访问位置的采集数据。数据采集点位可在该办公室内直接获取。本数据集共包含4个部分，所有数据集子集均可从捕获的pcap文件生成： ### Data 文件夹 该文件夹包含代表完整无线电环境地图的5个ESP32嗅探节点的pcap文件。文件夹名称对应5个ESP32嗅探采集节点，文件名则指向数据采集的参考位置。与参考位置网格名称匹配的坐标示例如下： | 数据点位 | X坐标 | Y坐标 | |----------|-------|-------| | A1 | 0.85 | 0.1 | | B1 | 1.85 | 0.1 | | A2 | 0.85 | 1.1 | | B2 | 1.85 | 1.1 | | A3 | 0.85 | 2.1 | | B3 | 1.85 | 2.1 | | ... | ... | ... | | A11 | 0.85 | 10.1 | | B11 | 1.85 | 10.1 | | ... | ... | ... | ### Data_Eval 文件夹 该文件夹包含来自5个ESP32采集节点的pcap文件，其数据采集自原始参考位置网格未覆盖的31个点位。文件名与数据采集的X、Y坐标一一对应。 ### Processed_Data 文件夹 此外，该文件夹包含3个存储已处理CSV文件的子文件夹：其一合并了所有无线电地图数据，其二合并了评估数据，其三存储了经线性插值得到的无线电地图数据。CSV文件格式如下： `X, Y, RSSI_1, RSSI_2, RSSI_3, RSSI_4, RSSI_5` 其中RSSI为接收信号强度指示（Received Signal Strength Indicator）。 ### Data_Scenarios 文件夹 该文件夹为便于实验复现，包含可直接复现论文中实验结果的完整数据集，共涵盖14种实验场景，如下表所示： | 场景编号 | 场景描述 | |----------|----------| | GPR00 | 仅使用实测数据，每个参考位置采集50个样本 | | GPR01 | 使用经线性插值填补空白点位的实测数据，每个参考位置采集50个样本 | | GPR02 | 仅基于实测数据训练的高斯回归模型，输出网格分辨率为1米，每个参考位置采集50个样本 | | GPR03 | 仅基于实测数据训练的高斯回归模型，输出网格分辨率为0.5米，每个参考位置采集50个样本 | | GPR04 | 基于线性插值数据训练的高斯回归模型，输出网格分辨率为1米，每个参考位置采集50个样本 | | GPR05 | 基于线性插值数据训练的高斯回归模型，输出网格分辨率为0.5米，每个参考位置采集50个样本 | | GPR06 | 基于筛选后的线性插值数据训练的高斯回归模型，输出网格分辨率为1米，每个参考位置采集50个样本 | | GPR07 | 基于筛选后的线性插值数据训练的高斯回归模型，输出网格分辨率为0.5米，每个参考位置采集50个样本 | | GPR08 | 仅基于实测数据训练的高斯回归模型，输出网格分辨率为1米，每个参考位置采集1个样本 | | GPR09 | 仅基于实测数据训练的高斯回归模型，输出网格分辨率为0.5米，每个参考位置采集1个样本 | | GPR10 | 基于线性插值数据训练的高斯回归模型，输出网格分辨率为1米，每个参考位置采集1个样本 | | GPR11 | 基于线性插值数据训练的高斯回归模型，输出网格分辨率为0.5米，每个参考位置采集1个样本 | | GPR12 | 基于筛选后的线性插值数据训练的高斯回归模型，输出网格分辨率为1米，每个参考位置采集1个样本 | | GPR13 | 基于筛选后的线性插值数据训练的高斯回归模型，输出网格分辨率为0.5米，每个参考位置采集1个样本 | 每个场景对应4个数据文件，文件名前缀对应数据场景名称，后缀则说明文件内包含的数据类型。后缀的含义如下表所示： | 文件后缀 | 含义说明 | |----------|----------| | _trncrd | 训练集标签 | | _trnrss | 训练集接收信号强度指示值 | | _tstcrd | 测试集标签 | | _tstrss | 测试集接收信号强度指示值 | 本数据集格式兼容支持除X、Y坐标外，还可检测建筑、楼层等附加信息的系统。接收信号强度指示数据格式为：`RSSI_1, RSSI_2, RSSI_3, RSSI_4, RSSI_5`。标签数据格式为：（由于本实验仅在单个办公室内开展定位任务，除X、Y坐标外其余字段均设为0）`X, Y, 0, 0, 0`。
