WiFi RSS & RTT dataset with different LOS conditions for indoor positioning

This is the second batch of WiFi RSS RTT datasets with LOS conditions we published. Please see https://doi.org/10.5281/zenodo.11558192 for the first release. We provide three real-world datasets for indoor positioning model selection purpose. We divided the area of interest was divided into discrete grids and labelled them with correct ground truth coordinates and the LoS APs from the grid. The dataset contains WiFi RTT and RSS signal measures and is well separated so that training points and testing points will not overlap. Please find the datasets in the 'data' folder. The datasets contain both WiFi RSS and RTT signal measures with groud truth coordinates label and LOS condition label. Lecture theatre: This is a entirely LOS scenario with 5 APs. 60 scans of WiFi RTT and RSS signal measures were collected at each reference point (RP). Corridor: This is a entirely NLOS scenario with 4 APs. 60 scans of WiFi RTT and RSS signal measures were collected at each reference point (RP). Office: This is a mixed LOS-NLOS scenario with 5 APs. At least one AP was NLOS for each RP. 60 scans of WiFi RTT and RSS signal measures were collected at each reference point (RP). Collection methodology The APs utilised were Google WiFi Router AC-1304, the smartphone used to collect the data was Google Pixel 3 with Android 9. The ground truth coordinates were collected using fixed tile size on the floor and manual post-it note markers. Only RTT-enabled APs were included in the dataset. The features of the dataset The features of the lecture theatre dataset are as follows: Testbed area: 15 × 14.5 m2
Grid size: 0.6 × 0.6 m2 Number of AP: 5
Number of reference points: 120
Samples per reference point: 60
Number of all data samples: 7,200
Number of training samples: 5,400
Number of testing samples: 1,800
Signal measure: WiFi RTT, WiFi RSS
Note: Entirely LOS
The features of the corricor dataset are as follows: Testbed area: 35 × 6 m2
Grid size: 0.6 × 0.6 m2 Number of AP: 4
Number of reference points: 114
Samples per reference point: 60
Number of all data samples: 6,840
Number of training samples: 5,130
Number of testing samples: 1,710
Signal measure: WiFi RTT, WiFi RSS
Note: Miexed LOS-NLOS. At least one AP was NLOS for each RP.
The features of the office dataset are as follows: Testbed area: 18 × 5.5 m2
Grid size: 0.6 × 0.6 m2 Number of AP: 5
Number of reference points: 108
Samples per reference point: 60
Number of all data samples: 6,480
Number of training samples: 4,860
Number of testing samples: 1,620
Signal measure: WiFi RTT, WiFi RSS
Note: Entirely NLOS
Dataset explanation The columns of the dataset are as follows: Column 'X': the X coordinates of the sample.
Column 'Y': the Y coordinates of the sample.
Column 'AP1 RTT(mm)', 'AP2 RTT(mm)', ..., 'AP5 RTT(mm)': the RTT measure from corresponding AP at a reference point.
Column 'AP1 RSS(dBm)', 'AP2 RSS(dBm)', ..., 'AP5 RSS(dBm)': the RSS measure from corresponding AP at a reference point.
Column 'LOS APs': indicating which AP has a LOS to this reference point.
Please note: The RSS value -200 dBm indicates that the AP is too far away from the current reference point and no signals could be heard from it. The RTT value 100,000 mm indicates that no signal is received from the specific AP. Citation request When using this dataset, please cite the following three items: Feng, X., Nguyen, K. A., & Zhiyuan, L. (2024). WiFi RSS & RTT dataset with different LOS conditions for indoor positioning [Data set]. Zenodo. https://doi.org/10.5281/zenodo.11558792 @article{feng2024wifi,
title={A WiFi RSS-RTT indoor positioning system using dynamic model switching algorithm},
author={Feng, Xu and Nguyen, Khuong An and Luo, Zhiyuan},
journal={IEEE Journal of Indoor and Seamless Positioning and Navigation},
year={2024},
publisher={IEEE}
} @inproceedings{feng2023dynamic,
title={A dynamic model switching algorithm for WiFi fingerprinting indoor positioning},
author={Feng, Xu and Nguyen, Khuong An and Luo, Zhiyuan},
booktitle={2023 13th International Conference on Indoor Positioning and Indoor Navigation (IPIN)},
pages={1--6},
year={2023},
organization={IEEE}
}

本数据集为我们发布的第二批包含视距（Line-of-Sight, LOS）条件的WiFi接收信号强度（Received Signal Strength, RSS）与往返时间（Round-Trip Time, RTT）数据集。首批数据集请访问：https://doi.org/10.5281/zenodo.11558192 查阅。

本数据集包含三个真实场景数据集，用于室内定位模型的选型工作。

我们将目标区域划分为离散网格，并为每个网格标注准确的真实坐标（ground truth coordinates）以及该网格对应的视距接入点（LOS APs）。本数据集的训练样本点与测试样本点严格分离，无重叠，包含WiFi RTT与RSS信号测量数据。数据集文件存放于'data'文件夹中。

所有数据集均包含WiFi RSS与RTT信号测量数据，同时标注了真实坐标与视距条件标签。

演讲厅场景：该场景为全视距（LOS）环境，部署5个接入点（Access Point, AP）。每个参考点（Reference Point, RP）均采集了60组WiFi RTT与RSS信号测量样本。

走廊场景：该场景为全非视距（Non-Line-of-Sight, NLOS）环境，部署4个接入点。每个参考点均采集了60组WiFi RTT与RSS信号测量样本。

办公室场景：该场景为混合视距-非视距（LOS-NLOS）环境，部署5个接入点，每个参考点至少存在1个非视距接入点。每个参考点均采集了60组WiFi RTT与RSS信号测量样本。

数据采集方法

本次实验使用的接入点为Google WiFi Router AC-1304，数据采集所用智能手机为搭载Android 9系统的Google Pixel 3。真实坐标通过地面固定瓷砖尺寸结合手工便利贴标记的方式获取。本数据集仅包含支持RTT功能的接入点相关数据。

数据集参数详情

#### 演讲厅数据集
测试区域：15 × 14.5 平方米
网格尺寸：0.6 × 0.6 平方米
接入点数量：5
参考点数量：120
每个参考点采样数：60
总样本数：7200
训练样本数：5400
测试样本数：1800
测量信号类型：WiFi RTT、WiFi RSS
备注：全视距（LOS）场景

#### 走廊数据集
测试区域：35 × 6 平方米
网格尺寸：0.6 × 0.6 平方米
接入点数量：4
参考点数量：114
每个参考点采样数：60
总样本数：6840
训练样本数：5130
测试样本数：1710
测量信号类型：WiFi RTT、WiFi RSS
备注：混合视距-非视距（LOS-NLOS）场景，每个参考点至少存在1个非视距接入点

#### 办公室数据集
测试区域：18 × 5.5 平方米
网格尺寸：0.6 × 0.6 平方米
接入点数量：5
参考点数量：108
每个参考点采样数：60
总样本数：6480
训练样本数：4860
测试样本数：1620
测量信号类型：WiFi RTT、WiFi RSS
备注：全非视距（NLOS）场景

数据集字段说明

数据集各字段说明如下：
字段'X'：样本的X轴坐标
字段'Y'：样本的Y轴坐标
字段'AP1 RTT(mm)'、'AP2 RTT(mm)'……'AP5 RTT(mm)'：对应接入点在该参考点处的RTT测量值（单位：毫米）
字段'AP1 RSS(dBm)'、'AP2 RSS(dBm)'……'AP5 RSS(dBm)'：对应接入点在该参考点处的RSS测量值（单位：dBm）
字段'LOS APs'：标注与该参考点存在视距连接的接入点

注意事项：当RSS测量值为-200 dBm时，表示该接入点距离当前参考点过远，无法接收到其信号；当RTT测量值为100000 mm时，表示未接收到该接入点的信号。

引用要求

使用本数据集时，请引用以下三项内容：
1. 数据集：Feng, X., Nguyen, K. A., & Luo, Z. (2024). WiFi RSS & RTT dataset with different LOS conditions for indoor positioning [Data set]. Zenodo. https://doi.org/10.5281/zenodo.11558792
2. 期刊论文：Feng, Xu, Nguyen, Khuong An, & Luo, Zhiyuan. A WiFi RSS-RTT indoor positioning system using dynamic model switching algorithm[J]. IEEE Journal of Indoor and Seamless Positioning and Navigation, 2024.
3. 会议论文：Feng, Xu, Nguyen, Khuong An, & Luo, Zhiyuan. A dynamic model switching algorithm for WiFi fingerprinting indoor positioning[C]//2023 13th International Conference on Indoor Positioning and Indoor Navigation (IPIN). IEEE, 2023: 1-6.

搜集汇总

背景与挑战

背景概述

该数据集包含三种不同LOS条件下的WiFi RSS和RTT信号测量数据，适用于室内定位研究。数据覆盖讲座厅、走廊和办公室三个场景，每个场景有详细的信号测量和地面真实坐标，适合模型训练和测试。