👉 To download the full dataset, please visit Indoor Location Competition 2.0 Dataset.

👉 For more details on the competition including winning solutions, please visit our Kaggle competition site.

Indoor Location Competition 2.0 (Sample Data and Code)

This repository contains sample data and code for Indoor Location Competition 2.0, a continuation of Microsoft Indoor Location Competition. Competition this year will be completely virtual and evaluated on large-scale real indoor location datasets. The dataset to be released consists of dense indoor signatures of WiFi, geomagnetic field, iBeacons etc., as well as ground truth collected from hundreds of buildings in Chinese cities.

Webinar Video

We held a webinar in July, the video is here.

Sample Data

data folder contains indoor traces from two sites. Each trace (*.txt) corresponds to an indoor path between position p₁ and p₂ walked by a site-surveyor. During the walk, site-surveyor is holding an Android smartphone flat in front of his body, and a sensor data recording app is running on the device to collect IMU (accelerometer, gyroscope) and geomagnetic field (magnetometer) readings, as well as WiFi and Bluetooth iBeacon scanning results. A detailed description of the format of trace file is shown below. In addition to raw traces, floor plan metadata (e.g., raster image, size, GeoJSON) are also included for each floor.

Trace File Format（*.txt）

Time	Data Type	Value
1574659531598	TYPE_WAYPOINT	196.41757	117.84907
	Location surveyor labeled on the map	Coordinate x (meter)	Coordiante y (meter)
1574659531695	TYPE_ACCELEROMETER	-1.7085724	-0.274765	16.657166	2
	Android Sensor.TYPE_ACCELEROMETER	X axis	Y axis	Z axis	accuracy
1574659531695	TYPE_GYROSCOPE	-0.3021698	0.2773285	0.107543945	3
	Android Sensor.TYPE_GYROSCOPE	X axis	Y axis	Z axis	accuracy
1574659531695	TYPE_MAGNETIC_FIELD	20.181274	16.209412	-32.22046	3
	Android Sensor.TYPE_MAGNETIC_FIELD	X axis	Y axis	Z axis	accuracy
1574659531695	TYPE_ROTATION_VECTOR	-0.00855688	0.051367603	0.362504	3
	Android Sensor.TYPE_ROTATION_VECTOR	X axis	Y axis	Z axis	accuracy
1574659531695	TYPE_ACCELEROMETER_UNCALIBRATED	-1.7085724	-0.274765	16.657166	0.0	0.0	0.0	3
	Android Sensor.TYPE_ACCELEROMETER_UNCALIBRATED	X axis	Y axis	Z axis	X axis	Y axis	Z axis	accuracy
1574659531695	TYPE_GYROSCOPE_UNCALIBRATED	-0.42333984	0.20202637	0.09623718	-7.9345703E-4	3.2043457E-4	4.119873E-4	3
	Android Sensor.TYPE_GYROSCOPE_UNCALIBRATED	X axis	Y axis	Z axis	X axis	Y axis	Z axis	accuracy
1574659531695	TYPE_MAGNETIC_FIELD_UNCALIBRATED	-29.830933	-26.36261	-300.3006	-50.012207	-42.57202	-268.08014	3
	Android Sensor.TYPE_MAGNETIC_FIELD_UNCALIBRATED	X axis	Y axis	Z axis	X axis	Y axis	Z axis	accuracy
1574659533190	TYPE_WIFI	intime_free	0e:74:9c:a7:b2:e4	-43	5805	1574659532305
	Wi-Fi data	ssid	bssid	RSSI	frequency	last seen timestamp
1574659532751	TYPE_BEACON	FDA50693-A4E2-4FB1-AFCF-C6EB07647825	10073	61418	-65	-82	5.50634293288929	6B:11:4C:D1:29:F2	1574659532751
	iBeacon data	UUID	MajorID	MinorID	Tx Power	RSSI	Distance	MAC Address	same with Unix time, padding data

The first column is Unix Time in millisecond. In specific, we use SensorEvent.timestamp for sensor data and system time for WiFi and Bluetooth scans.

The second column is the data type (ten in total).

• TYPE_ACCELEROMETER
• TYPE_MAGNETIC_FIELD
• TYPE_GYROSCOPE
• TYPE_ROTATION_VECTOR
• TYPE_MAGNETIC_FIELD_UNCALIBRATED
• TYPE_GYROSCOPE_UNCALIBRATED
• TYPE_ACCELEROMETER_UNCALIBRATED
• TYPE_WIFI
• TYPE_BEACON
• TYPE_WAYPOINT: ground truth location labeled by the surveyor

Data values start from the third column.

Column 3-5 of TYPE_ACCELEROMETER、TYPE_MAGNETIC_FIELD、TYPE_GYROSCOPE、TYPE_ROTATION_VECTOR are SensorEvent.values[0-2] from the callback function onSensorChanged(). Column 6 is SensorEvent.accuracy.

Column 3-8 of TYPE_ACCELEROMETER_UNCALIBRATED、TYPE_GYROSCOPE_UNCALIBRATED、TYPE_MAGNETIC_FIELD_UNCALIBRATED are SensorEvent.values[0-5] from the callback function onSensorChanged(). Column 9 is SensorEvent.accuracy.

Values of TYPE_BEACON are obtained from ScanRecord.getBytes(). The results are decoded based on iBeacon protocol using the code below.

val major = ((scanRecord[startByte + 20].toInt() and 0xff) * 0x100 + (scanRecord[startByte + 21].toInt() and 0xff))
val minor = ((scanRecord[startByte + 22].toInt() and 0xff) * 0x100 + (scanRecord[startByte + 23].toInt() and 0xff))
val txPower = scanRecord[startByte + 24]

Distance in column 8 is calculated as

private static double calculateDistance(int txPower, double rssi) {
  if (rssi == 0) {
    return -1.0; // if we cannot determine distance, return -1.
  }
  double ratio = rssi*1.0/txPower;
  if (ratio < 1.0) {
    return Math.pow(ratio,10);
  }
  else {
    double accuracy =  (0.89976)*Math.pow(ratio,7.7095) + 0.111;
    return accuracy;
  }
}

References:

https://developer.android.com/guide/topics/sensors
https://developer.android.com/reference/android/net/wifi/ScanResult.html
https://developer.android.com/reference/android/bluetooth/le/ScanRecord

Sample Code

Along with sample data from two sites, this repo also provides several scripts on parsing and analyzing indoor traces. All scripts are tested with Python 3.6.9 on both Windows 10 and Mac OS 15.

How to run the code

python main.py

Main functions

Functions	Output
Ground truth location visualization	output/site1/F1/path_images
Sample step detection and visualization	output/site1/F1/step_position.html
Geo-magnetic field intensity visualization	output/site1/F1/magnetic_strength.html
WiFi RSSI heatmap generation	output/site1/F1/wifi_images
iBeacon RSSI heatmap generation	output/site1/F1/ibeacon_images
WiFi SSID counts visualization	output/site1/F1/wifi_count.html

Contents

indoor-location-competition-20
│   README.md
│   main.py                                                      //main function of the sample code
|   compute_f.py                                                 //data processing functions
|   io_f.py                                                      //data preprocessing functions
|   visualize_f.py                                               //visualization function
│
└───data                                                         //raw data from two sites
      └───site1
      |     └───B1                                               //traces from one floor
      |     |    └───path_data_files                             
      |     |    |          └───5dda14a2c5b77e0006b17533.txt     //trace file
      |     |    |          |   ...
      |     |    |
      |     |    |   floor_image.png                             //raster floor plan
      |     |    |   floor_info.json                             //floor size info
      |     |    |   geojson_map.json                            //floor plan in vector format (GeoJSON)
      |     |
      |     └───F1
      |     │   ...
      |
      └───site2
            │   ...

License

This repository is licensed with the MIT license.