Automatic radio map construction for CSI-based fingerprinting with LTE

Fingerprint-based indoor localization methods require the construction of a radio map. This is a tool automating this tedious task, by having a robot follow a line, and stop every x cm to gather location-dependent characteristics (i.e. a fingerprint).

Summary

Fingerprint-based indoor localization methods require the construction of a radio map. This is a tedious task which can be automated. The presented solution consists of a Software-Defined Radio (SDR) mounted on a wheeled-robot (Thymio II). The SDR is connected to a LTE tower using a modified version of the srsLTE software. The robot follows a line, stopping every x cm to let the SDR gather characteristics (Channel State Information (CSI), Received Signal Strength Indicator (RSSI), Reference Signal Receive Power (RSRP), ...) from its communication with the LTE tower. Those characteristics are saved, along with the robot's location (estimated through dead-reckoning), and are ready to be used as fingerprints.

---

Requirements

• Thymio II
• USRP B200mini
• PC/SC reader with a SIM card

---

How to install

1. Install a specific fork of srsLTE following this

   This fork saves the channel characteristics (CSI, RSSI, RSRP, ...) to a file every ms

   Make sure SRSUE_FOLDERPATH in run.py points to the correct location (of the srsue folder, from the srsLTE installation)

2. Install the Thymio Suite and prepare the connection with the Thymio II following this

3. Install Aseba following this

   Run asebamedulla --version to ensure it is correctly installed

4. Ensure that you're able to connect to the Thymio II by launching asebastudio through some terminal. You'll be prompted to select a target: select the one called Thymio-II. This should establish the connection.

   If clicking on the Thymio-II target simply re-shows the same popup, this is a known bug. To fix it, add yourself to the dialout group (sudo adduser <USERNAME> dialout), replacing <USERNAME> by your Ubuntu username. Log out then log in for the change to take effect.

   Make sure the Thymio II is connected to the computer by USB

5. Create conda environment (called thymio) and activate it:

   conda env create -f environment.yml -n thymio
   conda activate thymio

6. Install dbus, gobject and sox:

   pip install dbus-python

   To install dbus-python, you might need to first run sudo apt install git virtualenv build-essential python3-dev libdbus-glib-1-dev libgirepository1.0-dev libcairo2-dev

   pip install PyGObject

   sudo apt install sox

---

How to run

There are two ways to run the software.

• Continuous: Establish once the connection between the SDR and the eNodeB, then keep it alive by pinging its IP address.

• Non-Continuous: Re-establish the connection between the SDR and the eNodeB everytime the Thymio II stops.

Continuous (Connect once)

Setup

1. Modify the constants defined in run_continuous.py to modify e.g.:
   • Distance travelled by the Thymio II between each stop
   • Amount of stops to make before fully stopping the recording
   • Initial position of the Thymio II
   • ...

Run

1. Connect by USB the Thymio II, the Software-Defined Radio and the PC/SC reader (with a SIM card plugged in) to the computer

2. Open a terminal and establish the connection to the Thymio II by running:

   asebamedulla "ser:name=Thymio-II"

   This opens a connection to the Thymio II. The terminal should read Found Thymio-II on port /dev/ttyACM0 and keep running.

3. Open another terminal and launch srsUE from this repo's root directory

   sudo srsue <PATH_TO_SRSLTE>/srsue/ue.conf

   Replace <PATH_TO_SRSLTE> by the path to the fork of the srsLTE software, and ue.conf by the configuration file you want to run

4. Wait for the SDR to connect (the terminal will somewhere output Network attach: 10.xx.xx.xx)

   The fingerprint files (i.e. ce.txt, else.txt and info.txt) should appear in this repo's root directory. Watch out, because ce.txt will grow quickly in size.

5. Open another terminal and ping the above IP address by running

   ping 10.xx.xx.xx

   This maintains the connection between the SDR and the eNodeB. Otherwise, no traffic is sent and the connection eventually times out.

6. Open another terminal and launch the run script by running:

   conda activate thymio
   python run_continuous.py

Non-continuous (Reconnect to the eNodeB at every stop) [DEPRECATED]

1. Connect by USB the Thymio II, the Software-Defined Radio and the PC/SC reader (with a SIM card plugged in) to the computer

2. Open a terminal and establish the connection to the Thymio II by running:

   asebamedulla "ser:name=Thymio-II"

   This opens a connection to the Thymio II. The terminal should read Found Thymio-II on port /dev/ttyACM0 and keep running.

3. Open another terminal and launch the run script by running:

   conda activate thymio
   python run.py

---

Test files

To ensure that the Thymio and SDR are working, several files are available:

• test_thymio.py
• test_fingerprint.py

test_thymio.py

To ensure that the Thymio II is able to work as intended:

1. Plug in the Thymio II through USB and place the Thymio II on the line to follow

2. Open a terminal and establish the connection to the Thymio II by running:

   asebamedulla "ser:name=Thymio-II"

3. Open another terminal and launch the test script by running:

   conda activate thymio
   python test_thymio.py

   This should instruct the Thymio to follow the black-line for x cm, then wait, then repeat y times

test_fingerprint.py

To ensure that the Software Defined Radio works as intended:

1. Connect via USB the Software-Defined Radio (SDR) and the PC/SC reader (with a SIM card plugged in) to the computer

   For the SDR, you have to used USB 3.0, else the datarate is too high for the connection.

2. Open a terminal and launch the test script by running:

   conda activate thymio
   python test_fingerprint.py

   This should instruct the SDR to connect to the eNodeB, gather a fingerprint and save it in a folder called test_fingerprint

---

How to handle the generated files

Some python scripts were written to facilitate the data gathering step, e.g.:

• move_fingerprint.py:

  • Move the files generated by sudo srsue ue.conf (i.e. ce.txt, else.txt and info.txt) into their own folders (ce, else, info)
  • Record the (x,y) coordinates where those files were gathered into a JSON file (i.e. locations.json)

• clean_fingerprints.py

  • Clean the fingerprint files (i.e. ce_0_raw.txt, else_0_raw.txt and info_0_raw.txt)
  • Save the cleaned files, e.g.:
    • ce_0_raw.txt -> ce_0.parquet
    • else_0_raw.txt -> else_0.pkl
    • info_0_raw.txt -> info_0.pkl

• offset_origin_locations.py

  • Offset the (x,y) coordinates of the origin after the recording.

• plot_fingerprints.py
  • Plot the amplitude and phase held in the cleaned ce.txt (e.g. ce_0.parquet)
  • Save the plot, e.g.:
    • ce_0.parquet -> ce_0.png

---

Misc

Tested with

• Ubuntu 18.04
• Aseba medulla 1.6.1