Open6GHub and WiFi Sensing at Hannover Messe 2023

Introduction to Channel Charting

We use wireless communication signals to reconstruct a map of the environment in a self-supervised manner.
First, Channel State Information (CSI) is collected at a massive MIMO antenna array.
CSI contains information about angles, reflections, powers and more.
In Stuttgart, we collect CSI with a distributed prototype base station or with a standard-compliant WiFi antenna array ("ESPARGOS").
Here, we visualize CSI from two WiFi transmitters.
We can estimate the signal's angle of arrival in realtime.
We collect so much data that we cannot visualize all of it here.

From CSI to Channel Chart
The propagation channel maps each physical position into CSI space.
This is like crumpling up a map into a high-dimensional space.
In Channel Charting, we use similarity relations to un-crumple the map.
Here, a neural network learns to reconstruct the map.

The data for this experiment was collected in a 20×20m area in a factory.
The antenna arrays are tracking a robot with a transmitter.

Applications of Channel Charting
We can use the reconstructed map for real-time localization.
The red dot is the true transmitter position, the neural network estimates its location from the wireless signal.
In the 20×20m area, we achieve around 35cm mean absolute error.

Applications of Channel Charting are:
* Localization
* Pilot assignment in massive MIMO
* Handover prediction for mobile networks
* Network planning and other service enhancements

For more information, visit https://dichasus.inue.uni-stuttgart.de/