Channel Charting in Real-World Coordinates with Distributed MIMO

This is the code for the results in the paper "Channel Charting in Real-World Coordinates with Distributed MIMO", S. Taner, V. Palhares, and C. Studer (c) 2025 Sueda Taner

email: taners@ethz.ch

Important Information

If you are using this code (or parts of it) for a publication, then you must cite the following paper:

S. Taner, V. Palhares and C. Studer, "Channel Charting in Real-World Coordinates with Distributed MIMO," in IEEE Transactions on Wireless Communications, 2025.

How to use this code...

Step 1: Download and save the data

• From dichasus-cf0x Dataset: Distributed Antenna Setup in Industrial Environment, Day 1, download the specifications file spec.json, the channel files dichasus-cf02, dichasus-cf03 and dichasus-cf04 (which should have the .tfrecord extension), and their offset estimates (which should be reftx-offsets-dichasus-cf0x.json) into a folder called data_raw.
• Run preprocess_dichasus.py. This will store .np versions of the CSI, timestamps, and ground-truth positions extracted from the .tfrecord files in a folder called data along with AP positions and LoS bounding boxes for each AP.

Step 2: Channel charting

• Set your training parameters as explained on top of main.py and run for the results in the paper. This code does the following:
  • We separate the complete dataset into training and testing samples.
  • We train a neural network for channel charting or positioning according to the settings.
  • We test the channel charting neural network on the test set.

Version history

Version 0.1: taners@ethz.ch - initial version for GitHub release.

Acknowledgments

This project makes use of the following external data and code:

• [dichasus-cf0x Dataset: Distributed Antenna Setup in Industrial Environment, Day 1](https://dichasus.inue.uni-stuttgart.de/datasets/data/dichasus-cf0x/, accessed on 1/3/2024: Our code uses the cf02, cf03, and cf04 datasets.
• Dissimilarity Metric-Based Channel Charting by F. Euchner, accessed on 1/3/2024: We use this code for (i) pre-processing the data from tfrecords and json files, and (ii) finding an affine transform that maps a channel chart to real-world positions using ground-truth position labels for our baseline B2.