# Neural RF SLAM for unsupervised positioning and mapping with channel state information

arXiv: https://arxiv.org/abs/2203.08264 · HTML: https://mlsanity.com/paper/2203.08264/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 6 of its 25 references that are in the map.
- First posted on arXiv 15 Mar 2022.

## Builds on (most cited first, 6 of 6)

- [Deep Sets](https://mlsanity.com/paper/1703.06114/) (2017)
- [A Belief Propagation Algorithm for Multipath-Based SLAM](https://mlsanity.com/paper/1801.04463/) (2018)
- [5G mmWave Cooperative Positioning and Mapping using Multi-Model PHD Filter and Map Fusion](https://mlsanity.com/paper/1908.09806/) (2019)
- [Map-Assisted Millimeter Wave Localization for Accurate Position Location](https://mlsanity.com/paper/1908.09773/) (2019)
- [5G Positioning and Mapping with Diffuse Multipath](https://mlsanity.com/paper/1912.08697/) (2019)
- [Exploiting Diffuse Multipath in 5G SLAM](https://mlsanity.com/paper/2006.15603/) (2020)

## Built on by (most cited first, 3 of 3)

- [Benchmarking Learnt Radio Localisation under Distribution Shift](https://mlsanity.com/paper/2210.01930/) (2022)
- [Neural 5G Indoor Localization with IMU Supervision](https://mlsanity.com/paper/2402.09948/) (2024)
- [CSIYOLO: An Intelligent CSI-based Scatter Sensing Framework for Integrated Sensing and Communication Systems](https://mlsanity.com/paper/2509.19335/) (2025)

## Abstract (arXiv)

We present a neural network architecture for jointly learning user locations and environment mapping up to isometry, in an unsupervised way, from channel state information (CSI) values with no location information. The model is based on an encoder-decoder architecture. The encoder network maps CSI values to the user location. The decoder network models the physics of propagation by parametrizing the environment using virtual anchors. It aims at reconstructing, from the encoder output and virtual anchor location, the set of time of flights (ToFs) that are extracted from CSI using super-resolution methods. The neural network task is set prediction and is accordingly trained end-to-end. The proposed model learns an interpretable latent, i.e., user location, by just enforcing a physics-based decoder. It is shown that the proposed model achieves sub-meter accuracy on synthetic ray tracing based datasets with single anchor SISO setup while recovering the environment map up to 4cm median error in a 2D environment and 15cm in a 3D environment