Karim presents OpenCSI at IEEE LCN 2026 in Coimbra

Karim Khamaisi successfully presented both of his OpenCSI contributions at the 51st IEEE Conference on Local Computer Networks (LCN 2026) in Coimbra, Portugal: the full paper, which introduces a self-calibration layer for WiFi sensing on heterogeneous hardware, and a live demonstration of that layer running on an ESP32 mesh.

The full paper: a calibration layer for cross-hardware WiFi sensing

OpenCSI: Self-Calibration Layer for Heterogeneous Mesh Wireless Sensor Networks, by Karim Khamaisi, Simon Sigg, and Bruno Rodrigues. Paper (IEEE, DOI) · Preprint (arXiv) · Code and dataset

Abstract. WiFi CSI sensing models trained in one environment usually fail in another, because standard normalization embeds chip- and room-specific artifacts into the feature space. We propose OpenCSI, an abstraction layer that hides those artifacts from the sensing logic above by exposing each link as a dimensionless Z-score against its own quiet-period temporal standard deviation. The denominator is learned online from a 120-second empty-room bootstrap and reported with a maturity tag, so downstream code can detect drift and abstain when the baseline is unreliable. We evaluate OpenCSI across three rooms and three ESP32 generations (S3, C3, C6 at HT20 and HE20), with two deployments varying chip and room together and one controlled S3↔C6 swap that holds positions fixed. A model trained on one cell holds binary F1 between 0.896 and 0.993 in seven of eight transfer cells without retraining, against 0.87 for warmup-normalized features; the eighth cell collapses on the noisiest silicon (ESP32-C3), whose compressed dynamic range breaks the ratio. The transfer is binary by construction: separating static from moving needs the magnitude that the temporal standard deviation absorbs. We release the source code and dataset to support reproducible cross-environment CSI research.

Cite this paper: the BibTeX entry, with a copy button, is on the paper page.

The demo: making zero-shot transfer observable

Demo: OpenCSI: A Self-Calibrating, Cross-Hardware WiFi Sensing Mesh, by Karim Khamaisi, Simon Sigg, and Bruno Rodrigues. Paper (PDF) · Code

Where most WiFi sensing demos show a downstream task on fixed hardware, this one exposes the calibration layer itself. Attendees can watch each link’s baseline mature in real time, step into the sensing volume, and see a classifier trained on an ESP32-C6 mesh keep detecting occupancy after the mesh is physically rebuilt from ESP32-S3 nodes, with no retraining and no target-domain data. Zero-shot cross-hardware transfer becomes something to watch happen rather than a number in a table.

Abstract. WiFi Channel State Information (CSI) enables privacy-preserving, camera-free occupancy detection on commodity hardware as cheap as a 3 USD ESP32 microcontroller. A model trained on one deployment, however, rarely survives a move to a different room or a different chip generation, because standard normalization embeds chip- and room-specific artifacts into the sensing features. This demonstration deploys OpenCSI, a self-calibration layer that exposes each mesh link as a dimensionless Z-score learned online from a per-link quiet-period baseline. Attendees interact with a live four-node ESP32 mesh and a real-time dashboard that shows the mesh topology, each link’s calibration maturity as its baseline forms, and the occupancy decision as a visitor steps into the sensing volume. A classifier trained beforehand on an ESP32-C6 mesh then continues to detect occupancy on a physically different ESP32-S3 mesh in the same room, without retraining and without target-domain data. Following the full paper, the demonstration showcases the OpenCSI calibration layer itself rather than a specific sensing task.

Cite this demo: the BibTeX entry, with a copy button, is on the demo page.

At the conference

The work also builds on contributions from Simon Sigg’s Bachelor’s thesis at HSG. Congratulations, Karim!