Edge AI

PanLUNA: An Efficient and Robust Query-Unified Multimodal Model for Edge Biosignal Intelligence

5.4M-parameter multimodal biosignal foundation model (EEG/ECG/PPG) deployable on a RISC-V microcontroller.

Marija Zelic, Anna Tegon, Yawei Li, Thorir Mar Ingolfsson, Luca Benini

PanLUNA: An Efficient and Robust Query-Unified Multimodal Model for Edge Biosignal Intelligence

TinyMyo: A Tiny Foundation Model for Flexible EMG Signal Processing at the Edge

3.6M-parameter EMG foundation model, first of its kind deployed on an ultra-low-power microcontroller.

Matteo Fasulo, Giusy Spacone, Thorir Mar Ingolfsson, Yawei Li, Luca Benini, Andrea Cossettini

TinyMyo: A Tiny Foundation Model for Flexible EMG Signal Processing at the Edge

Quantized Tiny Recursive Models for Edge Deployment

Quantize TRMs to low precision and deploy them on GAP9 or Cortex-M, exploring accuracy–energy trade-offs and adaptive compute.

Last updated on 5 Jul 2026 MSc Thesis

Tiny Recursive Models for Time-Series

Adapt Tiny Recursive Models (TRMs) to non-visual time-series domains and analyse recursion, deep supervision, and adaptive compute.

Last updated on 5 Jul 2026 MSc Thesis

BrainFuseNet: Enhancing Wearable Seizure Detection Through EEG-PPG-Accelerometer Sensor Fusion and Efficient Edge Deployment

EEG-PPG-accelerometer sensor fusion for wearable seizure detection with efficient edge deployment.

Thorir Mar Ingolfsson, Xiaying Wang, Upasana Chakraborty, Simone Benatti, Adriano Bernini, Pauline Ducouret, Philippe Ryvlin, Sándor Beniczky, Luca Benini, Andrea Cossettini

12 mJ per Class On-Device Online Few-Shot Class-Incremental Learning

On-device online few-shot class-incremental learning at 12 mJ per class for extreme-edge systems.

Yoga Esa Wibowo, Cristian Cioflan, Thorir Mar Ingolfsson, Michael Hersche, Leo Zhao, Abbas Rahimi, Luca Benini

Real-Time Drone Navigation with a Wearable EEG Headband

We built a non-stigmatizing EEG headband with an ultra-low-power PULP processor that recognizes eye blinks, eyebrow movements, and alpha waves in real time and uses them to control a flying nano-drone — all computation happens on-board the wearable, without a laptop in the loop.