# Ohio State's Scentronic system transmits smells over internet using ESP32 and gas sensors

> **Key takeaway:** Researchers built an open-hardware electronic nose and wearable actuator system that detects odors with metal oxide sensors, classifies them via machine learning, and reproduces them remotely using ESP32-controlled piezoelectric atomizers.

**Published:** 2026-10-26T04:10:24+00:00 
**Source:** Hackster.io News 
**Category:** open-hardware 
**Canonical URL:** https://fosswire.org/news/ohio-states-scentronic-system-transmits-smells-over-internet.html 

## Summary
Researchers at The Ohio State University have developed Scentronic, a system designed to digitize and transmit smells across the internet. The platform combines an electronic nose built from six metal oxide semiconductor gas sensors with wearable actuators driven by ESP32 microcontrollers to recreate detected odors at distant locations. A K-nearest neighbors machine learning classifier achieved perfect accuracy identifying five test odors—beer, cake, coffee, oranges, and red wine—though the training set was limited to single brands. The actuator subsystem uses piezoelectric atomizers vibrating at ultrasonic frequencies to disperse scented liquid droplets, with PWM-based intensity control enabling variable output strength.

The system demonstrated end-to-end sensor-to-actuator latency of 0.6 seconds over Wi-Fi through a cloud server, with network latency averaging 65 milliseconds. Human trials showed participants correctly identified individual reproduced odors 72.86% of the time and two-odor combinations 71.43% of the time. The researchers also integrated five odor generators and five liquid dispensers into a cup paired with a Meta Quest 3 headset, enabling synchronized virtual beverage selection with physical smell and taste reproduction. Multiple wearable form factors were demonstrated, including wristbands with two channels and handheld or necklace devices with five independent channels.

## Hardware architecture and sensing pipeline
- **Sensor array:** The electronic nose employs six metal oxide semiconductor gas sensors, each producing distinct electrical responses to volatile organic compounds, enabling pattern-based odor identification without resolving individual chemicals.
- **Machine learning classification:** A K-nearest neighbors classifier trained on sensor response patterns achieved 100% accuracy on a 50-sample held-out test set covering five food and beverage odors, though generalization across brands remains unverified.
- **ESP32 actuator control:** Each scent actuator uses an ESP32 microcontroller with pulse-width modulation to drive a piezoelectric atomizer, enabling electronic intensity adjustment and independent or simultaneous multi-channel operation.
- **Piezoelectric atomization:** Scented liquid is forced through microscopic openings via ultrasonic vibration, producing airborne droplets without heating or pressurized gas systems.
- **Network architecture:** Wi-Fi connectivity and cloud server relay enable remote operation with measured communication latency of approximately 65 ms and total system response latency of 0.6 seconds from sensor detection to actuator output.

## Why it matters
Scentronic represents a rare open-hardware entry in the largely proprietary digital scent space, with its ESP32-based actuator platform and sensor array design offering a reproducible foundation for community extension. The system's integration with VR headsets and its potential applications in remote medical evaluation, e-commerce, and immersive gaming align with growing open-source interest in haptic and sensory feedback for Linux-based VR platforms. If the researchers release schematics, firmware, and training datasets, the project could catalyze open-source development in an area currently dominated by closed commercial systems with expensive consumables.