Innovative Radar and Machine Learning System Enables Non-Lethal Identification of Essential Pollinating Insects

New radar technology identifies bees and wasps with 96% accuracy by analyzing wingbeats. A breakthrough for non-invasive biodiversity and ecological monitoring.

By: AXL Media

Published: Apr 29, 2026, 10:27 AM EDT

Source: Information for this report was sourced from EurekAlert!

A Breakthrough in Non-Invasive Ecological Monitoring

Monitoring pollinating insects is a vital task for both global agriculture and ecological health, yet traditional methods have long been hampered by significant logistical hurdles. Conventional identification is typically labor intensive and often requires the lethal sampling of specimens to confirm their species. However, a new study published in PNAS Nexus reveals a transformative approach led by Adam Narbudowicz and a team of researchers. By utilizing millimeter-wave radar and machine learning, scientists can now identify insects in mid-air without causing them harm or interrupting their natural behaviors.

Decoding the Language of Wingbeat Harmonics

The technology works by capturing the subtle changes in radar reflections caused by the rapid flapping of an insect's wings. These reflections, known as micro-Doppler signatures, contain a wealth of data that the human eye cannot perceive. To process this information, the research team developed a machine learning model capable of extracting more than 70 distinct harmonic, spectral, and temporal features. These features provide a digital fingerprint for each insect, allowing the system to distinguish between different types based on the specific physics of their flight.

From Campus Collection to Machine Learning Training

To build and train the model, researchers collected various insects on the campus of Trinity College Dublin. Each insect was temporarily placed in a small cylindrical plastic container positioned directly above a millimeter-wave antenna. Once the radar signatures were recorded, the insects were released back into the environment. This data collection process allowed the model to learn the fundamental wingbeat frequencies and the specific rates at which an insect’s wing movements change, which the authors identified as key distinguishing features for accurate taxonomic classification.