News » 27.08.2025 - Researchers develop sound-based crop sensing tool
Researchers at Carnegie Mellon University's Robotics Institute (RI) have developed a new sensing technology, called SonicBoom, designed to help robots locate and identify crops such as apples using sound rather than cameras. The system, still in early development, could eventually assist agricultural robots with tasks such as pruning vines or detecting ripe fruit hidden by foliage.
Moonyoung (Mark) Lee, a Ph.D. student in robotics, explained that the technology can determine the three-dimensional shape of objects through touch, without relying on a camera. This approach addresses a common problem in agricultural robotics, where visual sensors struggle to operate effectively in environments with heavy leaf cover that obscures both the fruit and the paths robot arms must take to reach it.
SonicBoom differs from existing tactile sensors by using contact microphones to detect vibrations when an object is touched. Unlike camera-based tactile sensors, which can be fragile, or pressure sensors, which can be expensive to install over large areas, contact microphones are more durable and can be housed in protective casings.
The prototype consists of six contact microphones mounted inside a PVC pipe. When the pipe makes contact with an object such as a branch, the microphones record vibrations, and the system triangulates the point of contact with a precision between 0.43 and 2.2 centimeters. The PVC housing also protects the microphones from damage.
To train the system, the team collected data from 18,000 contacts between the sensor and a wooden rod. A machine learning model was then used to map the audio signals to specific contact points. While the current version is designed to detect hard or rigid objects, adjustments to the setup could enable it to sense softer targets, such as fruit.
Lee said further research is exploring the potential for the microphone array to identify objects in addition to locating them. Possible uses outside agriculture could include safety systems for robots working near people, robots designed for human interaction, or operations in low-light environments.
The research team includes Lee, RI Associate Professor Oliver Kroemer, Ph.D. student Uksang Yoo, and RI faculty members Jean Oh, Jeffrey Ichnowski, and George Kantor.
Source: www.floraldaily.com
« Back