Virginia Tech researchers use vibrations and artificial intelligence to estimate buried rock depth

On September 18, 2026, Virginia Tech researchers Erik Komendera and Yiyan Ruan published work in the IEEE Sensors Journal describing how vibrations and artificial intelligence can estimate burial depth.

The study is titled Burial Depth Estimation for Partially Embedded Rocks Using Scanning Laser Doppler Vibrometry and Neural Networks. Erik Komendera, an assistant professor at Virginia Tech who previously served as an engineer at the NASA Langley Research Center, conducted the work with Yiyan Ruan. The work centers on vibrations, laser vibrometry and neural networks to estimate burial depth for robotic excavation on other planets.

The researchers tested their technique by burying 18 concrete blocks of varying shapes and sizes at different depths within sand, using the blocks to simulate rocks in an Earth-based testing environment. They struck exposed rocks with an instrumented hammer and used a scanning laser Doppler vibrometer to remotely measure the resulting vibrations. This non-contact measurement approach allows for data collection without requiring a sensor to touch the rock itself, which is practical for robotic exploration where power and time are limited.

The team trained a lightweight neural network with two hidden layers to determine burial depth by analyzing vibration characteristics and visible rock geometry. The network was deliberately kept compact so it could eventually run on the restricted computing hardware found aboard planetary rovers. Researchers taught the network to pair a robot’s view of a rock with the rock’s reaction to a strike, then use that link to infer the hidden part.

Soil stiffness, rock geometry, and the angle of contact with the sand all influence how accurately the vibration readings reveal burial depth, so the AI was specifically taught to account for these complex variables to ensure reliable depth estimations. Strikes send vibrations through rock and sand, and those vibrations change based on how deeply the rock sits and how tightly the soil holds it.

A rover could hit a rock, scan the vibrations, and decide whether to move a loose stone or avoid a boulder anchored deep in the ground. A scanning laser Doppler vibrometer reads vibrations from a distance, so a rover can study a rock without mounting anything on it or digging nearby. Choosing correctly would shield the robot’s equipment, save energy, and keep excavation work on schedule.

Mars mission planners weigh terrain elevation, surface roughness, and rock height when choosing where to land a spacecraft. The Virginia Tech news release frames the research as a step toward letting astronauts assess what is on and under uncharted planetary surfaces before landing.

The research aims to help robotic excavators and rovers on planets like Mars or the Moon identify whether obstacles are loose rocks or immovable buried boulders. Terrain roughness, rock height, and elevation are critical factors in determining spacecraft landing sites during Mars missions. Erik Komendera said, “How can robots make those judgments? Now we have an answer.”

Reading vibrations without contact suits robotic work because extra motions cost power, time, and mechanical wear. Robotic excavators risk wasting energy or damaging tools if they attempt to remove large, immovable boulders instead of small, loose rocks.

Beyond other planets, the vibration-based technology could identify unstable rock formations on Earth, map underground caverns, or assist in disaster cleanup, and Komendera suggests the tool could help analyze construction sites or disaster scenarios that require identifying buried hazards before moving rubble. Yiyan Ruan said that estimating a rock’s burial depth using vibrations can be tricky.

The researchers plan to keep building datasets and testing the method on Earth alongside potential space applications in unstable rock assessment, cavern measurement, and disaster cleanup.