12 Aug 2025

Acoustic Winter Terrain Classification for Offroad Autonomous Vehicles

Abstract: Autonomous vehicles can experience extreme changes in performance when operating over winter surfaces, and require accurate classification to transit them safely. In this work we consider acoustic classification of winter terrain, and demonstrate that a simple and efficient frequency-space analysis exposed to a small convolutional neural network, rather than recurrent architectures or temporally-varying spectrogram inputs, is sufficient to provide near-perfect classification of deep snow, hardpacked surfaces and ice. Using a dual-microphone configuration, we also show that acoustic classification performance is due to a combination of vehicle noises and vehicle-terrain interaction noises, and that engine sounds can serve as a particularly powerful classification cue for offroad environments.

1 Oct 2021

Characterizing Snow Surface Properties Using Airborne Hyperspectral Imagery for Autonomous Winter Mobility

Abstract: With changing conditions in northern climates it is crucial for the United States to have assured mobility in these high-latitude regions. Winter terrain conditions adversely affect vehicle mobility and, as such, they must be accurately characterized to ensure mission success. Previous studies have attempted to remotely characterize snow properties using varied sensors. However, these studies have primarily used satellite-based products that provide coarse spatial and temporal resolution, which is unsuitable for autonomous mobility. Our work employs the use of an Unmanned Aerial Vehicle (UAV) mounted hyperspectral camera in tandem with machine learning frameworks to predict snow surface properties at finer scales. Several machine learning models were trained using hyperspectral imagery in tandem with in-situ snow measurements. The results indicate that random forest and k-nearest neighbors models had the lowest Mean Absolute Error for all surface snow properties. A Pearson correlation matrix showed that density, grain size, and moisture content all had a significant positive correlation to one another. Mechanically, density and grain size had a slightly positive correlation to compressive strength, while moisture had a much weaker negative correlation. This work provides preliminary insight into the efficacy of using hyperspectral imagery for characterizing snow properties for autonomous vehicle mobility.