ATLAS, AN ALL-TERRAIN LABELSET FOR AUTONOMOUS SYSTEMS

All-Terrain off-road environments are the next frontier for autonomous vehicles to overcome. However, there are many obstacles in the way of this goal. Artificial intelligence has proven to be an invaluable asset in developing perception and path planning systems capable of overcoming these obstacles, but these AI systems fundamentally rely on the availability of data related to the operational environment in order to succeed. Currently, there is no unifying ontology for this data. This has inhibited progress on training AI by reducing the availability of cross-integrable datasets. We present ATLAS: A labeling ontology composed of over 200 labels specifically designed to encompass all-terrain off-road environments. This ontology will lay the ground work for creating scalable standardized all terrain off-road data and will enable future AI by providing an expansive and well labeled ontology that can push the field of autonomous vehicles to new heights.

Citation: W. Smith, D. Grabowsky, D. Mikulski, “ATLAS, an All-Terrain Labelset for Autonomous Systems,” In Proceedings of the Ground Vehicle Systems Engineering and Technology Symposium (GVSETS), NDIA, Novi, MI, Aug. 16-18, 2022.