In autonomous vehicle development, designs for unstructured, off-road environments can be very different than those designed for structured, on-road applications. A typical on-road deep learning system must focus on identifying cars, pedestrians and traffic signals, and are built to respond to highway markings, such as signage and lane markings. In contrast, off-road autonomous vehicles must be able to detect trees, rocks, abrupt changes in elevation and other obstacles. Active sensors, such as LiDAR, also respond differently to organic materials, like foliage, than to man-made materials, like steel and glass.
The CAVS vehicle proving ground, located on 55 acres of land adjacent to the CAVS building, provides controlled-access testing capabilities for both autonomous vehicles and vehicle mobility in an off-road environment. The proving ground features various terrains, including rocks, tall grass, wooded trails and lowlands. Varying courses provide transition points between different terrain and lighting scenarios. Future improvements to the site include additional surfaces and ride courses.
CAVS is collecting data from a variety of off-road environments in order to build a comprehensive data set.
Developing and evaluating new vehicles and algorithms requires data taken from relevant environments. CAVS is assembling autonomous vehicle sensor data taken from a variety of off-road environments in order to allow researchers (within both MSU and the general scientific community) to advance autonomous vehicles. This includes high-resolution lidar data, HD camera data, precise GPS data, and IMU data, with up to six time-synced camera perspectives in raw HD frames.
The next step in order to create useful data is to annotate classifications of objects and regions in the data. CAVS is involved in this process, while also researching and standardizing appropriate annotation methods for data taken in unstructured environments.
The CAVS proving ground allows for testing of autonomous vehicles sensors and algorithms in a controlled, repeatable environment. Key natural features have been digitized to allow correlation between simulations and real-world testing. The proving ground also provides for evaluation of sensor performance with organic surfaces, which may have different reflective properties.
Autonomous vehicles must be resilient against degradation of independent sensors, such as GPS loss due to tree canopy, camera obstructions from dirt or mud, and reduced performance due to weather. The CAVS proving ground provides natural resources for testing such scenarios.
In imperfect terrain, wheel slip affects autonomous vehicle performance. Off-road capable autonomous vehicles should include vehicle-terrain interaction effects when planning and executing paths. The CAVS proving ground allows evaluation of the effects and tools for development of intelligent off-road path planning algorithms and vehicle models for incorporating the effects of vehicle-terrain interaction.