Researchers at the US-based Southwest Research Institute (SwRI) have developed an innovative camera-based autonomous driving system called Vision for Off-Road Autonomy (VORA), which offers a stealthy alternative to traditional LIDAR and RADAR technologies. This innovative suite of tools can passively sense objects, create environmental models and locate units on a map, even in challenging off-road environments, without emitting detectable signals.
VORA was initially conceived as a solution for SwRI’s defense customers, as modern autonomous driving solutions rely heavily on LIDAR sensors, which emit high-energy lasers that can easily compromise the location of military units. RADAR technology, which uses radio waves, and Global Positioning System (GPS), which uses satellites, are also vulnerable to detection and interference.
“We thought about the toughest computer vision challenges and then focused on achieving dense, robust modeling for off-road navigation,” said Abe Garza, research engineer in SwRI’s Intelligent Systems Division.
At the heart of VORA is a deep learning stereo matcher (DLSM), a neural network-based tool that creates dense maps using camera data. By analyzing disparity or differences between motion in stereo images, DLSM can generate highly accurate maps. Additionally, the researchers developed a factor graph algorithm that combines sparse data from stereo images, landmarks, and inertial measurement unit (IMU) readings to provide precise localization data.
Although VORA was initially designed for defense applications, its capabilities extend far beyond the military. For example, the system could be invaluable for planetary exploration, where autonomous vehicles face limitations in power, payload and connectivity, making LIDAR systems impractical. Additionally, VORA could help automate agriculture in regions where GPS signals are unreliable, such as canyons and mountain slopes.
“We are excited to show our customers a plug-and-play stereo camera solution integrated into an industry-leading autonomy stack,” said Meera Towler, VORA project manager at SwRI.
SwRI plans to integrate VORA into other autonomy systems and conduct extensive testing at its off-road facility in San Antonio, paving the way for a future where stealthy, camera-based autonomous driving becomes a reality for military operations, space exploration and applications agricultural.