The inhospitable terrain of the Mojave Desert has become a major challenge for embryonic autonomous vehicles during DARPA problems of the mid-2000s.
Now the federal agency behind these molding competitions is planning further testing of AV technology in similar off-road conditions with the latest developments in modeling.
The Defense Advanced Research Agency, better known by its acronym DARPA, on Wednesday gave Intel and its staff a four-year contract to conduct research and model unusual topography.
“The off-road environment creates different types of problems, including uneven terrain, rocks, mud, large rocks, bushes,” said Herman Ross, director of the autonomous agents’ laboratory. Intel Labs. “In this project, we are creating a modeling platform that can accurately generate all types of emergencies, helping autonomous land vehicles learn to move effectively off-road at high speeds.”
This aspect of DARPA’s work, part “Robotic autonomy in challenging environments with resilience“or the” RACER “program, brings together researchers from Intel’s labs along with researchers from the Center for Computer Vision in Barcelona and the University of Texas.
Intel says it has nothing to do with the autonomy work on the road carried out by the company and its subsidiary Mobileye, which are focused on the upcoming deployment of the robot taxi. But overall, the lessons learned during road testing will help manage the off-road project for further future use.
“This project is looking for new ways to convey key ideas used in road conditions, such as modeling extreme situations, situations that can be challenging for an autonomous system adapted to off-road,” said Ross, who previously worked at Toyota. Research Institute.
Neither team completed the 132-mile desert trail between Barstow, California, and Primm, Nevada, during the 2004 DARPA Grand Challenge. A year later, five teams completed a feat on a similar route.
In the simulation process, Intel Labs says the “big” new set of environments will allow researchers to cover more than 100,000 square miles, bridge the gap between deploying cars on the road and SUV, and reduce the costs typically associated with real-world testing.
While DARPA’s first set of tasks for autonomous vehicles has helped hone the core technology, the current RACER program is designed to increase AV’s ability to run at the limits of its performance envelopes and, in some cases, beyond what is possible at human-controlled speeds.
Real-world testing is also part of the program. Last October, DARPA selected Carnegie Mellon University, NASA’s Jet Propulsion Laboratory and the University of Washington to participate in data collection and the integration of autonomous technologies into high-performance ATVs.