Several companies and universities have teamed up to develop mathematical approaches and tools for incorporating data-based machine learning algorithms into cyber physical systems as part of a Defense Advanced Research Projects Agency program aimed at ensuring the safety of autonomous technology.
Participants have demonstrated signs of progress in their 18-month research and development efforts under the first phase of DARPA's Assured Autonomy program, the agency said Wednesday.
Boeing collaborated with the University of California at Berkeley, Collins Aerospace and SGT Inc. to create an ML-based system design and analysis toolkit, dubbed VerifAI, that seeks to increase safety of aircraft systems during operations on ground.
The team also created a safety kernel method for an autonomous platform to detect input anomaly and determine a proper response behavior.
DARPA noted the tools were tested with Boeing-made evaluation platforms such as the Iron Bird X-Plane simulator and a testbed aircraft.
A group of HRL Laboratories researchers worked with the U.S. Army Combat Capabilities Development Command's Ground Vehicle Systems Center to demonstrate an assurance toolkit on a Polaris-built autonomous military vehicle.
Northrop Grumman partnered with research teams at University of Pennsylvania and Vanderbilt University to develop a learning enabled model for an autonomous undersea vehicle to monitor its operating status and identify safe courses of action to address mission objectives.
“While the Phase 1 tests demonstrate significant program progress, important work needs to be done in subsequent phases before these technologies become eligible for real-world deployment,” said Sandeep Neema, a DARPA program manager.
“Work in Phase 2 will focus on maturation and scalability, improving coverage for hazard scenarios, adding robustness to environmental changes, and optimizing mitigating behavior in contingencies."