As part of a U.S. Army Research Laboratory program, General Electric’s Research Laboratory designed an all-terrain robotic vehicle that demonstrated the ability to think in flight in an off-road environment. (General Electric)
WASHINGTON: General Electric’s research lab demonstrated a small autonomous robot for the U.S. Army on a wooded road in upstate New York that drove fast, avoiding fallen branches and crushing piles of leaves, only to trip over Once it fell, it was caught between two trees.
The robot stopped, tried a new path, and crashed into one of the trees. It stood still again for about 10 seconds, seeming to think about how to get rid of the pickle, then quickly backed up a bit to squeeze in through the narrow tree goalposts.
Developing artificial intelligence and autonomous driving in the field of self-driving cars has become easier in the commercial world where vast amounts of data on maps, roads, infrastructure can be plugged into systems.
But where the U.S. Army routinely operates, there is little such predictability.
“Basically being able to go into a new space, assess where you are, what you’re looking at, understand the uncertainty in your actions, and then act accordingly, which is basically what we’re trying to achieve with the SARA program,” John Lizzi, head of robotics and autonomous technologies at GE’s research lab, said in an Aug. 3 interview.
The Scalable Adaptive Resilient Autonomy Program, as SARA calls it, is a U.S. Army program designed to demonstrate “risk-aware” autonomous ground vehicles that can navigate safely in complex off-road test conditions.
GE is one of eight U.S. Army Research Laboratory-funded projects, all of which are in academia, to enable autonomous vehicle navigation in complex terrain, where lessons learned and technologies could impact optional next-generation combat vehicles From the manned chariot to the robotic chariot series.
GE has been developing technology in a U.S. Army-sponsored project for the past year.
“In future U.S. Army scenarios, autonomous systems must plan reliably while encountering challenging features while maneuvering through complex terrain,” said Eric Spero, U.S. Army SARA program manager. “Incorporating risk and uncertainty into the autonomous decision-making process allows our testbed to show us what it looks like to plan a direct path rather than a detour.”
GE uses its “Humble AI” technology to make artificial intelligence more human by programming robots to understand their capabilities and limitations, enabling machines to back off and assess uncertain situations.
Robots have the ability to decipher known and unknown paths as they navigate; Lizzi says they use camera data, lidar sensing capabilities, odometers and other measurements to gather information to decide which path to take.
Humble AI has been tested, for example, on how to safely optimize the energy output of wind turbines. The AI might recognize certain wind patterns, but if it encounters new wind or weather, it will enter a safe mode while deciding how to respond to the new situation.
The company’s work on AI technologies, including its development work with the U.S. Army, aims to enable decision-making at human speed or faster, Lizzi said. “I think that’s going to be absolutely critical.”
These systems also need to learn immediately, he added. “The paradigm today is, I go out and collect a lot of data, and then I program my system to operate in that domain of understanding. I think we’re going to get to the point where this paradigm no longer applies, so dynamic learning, Learning from limited data, learning from demonstrations may be one way we handle this situation.”
The AI development in this project did not happen in a vacuum either. As the program progresses, the technology is being integrated with the U.S. Army’s core autonomous stack and will continue to evolve in the future.
Lizzi said that while ground robotics has strong applicability, GE’s advances in AI technology through the U.S. Army Research Laboratory (ARL) program could be applicable to a variety of spaces, from industrial to commercial, across energy, aviation and healthcare.