Microsoft Research creating an AI to help autonomous gliders stay up indefinitely

Microsoft is testing an artificial intelligence agent which can control a glider much like a hawk does, remaining in the sky indefinitely by predicted air patterns and planning a route forward by seeking out columns of rising hot air and using them to stay aloft.

“Birds do this seamlessly, and all they’re doing is harnessing nature. And they do it with a peanut-sized brain,” says Ashish Kapoor, a principal researcher at Microsoft.

Microsoft researchers tested the technology in Nevada last week using two gliders guided by sensors and onboard computers.

The effort is in part to help machines make decisions when faced with uncertainty and specifically the hope is that the autonomous aircraft can eventually ride the air for hours or even days at a time while consuming very little power, helping to, say, track weather patterns, monitor farm crops or even deliver the internet to places where it’s otherwise unavailable.

The glider is testing a mathematical technique called the Markov decision processes as a way of identifying and responding to uncertainty in predicting the future in a relatively safe environment. With the sailplane, the team combined that model with another AI approach, called Bayesian reinforcement learning, to create a way for the system to learn what it needs to know about its environment as quickly as possible, in order to make the right decisions.

The team also is using what’s called Monte Carlo tree search, which is a way for AI to look for the most promising course of action.

“The core problem for robotics is uncertainty,” said Ken Goldberg, a professor at University of California, Berkeley. “This is what differentiates robotics from a game like Go or chess.”

“With a glider, you can test these algorithms with minimal risk to people and property,” Mr. Kochenderfer, a Stanford University professor of aeronautics and astronautics.

Kapoor says their glider is probably one of the few AI systems operating in the real world that’s not only making predictions but also taking action based on those predictions.

Interestingly the techniques are already being used by Bing and Windows but will eventually also be relevant in self-driving cars and other dynamic environments.

“The A.I. systems of tomorrow will face all the same challenges,” said Mr. Kolobov, a member of the Adaptive Systems and Interaction group (ASI) at MSR Redmond.  “The number of applications where these methods are used is growing.”

“These can be your cellular towers someday,” Kapoor says. “You don’t need any ground infrastructure.”
Eventually, the team says, the sailplane could even use solar or wind power to gather energy, theoretically making it possible for it to stay aloft indefinitely.

See the researchers in action below:


Read more about the trails at the New York Times here.