Alphabet’s Loon has been using algorithmic processes to optimize the flight of its stratospheric balloons for years now — and setting records for time spent aloft as a result. But the company is now deploying a new navigation system that has the potential to be much better, and it’s using true reinforcement-learning AI to teach itself to optimize navigation better than humans ever could.
Loon developed the new reinforcement-learning system, which it says is the first to be used in an actual product aerospace context, with its Alphabet colleagues at Google AI in Montreal over the past couple of years. Unlike its past algorithmic navigation software, this one is devised entirely by machine — a machine that’s able to calculate the optimal navigation path for the balloons much more quickly than the human-made system could, and with much more efficiency, meaning the balloons use much less power to travel the same or greater distances than before.
How does Loon know it’s better? They actually pitted the new AI navigation against their human algorithm-based prior system directly, with a 39-day test that flew over the Pacific Ocean. The reinforcement learning model kept the Loon balloon aloft over target areas for longer continuous periods, using less energy than the older system, and it even came up with some new navigational moves that the team has never seen or conceived of before.
After this and other tests proved such dramatic successes, Loon actually then went ahead and deployed it across its entire production fleet, which is currently deployed across parts of Africa to serve commercial customers in Kenya.
This is one of few real-world examples of an AI system that employs reinforcement learning to actively teach itself to perform better being used in a real-life setting, to control the performance of real hardware operating in a production capacity and serving paying customers. It’s a remarkable achievement, and definitely one that will be watched closely by others in aerospace and beyond.