How AI retains Google’s loon balloons within the air
Only Google might think that the way to improve the flight of giant helium-filled balloons is to develop better algorithms. And to do justice to the Mountain View-based search leviathan, it seems to have worked.
For the past several years, Project Loon, a subsidiary of Google’s parent company Alphabet, has been working to bring Internet access to rural and remote parts of the world using high-altitude balloons in the stratosphere to create wireless air networks. Last year, Loon announced that it had completed 1 million hours of stratospheric flight with its combined balloon fleet. In late October, Loon set a new record for the longest stratospheric flight by staying in the air for a whopping 312 days and covering a distance of around 135,000 miles.
In a new article published in Nature magazine, Loon explains how his balloons can stay in the air for weeks without human intervention or complete knowledge of the surrounding winds. The secret? Some impressively cutting edge AI
“Loon balloons navigate by moving up or down in height to capture favorable wind currents that direct them in the desired direction,” Sal Candido, Loon’s chief technology officer, told Digital Trends. “The decision as to when to get on or off is determined by sophisticated algorithms. Traditionally, these algorithms were written by human engineers. With enhanced learning, we are using the AI to create these algorithms. In essence, we’ve built a machine that can build a better navigation system than we humans can. This machine can also build these navigation systems in a fraction of the time it takes us humans. “
Reinforcement learning is a variant of machine learning that is heavily inspired by behavioral psychology. The guiding principle of reinforcement learning is the idea that software agents can learn to take action based on maximizing a reward. As is known, reinforcement learning was used by Google DeepMind to teach an AI how to play classic Atari video games. No more information was used than just the pixels that make up each frame of the game and the score on the screen. By being instructed to maximize scores, the DeepMind AI learned to play the games through trial and error and gradually improved its skills until it was a master.
Flying a balloon in such a way that it does not deviate from the course is of course a completely different task than playing computer games. A successful balloon flight does not result in a high score that immediately shows that it was successful. Still, as Candido said, enhanced learning is a crucial part of Loon’s success.
“[Reinforcement learning] is able to process large amounts of information and use it to solve the problem, rather than having a human naturally understand how to react to that information or having a computer search the room of all possible outcomes, ”said he. “Because loon navigation improves by taking a variety of factors and information into account [or] Data has exceeded the complexity that engineers can easily do [with regards to] The former and the latter searches are computationally difficult to scale across an entire fleet. [That makes reinforcement learning] a great tool for the job. “
Make the right decisions
With the help of reinforcement learning, the artificially intelligent balloons can make optimal decisions about the movement based on historical wind knowledge, the observed and forecast winds and the projected future flight paths. All of this data is weighed and various scenarios simulated before the balloon decides how to behave.
Compared to the previous controllers used to control Loon, the new learning-based reinforcement method kept Loon’s balloons within range of their ground station more effectively so that they could send and receive signals effectively. Additionally, if they’re off course, it means they’ll get back into the right positions faster.
“Our new learning-based reinforcement algorithm is up today, helping our balloons stay above the users in Kenya we serve as part of our partnership with Telkom Kenya,” said Candido.
Alphabet has long been committed to the idea of technology forever. The more people Loon can give access to the Internet, the better the initiative will be. And to do that, it takes increasingly intelligent technology to drive it. As this latest milestone shows, all of the basics appear to be covered.