Only Google could think that the way to improve the flight of giant helium-filled balloons is to come up with better algorithms. And to be fair to the Mountain View-based quest leviathan, it seems to have worked.
For the past several years, Project Loon, a subsidiary of Alphabet, Google’s parent company, has been working to provide Internet access in rural and remote parts of the world by using high-altitude balloons in the stratosphere to create overhead wireless networks. . Last year, Loon announced that it had achieved 1 million stratospheric flight hours with its combined fleet of balloons. Then, in late October, Loon set a new record for the longest stratospheric flight by staying in the air for a whopping 312 days, covering a distance of about 135,000 miles.
In a new paper, published in the journal Nature, Loon explains how his balloons can stay in the air for weeks at a time, without human intervention or full knowledge of the surrounding winds. The secret? Some impressively cutting edge AI
“Loon balloons sail up or down in altitude to catch the favorable wind currents that carry them in the desired direction,” Sal Candido, Loon’s chief technology officer, told Digital Trends. “Decisions on when to ascend or descend are determined by sophisticated algorithms. Traditionally, these algorithms have been written by human engineers. With reinforcement learning, we are taking advantage of artificial intelligence to build these algorithms. In essence, we have built a machine that is capable of building a better navigation system than humans. That machine can also build these navigation systems in a fraction of the time it takes us humans. “
Reinforcement learning is a type of machine learning largely inspired by behavioral psychology. The guiding principle of reinforcement learning is the idea that software agents can learn to act on the maximization of a reward. Google DeepMind famously used reinforcement learning to teach an AI to play classic Atari video games, using no more information than just the pixels that make up each frame of the games and the on-screen score. When told to max out his score, DeepMind AI learned to play the games through trial and error, gradually honing his skills until he became a master.
Flying a balloon in such a way that it does not go off course is a very different task than playing computer games, of course. A successful balloon ride does not come with a high score that makes it immediately apparent that you have been successful. But, as Candido said, reinforcement learning is nonetheless a crucial part of Loon’s success.
“[Reinforcement learning] it is capable of processing large amounts of information and applying it to solve the problem, instead of a human being inherently needing to understand how to react to that information or have a computer searching space for all possible outcomes, “he said. “Because Loon’s navigation improves by considering a lot of factors and information [or] data, the complexity has exceeded what engineers can easily do [with regards to] the first and last search is computationally difficult to scale in an entire fleet. [That makes reinforcement learning] a great tool for the job. “
Making the right decisions
Through reinforcement learning, AI balloons can make optimal decisions about how to move, based on historical knowledge of the wind, observed and predicted winds, and projected future flight paths. All this data is weighed and different scenarios are simulated before the balloon decides how to act.
Compared to previous controllers used to control Loon, the new reinforcement learning-based methodology more effectively kept Loon’s balloons within range of their ground station so they could effectively send and receive signals. When they drifted off course, it also meant they returned faster to the correct positions.
“Our new Reinforced Learning-driven algorithm is active today, helping our balloons stay ahead of the users in Kenya, whom 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 provide Internet access, the better the initiative. And to do that, it needs increasingly smart technology to power it. As this latest milestone shows, it seems to have all bases covered.