Alphabet subsidiary Loon announced its internet-beaming balloons will now be piloted using reinforcement-learning (RL) AI, an advancement it claimed will simplify processes and provide reliable and constant connectivity.
In a blog post, Salvatore Candido, Loon chief technology officer said the company developed the RL system with Google AI Montreal. He explained using its new AI tool balloons will be able to learn by “trial and error in an interactive environment using feedback from its own actions and experiences”. Balloons are launched into the stratosphere, and rely on prevailing winds to stay afloat, Candido noted navigating one as “deceptively complex”.
Loon claimed this is the first deployment of RL in a production aerospace system. It flew a balloon in a test over Peru in 2019 for 39 days, testing against the human-made system, StationSeeker, where it matched in performance and was able to autonomously navigate without human intervention resulting in more consistent internet coverage.
Loon and operator partner Telkom Kenya finally launched commercial 4G services with its balloon-based technology in July this year, after being delayed with multiple regulatory delays.
It marked the first launch in a non-emergency situation for the balloons, which were previously used to connect locations that were affected by natural disasters. The balloons can beam download speeds up to 18.9Mbps and uplink speeds of 4.74Mbps.