After the launch of Project loon, Google has been using static algorithms to make its internet balloons change altitude and stay in position for several years. Those hard coded algorithms were confined to adapt to unexpected weather patterns, which are pretty common in a place like the stratosphere.
Bye bye control algorithm, welcome AI
Well, those situations are history now. The Project Loon team has revealed that their engineering team is moving away from those hard coded control algorithm, instead, they are using machine learning to alter those internet balloon’s behavior and make them stick to the desired flight path much longer. The company has already launched a test balloon into the stratosphere over Peru, which stayed there for 98 days, adapting to difficult wind conditions that might have sent it floating away.
According to Wired, those algorithms now explore over massive amounts of data, and, of course, learn from it. In one case, the balloon temporarily drifted over the Pacific Ocean to catch winds when it visualized the lack of gusts to stay over land. With the help of “reinforcement learning,” the balloon can refine its behavior even after making calculations as to what’s going to happen. As for the balloon over Peru stratosphere, it made just fewer than 20,000 alterations to its altitude over the course of those 98 days.
This Ai-based upgrade should help Google to trim cost while expanding their reach, but won’t help to keep those balloons in place for longer. It can either spread out over a wider area or scale back the size of the fleet instead of using so many balloons to cloak an area with aerial internet access. However, it’s really good news for those who soon may start depending on Project Loon to get internet access.