Project Loon sounds like something that crazy people will do but of course we know that if this continues to evolve, it might change the way Internet connectivity is brought to areas that don’t have the infrastructure. As they continue to experiment with things like artificial intelligence and machine learning, they also put them to the actual test by putting out the balloons. This summer, Google experimented with the Internet-carrying balloons over Latin America and they were able to stay in Peruvian airspace for 98 days.
Google has been conducting test flights for the past months in order to predict wind patterns at various altitudes and then inputting the data to create software algorithms in order to properly steer the balloons into areas where people will be able to access the connectivity. For this latest experiment, they launched the balloon from Puerto Rico with the aim of reaching Peru. It didn’t go as smoothly as expected and they had to make several adjustments. But at the end of the experiment, the balloon spent 14 weeks in Peruvian airspace but had to make almost 20,000 altitude adjustments all throughout.
One of the systems that they’ve been using is a simpler form of machine learning called the Gaussian process wherein they comb through previous data to make future predictions and adjust behavior. But they also use reinforcement learning since looking at past data doesn’t mean that it will happen again. This software adjusts behavior even after the process has made predictions.
If you’re interested in this kind of testing and process, Google is pretty open about how they are testing our Project Loon. But for those more interested in the results, it’s probably going to take some time before they will be able to apply this into something end-users will actually be able to use.
SOURCE: Google
