X-Labs’ Project Loon will use AI to adjust to weather and remain in position over areas. (via Loon)
Anyone who has ever had a helium-filled balloon can tell you, when you let go, it’s going to float away in the direction the wind is blowing. This often results in crying children and failed attempts by parents to try to snag the ever-rising colorful gasbag. The same can be said for scientists and engineers who lose giant weather balloons equipped with expensive gear due to weather.
In an effort to keep their balloons from floating away, researchers from X-Lab’s (Google’s offshoot research department) Project Loon is looking to use AI as a sort of pilot to adjust to errant weather and keep the balloons in a stable position over a certain area. While that may not sound like a big deal, it actually is, especially when you want to provide Wi-Fi to areas that don’t have it.
The research company was actually successful at keeping one of their internet balloons floating above Peruvian airspace for a three-month duration using a form of static AI that could react using sets of pre-programmed data including altitude, location and wind speed. Try as you might however, predicting weather isn’t an exact science and the only known predictable variable is its unpredictability.
It’s important to note that the internet balloons function the same as hot air balloons in terms of navigation, which is either up or down. In the case of the internet version, they take advantage of cross currents found high in the stratosphere, so it will float one way at lower altitude and the opposite at higher altitudes, thereby being able to be situated over certain areas alone. While using a static AI for predetermined weather patterns is a good start, having one that uses ‘machine learning’ is better for the unpredictable nature of weather, which is what the researchers plan to do on their next leg of Project Loon.
Machine learning AI will allow the balloons to acquire large amounts of atmospheric data, analyze it, then learn from it and change its behavior to better react to weather. In one test, the new AI (based on Gaussian processes) determined that there wasn’t enough wind gusts to remain over a land-based test area and subsequently moved out over the Pacific in order to get enough wind to sail back over the test area.
X-Labs’ Wi-Fi balloon is outfitted with the same equipment found on a cell tower and is powered by solar panels.
The hardware on X-Labs’ internet balloons is the same found in cell towers all over the globe, although they have been redesigned to be lighter, more energy efficient and small enough to fit inside of a small package. Onboard each balloon is LTE-based transceivers that taps connectivity from ground stations and then pumps it out over a designated area. To get to remote areas, balloons can be daisy-chained to bounce the LTE signal to wherever it needs to go.
A flight capsule contains the brains of the balloon, which is not only outfitted with the new AI but also controls navigation by pumping air into or out of the balloon. Powering the balloon is a pair of solar panels that provides energy to its systems during the day and charges a battery for use at night. It is even equipped with a parachute that will deploy incase the balloon encounters catastrophic damage.
The Loon balloons are still under development and probably won’t roll out on a massive scale anytime soon, however with their continued advancements in AI and the ability to manufacture them on a large scale; we could see them in the not too distant future.
Have a story tip? Message me at: cabe(at)element14(dot)com