When I did my high altitude balloon I was struck by the changing wind directions at different altitudes. In this blog there's a picture of the actual drifting flight path: http://blog.jgc.org/2011/04/gaga-1-flight.html
This flight path was pretty closely predicted 48 hours before using NOAA predicted wind data (see this blog: http://blog.jgc.org/2011/04/gaga-1-looking-good-for-sunday.h...). The difference in size between the two 'loops' is caused by a slow ascent and rapid descent.
If I'd had altitude control (like Loon) and uploaded data on wind speed it would have been possible to make the balloon hang around a target area.
That's interesting, it seemed like the most challenging problem to solve given the balloons are always moving. They mention using machine learning to perfect the coverage:
> This is the first time we have used our new machine learning powered algorithms to keep balloons clustered over Puerto Rico, so we’re still learning how best to do this. As we get more familiar with the constantly shifting winds in this region, we hope to keep the balloons over areas where connectivity is needed for as long as possible.
Definitely the most interesting part of this project in my opinion.
> This is the first time we have used our new machine learning powered algorithms to keep balloons clustered over Puerto Rico, so we’re still learning how best to do this. As we get more familiar with the constantly shifting winds in this region, we hope to keep the balloons over areas where connectivity is needed for as long as possible.
Apparently, back in 2016 they launched the balloons from Puerto Pico to do their first test of the autopilot system in Peru. Now they are flying the balloons from Nevada to implement it in Puerto Rico.
> we had figured out how to cluster balloons in teams, dancing in small loops on the stratospheric winds, over a particular region.
This is insane. They're using ML to predict wind speeds at different altitudes and how to move the balloons with their various speed and power restrictions to optimize on-the-ground connectivity while maintaining close enough clusters and decent backhaul. Incredible.
To be clear, it's not so much they are using ML to predict wind speeds at different altitudes; they use a forecast model like everyone else for that. The clever bit is being able to control your altitude such that you find favorable wind-altitude tuples to be in to keep you generally in one spot over days/weeks. It's sort of like what ocean/circumnavigating sailors do but in a higher dimensional space, both literally and in your search algorithm. Fascinating stuff.
The loons don't need to stay in one spot - for example, two loons could rotate around a common center point and provide good service. Allowing the loons to wander while optimizing ground connectivity is harder computationally but makes it possible to maintain good connections even if the wind isn't entirely cooperating.
That is pretty awesome. I didn't realize the scale and complexity of what they were doing. When I first heard reference to this, I for some reason assumed that these were simple tethered balloons, not stratospheric swarms of AI balloons!
"The 100G program is exploring high-order modulation and spatial multiplexing techniques to achieve the 100 Gb/s capacity at ranges of 200 km air-to-air and 100 km air-to-ground from a high-altitude (e.g. 60,000 ft.) aerial platform. The 100G program is developing the technologies and system concepts to project fiber optic class 100 Gb/s capacity via airborne data links anywhere within the area of responsibility (AOR).
Modern expeditionary military missions generate and exchange massive amounts of data that are used to produce situational awareness and guide decision-making. Much of the data must travel long distances along backbone communications networks composed of high-capacity links that interconnect command centers. While optical fiber services can provide a multi-gigabit data backbone in many parts of the world, modern expeditionary forces require a similar capability in places where fiber access does not exist. SATCOM services can provide some capacity to remote areas, but cannot provide the capacity needed to support the amount of data generated by emerging ISR systems."
https://www.darpa.mil/program/100-gb-s-rf-backbone
Haven't seen Loon in the headlines for a while. This is really cool. In hindsight, disaster relief seems like such an obvious application for a technology like Loon.
Google could catch some serious tax breaks as well as goodwill if they worked out deals for providing impromptu connectivity services in disaster zones.
Is SolarCity considered a quicker replacement to the damaged power grid and supply?
That a truck can carry far more energy in the form of petroleum than it can in the form of batteries is one of the arguments against wasting fossil fuels on non-extreme conditions, like regular transportation and non-emergency electric power generation.
For Puerto Rico right now cheap, even if relatively inefficient, generators and lots of fuel seems like a good way to help in the very short term. Military surplus diesel generators, for example, are really cheap for how much power they can output with enough fuel.
They aren’t directly comparable. You ship a truckload of PV and once installed it keeps providing daytime power without further delivery; you ship a battery to somwhere with PV and that power can now be made in the day and used at night, every day/night for a few years.
Even without batteries, shipping PV will at least keep the freezers cold (and food fresh).
AFAIK, many other big-name companies have been donating manpower and resources to help get PR back on its feet. What makes this announcement any more noteworthy? I'd get it if it included a bit more information, but the blog post reads as marketing fluff.
Does anyone know of any reliable sources showing in-depth sample numbers from previous deployments? I'd be really interested in getting an idea of what area size these balloons could cover, for how long it can be reliably maintained over an area, and their expected throughput. The Helping out in Peru [0] post has a bit of info, but it's very light on details.
If anyone is familiarized with these industries, I'd love to read up on the kinds of challenges faced with getting something like this deployed. Looking at it from an outsider's perspective, being without connectivity for over a month seems like a pretty long time. It's great that Project Loon is seeing improvements in their deployment time, but the article doesn't shed any light on the process, which is rather disappointing.
It’s newsworthy because an experimental Google X project is being used to solve a real world problem.
The US military might spend a billion dollars in aid, for example, but the world won’t be any different in the future, beyond the lives directly impacted, because of this.
You are definitely asking the right questions. It is way too easy for Google to declare victory here without any meaningful level of service delivered. "At least one of the balloons managed to station keep for a while" =/= "isolated people in Puerto Rico were actually able to communicate with the outside world"
Im sure there are a few people who will find Loon's service useful.. but 80% are without power, so they're mostly not going to use internet anyway -- no power means cell phones cant be charged, etc. Some have generators, but are they able to get fuel? Emergency personnel cant even get fuel to deliver food and medicine.
Internet and cell service is being restored at a similar rate to power.. probably because these two depend on each other.
So it seems like this is only really useful to people traveling from one of the areas with power to an area without power (first responders, workers doing repairs, etc).
This is a common misconception that many have tried to dispel. It is trivial to charge your phone from your car (when there is gas) and from small solar cells (of which there are many). Even at 2G speeds (56Kbps) there are tremendous benefits to having internet service, from co-ordinating relief efforts to monitoring beacon data (temperature, wind speed, rain fall, pressure, humidity, Etc.)
Two pieces of information that have come from the hurricane areas is that it would be useful to stock/supply solar powered phone recharging units that could be distributed, and very low bandwidth web sites (text or SVG only very few if any images) for co-ordination.
After Hurricane Harvey,* I ordered two RUIPU solar-powered USB auxiliary batteries [0] for about USD $40 each. Each is about the size of a very-thick iPhone 6S Plus. One whole side consists of solar cells. The battery has three USB "power out" jacks and one micro-USB "power in" jack, allowing the battery to be charged from an electrical outlet. My two batteries now live in a window in my study, where their solar cells get several hours of direct sun every morning to keep them charged up. They seem to work OK.
* We were unaffected personally, although a few nearby friends weren't as fortunate.
I wonder how many people had portable solar panels before the hurricanes though. It's not like they could order it after the hurricane. And the solar panels on people's houses are probably damaged... two hurricanes would do that.
If they decided they needed to get emergency power for radio communications to people and the choices were between generators w/ a supply of gas or portable solar charging kits I know which I would choose.
We are charging them at coffee shops, fast food restaurants, malls, with our cars, and anywhere else we can find a working plug. It's not easy but we do what we can because without communications there are no logistics, no deliveries, no banking, no news, and no jobs. I've got communications only a few days a week depending on where I go. My elderly parents who I help take care of have got no communications at all which is an added source of stress. Projects like Loon are essential for all of us (although I don't expect them to help much).
Fuel has been widely available for about two weeks now. Some businesses have been running with generators but it's not sustainable in the long run. Many others are trying to figure out whether they'll be able to reopen their doors in the future at all.
This flight path was pretty closely predicted 48 hours before using NOAA predicted wind data (see this blog: http://blog.jgc.org/2011/04/gaga-1-looking-good-for-sunday.h...). The difference in size between the two 'loops' is caused by a slow ascent and rapid descent.
If I'd had altitude control (like Loon) and uploaded data on wind speed it would have been possible to make the balloon hang around a target area.