Hacker News Reader

Comment by comboy | Hacker News Reader

I wonder when are we going to have some low resolution (good enough to spot a car) realtime recording of the full globe. I know it's tons of data but with some machine learning it should be possible to skip most of oceans forests and deserts (unless some unusual object appears in that area).

View on HN

The largest earth-imaging constellation in the world can image the planet every day. Getting the data to the ground is a challenge, so my understanding is that they don't bother transmitting images obscured by clouds, or images of empty ocean. https://www.planet.com/

Updating the cadence to an ~hourly image would require ~2000 satellites, and updating it to every minute would require another 1.5 orders of magnitude... approximately 50,000 satellites.

The downlink capabilities to pull that data off the satellites would be colossally huge. A tremendous number of groundstations with a huge array of antennae constantly coordinating and pulling data down.

I'm sure Planet is heading in that direction, but it's actually hard to imagine the commercial value of such a mind-blowingly massive infrastructure investment.

In contrast to some of the other commenters, I do _not_ think that the NRO or any other collection of government organizations has a planetary image datastream anywhere close to this. The number of satellites and quantity of data required is really mind blowingly immense. I think that the downlink stations would be impossible to conceal, so if we had such a capability, we'd know about it.

It's also not necessary for their work - they typically task a satellite with watching an area of interest if they need to. They don't need 24/7 planet-wide monitoring.

> A tremendous number of groundstations with a huge array of antennae constantly coordinating and pulling data down.

In the limit where you're talking about a satellite "instantaneously local" to every location on the planet, that becomes just one ground antenna per satellite to handle whichever one is nearest. Yeah, 50k stations are expensive, but not nearly as much as the satellites they are talking to.

Satellites could easily spoof cellular towers, there is literally billions of ground stations, willing or not.

Actually, no they couldn’t, assuming you’re talking about cell phones being the ground station. One, your cell phone doesn’t transmit with enough power to reach even a LEO satellite, and the existing protocols dictate a two way handshake. Second, what do you anticipate the phone to do with the data once it’s received?

Cellular towers aren't able to talk that far.

http://lteuniversity.com/ask_the_expert/f/59/p/3370/11922.as...

Or replace "cellular tower" with long-range wifi

https://en.wikipedia.org/wiki/Long-range_Wi-Fi#Notable_links

My point stands that there exists a possible path to get TB to PB of data down to earth through existing handsets. A collection of handsets could form a phased array receiver, everything is basically SDR anyways.

The satellite will likely have less than 30 seconds to connect to such a tower, upload the data and get new task data. And it looks like the connection in your example used aimed antennas, not just broadcasting the signal onto the surface of the planet, which means you loose a lot of signal strength.

50k ground stations would be multiple orders of magnitude more than we’ve ever had. Also, 75% of the earth’s globe is water, if it’s a 1:1 pairing as you suggest, where do you expect to put 75% of those ground stations?

A floating radio station in the middle of the ocean is still much cheaper than a flying radio station in orbit is all I'm saying.

You might assume that fairly, but depending on parameters involved that might not be true.

If you need to downlink data every minute, stations must be spaced around the earth such that every satellite passes over a groundstation every minute. LEO orbital period is about 100 minutes, so with a circumference of 25000 miles, ground stations would have to be spaced about every 250 miles over the surface of the earth. The earth has a surface area of about 200 million square miles, so divided into 250x250 mile chunks, one would only need about 3200 groundstation sites, but each would have to be able to communicate with ~20-30 satellites at any given moment.

Because the satellites are small and relatively low power, the ground stations used to communicate with a high speed link probably use a 3-5 meter dish. So each groundstation site needs either a large array of rapid-slew tracking antennas or a big ol' phased array with some seriously hardcore software on board. A bank of ~20-30 3-5 meter antennae capable of slew and pitch to track a LEO target is a pretty serious investment. Each tracking antenna probably costs more than 100k... they'll have to have chilled antennae for low noise, and some heavy duty mechanisms to move them around. so 3200 sites * 30 dishes per site * 100k per dish is about 10 billion dollars - not including the cost of land or humans.

A phased array capable of directing a tight beam to 30 individual targets simultaneously all with different movement parameters is a many 10's of million dollar piece of equipment, and will only inflate the above number.

If one builds 50,000 satellites similar to what Planet has already built, the cost comes down into the single digit thousands of dollars per unit. Call it 10k. At that point, launch costs dominate the economics. Each sat has a mass of (say) 5 KG. The falcon heavy can launch 50,000 kg, but for cubesats some significant fraction of the mass will be in the deployment mechanism, so call it 7500 satellites per launch. That's 7 launches, at ~100 million each, so call it a round billion dollars to build and launch the satellites.

I think there are a couple orders of magnitude difference between your idea and what it'd take to process.

My googling shows 148.94 trillion square meters, with 18.6 trillion

We'll being generous to say "good enough to spot a car" is 2 pixels with 1 meter pixel size.

To be moderately confident it's a car, I'd guesstimate we'd need quite a bit higher resolution.

With no compression and RGB 24 bit encoding, that's 3 bytes/pixel, or 446 TB of data per "frame" of our realtime recording. Of course that could be compressed down a TON, but regardless it's a staggeringly large quantity of data to work with.

> some machine learning it should be possible to skip most of oceans forests and deserts (unless some unusual object appears in that area).

As far as I can think through, you'd need to process the data in the oceans, forests, and deserts if you want to find unusual objects. We can't do a shortcut and say only 1/8 of the world is inhabited/developed in some way and that's all we're going to monitor. Perhaps you take 1/4 or 1/8 of the data from the vast amounts of uninteresting data and look for larger anomalies.

No matter how you look at it, there'd be a ridiculous amount of data and processing required to do object detection on that scale.

Excellent back-of-the-envelope estimation.

I've recently been doing some world simulation for games, and found that you quickly start pushing RAM limits if you want to represent an Earth-sized planet at square kilometer precision. Turns out the Earth is pretty big. Mars is more reasonable, with about a quarter of the surface area.

Googling suggests YouTube is using 10 PB/year.

With compression I don't think the storage per se would be a problem for the IC.

And if you're looking for that particular missing airplane, you could just look at the timestamped photo and see it there.

I don’t disagree with anything you said here, but it reminds me an argument about IRC privacy from the ’90s. People frequently accused IRC networks of secretly logging private conversations, and the usual argument was that this would require such an immense amount of storage that any such effort would be astronomically expensive and entirely infeasible.

I don’t know if that was really true even back then, but today it’s completely doable and done all the time (think Google Chat or Facebook Chat).

My point is, times change, it’s entirely possible that in 20 years storing that much data will be completely possible.

The problem is that it's not just storage. Satellite to ground station links are slow, and while they have become faster in recent years, we're still talking in the realm of Mb/s for most commercial services.

In 20 years, we might have relatively faster links, but the improvement rate of those is much, much slower than storage density.

I'd assumed much better bandwidth is available for the a certain class of satellite (gbits a second via laser to downlink) [1], but for cost effective surveillance high altitude drones [2] would be the tool of choice to monitor cities, if not the world.

[1] https://www.wired.com/story/how-to-build-a-space-communicati...

[2] https://www.theverge.com/2013/2/1/3940898/darpa-gigapixel-dr...

This is actually a pretty good argument for sparse data handling; push some intelligence out to the edges of your network (e.g. the spy satellites or what have you) that are collecting the data in the first place, and then only store "interesting" pixels. E.g. regions where there are movement; regions already determined to be of interest (densely populated areas, war zones, trading vessel corridors, etc....)

There is research going on right now to build native camera systems that emit data (at the hardware level) that is not dense but rather sparse; information is transmitted as "blocks of pixels that changed between frame two and three" rather than "here is the state of all pixels at frame two, and here is the state of all pixels at frame three". In essence, doing a small amount of delta compression within the hardware itself.

This seems like a really natural fit for an application such as this one; and could potentially cut the amount of data needed to be paid attention to down by orders of magnitude.

Many of these ideas are similar to what is used to compress video down in the first place; it would be interesting to see a combination of video compression technology and "region of interest" ML pruning, where the ML models are applied only to interesting regions, as identified by the compression codec. The codec already has done the analysis to figure out which portions of the video have changed, which portions of the video are simply shifted versions of information from the last frame, etc.... You could significantly speed up ML analysis by making use of that kind of information.

>In essence, doing a small amount of delta compression within the hardware itself.

This is already what hardware video codecs do well. If you’re talking about Chronocam-style differential sensors, they have their specialized use cases, but this is not one of them. Bandwidth from your sensor to your codec is already obscenely high, and you’re on a moving platform anyway.

As far as what video to actually store or do further processing on, ML ROI absolutely makes sense. This is basically what the Google Edge TPU is for: https://cloud.google.com/edge-tpu/

If anyone's interested to read more, I know of them as "event cameras" (as distinguished from "pixel cameras". [0] is one of the universities doing research into them. We're intrigued to use them one day for our automation solutions, but they're a bit pricey at this stage.

A huge advantage of them is that you have practically instant latency on the output, as opposed to restricted frame rates with motion blur from traditional cameras. Have a look at [1] for a few demos. The spinning bucket path reconstruction and drone visual odometry are my favourites.

[0] http://rpg.ifi.uzh.ch/research_dvs.html [1] https://www.youtube.com/watch?v=F3OFzsaPtvI

It's pretty clear that the US government needs to make this a reality. Continuous high resolution live video (visible + Thermal). Entire classes of geopolitical problems will become non-issues. Later, the service can be offered to the general public, like GPS.

If US gov doesn't do it, China will.

The problem is that planes, afaik, do not continuously broadcast their position and health. If you don’t find the black boxes, you can’t figure out what happened. The investigators of MH370 were so desperate for any information of the possible flight path, they had to turn to the pings the jet engines sent to the manufacturer as part of their monitoring [1]. All we need is for passenger planes to broadcast regularly and in short intervals. No indiscriminate high-resolution global monitoring and data processing needed.

[1] https://blogs.crikey.com.au/planetalking/2014/03/25/mh370-la...

Airlines will track planes every 15 minutes starting in November. http://time.com/5137172/mh370-missing-plane/ Some airlines have already met this goal.

I’m stumped as to why they would pick such a ridiculously low sampling rate. Crashes often play out in minutes, wouldn’t it make more sense to have sub-minute resolution? Or is sending a location ping more complicated than I think?

In places without terrestrial ADS-B receivers, you’re relying on space based bandwidth constrained networks to pipe the data to ground stations.

The goal is just to narrow down the search area after a crash. Even at 700 MPH you can only cover 175 miles in 15 minutes. That's a pretty tractable area to search for a blackbox ping. And that's if they have no other info about the crash.

100K square miles is a pretty big area to search. Actually more than that since a plane cruising at 30K feet could glide quite far if it maintained some airframe integrity.

Yeah the search area would be about the size of New Mexico. But a larger area than that was searched in one day, Wednesday March 19 2014. http://www.nationsonline.org/oneworld/map/malaysia-MH370-sea...

By way of counterexample, note we knew pretty well where AF447 crashed and it still took almost two years to locate the wreckage on the sea floor.

I'm completely spitballing because what few articles I looked at didn't really answer that question. I'd wager it comes down to money. If events like MH370 are extremely rare then how much money is worth spending on it? If an event like MH370 happened tomorrow and we knew where it crashed how likely is it to change the outcome?

I'm not trolling and I don't know the answer to that last one. But I'm assuming the general consensus is that it wouldn't change anything - a plane would be down and everyone would still be dead. So instead of mandating an expensive overhaul of everything that can feed more data constantly I assume they're probably trying to make some already in place tech fill the gap and it has limits. I am entirely speculating on that though. But when in doubt, something usually comes down to a cost-benefit analysis.

People need to "feel" safe is one answer. Dead, yeah but they'd know the reason and so on. No one wants their loved one lost without a trace.

Also the money spent is nothing in the grand scale of aviation.

The obvious advantage is post-crash analysis to prevent it from happening in the future.

"But I'm assuming the general consensus is that it wouldn't change anything - a plane would be down and everyone would still be dead"

When you know the exact crash site, you can send help more accurate and therefore faster to save people, as airplanes can sometimes make a emergency water landing...

And in general I don't quite get it. GPS exists, so does Sattelite communication. And sure, for one private person it is quite expensive, but for an Airliner??

> "airplanes can sometimes make a emergency water landing..."

But in those events the planes have transponders/beacons that will tell you where they are. Remember with MH370 the pilot disabled it and then, to the best our knowledge, pointed the plane at the water.

Considering that every android phone (literally billions of units) sends a location update every 5-10 minutes directly to Google, and nobody cares about the cost, why would airplanes transmitting continuous location updates be too costly?

Android phones stop sending location updates when they go out of cellphone reception.

A solution that can continue sending updates from any point on earth costs a lot more than one that only works in range of cellphone networks.

While true, if the airplanes already send 15 minutes updates, the infrastructure is already there. It seems logical simply increasing the ping frequency should not lead to a proportional increase in cost.

Someone up-thread said that it's a progressive target, it'll be brought down to 1 minute intervals by 2021.

The article said that as well. Although, oddly, it says "when they're in trouble.."

>New aircraft must broadcast their locations every minute when they’re in trouble, but only from January 2021. A gradual tightening of requirements starts in November, when airlines must track planes every 15 minutes under regulations adopted by the United Nations’ International Civil Aviation Organization.

There's already in many areas a requirement for aeroplanes to provide ADS-C via Inmarsat satellite updates at 15 minute intervals so that makes it easy to regulate (just increased cost making it global) https://www.aviationtoday.com/2015/09/25/ads-c-makes-increas...

The in distress technology isn't locked down yet but one option is increasing the capabilities of the existing distress beacons standards http://insidegnss.com/the-cospas-sarsat-meosar-system/

This is the only missing modern airliner in memory. Black-boxes and the existing electronic traces of their communications have been enough to find pretty much every major commercial crash for the last 50 years.

In reality tracking commercial airliners in real time is less useful for improving safety than better black boxes that collect more data.

The article says it is going down to once a minute starting in 2021.

It's in the article:

"Under the rules taking effect in 2021, a plane would switch to one-minute tracking automatically when systems detected it was in distress because of turbulence, mechanical difficulties or an unexplained change in course, such as during a hijacking or if the crew became unconscious.

Pilots couldn’t turn the system off after it activates automatically, ICAO said. The system would deactivate itself once the plane was flying safely again.

However, a pilot could turn off the system if it was manually activated.

The challenges tied to minute-by-minute tracking include adding computing power and internet bandwidth to process larger volumes of data. The tighter system also may require reserving more space on the flurry of satellites being launched to satisfy demands for constant internet connectivity."

You can't actually get that far in 15 minutes even at full speed, so 15 minutes is a high enough resolution to at least enable search parties to find the black box, which will hold much better data than can be transmitted anyway.

This sounds like space based ADS-B. https://aireon.com/ http://www.globeinvestor.com/servlet/ArticleNews/story/GAM/2...

The problem here is more that it was the middle of the ocean... gotta have line of sight to transmit.

The jet engines were using satellite communications to ping the manufacturer.

Or does some government actually already have this, but doesn't use it in these cases due to not wanting to reveal their advantage?

I bet the United States' National Reconnaissance Office already has satellites that have the capability. And it's fair to assume they won't be redirecting sensitive military hardware for civilian purposes.

https://en.wikipedia.org/wiki/National_Reconnaissance_Office

Assuming for a moment the NRO has developed suitable optics to image the Earth real-time from geo at say 1 FPS, along with a sensor, and way to pipe everything back to the ground. There are still a couple unavoidable major flaws with such a system:

1) Clouds

2) You can't see all latitudes from geo, the higher the latitude gets, the more marginal the images will be.

So I really doubt such a capability exists. Look a the FOV for a geo satellite like GOES-15: http://www.wmo-sat.info/oscar/satellites/view/151

I don't understand where #2 is coming from. There's plenty of places to go in space other than GEO.

How are you going to continuously monitor the same area on the Earth without being in GEO? We'd know about a constellation of NRO satellites with continuous coverage of an area of the Earth from the folks at satobs.org.

You can't see anything from GEO, it's just too far away to be useful. No way they'll put them out that far.

Yeah, I know. The whole point of my post was in response to abhiminator that I don't think the NRO has the capability to get a live stream of the entire Earth continuously. Are you agreeing with me, or are you saying you think there is an undiscovered & undetectable NRO LEO/MEO constellation continuously observing all parts of the planet?

> continuously observing all parts of the planet?

No there's no way this is happening right now. I'm confident that they can continuously observe specific arbitrary locations of interest, but not the entire planet all at the same time. That's crazy talk.

Thank you for the write-up. I stand corrected.

It's not unheard-of; https://www.airspacemag.com/daily-planet/spysat-and-shuttle-...

On a (much) smaller scale, there is ARGUS-IS [0]

[0] https://en.wikipedia.org/wiki/ARGUS-IS

>realtime recording of the full globe

The scale of infrastructure needed for something like that would be staggering.

Important question is, for what purpose?

Assuming it's the aviation safety domain we're talking about, I think improving aircraft connectivity to ground stations and enhancing real-time monitoring [0] would be a superior solution than to invest in array of video-recording infrastructure to keep tabs on the planet--this without even factoring in the privacy concerns that may arise out of continual recording and access to the said data.

[0] https://www.wsj.com/articles/how-to-avoid-another-malaysia-f...

The new Iridium Next satellites carry ADS-B receivers to enable global coverage. I leave it up to the reader to determine difficulty level of tasking nearby optical surveillance satellites when an anomalous flight path is detected.

https://aireon.com/resources/iridium-next/

Iridium NEXT satellites contain ADS-B receivers for tracking planes that are broadcast ADS-B.

https://aireon.com/resources/its-just-ads-b/

Check this out: https://earthnow.com/

Good old Nathan Myhrvold!

That solution is some time away. In the near term, I'd like to see the commercial aircraft engine OEM's start transmitting in flight GPS coordinates of their engines so that they can be tracked in real time. These engines currently transmit engine telemetry back to their OEM's so they can monitor their performance and condition, but I'm not sure of the time frequency it transmits the data back to the monitoring facility. In the case of Flight MH370, the Rolls Royce engines were transmitting engine telemetry, but it didn't contain the necessary information to be able to track its flight path and precisely determine where it ended up.

Wouldn’t it make more sense to mandate every commercial passenger plane transmit its location every second to a satellite (and be something that can’t be disabled from cockpit)?

1) the tech already exists (wouldn’t even require new satellites)

2) the costs involved are actually not very high

3) would help negate the security and reliability weaknesses of ADS-B.

You can already track most people by their radio tracking devices.

Except when they're in Airplane Mode.

Are we sure the device is 100% offline?

Yes.

There already are such systems...said too much.

It's almost certain the US government knows more than it lets on, but feels the disclosure of information might reveal sources and methods to a possible adversary.

Maybe, but if that were the case, I would have suspected a search ship to 'happen upon' the wreckage at some point.

I don't think they know where it is but they certainly are more informed than the other players. As I said, "more information", not omnipotent knowledge.

If you're so certain, the US could just point at the location of the wreck without actually giving you any new information.

Yea, like most conspiracy theories, it's a pretty silly contention.

Thanks to Elon Musk, we have real-time tracking for modern airplanes! Iridium NEXT has ADS-B receivers on each of their satellites which can see all ADS-B equipped airplanes around the entire globe. It's mostly up-and-running, you can follow the progress at: https://www.iridiumnext.com/

Background on the system for tracking: https://aireon.com/

I didn't realize ADS-B had that kind of range.

I’m able to track aircraft around 100 miles away using a cheap receiver and a 50’ wire antenna. About $50 total investment. I would guess commercial aircraft must have fairly high watt transmitters.

They have upward-facing transmit antennae for 1090MHz ADS-B/extended-squitter for TCAS reasons.

What does it have to with Elon? The owner is Iridium Communications and the satellites are produced by Thales. They're being launched on Falcon 9 but they could have been launched on anything.

Iridium was able to be resurrected only because the Falcon 9 slashed the cost to replace their satellites, it was a shell of a bankrupt company for nearly two decades.

Iridium was "resurrected" from their bankruptcy over 17 years ago, in 2001. Their resurrection has nothing to do with the Falcon 9, and they were planning on launching The NEXT constellation before Falcon 9 was even on the scene. If not F9, they had plans to use the Ukrainian Dnepr. The Falcon 9 ended up letting them do it for cheaper, but that still doesn't mean Musk or SpaceX were necessary (or worthy of thanks) for it to happen.

They went 15 years post-bankruptcy floundering around post bankruptcy without being able to replace a single one of their rapidly decaying satellites. They made numerous plans over those 17 years and none could get funded because they never made financial sense.

They could never get funding to use Ukrainian Dnepr rockets even if they were cheaper than SpaceX. The Ukrainians rockets payload capacity is far smaller than a Falcon 9, quality is questionable, and clearly would never have been able to hit the cadence Iridium needed (only 9 launches this decade, barely more than one per year).

Then SpaceX came along. No-one thought the Falcon 9 would be a success, because it was so ambitious compared to the Falcon 1. Then no-one thought SpaceX could be competitive on pricing, and SpaceX blew everyone away on pricing way before they even started re-using their rockets (because Musk and his team were smart enough to design the Falcon 9 to be the first rocket that could be mass assembled. He's not Tesla or Edison, in reality he's Henry Ford). SpaceX's success proving the Falcon 9 was a big reason iridium got the massive funding they needed for the new satellites.

Sure it could be launched on anything but the fact remain that it is the elon musk company that launch it, instead any other.

"Thanks to Elon Musk"? Seriously? This is exactly the type of Musk-worship that makes people angry.

Elon Musk has hardly anything to do with Iridium NEXT. SpaceX launched the satellites, but that's it. If it wasn't SpaceX, it would've been any other launch provider (just like how China, Russia, and McDonnell Douglas provided 23 Iridium launches, compared to SpaceX's 8 Iridium NEXT launches).

You really should be saying "Thanks to Bary Bertiger, Raymond J. Leopold and Ken Peterson", who were the three engineers who actually came up with, designed, and developed the entire Iridium constellation, as well as Motorola, the company that financed Iridium, Lockheed/Orbital ATK/Thales, the companies that built the satellites, and of course Iridium Communications, the company that actually runs it.

Give credit where credit is due. And it's not due to Musk.

You both are talking past each other. Sure Elon had nothing to do with developing iridium technology. The question is whether Iridium would have been brought back to life without SpaceX. Iridium went bankrupt because it's costs were so high, they went 15 years without replacing a satellite, even though they were only rated for 8 years. They had plans to replace them, but kept postponing. It really looked like they couldn't raise enough money for the huge launch costs and were were just going to ride them into the ground (literally).

Then they suddenly got new funding and started launching satellites again in 2017. Right at the same time Falcon 9s dramatically cut commercial launch costs by well over half compared to other launch providers.

So Elon deserves some credit here, just as Bary, Raymond and Ken do.

Iridium "got new funding" in 2001 when it was bought by PE firms and restructured to generate piles of cash from selling satellite communications to the government, long before your claim of 2017. It had nothing to do with the Falcon 9 other than coincidence. Trying to claim otherwise is misleading.

And what did they do for 15 years with that money? Made plans to replace their dying satellites and having every single plan refused to be funded because of the massive costs.

Suddenly SpaceX cuts launch costs by 4x and Iridium is able to get funded and launch again. Trying to claim coincidence is misleading.

I agree with you about Musk worship in general, but I bet the financial viability of a satellite constellation is at least somewhat dependent on launch costs, and SpaceX has undeniably brought those down significantly. IIRC, the original Iridium company went bankrupt and the constellation was only kept in service by a U.S. government bailout.

Reply on news.ycombinator.com