No more powerful than a car headlight
August 27, 2013 8:29 PM Subscribe
Finding a Way: The Future of Navigation (BBC Radio 4 program audio, 30 minutes) examines problems with our dependence on GPS and what can be done about it.
From a white van man who wants to take a 2 hour lunch to governments that don't like their neighbors, GPS signals are weak and easy to jam with a tiny electronic device. You can get one for less than $100 but using one can get you into a lot of trouble. Solar flares can cause problems, too. (Not to mention guiding ships off course with GPS spoofing.)
Before GPS, ships navigated using LORAN but LORAN transmissions ended in 2010 with the near universal proliferation of GPS. The UK is working on implementing eLORAN as a backup to GPS. South Korea is working faster.
(Previously)
From a white van man who wants to take a 2 hour lunch to governments that don't like their neighbors, GPS signals are weak and easy to jam with a tiny electronic device. You can get one for less than $100 but using one can get you into a lot of trouble. Solar flares can cause problems, too. (Not to mention guiding ships off course with GPS spoofing.)
Before GPS, ships navigated using LORAN but LORAN transmissions ended in 2010 with the near universal proliferation of GPS. The UK is working on implementing eLORAN as a backup to GPS. South Korea is working faster.
(Previously)
It seems like detecting GPS jammers should be fairly straightforward. You could put a GPS receiver into something like a speed camera — actually, I'd be surprised if speed cameras don't already have GPS receivers, for time synchronization; lots of portable devices do — and get a high-gain antenna so you could determine exactly which vehicle was the culprit. Fuck with GPS, lose your license. Done and done.
Now that said, over-reliance on a single navigational system is a real problem and one that we should work on from other angles as well. Shutting down LORAN is definitely looking a bit premature...
It seems like it ought to be possible to do something like LORAN for civilian land navigation applications using cell towers as known-position base stations. That would be much harder to jam, because the power levels would be a lot higher. These guys seem to be working on something like that, using the 2.4GHz ISM band. But as long as they keep it proprietary I don't think it'll get any traction.
posted by Kadin2048 at 9:42 PM on August 27, 2013 [1 favorite]
Now that said, over-reliance on a single navigational system is a real problem and one that we should work on from other angles as well. Shutting down LORAN is definitely looking a bit premature...
It seems like it ought to be possible to do something like LORAN for civilian land navigation applications using cell towers as known-position base stations. That would be much harder to jam, because the power levels would be a lot higher. These guys seem to be working on something like that, using the 2.4GHz ISM band. But as long as they keep it proprietary I don't think it'll get any traction.
posted by Kadin2048 at 9:42 PM on August 27, 2013 [1 favorite]
The Russians have one called GLONASS which they're starting to push for commercial use, and the EU are working on one called Galileo (which South Korea among others are a partner on) that's a long way off completion.
posted by kersplunk at 10:07 PM on August 27, 2013
posted by kersplunk at 10:07 PM on August 27, 2013
GPS is important for modern manned aviation, but it's absolutely critical for drones. Last year a Schiebel drone being tested in South Korea lost GPS and crashed into its control van, killing its operator. 10 miles away, on the same day, a South Korean Coast Guard helicopter had to abort takeoff due to GPS jamming by North Korea.
If the Schiebel drone lost GPS because it was being jammed, I think that would make it both the first fatality caused by GPS jamming and the first fatal drone accident.
NASA is holding a competition next year, the Unmanned Aircraft Systems Airspace Operations Challenge, focusing on technologies required to safely integrate drones into the national airspace. Besides sensing and avoiding other aircraft, there is a big emphasis on safe navigation even with degraded or lost GPS.
posted by jjwiseman at 11:04 PM on August 27, 2013 [2 favorites]
If the Schiebel drone lost GPS because it was being jammed, I think that would make it both the first fatality caused by GPS jamming and the first fatal drone accident.
NASA is holding a competition next year, the Unmanned Aircraft Systems Airspace Operations Challenge, focusing on technologies required to safely integrate drones into the national airspace. Besides sensing and avoiding other aircraft, there is a big emphasis on safe navigation even with degraded or lost GPS.
posted by jjwiseman at 11:04 PM on August 27, 2013 [2 favorites]
This is in danger of becoming a Bond movie. I can see North Korea using a vessel going off course to start another border skirmish or worse.
It already was.
posted by gjc at 2:11 AM on August 28, 2013
It already was.
posted by gjc at 2:11 AM on August 28, 2013
This is why Astro-Navigation is still a requirement in Canada for Nav Officers. (Probably is everywhere else as well). We don't trust GPS 100%. I'm not sure, but I think LORAN transmitters are being put back on line up here as well.
posted by SpannerX at 2:29 AM on August 28, 2013
posted by SpannerX at 2:29 AM on August 28, 2013
HERE BE DRAGONS: The Unstable Landscapes of GPS
posted by radwolf76 at 3:07 AM on August 28, 2013 [2 favorites]
posted by radwolf76 at 3:07 AM on August 28, 2013 [2 favorites]
It seems like detecting GPS jammers should be fairly straightforward.
It is, and in a way such a system has already been devised. Differential GPS was devised during the days of selective availability, and it works by measuring the error in the GPS signal at a known location and broadcasting that error so that DGPS receivers in the immediate area can apply the necessary correction.
DGPS was designed to correct for a wide-scale, consistent degradation of service, but I could see the the same idea used on a local level to at least detect that something is wrong and where possible, securely notify area receivers out-of-band so they at least know not to trust their current location. Imagine a network of GPS receivers on fixed infrastructure (cell phone towers, near-shore navigational beacons, etc) constantly monitoring the health of the system and pushing out cryptographically-signed warnings, or even corrections, via the cell phone data network. It might not protect the integrity of GPS everywhere, but it would at least guard against some jamming attacks.
posted by RonButNotStupid at 4:31 AM on August 28, 2013
It is, and in a way such a system has already been devised. Differential GPS was devised during the days of selective availability, and it works by measuring the error in the GPS signal at a known location and broadcasting that error so that DGPS receivers in the immediate area can apply the necessary correction.
DGPS was designed to correct for a wide-scale, consistent degradation of service, but I could see the the same idea used on a local level to at least detect that something is wrong and where possible, securely notify area receivers out-of-band so they at least know not to trust their current location. Imagine a network of GPS receivers on fixed infrastructure (cell phone towers, near-shore navigational beacons, etc) constantly monitoring the health of the system and pushing out cryptographically-signed warnings, or even corrections, via the cell phone data network. It might not protect the integrity of GPS everywhere, but it would at least guard against some jamming attacks.
posted by RonButNotStupid at 4:31 AM on August 28, 2013
Signal-based navigation is obsolete. As inertial nav systems miniaturize, they'll take the place of GPS, eLORAN and other nav systems that require signal reception. You'd use an available GPS signal to calibrate the device on startup - so long as your battery doesn't run down, you won't need it again. Inertial nav is already in widespread use on large ships.
posted by Slap*Happy at 4:59 AM on August 28, 2013
posted by Slap*Happy at 4:59 AM on August 28, 2013
Inertial nav systems also drift like hell and require occasional updates of actual position. Right now GPS is the best thing we've got, until terrain or feature-based navigation becomes more robust. And that usually requires that you know the terrain in advance.
posted by olinerd at 5:02 AM on August 28, 2013 [4 favorites]
posted by olinerd at 5:02 AM on August 28, 2013 [4 favorites]
Agreed, INS is not viable as a standalone navigation system right now. A combination GPS/INS solution (GINS) is very common at the moment and is a very good navigation solution.
This is why Astro-Navigation is still a requirement in Canada for Nav Officers. (Probably is everywhere else as well).
Not sure if it's still a requirement for nav officers in USAF, but (at least on the planes I work on) the sextant ports (yes, they were built with sextant ports) have all been crammed with antennas. There's no way to use astro navigation at the moment.
GPS jamming and spoofing is a very big concern for the Air Force and there are technologies that can prevent both.
posted by backseatpilot at 5:08 AM on August 28, 2013 [1 favorite]
This is why Astro-Navigation is still a requirement in Canada for Nav Officers. (Probably is everywhere else as well).
Not sure if it's still a requirement for nav officers in USAF, but (at least on the planes I work on) the sextant ports (yes, they were built with sextant ports) have all been crammed with antennas. There's no way to use astro navigation at the moment.
GPS jamming and spoofing is a very big concern for the Air Force and there are technologies that can prevent both.
posted by backseatpilot at 5:08 AM on August 28, 2013 [1 favorite]
Looking back at this thread again I think it's important to note that no single type of navigation really stands alone. Look back just a few decades ago, you had people using dead reckoning ("Well, we're going about this fast, and in roughly this direction, and it's been for about this much time, so... guess we're here") in conjunction with fixed navigation references (celestial navigation). Over time a few things new technologies came into play, like radar, but fundamentally it was the same thing -- dead reckoning and triangulation against a fixed point. Even most car GPS systems now have accelerometers in them to dead reckon themselves to some degree.
INS and GPS work in conjunction like this. INS is a fancier form of dead reckoning and GPS is your triangulation. The timing technology that is part of GPS, as mentioned, is pretty critical these days too; you can do some really nifty stuff with timing, though the increased availability of (relatively) cheap, small atomic clocks may help reduce the frequency needed for GPS updates simply to resynchronize time. But even as INS technology improves, you still need a fixed reference point to do an initial calibration, and to correct for your drift. Some sort of satellite-based triangulation is still one of the best options we have. Whether or not that's GPS or GLONASS or Galileo is beside the point. But no one right now is using INS without GPS, and I don't foresee a time that we will be using it without SOME fixed point reference.
posted by olinerd at 5:41 AM on August 28, 2013 [2 favorites]
INS and GPS work in conjunction like this. INS is a fancier form of dead reckoning and GPS is your triangulation. The timing technology that is part of GPS, as mentioned, is pretty critical these days too; you can do some really nifty stuff with timing, though the increased availability of (relatively) cheap, small atomic clocks may help reduce the frequency needed for GPS updates simply to resynchronize time. But even as INS technology improves, you still need a fixed reference point to do an initial calibration, and to correct for your drift. Some sort of satellite-based triangulation is still one of the best options we have. Whether or not that's GPS or GLONASS or Galileo is beside the point. But no one right now is using INS without GPS, and I don't foresee a time that we will be using it without SOME fixed point reference.
posted by olinerd at 5:41 AM on August 28, 2013 [2 favorites]
HA. Google can make a car, that can drive across the goddam country by itself using only slightly more than a few off-the-shelf cameras... The same Google that has mapping tools that can see the cracks in every driveway in the ENTIRE WORLD... The same Google that accepted millions of dollars of PRISM money...
And you guys think that the US military's multi-million dollar drones are relying on GPS?!?
I'm sorry, but just taking all of this at face value, and with no justifiable facts to back it up... I'm already assuming that GPS has already been relegated to a secondary if not tertiary navigation option.
posted by Blue_Villain at 6:04 AM on August 28, 2013
And you guys think that the US military's multi-million dollar drones are relying on GPS?!?
I'm sorry, but just taking all of this at face value, and with no justifiable facts to back it up... I'm already assuming that GPS has already been relegated to a secondary if not tertiary navigation option.
posted by Blue_Villain at 6:04 AM on August 28, 2013
But no one right now is using INS without GPS
We have a customer that does! It's a huge pain in the ass and they've resorted to flying with a handheld Garmin GPS in the cockpit.
I'm sorry, but just taking all of this at face value, and with no justifiable facts to back it up... I'm already assuming that GPS has already been relegated to a secondary if not tertiary navigation option.
This is what I do for a living. I assure you, I am not taking this at face value. GPS is what people use, there's nothing else.
posted by backseatpilot at 6:06 AM on August 28, 2013 [3 favorites]
We have a customer that does! It's a huge pain in the ass and they've resorted to flying with a handheld Garmin GPS in the cockpit.
I'm sorry, but just taking all of this at face value, and with no justifiable facts to back it up... I'm already assuming that GPS has already been relegated to a secondary if not tertiary navigation option.
This is what I do for a living. I assure you, I am not taking this at face value. GPS is what people use, there's nothing else.
posted by backseatpilot at 6:06 AM on August 28, 2013 [3 favorites]
there's nothing else.
/facepalm
posted by Blue_Villain at 6:19 AM on August 28, 2013
/facepalm
It seems like it ought to be possible to do something like LORAN for civilian land navigation applications using cell towers as known-position base stations. That would be much harder to jam, because the power levels would be a lot higher. These guys seem to be working on something like that, using the 2.4GHz ISM band. But as long as they keep it proprietary I don't think it'll get any traction.Apparently not. Not at least for the US military.
posted by Kadin2048 at 9:42 PM on August 27 [1 favorite +] [!]
The Russians have one called GLONASS which they're starting to push for commercial use, and the EU are working on one called Galileo (which South Korea among others are a partner on) that's a long way off completion.
posted by kersplunk at 10:07 PM on August 27 [+] [!]
And China has kind of a weird one called Beidou.
posted by jjwiseman at 10:47 PM on August 27 [+] [!]
posted by Blue_Villain at 6:19 AM on August 28, 2013
Apparently not. Not at least for the US military.
The military has always had their own flavor of GPS which is more secure than the civilian version and requires special receivers, but it's still GPS.
posted by RonButNotStupid at 6:28 AM on August 28, 2013
The military has always had their own flavor of GPS which is more secure than the civilian version and requires special receivers, but it's still GPS.
posted by RonButNotStupid at 6:28 AM on August 28, 2013
Correct, not for the US military. Even if it weren't decommissioned, Loran isn't accurate enough for drone navigation. And no way in hell is the DoD going to rely on another country's navigation satellites. The thing is, though, that GLONASS and Galileo are going to suffer from the same spoofing and jamming issues that GPS has now - the vulnerabilities are inherent in the medium (broadcasting non-encrypted signals across the globe).
posted by backseatpilot at 6:29 AM on August 28, 2013 [1 favorite]
posted by backseatpilot at 6:29 AM on August 28, 2013 [1 favorite]
And you guys think that the US military's multi-million dollar drones are relying on GPS?!?
The larger drones have ring-laser gyroscope driven IMUs which is what manned military planes also use as a supplement to GPS. Like all inertial navigation systems, they drift over time and are initialised when the plane is started up (one of the reasons why there's co-ordinates painted on the side of hangars). If you jam GPS, they can run exclusively on their inertial systems for a classified but finite length of time before the drift becomes unacceptable.
How long that time is, whether guided weapons will still work (and the degree to which their accuracy will be degraded), which drones and planes have which IMU, how GPS spoofing rather than jamming is detected and handled - these are all classified information.
What is obvious though is that persistent, area-wide GPS jamming and spoofing would cause a significant operational challenge for a modern military.
posted by atrazine at 6:46 AM on August 28, 2013 [1 favorite]
The larger drones have ring-laser gyroscope driven IMUs which is what manned military planes also use as a supplement to GPS. Like all inertial navigation systems, they drift over time and are initialised when the plane is started up (one of the reasons why there's co-ordinates painted on the side of hangars). If you jam GPS, they can run exclusively on their inertial systems for a classified but finite length of time before the drift becomes unacceptable.
How long that time is, whether guided weapons will still work (and the degree to which their accuracy will be degraded), which drones and planes have which IMU, how GPS spoofing rather than jamming is detected and handled - these are all classified information.
What is obvious though is that persistent, area-wide GPS jamming and spoofing would cause a significant operational challenge for a modern military.
posted by atrazine at 6:46 AM on August 28, 2013 [1 favorite]
I'm glad VOR still works, though it's a bit crap on the ground.
It's the timing aspect of GPS single-sourcing that's worrying. So much depends on that PPS signal: banking encryption, power line synch, …
posted by scruss at 6:53 AM on August 28, 2013
It's the timing aspect of GPS single-sourcing that's worrying. So much depends on that PPS signal: banking encryption, power line synch, …
posted by scruss at 6:53 AM on August 28, 2013
Seriously... nobody's picking up on the concept that I can tell car to drive across the country, and it will do this without any additional input from the driver, using only cameras... to you know, figure out where lanes and buildings and people are... and then just assuming the military doesn't have the same thing on a plane. Despite the fact that the company who built said car used military satellites to take pictures of the entire globe... and the military allowed them to do this... because the military just, I dunno, likes them a lot?
Riiiiight. And I'm the crazy one here.
posted by Blue_Villain at 6:56 AM on August 28, 2013
Riiiiight. And I'm the crazy one here.
posted by Blue_Villain at 6:56 AM on August 28, 2013
The google self-driving car uses GPS as well. There's a video floating around that's about half an hour long that explains a lot about how it works. Maybe you should watch it. The car definitely does not rely solely on cameras for navigation.
posted by tylerkaraszewski at 7:20 AM on August 28, 2013 [4 favorites]
posted by tylerkaraszewski at 7:20 AM on August 28, 2013 [4 favorites]
No, but it relies on cameras to stay in the lanes and not hit moving objects in front of it. No?
posted by Blue_Villain at 7:23 AM on August 28, 2013
posted by Blue_Villain at 7:23 AM on August 28, 2013
Seriously... nobody's picking up on the concept that I can tell car to drive across the country, and it will do this without any additional input from the driver, using only cameras...
As tylerkaraszewski just mentioned, the Google car uses GPS to locate itself within several meters, and then uses its stored maps and sensors to fine-tune the location from there. Doing the same thing reliably without GPS is a much harder problem. Maybe it would work, or maybe it would mostly work and occasionally get confused and drive off a bridge, but either way it's not a problem that we know has been solved.
And for drones in particular, consider the problem of a computer visually identifying where it is taking into account weather, clouds, change of seasons, high speeds, vibrations, and the fact that it is expected to hit targets with high accuracy and crash into things relatively rarely.
I also think it's also a mistake to assume that the military is using more advanced technology than civilians, even if these problems are theoretically possible to solve. Military procurement isn't like James Bond picking up the latest toys from Q; it's more into predictability and bureaucracy. As of last October, over half of US military drones still broadcast unencrypted video feeds that can easily be accessed with off-the-shelf hardware. Think about the process that allowed that to happen, and whether the default assumption should be that the same planes are using more advanced visual processing tech than anyone else knows about.
posted by jhc at 7:27 AM on August 28, 2013 [2 favorites]
As tylerkaraszewski just mentioned, the Google car uses GPS to locate itself within several meters, and then uses its stored maps and sensors to fine-tune the location from there. Doing the same thing reliably without GPS is a much harder problem. Maybe it would work, or maybe it would mostly work and occasionally get confused and drive off a bridge, but either way it's not a problem that we know has been solved.
And for drones in particular, consider the problem of a computer visually identifying where it is taking into account weather, clouds, change of seasons, high speeds, vibrations, and the fact that it is expected to hit targets with high accuracy and crash into things relatively rarely.
I also think it's also a mistake to assume that the military is using more advanced technology than civilians, even if these problems are theoretically possible to solve. Military procurement isn't like James Bond picking up the latest toys from Q; it's more into predictability and bureaucracy. As of last October, over half of US military drones still broadcast unencrypted video feeds that can easily be accessed with off-the-shelf hardware. Think about the process that allowed that to happen, and whether the default assumption should be that the same planes are using more advanced visual processing tech than anyone else knows about.
posted by jhc at 7:27 AM on August 28, 2013 [2 favorites]
I don't like to make a habit out of this, but you asked for it.
There are several problems with your "argument" that Google can drive cars autonomously, therefore drones don't require GPS. Kind of an ad hominem to begin with, but anyway. Here are some points to consider:
Cameras are great for when you're close to stuff. In fact, lots of planes use cameras (especially infrared cameras) for "synthetic vision" applications, which allow you to "see" through clouds and see whatever airport you're aiming for. Cameras are not good for area navigation. You need to compare the image you capture against a known-good image of the area to identify where you are, which means keeping an up-to-date database of satellite imagery on board at all times. There are several problems with this approach, but the top several are -
1) Ground features change frequently. Civilian navigational charts get updated several times a year because of this, and they don't show Google Earth-like imagery, just iconography pointing out major items of interest on the ground. I am not aware of any software that exists currently that can take an aerial photograph and compare it to a map and determine location. Navigating by comparing "your" (or the drone's) view to a satellite image would require way more imaging satellites taking way more images of the ground than currently exist. Plus, keeping that updated on your aircraft would be a daunting task.
2) Aircraft exist in more dimensions than automobiles. Taking photos at odd angles at different times of day (due to the sun casting shadows) will confuse image recognition software. Self-driving cars just need to worry about keeping the white lines on either side of themselves, for the most part.
3) All of this requires processing power which drones can't accommodate. They're relatively small, and they have limited space for hardware, power, and cooling. There's no point in flying if you don't have any space for your mission equipment.
As far as "figure out where lanes and buildings and people are", in aviation parlance that's known as "sense and avoid" and it's still a very active topic. It has not been solved satisfactorily yet, which is why drones are still very highly controlled in US airspace by the FAA. Cameras (or whatever you think Google Cars are using) do not have enough resolution to a) identify a threat, b) determine its vector, and c) get out of its damned way. Avoiding pedestrians, sure, but at comparable distances in the air you're already committed to a mid-air collision. Take a look at this accident report from the NTSB (the pdf has the relevant info I want to point out), involving a mid-air over the Hudson River between a helicopter and a Piper. The NTSB did a simulation to determine when the two pilots were able to see each other, and they found that the aircraft were almost on top of each other before they could be distinguished by human eyes. Cameras won't save you here. There are other technologies to help with sense and avoid, but they often require each aircraft to have active transponders operating which won't help avoid small planes with no avionics, birds, unlit towers, or other hazards. Also, how do you teach a computer to differentiate a plane far out on a collision course from a speck of dirt on the lens? They look basically the same. I've had days in the cockpit where I thought sunlight glinting off a skyscraper was actually the anti-collision lights from another plane.
The point is that radio- and satellite- based navigation (and inertial systems, as well) do a hell of a lot better of a job of determining your position and speed than even your own eyes can, so I'm having trouble imagining a camera-based system that can do even as good a job as my two squishy eyeballs. That being said, here's where cameras come in handy on an aircraft:
1) "Synthetic Vision" systems for landing (for both piloted and unmanned aircraft)
2) Nothing Else.
Navigation and obstacle avoidance are completely different in an aircraft compared to driving on a road and the two are not worth comparing.
posted by backseatpilot at 7:32 AM on August 28, 2013 [6 favorites]
There are several problems with your "argument" that Google can drive cars autonomously, therefore drones don't require GPS. Kind of an ad hominem to begin with, but anyway. Here are some points to consider:
Cameras are great for when you're close to stuff. In fact, lots of planes use cameras (especially infrared cameras) for "synthetic vision" applications, which allow you to "see" through clouds and see whatever airport you're aiming for. Cameras are not good for area navigation. You need to compare the image you capture against a known-good image of the area to identify where you are, which means keeping an up-to-date database of satellite imagery on board at all times. There are several problems with this approach, but the top several are -
1) Ground features change frequently. Civilian navigational charts get updated several times a year because of this, and they don't show Google Earth-like imagery, just iconography pointing out major items of interest on the ground. I am not aware of any software that exists currently that can take an aerial photograph and compare it to a map and determine location. Navigating by comparing "your" (or the drone's) view to a satellite image would require way more imaging satellites taking way more images of the ground than currently exist. Plus, keeping that updated on your aircraft would be a daunting task.
2) Aircraft exist in more dimensions than automobiles. Taking photos at odd angles at different times of day (due to the sun casting shadows) will confuse image recognition software. Self-driving cars just need to worry about keeping the white lines on either side of themselves, for the most part.
3) All of this requires processing power which drones can't accommodate. They're relatively small, and they have limited space for hardware, power, and cooling. There's no point in flying if you don't have any space for your mission equipment.
As far as "figure out where lanes and buildings and people are", in aviation parlance that's known as "sense and avoid" and it's still a very active topic. It has not been solved satisfactorily yet, which is why drones are still very highly controlled in US airspace by the FAA. Cameras (or whatever you think Google Cars are using) do not have enough resolution to a) identify a threat, b) determine its vector, and c) get out of its damned way. Avoiding pedestrians, sure, but at comparable distances in the air you're already committed to a mid-air collision. Take a look at this accident report from the NTSB (the pdf has the relevant info I want to point out), involving a mid-air over the Hudson River between a helicopter and a Piper. The NTSB did a simulation to determine when the two pilots were able to see each other, and they found that the aircraft were almost on top of each other before they could be distinguished by human eyes. Cameras won't save you here. There are other technologies to help with sense and avoid, but they often require each aircraft to have active transponders operating which won't help avoid small planes with no avionics, birds, unlit towers, or other hazards. Also, how do you teach a computer to differentiate a plane far out on a collision course from a speck of dirt on the lens? They look basically the same. I've had days in the cockpit where I thought sunlight glinting off a skyscraper was actually the anti-collision lights from another plane.
The point is that radio- and satellite- based navigation (and inertial systems, as well) do a hell of a lot better of a job of determining your position and speed than even your own eyes can, so I'm having trouble imagining a camera-based system that can do even as good a job as my two squishy eyeballs. That being said, here's where cameras come in handy on an aircraft:
1) "Synthetic Vision" systems for landing (for both piloted and unmanned aircraft)
2) Nothing Else.
Navigation and obstacle avoidance are completely different in an aircraft compared to driving on a road and the two are not worth comparing.
posted by backseatpilot at 7:32 AM on August 28, 2013 [6 favorites]
This is why Astro-Navigation is still a requirement in Canada for Nav Officers. (Probably is everywhere else as well).
I can't speak to the USAF, but I heard that at the Naval Academy, traditional celestial navigation was moved from a required to an optional course as of a couple of years ago. One of my sailing buddies was an occasional instructor / guest lecturer in the course. He suggested that if it was something you wanted to learn, that it was time to take a course in it before everyone who knew it well retired.
Perhaps that will change if the vulnerabilities of satellite navigation become obvious.
One system that still uses celestial navigation, somewhat menacingly, is ICBMs. In order to be a viable second-strike system they can't have a hard dependency on GPS. They do a lot of inertial navigation but -- based on what I've heard -- perform a mid-flight celestial position fix in order to eliminate drift, and then glide to their targets on INS. So, at least if you are at altitude and can guarantee an unobstructed view of the stars, fully automated celestial navigation to some high degree of precision is possible.
posted by Kadin2048 at 7:54 AM on August 28, 2013 [1 favorite]
I can't speak to the USAF, but I heard that at the Naval Academy, traditional celestial navigation was moved from a required to an optional course as of a couple of years ago. One of my sailing buddies was an occasional instructor / guest lecturer in the course. He suggested that if it was something you wanted to learn, that it was time to take a course in it before everyone who knew it well retired.
Perhaps that will change if the vulnerabilities of satellite navigation become obvious.
One system that still uses celestial navigation, somewhat menacingly, is ICBMs. In order to be a viable second-strike system they can't have a hard dependency on GPS. They do a lot of inertial navigation but -- based on what I've heard -- perform a mid-flight celestial position fix in order to eliminate drift, and then glide to their targets on INS. So, at least if you are at altitude and can guarantee an unobstructed view of the stars, fully automated celestial navigation to some high degree of precision is possible.
posted by Kadin2048 at 7:54 AM on August 28, 2013 [1 favorite]
Kadin2048: "They do a lot of inertial navigation but -- based on what I've heard -- perform a mid-flight celestial position fix in order to eliminate drift,"
Yep.
posted by ArgentCorvid at 8:16 AM on August 28, 2013
Yep.
posted by ArgentCorvid at 8:16 AM on August 28, 2013
Seriously... nobody's picking up on the concept that I can tell car to drive across the country, and it will do this without any additional input from the driver, using only cameras... to you know, figure out where lanes and buildings and people are... and then just assuming the military doesn't have the same thing on a plane. Despite the fact that the company who built said car used military satellites to take pictures of the entire globe... and the military allowed them to do this... because the military just, I dunno, likes them a lot?
First of all you can't, in fact, tell your car to do that just yet. Google's self driving car prototypes use active sensing (LIDAR) to super-impose distance data on top of the camera images. That is harder to do from altitude, it can also be jammed, and worst of all active sensor systems are easily detectable. Building a three dimensional view of the world from camera data is actually very difficult, especially over terrain with relatively few features resolvable or in poor weather.
Second, those systems do exist, ground following RADAR systems exist and cruise missiles have used them in the past.
Third, Google didn't use military satellites to take pictures of the entire globe. Their high-res imagery is from low-flying planes and they purchased it commercially.
Fourth, the detailed LIDAR mapping Google uses for the self driving cars is done from ground level, you cannot get that kind of centimetre precision in positioning from very far away because of diffraction limits.
posted by atrazine at 8:36 AM on August 28, 2013 [3 favorites]
First of all you can't, in fact, tell your car to do that just yet. Google's self driving car prototypes use active sensing (LIDAR) to super-impose distance data on top of the camera images. That is harder to do from altitude, it can also be jammed, and worst of all active sensor systems are easily detectable. Building a three dimensional view of the world from camera data is actually very difficult, especially over terrain with relatively few features resolvable or in poor weather.
Second, those systems do exist, ground following RADAR systems exist and cruise missiles have used them in the past.
Third, Google didn't use military satellites to take pictures of the entire globe. Their high-res imagery is from low-flying planes and they purchased it commercially.
Fourth, the detailed LIDAR mapping Google uses for the self driving cars is done from ground level, you cannot get that kind of centimetre precision in positioning from very far away because of diffraction limits.
posted by atrazine at 8:36 AM on August 28, 2013 [3 favorites]
It already was.
I've seen the movie, which is why it came to mind.
posted by arcticseal at 9:02 AM on August 28, 2013
I've seen the movie, which is why it came to mind.
posted by arcticseal at 9:02 AM on August 28, 2013
Yeah, a weird cold war technology is Digital Scene Matching Area Correlation (DSMAC), a method of guiding cruise missiles that were almost to their target by comparing a stored image of the target to the view through a camera. The missiles even had strobe lights so they could navigate at night by lighting up the terrain below.
backseatpilot, I agree that vision systems are not yet at the point where they can help with navigation or sense-and-avoid, but I believe they're much closer than you think (I've worked on computer vision applications). From what I've heard, multiple teams are experimenting with vision based systems to handle the sense-and-avoid component of the NASA Unmanned Aircraft Systems Airspace Operations Challenge. This is an area where I think robots will inevitably outperform humans.
I also think vision based navigation could probably be used today, not on a large scale, but in conjunction with INS to deal with a temporary GPS outage (or just to get down on the ground during a longer outage).
posted by jjwiseman at 12:04 PM on August 28, 2013
backseatpilot, I agree that vision systems are not yet at the point where they can help with navigation or sense-and-avoid, but I believe they're much closer than you think (I've worked on computer vision applications). From what I've heard, multiple teams are experimenting with vision based systems to handle the sense-and-avoid component of the NASA Unmanned Aircraft Systems Airspace Operations Challenge. This is an area where I think robots will inevitably outperform humans.
I also think vision based navigation could probably be used today, not on a large scale, but in conjunction with INS to deal with a temporary GPS outage (or just to get down on the ground during a longer outage).
posted by jjwiseman at 12:04 PM on August 28, 2013
BTW, it's interesting to imagine how the world would be different today if Bill Clinton hadn't turned off Selective Availability. GPS is a lot less useful with 300 foot random error...
posted by jjwiseman at 12:10 PM on August 28, 2013 [1 favorite]
posted by jjwiseman at 12:10 PM on August 28, 2013 [1 favorite]
Just coming back to say that as a human, I am really good at not hitting other cars on the highway, but I don't have a fucking clue where I am most of the time. (seriously, once accidentally went to Maine while aiming for Vermont)
Obstacle avoidance + rules of the road =/= navigation.
posted by olinerd at 12:14 PM on August 28, 2013 [2 favorites]
Obstacle avoidance + rules of the road =/= navigation.
posted by olinerd at 12:14 PM on August 28, 2013 [2 favorites]
examines problems with our dependence on GPS
What's this "we" shit, Kemo Sabe?
May your GPS plot you a route off my lawn.
posted by Spatch at 2:46 PM on August 28, 2013
What's this "we" shit, Kemo Sabe?
May your GPS plot you a route off my lawn.
posted by Spatch at 2:46 PM on August 28, 2013
I have an old GPS kicking around from the days before SA was turned off.
GPS failure bad enough to require LORAN seems like an example of "very low probability, very high impact" change. I think that taking LORAN offline was an unwise choice. It's old and crufty but it works under an entirely different mechanism, so it's belt and suspenders, and the entire LORAN system probably costs a tiny fraction of GPS, and probably a fraction of the cost in life and limb of one bad GPS-failure-related crash. A bad solar flare takes out a couple of satellites, and suddenly things are challenging in a section of sky, I'd think.
posted by rmd1023 at 6:25 PM on August 28, 2013 [1 favorite]
GPS failure bad enough to require LORAN seems like an example of "very low probability, very high impact" change. I think that taking LORAN offline was an unwise choice. It's old and crufty but it works under an entirely different mechanism, so it's belt and suspenders, and the entire LORAN system probably costs a tiny fraction of GPS, and probably a fraction of the cost in life and limb of one bad GPS-failure-related crash. A bad solar flare takes out a couple of satellites, and suddenly things are challenging in a section of sky, I'd think.
posted by rmd1023 at 6:25 PM on August 28, 2013 [1 favorite]
« Older To make journalism harder, slower, less secure | “It just got very, very old and all of us felt... Newer »
This thread has been archived and is closed to new comments
This is in danger of becoming a Bond movie. I can see North Korea using a vessel going off course to start another border skirmish or worse.
posted by arcticseal at 8:43 PM on August 27, 2013