5 ms·

>Self-driving cars, for example, must have infinitesimally precise location data to avoid accidents. I'm sure surveyors, machinists, and physicists would be in

>Self-driving cars, for example, must have infinitesimally precise location data to avoid accidents. I'm sure surveyors, machinists, and physicists would be interested in how "infinitesimal" location data can be obtained. This article doesn't really spell out what is going on. Did Australia use some other datum than WGS84 and now they changed it?

SilasX 10y ago
I think they meant to say (or should have said) that SDCs need more locational precision than GPS provides. I was going to justify their phrasing as being easier to understand, and accurate enough for a lay audience, but ... mine is still better, IMHO. I would have said: "Self-driving cars, for example, need more precision than GPS has to avoid accidents."
- verandaguy 10y ago
  Isn't this why many self-driving cars employ cameras, IR, LIDAR/radar etc.?
  - jonknee 10y ago
    Yes, I don't know of any self-driving car that uses GPS alone. That sounds problematic for many reasons (enter a tunnel and you're dead!).
    - komali2 10y ago
      Just take a look at your google fitness tracker map after a run to see why this could be bad. Oops, we're suddenly a block north for some reason? Let me just veer sharp right to fix this!
    - verandaguy 10y ago
      Clearly, you can just avoid tunnels -- nobody really _needs_ to leave Switzerland when they're already there, right? :P
      - kwhitefoot 10y ago
        Bit difficult to drive in Norway too. Hundreds of tunnels and an average length of about 800m, longest about 25km. And most of those over a kilometre are not straight so dead reckoning would be risky.
ghaff 10y ago
As I understand it, WGS84 is compatible with various regional terrestrial datums (data?). This is the first stage of a modernization effort in Australia for the datum used there. A lot more information here: http://www.icsm.gov.au/geodesy/modern.html http://www.icsm.gov.au/geodesy/modern.html
- Stratoscope 10y ago
  Interestingly, "datums" (not data) is the correct plural of datum in this context.
jonknee 10y ago
I don't think this is even true though, it's not like self driving cars are driving on GPS alone. They use GPS to know generally where they are and then sensor data to actually navigate in the space. That said there aren't a ton of mapping providers so rolling out a change should not be very hard.
- adrianratnapala 10y ago
  But why would you bet the farm on the assumption that there is no bug anywhere in important code that takes GPS too literaly. True, the GPS coordinates might have a errors much greater than this 1.5 metres. But: (1) bugs can be devious, and I don't trust them to respect this particular insurmountable barrier, and (2) this error accumulates roughly linearly over time and it is better to make frequent small changes than to wait until you have to make a big, panicked one.
  - jonknee 10y ago
    Sure, but this particular example is pure FUD in regards to self-driving cars. That's not to say there are no bugs, but this sort of GPS inaccuracy is not an issue for safe autonomous travel.
xj9 10y ago
That isn't how self-driving cars work. They use GPS much like a person would and drive much like a person would (using various senses and knowing how to maneuver the vehicle).
ChrisFoster 10y ago
No country that I know of uses one of the WGS84 frames for their official datum, and IMO they'd be crazy to do so if they cared about regional accuracy and stability of the frame, and having some kind of say in how it's realized. The reason I say this is that the WGS84 frame is maintained by the NGA (US National Geospatial Intelligence Agency) based on a relatively sparse and worldwide GPS base station network, and updated to a new version whenever they deem necessary. (So far we're up to the sixth version of the frame, officially known as "WGS 84 (G1762)" - see http://earth-info.nga.mil/GandG/publications/NGA_STND_0036_1_0_0_WGS84/NGA.STND.0036_1.0.0_WGS84.pdf http://earth-info.nga.mil/GandG/publications/NGA_STND_0036_1... for the gory details) Note that "using the (current) WGS84 datum" is a very different thing from "using GPS measurements to define a national datum". At the most basic level, GPS measurements just give you timing information which allow you to work out how far you are from the current set of visible satellites. However, to figure out where you are in some well defined coordinate system, you also need the position of the satellites themselves as a function of time (their /ephemerides/, in the jargon). But wait, there are no absolute physical reference frames, so the positions of the satellites themselves need to be measured with respect to something. This "something" is generally a network of GPS base stations attached to the ground in geophysically stable locations. Depending on which set of ground stations you choose and your processing conventions, you'll compute different ephemerides for the satellites. The NGA have one such set of base stations, and they use it to compute the WGS84 /broadcast ephemerides/ which are broadcast by GPS satellites every 30 seconds. But you can make your own! For example the IGS (International GNSS Service, ugh, acronyms) use a much denser set of base stations all around the world to compute their own ephemerides, giving datum IGS14 (previously IGS08) and I believe there's a public transparent process for nations to contribute to this frame. So if you want to define a stable and accurate national datum using GPS measurements, I believe the procedure goes something like this: 1) Build a robust set of GPS ground stations, well distributed across the land mass 2) Observe GPS timing signals continuously at all base stations 3) Solve for the positions and velocities of the base stations to realize a terrestrial reference frame defined by your base stations. Some arbitrary but conventional choices must be made here - for example, what's the orientation of the cartesian coordinate axes? 4) Process the GPS observations each day/week/whatever, along with a (potentially less dense) set of measurements contributed by the rest of the world to compute satellite ephemerides. 5) Make the resulting ephemerides publicly available, so people can realize high precision measurements in your datum, by combining with their own GPS measurements. tl;dr - All measurement is relative. The GPS satellite constellation is used to transfer a coordinate frame defined by some ground control stations (you pick which ones!) back to any other point on the ground. [edit - formatting]
techdragon 10y ago
This article explains why it's off by 1.5 meters much better. http://www.abc.net.au/news/2016-07-28/aust-latitude-longitude-coordinates-out-by-1-5m-scientists/7666858 http://www.abc.net.au/news/2016-07-28/aust-latitude-longitud...

SilasX 10y ago

I think they meant to say (or should have said) that SDCs need more locational precision than GPS provides. I was going to justify their phrasing as being easier to understand, and accurate enough for a lay audience, but ... mine is still better, IMHO. I would have said: "Self-driving cars, for example, need more precision than GPS has to avoid accidents."

verandaguy 10y ago
Isn't this why many self-driving cars employ cameras, IR, LIDAR/radar etc.?
- jonknee 10y ago
  Yes, I don't know of any self-driving car that uses GPS alone. That sounds problematic for many reasons (enter a tunnel and you're dead!).
  - komali2 10y ago
    Just take a look at your google fitness tracker map after a run to see why this could be bad. Oops, we're suddenly a block north for some reason? Let me just veer sharp right to fix this!
  - verandaguy 10y ago
    Clearly, you can just avoid tunnels -- nobody really _needs_ to leave Switzerland when they're already there, right? :P
    - kwhitefoot 10y ago
      Bit difficult to drive in Norway too. Hundreds of tunnels and an average length of about 800m, longest about 25km. And most of those over a kilometre are not straight so dead reckoning would be risky.

ghaff 10y ago

As I understand it, WGS84 is compatible with various regional terrestrial datums (data?). This is the first stage of a modernization effort in Australia for the datum used there. A lot more information here: http://www.icsm.gov.au/geodesy/modern.html http://www.icsm.gov.au/geodesy/modern.html

Stratoscope 10y ago
Interestingly, "datums" (not data) is the correct plural of datum in this context.

jonknee 10y ago

I don't think this is even true though, it's not like self driving cars are driving on GPS alone. They use GPS to know generally where they are and then sensor data to actually navigate in the space. That said there aren't a ton of mapping providers so rolling out a change should not be very hard.

adrianratnapala 10y ago
But why would you bet the farm on the assumption that there is no bug anywhere in important code that takes GPS too literaly. True, the GPS coordinates might have a errors much greater than this 1.5 metres. But: (1) bugs can be devious, and I don't trust them to respect this particular insurmountable barrier, and (2) this error accumulates roughly linearly over time and it is better to make frequent small changes than to wait until you have to make a big, panicked one.
- jonknee 10y ago
  Sure, but this particular example is pure FUD in regards to self-driving cars. That's not to say there are no bugs, but this sort of GPS inaccuracy is not an issue for safe autonomous travel.

xj9 10y ago

That isn't how self-driving cars work. They use GPS much like a person would and drive much like a person would (using various senses and knowing how to maneuver the vehicle).

ChrisFoster 10y ago

No country that I know of uses one of the WGS84 frames for their official datum, and IMO they'd be crazy to do so if they cared about regional accuracy and stability of the frame, and having some kind of say in how it's realized. The reason I say this is that the WGS84 frame is maintained by the NGA (US National Geospatial Intelligence Agency) based on a relatively sparse and worldwide GPS base station network, and updated to a new version whenever they deem necessary. (So far we're up to the sixth version of the frame, officially known as "WGS 84 (G1762)" - see http://earth-info.nga.mil/GandG/publications/NGA_STND_0036_1_0_0_WGS84/NGA.STND.0036_1.0.0_WGS84.pdf http://earth-info.nga.mil/GandG/publications/NGA_STND_0036_1... for the gory details) Note that "using the (current) WGS84 datum" is a very different thing from "using GPS measurements to define a national datum". At the most basic level, GPS measurements just give you timing information which allow you to work out how far you are from the current set of visible satellites. However, to figure out where you are in some well defined coordinate system, you also need the position of the satellites themselves as a function of time (their /ephemerides/, in the jargon). But wait, there are no absolute physical reference frames, so the positions of the satellites themselves need to be measured with respect to something. This "something" is generally a network of GPS base stations attached to the ground in geophysically stable locations. Depending on which set of ground stations you choose and your processing conventions, you'll compute different ephemerides for the satellites. The NGA have one such set of base stations, and they use it to compute the WGS84 /broadcast ephemerides/ which are broadcast by GPS satellites every 30 seconds. But you can make your own! For example the IGS (International GNSS Service, ugh, acronyms) use a much denser set of base stations all around the world to compute their own ephemerides, giving datum IGS14 (previously IGS08) and I believe there's a public transparent process for nations to contribute to this frame. So if you want to define a stable and accurate national datum using GPS measurements, I believe the procedure goes something like this: 1) Build a robust set of GPS ground stations, well distributed across the land mass 2) Observe GPS timing signals continuously at all base stations 3) Solve for the positions and velocities of the base stations to realize a terrestrial reference frame defined by your base stations. Some arbitrary but conventional choices must be made here - for example, what's the orientation of the cartesian coordinate axes? 4) Process the GPS observations each day/week/whatever, along with a (potentially less dense) set of measurements contributed by the rest of the world to compute satellite ephemerides. 5) Make the resulting ephemerides publicly available, so people can realize high precision measurements in your datum, by combining with their own GPS measurements. tl;dr - All measurement is relative. The GPS satellite constellation is used to transfer a coordinate frame defined by some ground control stations (you pick which ones!) back to any other point on the ground. [edit - formatting]

techdragon 10y ago

This article explains why it's off by 1.5 meters much better. http://www.abc.net.au/news/2016-07-28/aust-latitude-longitude-coordinates-out-by-1-5m-scientists/7666858 http://www.abc.net.au/news/2016-07-28/aust-latitude-longitud...