Hacker News new | past | comments | ask | show | jobs | submit login

A "fun" side effect of this rapid movement rate is that Australia drifts noticeably relative to GPS-based coordinates. Many programmers made the lazy assumption that coordinates remain fixed and don't need to be adjusted based on the current date.

The effect of this is that the maps in my car are already out by several meters. It's sufficiently bad that it'll sometimes jump to a different road.

Geoscience Australia developed the new "Geocentric Datum of Australia 2020" (GDA2020) to fix this, but not a lot of software has been updated with it as of 2021.

See: https://www.ga.gov.au/scientific-topics/positioning-navigati...




Several metres is a slight exaggeration, not enough to matter for such purposes anyway, and may even already be taken into account.

From 1994 to 2020, the physical difference was about 1.8m (that’s comparing it with another map projection, but what else can you do?), so in what should be the worst case scenario (if they were actually using accurate maps which I think would reasonably imply nothing worse than GDA94), you should have less than two metres of drift, and the only time that’s ever going to matter is when you want to identify service roads distinctly from the adjacent main road. And continental drift is probably still even within your car’s GPS sensor’s margin of error. No, I don’t think you can reasonably blame continental drift for road jumping; it’s typically going to be bad map data, with roads simply being marked in the wrong place. I have a 2010 Mazda 3 with maps from 2010, and out in the country its maps are commonly out by 10–20 metres, with the largest deviation I’ve observed where I know the route hasn’t changed in at least the last couple of decades being about 150m.

To my mind, the real doozy about these geodetic datums is that they’re time-dependent, including a prediction of continental drift. I have no idea whether the car’s maps system does do that, but it should be doing something like storing all coordinates in GDA94, rendering in MGA94, and projecting GPS coordinates into that coordinate space, so that continental drift’s effects are negligible.

(I’ve wondered casually about continental drift in map projections before, but never gotten round to investigating the reality of it. I didn’t realise they could be time-dependent until today. I must get round to asking my dad about it, as he’s done quite a lot of maps and projections stuff.)


The main issue is for navigation that supports lane-level precision. The 2m drift is still not enough by itself, but there is a base error of a meter or two in the GPS signal itself. So the two together is enough to confuse the navigation and have it show you driving in an adjacent lane, which then gives you bad directions.


Huh, didn’t know anything tried to achieve or do anything with lane-level precision.

But if they’re genuinely needing that level of precision, I’d be very surprised if they weren’t taking continental drift into account.


From memory the opening of a military technology to the general public forbid using it to this level of accuracy? Maybe it was only sub metre accuracy?


That was removed a long time ago.

During the 1990s, GPS quality was degraded by the United States government in a program called "Selective Availability"; this was discontinued on May 1, 2000 by a law signed by President Bill Clinton.

GPS can get extreme precision as in mm when given a long tine and a fixed location, which is mostly useful for geologists. However, you can get close to that using extreme precision accelerometers and GPS, but it’s rarely worth the cost.


A rough guide: The GPS signal works on two codes: coarse acquisition (C/A) and precision (P) codes. The P codes are 1 week long and require a secret key even to this day. C/A is I think a few hundred milliseconds and easily locked onto.

They used to use C/A for what it says: coarse acquisition of where you are (space and time), to work out where you are in the P code signal to start looking. Apparently a difficult problem without the C/A to start you in the right place.

To stop baddies using GPS for anything evil, the C/A had an error built in. This error in C/A is what Bill Clinton removed.

To this day military still can use P codes and it is still possible for the C/A to go out to lunch. But practically this is unlikely to happen given the amount of civil applications dependent on accurate gps.


Fascinating replies from both of you, thanks.


That's cool tech; what devices support that sort of navigation? I'd love to put one in my car.


This doesn’t have anything to do with earthquakes.

GPS doesn’t have anywhere near the precision you think.


Mobile phones typically have a 50cm precision in most places, especially newer models that also support GLONASS, Galileo, and QZSS.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: