The Car Navigation System / 1981 | Hacker News Reader

The Car Navigation System / 1981

global.honda

14 pointsT-A16 comments

There was a recent Japan's Top Inventions episode on this:

https://www3.nhk.or.jp/nhkworld/en/ondemand/video/2091030/

It's very much the same content, but in video form.

Summary: current location of car is tracked using dead reckoning[1], then plotted against maps that were printed on transparencies. Highlights include using the rocket-inspired gas-rate gyro for finding car heading, and figuring out that the car's location tracking was more accurate than the maps they were testing with.

[1] https://en.wikipedia.org/wiki/Dead_reckoning

Were these "mechanical" gyros somehow vastly superior to the various MEMS contraptions, even the expensive ones? From my experience, the latter ones would dead reckon you to the moon within seconds.

I think so, yeah. The space age made precision manufacturing relatively accessible. The U.S. military had a pretty insatiable appetite for these kinds of things for all sorts of stuff. submarines, torpedos, aircraft, drones, and on and on. https://en.wikipedia.org/wiki/Inertial_navigation_system

a quick search didn't find any hard numbers. but, yeah, I'd bet a few dollars they they were (are?) very good.

> For example, on a map of a 1:100,000 scale, a road of 10 meters wide is shown via a line only 0.1 mm thick. However, in an area where many roads run close together, a number of thin lines will necessarily overlap, making the map difficult to read. In such a case the roads are shown by ignoring actual distances to a certain degree. This practice, called "deformation," has long been common in cartography.

Surely it's Engrish. The proper term is generalization, or in this particular case displacement. (ArcGIS calls this "conflict resolution").

Great podcast that touches on some of these early systems

https://www.youtube.com/watch?v=0HCGn9_aVu0

It blows my mind that they used dead reckoning instead of GPS and it (supposedly) worked!

It blows my mind that dead reckoning ever works, at all. The very definition of "compounding error."

I have seen it said that military/aviation-grade inertial navigation gyroscopes are among the most impressive bits of technology ever manufactured.

when MacBooks got the inertial sensor for parking the hard disk, I hacked up a script to estimate velocity and precision. It was bad. It was fun to have friends drive around, and see how well it did.

It was great for screwing around. But yeah, getting real precision would be a massive undertaking.

__edit__ now that I think about it, they could get massive improvements by counting how many times the wheel rotated. The tire would change size as it wore down, but that would have been a really helpful datapoint.

    I hacked up a script to estimate velocity and precision

Hahahaha that rules!

    they could get massive improvements by counting how 
    many times the wheel rotated. The tire would change 
    size as it wore down

I think the change in size due to rubber wear would be quite dwarfed by the change in size due to tire pressure changes and tire temperature changes (or somebody just replacing the tires with a slightly different model with different dimensions) Also what happens when you're doing sick drifting and the rotation of your wheels becomes entirely decoupled from your car's motion vector????

still, that totally makes sense. tire rotation is another datapoint (or, four datapoints) and a smart enough system could integrate that.

you could also recalibrate the tire data following a complete stop, when the car starts moving again, kind of a zero velocity update thing if I am understanding the wikipedia article on INS in a halfway competent way

The navigation app that I use (NAVITIME) supports connecting to one of those generic Chinese Bluetooth ODB2 dongles and then when you are in a tunnel it will use the car’s speedometer to figure out how far into the tunnel you are

That's awesome!

I'm wondering, though, how necessary it is? Have you found it to be an upgrade over standard GPS?

I don't drive through a lot of tunnels, but I've never noticed my various GPS apps and devices struggling with them over the years. I always assumed they just more or less accomplished the same thing via accelerometer and/or extrapolating your last known heading+velocity when entering the tunnel and/or some handy heuristics (the software assumes the driver is following the path of the tunnel and not boring his own path through the tunnel wall)

If NAVITIME is superior I will totally check that out just for the geekery of it, been meaning to get one of those ODB2 dongles just for the heck of it

Wheel slippage is probably far more significant.

.honda is a TLD now? Whoa.

The list of brand-name TLDs is large.

https://en.wikipedia.org/wiki/List_of_Internet_top-level_dom...

If you have enough money, or your own country, I think anybody can buy .tld these day. Last time I looked into it, the application was well over $100k, with similar annual maintenance fees.

Not a bad moneymaker. It's such a pity that the list of internationalized domain names seems to be empty – Han characters must be an absolute goldmine.

Maybe Elon can have x.x.x

Reply on news.ycombinator.com