Live Roads promises 1.5m-accuracy with standard cellphone hardware | Hacker News Reader

Live Roads promises 1.5m-accuracy with standard cellphone hardware

arstechnica.com

17 pointscpeterso10 comments

I think the key is "probabilistic models" and "sensor fusion". Not more accurate GPS.

Let's say you know a guy is traveling at about 60 mph on a four lane road, two lanes in each direction.

Then the phone's acceleration sensors tell you the guy just moved "left" by 10 feet. A little while later you decide he just moved "right" by 10 feet. He goes back and forth a few times. You also make the assumption that he's not driving on either shoulder. So, you can guess that after he moves left he's in the fast lane; after he moves back right he's in the slow lane.

Phones are so sophisticated these day. Here's something from Apple I found after a quick search:

An accelerometer measures changes in velocity along one axis. All iOS devices have a three-axis accelerometer, which delivers acceleration values in each of the three axes shown in Figure 1.

https://developer.apple.com/documentation/coremotion/getting...

Edit: this can get a lot more sophisticated if you can combine data from multiple users in the same area. Let's say that people in the left lane are moving at 40 mph and in the right lane are moving at 30 mph. You can easily refine your guesses. All your left lane guys should be traveling about 10 mph faster than your right lane guys.

In fact that info is so widely available that I'd be surprised if Google and Apple weren't already using it.

There are plenty of startups doing this. I spoke to Navisens at TreeHacks and they weren't willing to describe their methods outside of typical marketing technobabble.

Sounds like they using weather data to determine slight changes in the speed of light. I'd be surprised if this hasn't been done before though, it seems pretty obvious. So there's probably another component I'm not thinking of.

>slight changes in the speed of light

what could this possibly mean? are you trying to say they're measuring the differences in refractive index of air as a function of humidity?

I think they're talking about environmental factors that affect the propagation speed of the GPS radio signals from the satellites to the surface. This would give them higher resolution timings on the arrival of the GPS signals and thus better accuracy.

I don't think this has anything to do with the actual speed of light as that is constant (in empty space).

On a related note, one of the recent SpaceX launches caused some perturbations in the ionosphere and that had a measurable effect on GPS accuracy:

https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1002/2017...

I think they are using the accelerometer (or even microphone) to measure road noise. What appears to us as mostly white noise probably reveals unique road flaws such as cracks, concrete joints, potholes, dips, bumps.

Sample this data and feed it into a signal matching algorithm like Shazam...

I am not an Android dev, so I don't know if it is possible, but maybe they are accessing the GPS radio directly and doing RTK?

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

This is duscussed in the article. The author states that RTK doesn't scale because of the need for base stations spread about. Instead he guesses that they use Trimble's RTX, which is used in Cadillac's Super Cruise system.

Is Trimble’s RTK just a network of base stations providing corrections to receivers over satellite or IP?

It has been awhile since I researched this, but I think you can just have n RTK receivers within some distance and get quality data, you don't need any actual base stations, you just need multiple simultaneous observations.

The S7 gnss doesnt provide raw observation data so they are likely using the GNSS solution with the phones IMU through a kalman filter.

Reply on news.ycombinator.com