If Google is able to give my position with surprising accuracy using only WiFi, would if be possible to build accurate (say to 10 cm) indoor positioning system, if I had 4 (or more) access points and measured their exact position?
Certainly not by timing signal propagation times (the way GPS operates) - you need absolute timing with sub-nanosecond accuracy on both ends, which you don’t have. Even if you did have atomic clocks attached to your WiFi access points, there’s too much jitter in commodity hardware and software to keep the accuracy.
Measuring signal strength would be an alternative, but I doubt it would work either - I suspect that the signal-strength measurement data available from WiFi adapters does not have nearly enough resolution for that accuracy. You’d also have to contend with varying interference (cordless phones, microwaves, moving bodies, etc.) affecting your readings.
Dunno about 10cm accuracy, but Google already does it at certain malls to an accuracy of several meters: http://www.newscientist.com/article/mg21228425.200-brave-the-mall-maze-with-indoor-positioning-systems.html#.UrrY9hwy_Nw
And this article has various ways you can replace or supplement wifi positioning:
From your site:
I don’t buy that. I tested my smartphone, and after just 30 seconds velocity readings diverge to as much as 10 m/s for phone lying still (due to accelerometer noise). So you can’t measure position like that.
The specifics probably differ between implementations and devices, but dead reckoning is a pretty popular technique at least in GPS devices. As far as I know it’s usually not used as a standalone technique, but in conjunction with other navigational sensors/data.
I question the assertion that Google can locate us accurately using wifi. My understanding is that the best it can do is locate us based on the traceroute information along with metadata that identifies the physical location of the ISP’s gear. (How Google gets access to the ISP’s data is a mystery to me, but clearly, they do.)
If I turn off GPS, Google can locate me to the neighborhood, but not to my home or within my home or yard. At least, that’s what I remember; I’ll test it again and se.
Dead reckoning works great as an update tool in autos, trains, or speedboats, where you’re moving faster than GPS updates. It’s based on previous GPS updates, not accelerometer readings. I seriously doubt that the accuracy of dead reckoning using accelerometer readings would be accurate enough to be useful.
It’s a testable hypothesis, though, if anyone’s written the software. I’d be interested to see the results!
There are systems that use accelerometers:
But they’re probably a lot more useful when your vector is a constant speed and direction, as opposed to being a crazy pedestrian wandering around inside a mall towards shiny things.
Google can cross-reference the wi-fi system you are on with the location of people who have connected in the past when using GPS on smart phones. They also collected loads of wi-fi locations when they did streetview.