[LINK] Google's WiFi bungle

Martin Barry marty at supine.com
Wed May 19 21:44:01 AEST 2010

$quoted_author = "Jan Whitaker" ;
> I find this interesting, if this is the actual use of the collected 
> data, taken in conjunction with Kim's point about the broadcast (wide 
> area, longish distances) and the value of this for locating wifi 
> hotspots in relation to a particular house. At the moment I only see 
> 2 networks near my place (one being mine). But in large buildings 
> there can be lots. So the location aspect of the Streetview 
> collection is not very good to start with. In a city, you have no 
> idea where those networks are. So what's the point? It can't be GPS 
> because the actual location is suspect. Are they using relative 
> signal strength? Are they using triangulation? Has anyone dug deeper 
> on this beyond what is being said on the surface (and has been shown 
> to be misleading at first depending on who was answering)?

Actually it is. The car knows exactly where it is thanks to a combination of
GPS, dead reckoning and mapping. Detecting many WiFi networks just makes the
data for any particular location more durable.

The data is then used in a triangulation calculation based on the strength
of signal from surrounding WiFi access points. It's accuracy would be
questionable on it's own but it could be combined with questionable GPS (not
enough satellites or weak signal) to refine a more accurate position.

This only requires detecting three details about each access point in each
sampling period:

- signal strength

This project did not require collection of payload data but Google has done
this and hence the problem.

Oh, there are other companys who have already collected WiFi data for
exactly the same purpose. 


More information about the Link mailing list