I have been involved with designing GPS based monitoring systems for over 15 years. GPS can be very accurate - but only in perfect conditions. By perfect, I mean where the receiver has a clear view of the sky horizon to horizon with no objects that can reflect the signals. Place a device inside a car cockpit and it is immediately compromised. Tall buildings nearby cause what's known as urban canyon effect. Not only does it severely limit the visibility of the GPS constellation, but multipath reflections lead to inaccurate measurements from different satellites. Modern consumer devices and their software have evolved to give quick and constantly available fixes and the price they pay is they rely on all manner of "dead reckoning" techniques, including "assisted" GPS where the fix is not via the GPS satellite system but often by nearby WiFi and mobile telephone triangulation. Some even use their internal accelerometers, which are inherently inaccurate due to the need to mathematically integrate acceleration to deduce speed, also integrating any errors which grow over time. You have to integrate that again to get position, trust me, it doesnt work in practice. Systems integrated into cars use their on board sensors, speed from wheel measurement etc, so that speed indicated is not GPS derived.
In short, when they work well, they can work very well, to within 1 mph in most cases but there are many times when they are little better than taking educated guesses when it comes to speed measurement. As you have observed stood a traffic lights.
For the technical among us, speed is usually derived by measuring Doppler shift of the satellite signals and not by looking at geolocation change.
Note: cheap phones may not even have true GPS but use WiFi and mobile masts for location. These are inherently much more inaccurate than true GPS particularly if you are not stationary.
Edited by brum on 28/03/2015 at 14:47
|