[OSM-talk] Highway tags and other junk

Richard Fairhurst richard at systemeD.net
Thu Dec 14 17:29:29 GMT 2006


Ben wrote:

>> Slightly off the point, but once we have a critical mass of GPS data,  we
>> can start to derive this automatically...
>
> How would this work, as If i walk along the side of a dual carrageway the
> data would say average speed 3.5mph.  In contrast a car may say 70mph,
> therefore the average speed of the road is about 37mph

Nonono.

The _mean_ speed (simple mathematical average) of all traffic on that  
road is 36.5mph.

But if you draw a distribution graph, you'll see (for a completely  
fictitious road) a clump of tracks averaging 3mph; these are the  
walkers. Then there'll be another clump at 10mph-15mph; these are the  
cyclists. Then there'll be a more general distribution from 40mph to  
70mph.

 From this, Watson, we can deduce all manner of interesting things. If  
there's a decent clump at 70mph, for example, we can judge that it's  
quite feasible (at that time of day) to drive at that speed. If they  
all suddenly slow down to 60mph at a certain point and speed up again  
immediately after, there's probably a speed camera there.

If the cycling tracks give up halfway where there's no junction,  
they've probably been knocked over by the speeding drivers. Especially  
if the track then speeds up to 85mph down the middle of the road and  
ends at a hospital. And so on.

You don't just take a simple mean, you do interesting statistical  
analysis on it. And I will shut up there because my statistical sk1llz  
are far from l337 and there are doubtless much better qualified people  
on the OSM lists who can do the maths.

cheers
Richard





More information about the talk mailing list