[Tagging] Tagging average speed [Was: Re: Residential roads]

[Colin Smale]

 From the "better late than never" department:

I just discovered a description of how Inrix work with speeds to assist 
with predicted journey times.

Inrix are a leading supplier of this kind of information to 
organisations such as ITIS (in the UK) and TomTom, so I think they have 
probably thought it through...

http://euscorecard.inrix.com/scorecard_eu/UK/methodology.asp

In short they have a "reference speed" for each road segment, which 
represents the speed in the absence of traffic (capped to the legal 
maximum). Beyond that they build a database of average speeds for each 
hour of each day of the week, giving 1+7*24=169 values for each segment. 
They can then do some mathematical wizardry to produce statistics to 
show the effects of congestion.

The "reference speed" is definitely something that could have a place in 
OSM and as it is based on actual journey times it would surely be 
beneficial to routing decisions - better anyway than blindly taking the 
legal maximum speed or attempting to predict the driving speed based on 
attributes of the way in OSM.

If we could agree a tag for this it would at least give a place to put 
the "reference speed". Routers etc can then use this instead of maxspeed 
where it is available.

Colin

On 04/10/2010 11:57, Woll Newall wrote:
> I'm talking about speeds that can be consistently measured, whether 
> because of consistent rush hour conditions or other factors.
>
> I wouldn't use 'average speed' for the tag, because it implies 
> something else, but that's what the OP chose for this thread. 'traffic 
> speed' or something like that would be better.
>
> I'm not talking about "the fastest speed you can drive down this 
> curving country road", which I agree is going to vary depending on the 
> driver etc.
>
> If I go to the nearest main road to my house in between the hours of 
> 16:00 and 18:00 every weekday, I can measure the speed at which the 
> traffic is moving. It will be consistent every day. It will be 
> significantly lower than the 'maxspeed' (something like 1/5th 
> maxspeed, like 8mph). All of the motor traffic will be travelling at 
> this speed (it's so slow that 'slow' drivers will not be left 
> behind!). For the rest of the weekday daytime the speed is more 
> variable but is also significantly slower than the maxspeed. Only late 
> at night/early morning is it physically possible for the traffic to 
> get up to the maxspeed.
>
> In UK cities, I expect many roads are like this. As you may imagine, 
> using the raw maxspeed for routing (as all the routing systems and 
> online routers I have ever used seem to do) is useless in this 
> situation. Not only is the 'time to destination' totally incorrect, 
> the route chosen is often wrong as well, because choosing a slightly 
> different route would make journey times much quicker.
>
> Routing programs can't use heuristics to work out these speeds, they 
> are too dependent on local micro-conditions. But we can measure them.
>
> Woll
>
> _______________________________________________
> Tagging mailing list
> Tagging at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/tagging