[Routing] "SatNav warriors invade Somerset village"

[Chris Fleming]

Marcus Wolschon wrote:
> Chris Fleming schrieb:
> |
> http://maps.google.co.uk/maps?f=d&hl=en&geocode=&saddr=Edinburgh&daddr=Newton+Stewart&sll=54.162434,-3.647461&sspn=17.680923,36.079102&ie=UTF8&ll=55.218057,-3.825989&spn=1.073254,2.254944&z=9
> | (The route was basically follow the A702 and then the A712, picked
> | because it was the most direct route and also because the A712 was
> unmapped)
> |
> | However having traveled down the A712, I can now see why it was
> | unmapped, the road was narrow windy and very slow. In fact a friend
> | traveling the same route  took the A74(M), A701 and the A75 in a quicker
> | time.
> |
> | Now we need some way of tagging "slow" roads, so that we don't pick
> | then. I'm not sure if this would be an "average speed" tag or a you'll
> | only go 3/4 or half the speed on this road compared to your speed on a
> | "normal" A road?
>
> a)
> We had multiple suggestions for using collected real-world travel-times.
> I'm still up for setting up a web-service to collect these. Maybe
> between OpenExpo next week and FOSSGIS.
>   
We'll I have a gps-track so this can be done.

> b)
> Does the A702 have fewer lanes?
> Is it steeper ("alt"-tag)?
> Does it have more curves?
>   
Yes, Yes, Yes !
> For the "windy" I have though of a simple physics-model a few times
> that not only uses the sum of segment-length and the "highway"-tag
> but estimates:
> * maximum speed a curve with a radius R can be traveled
> * maximum speed on differences in altitude
> * expected acceleration after a slow segment into a fast segment following.
>
> These speeds and accelerations are user-selected by vehicle-type
> ~ "car-efficient"=slow acceleration, taking curves fast
> ~              limited top-speed
> ~ "car-sporty"=accelerate more but also decellerate in curves,
> ~              better top-speed
> ~ "lory"=accelerate very slow and take  hills slowly,
> ~        be very slow ans extreme curves)
> ~ ...
>
> I think this will break the usual routing-algorithms if applied to
> the complete path but between relevant intersections this
> could improve a usual time-metric considerably on windy or
> steep roads. (Think serpentines up or down a hill)
> After calculating the route it can be applied to the complete
> path to get a better estimate of an ETA.
>   
I like this approach. For me anyway, I know the road was very windy 
trying to consistently apply a windy=none/some/very type tag is tricky.

It would be good to compare results of the "algorithmic" approach with 
real average speeds. If this could be done consistently then this would 
be a big leap forward.

> c)
> Do we have a tag for the state of the road-surface that
> is in actual use?
>   
I'm not sure, I tend to tag cycle tracks and pathways. But on this kind 
of road it would be good. Again things like covered in mud, bad surface 
would apply to this road :)

Cheers
Chris