[Tagging] Opinion on meaning of tracktype, smoothness and surface for routing

Paul Johnson baloo at ursamundi.org
Fri Mar 21 18:29:12 UTC 2014


On Fri, Mar 21, 2014 at 11:24 AM, André Pirard <A.Pirard.Papou at gmail.com>wrote:

   The problem comes with such roads as Belgian
N674<http://www.openstreetmap.org/way/72502906#map=17/50.53927/5.64976>which
is uniformly classified as national on IGN official maps. On the
> eastern part, it's worth the primary status for through, heavy traffic.  On
> the center part, it's certainly only secondary.  But on the part going NW
> it's a dangerous road. And, despite its official status, only its 5 m width
> and bends can show that : 2.5 m wide lorries can't cross each other and
> they step on the verge. The road once crumbled down into the meadow below
> and is waiting for the next turn. Even cars must break and make a brisk
> turn.  It would be nice that  OSM routing avoided that road.
>

Sounds like a reasonably well equipped routing engine would naturally avoid
it if there's a better alternate available just based on the number and
radius of curves, classification of the roadway, surface type, and the
vehicle being driven.  Bonus if horizontal and vertical limitations of the
roadway are mapped and the routing engine is aware of the x, y and z
dimensions of the vehicle in question (which is why I see a lot of
"driverless vehicle" technology being more practical for navigation
assistance for human drivers than as a fully autonomous vehicle).

>   Of course everybody is free to add a road width as well, there is the
> tag "width" for this, and also the tag "lanes". Unfortunately until now,
> only 5% of all highway-elements (admittedly not only roads) have the tag
> lanes and 1% has the tag width.
>
> The width would come as a complementary information: avoid it despite a
> gentle official classification apparency.
>

Does anyplace outside North America subjectively tag highway=* or is it
formally tied to the official classification in most places, such as it is
in the UK?

>      Moreover, the width can be very easily measured with JOSM on Bing.
>>
>
>  you should be careful with the spherical mercator projection though, you
> might end up with different widths for the same width due to different
> latitudes, I am not sure how precise those measurements in JOSM actually
> are (some time ago they weren't but maybe this is fixed now).
>
> Good point that would have to be analyzed. Especially if there's a
> difference between NS and EW measurements!
>

I would also be interested in hearing if the JOSM measurement tools are
currently maintained to be accurate.  Granted, I'm not expecting it to be
as accurate as finding some survey marks on either verge and start running
survey lines (for a couple reasons, 1) A digital theodolite, it's
accessories, getting licensed as a traffic control engineer, filing permits
and putting out traffic control devices to do so safely and accurately is
probably beyond the reach of almost everyone here besides the parallel
universe me that won the Powerball jackpot and now has nothing better to
do; and 2) most available aerial imagery, when properly aligned, seems to
be getting us a horizontal dilution of precision within the resolution of
the imagery or better on objects created by relatively skilled and
experienced mappers (based on triangulation from known survey markers to
known points and lines when I've had the access to professional surveying
equipment).

>    Of course, the closely related parameter is speed.
>>
>
>  related to width? I do not think there is a close relation, at least not
> a reliable one.
>
> Speed (to drive safely) is not intrinsic but in fact a consequence of
> other factors, including narrow width. Or it can be enforced.
>

In some places, quite strictly.  Oklahoma will write you up for 76 in a 75
maxspeed or 39 in a 40 minspeed.  The resulting speed uniformity does make
Oklahoma one of the easier states to drive in, assuming your driving
personality is more positive than chaotic neutral on either or both axes
(otherwise, it tends to be expensive, dangerous or both).

>    While reading your texts, I've had a crazy idea:  measuring vibration
>> in the car. There are Android vibration measuring programs like Vibration
>> Monitoring.  Alas, car vibration is very much dependent on car suspension.
>> But would some of us experiment this or another idea and come up with a
>> solution?
>>
>
>  this sounds interesting indeed, while I agree that it mostly depends on
> the car suspension. With (unsuspended) bicycles this would be more reliable
> I guess, but still the ability of the driver / rider to avoid holes in the
> surface might make a huge difference (e.g. in Rome there are some very bad
> roads with profund holes that get tapped every now and then but later
> reopen due to the heavy traffic. If you are on roads that you drive often
> you almost automatically get the habit of avoiding them, also at higher
> speeds, because you know their exact locations by mind).
>
> Yes, dodging the suspension would be the idea, see next.
>

Even on a fixed-fork hardtail bicycle, ride quality is still drastically
affected by tire pressure, diameter, width, and tread pattern, as well as
the length of frame and load balance.  For the sake of finding the most
middle-of-the-road measurements, I propose our "perfectly average vehicle"
would be a 1950 Flying Pigeon PA-02 (this is still the current model year
for this model, still being built) with a statistically average height and
weight human driver, with the stock tires at the recommended pressure
(assuming it has pneumatic tires; can't remember if it does or not).  I
picked this because if you took every single vehicle on the planet and
randomly picked one, your odds are nearly 1 in 2 that it will be a 1950
Flying Pigeon PA-02.


> On 2014-03-20 21:15, Fernando Trebien wrote :
>
> Even so, we would still have to presume things about the driver's
> personality (an adventurous person would not care much about rougher
> surfaces, while a precaucious one would probably rather avoid them).
> We can pick a "standard" personality (we don't even know that very
> well without some statistics, do we?) or we can probe other people and
> then apply statistics on the results.
>
>  There would of course be a protocol such as driving at a 40 km/h constant
> speed and in straight line, unless, of course, it would break something, a
> fact that should of course intervene in the classification.
>

For the sake of automation, 45 or 50 km/h might be a better speed choice,
as most North America market vehicles lock out cruise control at speeds
less than 25 or 30 MPH (which is lame, since I would like to lock in at 10
or 15 MPH for parking lot and living street situations, freeing my foot to
be waiting shy of the brake pedal).
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tagging/attachments/20140321/7ae0cdab/attachment-0001.html>


More information about the Tagging mailing list