Re: [Talk-transit] Mapping bus routes with doglegs and loops

[Paul Johnson]

On Sun, Feb 22, 2015 at 12:55 AM, Jo  wrote:

> Hi Paul,
>
> How do passengers know where to go and stand if there are no physical
> markers? I think a bus stop should be able to be defined by the fact the
> driver knows where to halt and the passengers know where to wait.
>

In our case, this is published in the schedules and (occasionally) on the
billboards on the outside of the bus.  You can flag a bus about a bus
length after any intersection with no marked bus stop within a one-block
radius.


> Isn't that 'convention' also some sort of ground truth? I"m sure this case
> happens in many countries around the world, although in some of those
> countries it may be the case a bus can be flagged at any point on its
> itinerary.
>

Does word-of-mouth essentially count as ground truth?  I'd like to know if
there's some accepted key=value for this situation that can be used with
highway=bus_stop, if one exists.


> Of course, as soon as roads become busier, that's not possible/practical
> anymore.
>

Despite the fact that Tulsa ended up largely flat enough to put major
thoroughfares along the lines of survey for range and township, the side
streets within those sections often wind around and dead end uselessly for
no reason, even where there's no creek or other obstruction.  This makes
many of the section line thoroughfares rather congested, particularly
during key times (the Advent shopping season, rush hour in general but
especially on a Friday evening, etc), but in our case, this meme applies
.
Tulsans drive obnoxiously close to each other but you can tell who is from
out of town (we have too many different license plates for that to be a
reliable indicator) based on who follows close to a bus.


> I've been working a few years adding almost 7 stops for a small
> country. That was a lot of work, of course. But now I notice that adding
> and maintaining the routes is even more work, hence the creation of the
> script to automate the process where possible.
>
> One of the problems I faced is that when I needed to fix a route, I had to
> apply the same fix to all the variations of that route, over and over
> again. Now I do it once, creating a 'golden  route', then letting the
> script take care of the others. It's still some work, as I need to check
> manually if the code got it right, route by route.
>

Still, I'm laying the initial groundwork to get to the point where we have
a "golden route" for each route (or in the 101 Suburban Acres case,
actually eight routes).


> Concerning the roles, I guess they may help JOSM and iD when people split
> ways, although I think JOSM gets that right without them already. A bigger
> problem is people joining ways, which results in stumps that are not
> connected to the next way anymore. And of course, deleting ways,
> potentially replacing them by new ones.
>

I've been noticing a trend towards shorter ways, to the point where people
are a little worried about merging ways now.  This is generally a Good
Thing, since joining ways generally breaks a lot more than relations (lane
tagging comes immediately to mind, and because of the tag conflict, this
often sets off warning bells).
___
Talk-transit mailing list
Talk-transit@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-transit

Re: [Talk-transit] Mapping bus routes with doglegs and loops

[Jo]

Hi Paul,

How do passengers know where to go and stand if there are no physical
markers? I think a bus stop should be able to be defined by the fact the
driver knows where to halt and the passengers know where to wait. Isn't
that 'convention' also some sort of ground truth? I"m sure this case
happens in many countries around the world, although in some of those
countries it may be the case a bus can be flagged at any point on its
itinerary. Of course, as soon as roads become busier, that's not
possible/practical anymore.

I've been working a few years adding almost 7 stops for a small
country. That was a lot of work, of course. But now I notice that adding
and maintaining the routes is even more work, hence the creation of the
script to automate the process where possible.

One of the problems I faced is that when I needed to fix a route, I had to
apply the same fix to all the variations of that route, over and over
again. Now I do it once, creating a 'golden  route', then letting the
script take care of the others. It's still some work, as I need to check
manually if the code got it right, route by route.

Concerning the roles, I guess they may help JOSM and iD when people split
ways, although I think JOSM gets that right without them already. A bigger
problem is people joining ways, which results in stumps that are not
connected to the next way anymore. And of course, deleting ways,
potentially replacing them by new ones.

Jo





2015-02-22 6:10 GMT+01:00 Paul Johnson :

> On Sat, Feb 21, 2015 at 3:48 PM, Jo  wrote:
>
>> Hi Paul,
>>
>> I see what you mean now. I dropped the forward/backward roles on the ways
>> a few years ago. Recently I thought that they should be a help for the
>> sorting algorithm, but your example proves they aren't.
>>
>
> Well, they do help for sorting out which way a route should be going and
> in which order to some extent.  Though I'm starting to wish we had a way to
> number the sequence.  Role wouldn't do it since we still need forward and
> backwards...
>
>
>> I've been developing a script, which tries to use other good relations to
>> fix the one it's currently run on.
>>
>> For that script to be able to work, you'd need the stops in the correct
>> order though.
>>
>> It works like this:
>>
>> For a sequence of stops it tries to find other relations which have the
>> same sequence. The other relation with the longest sequence in common
>> 'wins'.
>>
>> Then it finds the ways adjacent to the stops on each end of the sequence,
>> then uses the sequence of ways that connects the stops.
>>
>> We can work that way, because we have received the stops and the
>> timetables from the operators, but it's the opposite of what you start
>> with, when you have to get on the
>> bus to create a GPX to get an idea about one the variation routes of a
>> line. (After that you'd use the unstable plugin to add the stops that are
>> already mapped).
>>
>
> Well, the GPX would gather the itinerary.  I still need to go back through
> and doublecheck about 1600 stops, since there's a very high number of stops
> that aren't signed in any way, shape or form that Code for America Tulsa
> received from Tulsa Transit.  And, as far as I'm aware, we expect some kind
> of permanently fixed marker recognizable as such to be able to map it as a
> bus stop.  If we *do* have some way to tag this situation despite a lack
> of ground truth in the physical sense, then, by all means, someone please
> let me know now, so I can back out of tagging stops for a minute, revert
> and repull.  In which case, I'll have the opposite problem I do now, which
> would be *adding* a large number of stops that *aren't* in the data we
> got from Tulsa Transit (which, IMO, is the less worse problem to have, even
> though that's a bigger project).
>
> The script works quite well, as long as you have some 'golden' routes it
>> can grab way sequences from.
>>
>
>  Ouch!  Yeah, I'm not entirely sure that's going to be readily done given
> Tulsa's situation.  I wish this situation were unique, but I somehow think
> I'm going to be beating my head against the wall when I start working with
> the OK Coders to pull in Oklahoma City's transit systems.  And maybe in the
> future, the Iowa Pacific Railroad's upcoming regional transit system, the
> Eastern Flyer Express and it's associated bus network...
>
___
Talk-transit mailing list
Talk-transit@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-transit

Re: [Talk-transit] Mapping bus routes with doglegs and loops

[Paul Johnson]

On Sat, Feb 21, 2015 at 3:48 PM, Jo  wrote:

> Hi Paul,
>
> I see what you mean now. I dropped the forward/backward roles on the ways
> a few years ago. Recently I thought that they should be a help for the
> sorting algorithm, but your example proves they aren't.
>

Well, they do help for sorting out which way a route should be going and in
which order to some extent.  Though I'm starting to wish we had a way to
number the sequence.  Role wouldn't do it since we still need forward and
backwards...


> I've been developing a script, which tries to use other good relations to
> fix the one it's currently run on.
>
> For that script to be able to work, you'd need the stops in the correct
> order though.
>
> It works like this:
>
> For a sequence of stops it tries to find other relations which have the
> same sequence. The other relation with the longest sequence in common
> 'wins'.
>
> Then it finds the ways adjacent to the stops on each end of the sequence,
> then uses the sequence of ways that connects the stops.
>
> We can work that way, because we have received the stops and the
> timetables from the operators, but it's the opposite of what you start
> with, when you have to get on the
> bus to create a GPX to get an idea about one the variation routes of a
> line. (After that you'd use the unstable plugin to add the stops that are
> already mapped).
>

Well, the GPX would gather the itinerary.  I still need to go back through
and doublecheck about 1600 stops, since there's a very high number of stops
that aren't signed in any way, shape or form that Code for America Tulsa
received from Tulsa Transit.  And, as far as I'm aware, we expect some kind
of permanently fixed marker recognizable as such to be able to map it as a
bus stop.  If we *do* have some way to tag this situation despite a lack of
ground truth in the physical sense, then, by all means, someone please let
me know now, so I can back out of tagging stops for a minute, revert and
repull.  In which case, I'll have the opposite problem I do now, which
would be *adding* a large number of stops that *aren't* in the data we got
from Tulsa Transit (which, IMO, is the less worse problem to have, even
though that's a bigger project).

The script works quite well, as long as you have some 'golden' routes it
> can grab way sequences from.
>

 Ouch!  Yeah, I'm not entirely sure that's going to be readily done given
Tulsa's situation.  I wish this situation were unique, but I somehow think
I'm going to be beating my head against the wall when I start working with
the OK Coders to pull in Oklahoma City's transit systems.  And maybe in the
future, the Iowa Pacific Railroad's upcoming regional transit system, the
Eastern Flyer Express and it's associated bus network...
___
Talk-transit mailing list
Talk-transit@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-transit

Re: [Talk-transit] Mapping bus routes with doglegs and loops

[Jo]

Hi Paul,

I see what you mean now. I dropped the forward/backward roles on the ways a
few years ago. Recently I thought that they should be a help for the
sorting algorithm, but your example proves they aren't.

I've been developing a script, which tries to use other good relations to
fix the one it's currently run on.

For that script to be able to work, you'd need the stops in the correct
order though.

It works like this:

For a sequence of stops it tries to find other relations which have the
same sequence. The other relation with the longest sequence in common
'wins'.

Then it finds the ways adjacent to the stops on each end of the sequence,
then uses the sequence of ways that connects the stops.

We can work that way, because we have received the stops and the timetables
from the operators, but it's the opposite of what you start with, when you
have to get on the
bus to create a GPX to get an idea about one the variation routes of a
line. (After that you'd use the unstable plugin to add the stops that are
already mapped).

The script works quite well, as long as you have some 'golden' routes it
can grab way sequences from.

Polyglot

2015-02-21 21:40 GMT+01:00 Paul Johnson :

> On Sat, Feb 21, 2015 at 2:31 PM, Jo  wrote:
>
>> Hi Paul,
>>
>> It helps, when you select a subset of ways and only let JOSM sort those
>> automatically.
>>
>
> True, but if you've had to edit a section of a dogleg to add another, sub
> dogleg, this breaks, too.  The ground truth is breaking the tool.
>
>
>> Can you select one of the relations, then do Ctrl-Shft-h, then copy that
>> url.
>>
>
> No problem, one such example is the 101 Suburban Acres southbound via
> Denver & 49th and Westview
> .  Towards where
> it heads off MLK to Osage Casino along East 63rd North Street, it has a
> dogleg with two branching doglegs, two of those doglegs also have a loop.
>
___
Talk-transit mailing list
Talk-transit@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-transit

Re: [Talk-transit] Mapping bus routes with doglegs and loops

[Paul Johnson]

On Sat, Feb 21, 2015 at 2:31 PM, Jo  wrote:

> Hi Paul,
>
> It helps, when you select a subset of ways and only let JOSM sort those
> automatically.
>

True, but if you've had to edit a section of a dogleg to add another, sub
dogleg, this breaks, too.  The ground truth is breaking the tool.


> Can you select one of the relations, then do Ctrl-Shft-h, then copy that
> url.
>

No problem, one such example is the 101 Suburban Acres southbound via
Denver & 49th and Westview
.  Towards where it
heads off MLK to Osage Casino along East 63rd North Street, it has a dogleg
with two branching doglegs, two of those doglegs also have a loop.
___
Talk-transit mailing list
Talk-transit@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-transit

Re: [Talk-transit] Mapping bus routes with doglegs and loops

[Jo]

Hi Paul,

It helps, when you select a subset of ways and only let JOSM sort those
automatically. Can you select one of the relations, then do Ctrl-Shft-h,
then copy that url.

Jo

2015-02-21 21:12 GMT+01:00 Paul Johnson :

> Is there an easy way to map these?  In JOSM, I run into problems with
> trying to sort when I edit that pretty much complicates the situation to
> the point where I end up having to start over if I get the order wrong, and
> the public transport plugin isn't the most stable thing in the world.
> Starting to bang my head into the wall with this.
>
> Several of many examples I'm running into can be found (as I've mapped
> them so far) with [route=bus][ref=101] in bounding
> box -96.0084915,36.1395383,-95.9528732,36.2658677
>
> ___
> Talk-transit mailing list
> Talk-transit@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-transit
>
>
___
Talk-transit mailing list
Talk-transit@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-transit

[Talk-transit] Mapping bus routes with doglegs and loops

[Paul Johnson]

Is there an easy way to map these?  In JOSM, I run into problems with
trying to sort when I edit that pretty much complicates the situation to
the point where I end up having to start over if I get the order wrong, and
the public transport plugin isn't the most stable thing in the world.
Starting to bang my head into the wall with this.

Several of many examples I'm running into can be found (as I've mapped them
so far) with [route=bus][ref=101] in bounding
box -96.0084915,36.1395383,-95.9528732,36.2658677
___
Talk-transit mailing list
Talk-transit@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-transit