[Talk-transit] Public Transport routing

Sarah Hoffmann lonvia at denofr.de
Sat Sep 5 14:50:10 BST 2009


On Sat, Sep 05, 2009 at 09:57:26AM +0100, Peter Miller wrote:
>
> On 5 Sep 2009, at 07:50, Sarah Hoffmann wrote:
>
>> On Fri, Sep 04, 2009 at 07:42:58PM +0100, Péter Connell wrote:
>>> To not use timetables would case various aberrant behaviours like
>>> routing Yorkshire-Scotland traffic via Settle (side-note: does NRES
>>> still need to be forced to route this way?)
>>>
>>> Speed differentials are such that you'd need scheduling information  
>>> for
>>> any meaningful set up, e.g. here http://osm.org/go/evjiG2F_- where  
>>> the
>>> fastest buses run along the bypass
>>
>> This problem might be solved by adding approximate travel times to
>> the routes. While timetables themselves are far to complex to add
>> to OSM, the time it takes to travel between two stations stays
>> fairly constant. It might be a nice use for the role in route
>> relations.
>>
>> The routing still wouldn't be perfect because change times are
>> also hard to predict when there is no timetable information.
>> Nonetheless, it would be immensily helpful to plan travels that
>> involved multiple modes of transport and multiple operators.
>> If you ever tried to plan a trip from a small town in France
>> to a small town in Spain, you will know what I'm talking about.
>
> I see huge mission creep[1] here. Wikipedia defines Misson Creep like  
> this:

I'm afraid you have read far too much into my mail. Like you I am
absolutely convinced that timetable information should not be included
into the OSM database.

What I was thinking about (and how I understood Frankie) is 
a journey planer that tells you how to get from A to B without 
too much details on timetables. For example:

<snip>
Travel information Rendlesham,GB to Trets,FR:
take the bus to Melton, then the train to London, Eurostar to Paris, 
TGV to Aix-en-Provence TGV, the shuttle bus to Aix-en-Provence and the
regional bus to Trets. 
Approximate travel time: 13 hours (not including waiting for connections), 
All services served regularly except bus to Trets which doesn't run on Sundays.
Websites with time tables: ....
<snap>

Just gathering this kind of information takes hours nowadays. There
may be integrated systems like Travelline for some countries, but as
soon as you try to cross their boundaries you are lost. This is exactly where
OSM could show its strengths.

> A) A point where all the world's public transport access points (bus  
> stops,station etc) are in OSM (but not the routes) and then all the  
> route and other schedule information is available from some other DB
> B) A point where all the world's public transport access points (bus  
> stops,station etc) and all the routes are in OSM and all the other  
> schedule information is available from some other DB
> C) A point where all the world's transport schedules are in OSM and get 
> updated on a regular basis by Bots?

Personally, I would have stopped with A and put everything else in an
external database. But in the last months it
has become quite obvious that there are a lot of people interested
in this topic and are willing to invest quite a bit of time to get
routes added to the database. So, it looks like B is crowdsourcable, too.
I was only proposing to add frequencies and travel times (not timetables!)
to those routes that are being added to OSM anyways. They should be
as stable (or unstable, in the case of GB) as the routes themselves.

> I suggest that we focus for now on getting all the bus stops and public 
> transport routes from all of the 85 Google Transit feeds that are already 
> available for reuse[2]  (but only the ones where the licensing in 
> compatible with OSM)  into OSM and then think about setting up 
> OpenStreetSchedules, OpenTimetableService, OpenStreetTimetables or 
> whatever to contain all the details we need? For the time being people 
> can dip into these GTFS data feeds individually, but in time a single big 
> DB might be useful which can get to chopped up by country or state as is 
> happening with OSM data.

GTFS is a nice start but at the moment very US-centric. I don't see that
we will have timetables available on a global scale within the next years.
Thus, another approach to public transport routing might be interesting.

Sarah





More information about the Talk-transit mailing list