Support routing through OSM areas for pedestrian travel

[barbeau]

Adding this issue to document some conversation related to this improved feature of routing, which would support routing on OSM areas for pedestrian travel. Areas are typically large paved (or possibly unpaved) portions of ground that are commonly used by pedestrians. Current behavior from our instance at USF (circa end of 2011 code) is that OTP routes around the edge of OSM areas, which looks strange on smaller areas, and provides unnecessarily long paths on larger areas.

Our thoughts at USF are to process areas during the GraphBuilder process and transform them from 2D areas to a cross-area network that would interconnect all joining pedestrian ways that enter/exit the area. This would allow routing library to be unmodified and A* should route through nearest entrance/exit as if the area were simply existing ways. This process would generate (n^2-n)/2 new links, where n is the number of pedestrian ways entering the area, and should be feasible for moderately-sized areas, although scalability becomes an issue at some point for huge areas with many entering ways.

Conversation about this is below, with first statement by Ed Hillsman from USF.

Tampa version of OTP seems to route along the edges of pedestrian areas (see below). But for something as large as St. Mark's Plaza, that would be far from ideal. I would still lean toward augmenting the graph by automatically generating a cross-area network for each pedestrian area and plugging it into the rest of the footway graph, but I'm not wed to this (for St. Marks, it looks like this would create about [18^2-18]/2=153 links) and I would be curious to see what others come up with.

http://opentripplanner.usf.edu/index.html?submit&fromPlace=28.059139200169,-82.41646717986&toPlace=28.059309620362,-82.412910570707&mode=WALK&min=QUICK&maxWalkDistance=840&time=8:26%20am&date=04/03/2012&arr=Depart&itinID=1

-----Original Message----- From: opentripplanner-dev@googlegroups.com [mailto:opentripplanner-dev@googlegroups.com] On Behalf Of David Turner Sent: April 02, 2012 3:41 To: gazzu Cc: opentripplanner-dev@googlegroups.com Subject: Re: [OpenTripPlanner] Venice squares missing from otp graph

Piazza San Marco appears to be an area, and we don't presently route on areas. We would like to do this at some point -- patches accepted!

On Mon, 2012-04-02 at 09:07 -0700, gazzu wrote:

Hi, i've builded the otp graph from some downloaded osm data of the Venice area, but when searching walk paths it doesn't pass through the squares, Piazza San Marco included! (often known in English as St Mark's Square) I don't know if i'm missing some data from osm or it doesn't build the graph using the squares data... Can someone give me some hint please?? Thanks in advance ;)

gazzu

[novalis]

I have a theory on this that is too horrible to contemplate, yet too awesome not to try. So I'm working on it.

[mattwigway]

I had written up my thoughts on this a while back. Sounds like they're about what Ed Hillsman recommended. I can't wait to see what @novalis has up his sleeve, though.

[grant-humphries]

Here are a few test cases from the Portland area:

http://www.openstreetmap.org/?lat=45.53259&lon=-122.668244&zoom=18&layers=M http://www.openstreetmap.org/?lat=45.507937&lon=-122.525846&zoom=18&layers=M http://www.openstreetmap.org/?lat=45.509298&lon=-122.540131&zoom=18&layers=M

[grant-humphries]

and here's one more (pioneer square):

http://www.openstreetmap.org/?lat=45.518767&lon=-122.678753&zoom=18&layers=M