[OSRM-talk] 'penalty' for snapping

Steven M. Ottens steven at minst.net
Tue May 10 22:04:04 UTC 2016

HI all,

I've ported my application to osrm 5.1 and added the travel distance 
from village to road to my results using osrm.nearest and assuming that 
the resulting waypoint is the same as the source point osrm.table is 
using for its calculations.The percentage of the population that could 
reach a hospital in 60 minutes dropped from 95.8% to 92.3% when assuming 
a 5km/h speed and further down to 68.8% when assuming a 1km/h speed for 
the source to road distance. This is definitely an improvement in 
accuracy over the previous result, though if this could somehow be 
incorporated with the osrm.tables function that would be even better, 
since the source -to-road traveltime is a big part of the total 
traveltime here. Obviously the same situation might apply to the 
destination points (in my case they are on or close by the road so it 
doesn't matter that much, but that doesn't need to be the case in other 



BTW the osrm.nearest syntax is a bit weird; the documentation states 
"Where|coordinates|only supports a single|{longitude},{latitude}|entry." 
which apparently meant { coordinates: [ [ lon,lat ] ] }  in stead of { 
coordinates: [ lon, lat] } but that might be me reading it wrong.

On 4/19/2016 6:00 PM, Steven M. Ottens wrote:
> Hi Daniel,
> Your idea is pretty much what I used for my travel time map: 
> http://stvno.github.io/page/bartholomap/ though N was easier to 
> determine: nearest non-stop airport, nearest one stop airport etc.
> For overland speed: I like the wiggle factor Bjorn was suggesting, 
> however it supposes we know the landuse, which as far as I know OSRM 
> doesn't take into account
> For N if we can have a default of 4 on separate roads that would 
> probably help a lot already without too high a cost. (separate roads 
> being at least one fork in the road away) Does that sound reasonable?
> I guess the strategy to determine the N points could be made 
> configurable, different use cases require different strategies I'd say.
> Steven
> On 4/19/2016 4:44 PM, Daniel Patterson wrote:
>> Hi Steve,
>>   I was thinking about exactly this just the other day when 
>> considering https://github.com/Project-OSRM/osrm-backend/issues/2277
>>   What I came up with was roughly:
>>     Find N nearest neighbours, and calculate the walking duration to each
>>     Insert all neighbours into the search heap, and set the initial 
>> cost to the walk duration
>>     Route as normal, but include the walk duration in the total time 
>> at the end
>>   The question is, how big do we make N, and how do we determine the 
>> overland speed?  Overland speed I suppose could be a profile 
>> parameter, and it feels like walking speed would usually be the most 
>> appropriate thing.  But I'm not sure how many possible start points 
>> we should reasonably consider.  Everything within a radius?  Closest 
>> N points?
>> daniel
>> On Tue, Apr 19, 2016 at 12:39 PM, Steven M. Ottens <steven at minst.net> 
>> wrote:
>>     Hi all,
>>     I am working on a tool that will calculate the travel time from
>>     villages to for instance the nearest hospital for rural
>>     accessibility analysis. So far I have build a tool which will
>>     calculate this for 215.000 villages in a few minutes, thanks to
>>     the help of Daniel P. However I realized that not all villages
>>     are on a road, actually quite a few are at least an hour walk
>>     from the nearest drivable road see:
>>     http://stvno.github.io/img/OSRM-snapping.png
>>     The villages are typically connected to the roads by informal
>>     trails, which I do not have as a road network. So I was wondering
>>     if it would be possible to add a basic penalty to the distance of
>>     the point to the nearest road.
>>     I am aware that the nearest service will provide me with the
>>     distance:
>>     https://github.com/Project-OSRM/osrm-backend/wiki/New-Server-api#service-nearest
>>     but before I am going to calculate the extra traveltime for each
>>     village, I was wondering if it is possible to integrate this with
>>     an osrm.table request. I am aware this opens quite a can of
>>     worms: if there is a penalty to snapping, it might be worth it to
>>     snap to a road which is further away, but in the end quicker to
>>     get to the destination. But for the moment I'm happy with just
>>     the added traveltime for nearest road snapping.
>>     Any suggestions on how to handle this?
>>     Steven
>>     _______________________________________________
>>     OSRM-talk mailing list
>>     OSRM-talk at openstreetmap.org <mailto:OSRM-talk at openstreetmap.org>
>>     https://lists.openstreetmap.org/listinfo/osrm-talk
>> _______________________________________________
>> OSRM-talk mailing list
>> OSRM-talk at openstreetmap.org
>> https://lists.openstreetmap.org/listinfo/osrm-talk
> _______________________________________________
> OSRM-talk mailing list
> OSRM-talk at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/osrm-talk

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/osrm-talk/attachments/20160510/20951a08/attachment.html>

More information about the OSRM-talk mailing list