[Talk-us] Routing tests
the Old Topo Depot
oldtopos at novacell.com
Fri Jul 20 20:39:15 BST 2012
It may appropriate to consider several metrics, but I also think it
important to get a first version running. Another way of saying this is
iterate and add refinements over time.
Metrics to consider (pre vs post redaction) in roughly order of complexity:
1. Distance of route (using segment lengths, not end point to end point)
2. Number of ways/way segments traversed (rough but perhaps useful)
3. Actual route. For example, in the case of NY to LA, it is reasonable to
expect that once the route enters a motorway at one end, it will stay on
motorways or motorway_links until near the endpoint. It might be
appropriate to run both directions and use the first entry into a motorway
at either end as the "near the endpoint" locations. This might be computed
using pre-redaction data.
4. Over time, as fixes are applied, tighten difference checks to finer
On Fri, Jul 20, 2012 at 12:24 PM, Toby Murray <toby.murray at gmail.com> wrote:
> I think this topic deserves its own thread.
> On Fri, Jul 20, 2012 at 11:20 AM, Kai Krueger <kakrueger at gmail.com> wrote:
> > OK, I have had a first stab at it.
> > http://apmon.dev.openstreetmap.org/us_routing_grid.html
> > It is only 25 cities so far, but if it turns out to be useful it can
> > be extended to more cities.
> > The big problem though is automatically figuring out if routes are
> broken or
> > not and then colour them red or green accordingly. At the moment it
> > do this at all.
> > Anyone have some good ideas for this?
> > The routing information is obtained from "Open Source Routing Machine",
> > which is amazingly fast at calculating all of these long distance routes.
> > OSRM updates its data daily, so I think the routing data is based on OSM
> > data from yesterday. I.e. post redaction. If there is interest I could
> > regenerate this routing grid daily too.
> According to the wiki page, the old system used the as-the-crow-flies
> distance between two cities as a basis. If it was more than 50%
> longer, something is wrong. Not sure if that would still give
> appropriate results or not though... There may be enough connectedness
> now on secondary roads that the distance isn't off by quite that much
> any more even with routing problems. I just did a route on OSRM from
> NY to LA and it came out to just under 3,000 miles which seems about
> right but there are some obvious routing errors. For example it gets
> off of I-70 and goes through town in Saint Clairsville, Ohio but it is
> about the same distance so things like this won't show up in the grid.
> Talk-us mailing list
> Talk-us at openstreetmap.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Talk-us