I am curious whether and how routing systems handle OSM roundabout/ordinary road/street related anomalies. As a rule, these anomalies pop up long before the graph, the network is prepared for routing. No doubt, these anomalies may confuse any routing and line-following system.
I would appreciate help/references to similar subjects. I was through "Routing Archives" for the last three years but the closest I came is the "Tools for getting from OSM to a routable network" related discussions. Yet, the subjects there are targeting something else. Here are some of these anomalies as I see them. Working in another OSM data based application area I need to detect, in any of the road/street classes, the roundabouts (ra) and the ordinary sections (rb) and their relations. The tags are used for extraction. The rest is topology/geometry recognition based. During this process, besides the safe cases, I meet typical cases like: -Unsafe roundabouts. These ways are tagged as ra but some of the usual parameters are damaged. E.g. the ra contains self-crossings, before the end point/node meets the start point the way has made several loops, the consecutive edge vectors go in zigzag shape (with sharp angles), the convex nature on a ra section is considerably destroyed, too large to be a ra and so on. -Incomplete roundabouts. E.g. a ra is missing one or more sections/ways to form a circular way, a ra circular way has one or more sections not tagged as ra, between two points/nodes there are more than one segments/ways, isolated ra (not a junction) and so on. -Ordinary or connecting road segments tagged as ra. Obviously errors in ra class and missing in rb class. -Safe roundabouts not tagged as ra. These are circular ways in the rb class (not tagged as ra) who's parameters perfectly math the usual ra parameters in that road/street class. -Ordinary road/street sections from rb that might be connected/recognized as ra with perfect ra parameters. Just to mention some. Note that the number of these mentioned cases is tens of thousands and spread all over the Planet. Their recognition and eventual correction is not just a simple exercise. Especially control/checking the cases is a tedious and complicated work (I have used a large sample and compared to data/maps based on TA, NT, AND, G-maps, B-maps, satellite/aerial picture maps.). Therefor I would like to see, to learn, how others cope with the similar issues/anomalies. Thanks for your time, Sandor.
_______________________________________________ Routing mailing list Routing@openstreetmap.org https://lists.openstreetmap.org/listinfo/routing
