Right, the user brain should be engaged. I think that one should sometimes fine-adjust the location that you get from address search. Do you want the goods delivery entrance, the mailbox, the main entrance, some minor entrance, some specific corner of a parking lot, a back yard, or what? These could make a difference also for car routing.
I would argue that it is good mapping practice to attach the address to the entrance=main node of the building, and to map access paths all the way to that entrance=main. Then, by default, the address search should perhaps return the entrance=main node. If there are multiple nodes associated with an address, Osmand could ask the user to choose the exact node. Perhaps even the map data could be improved further. Here is a related example, a shop entrance in a fenced area whose only entrance is from the west: http://www.openstreetmap.org/node/458418466 Some years ago when the parking aisles were not yet mapped, Garmin would tell me to enter the area from the east and then climb the fence. I don’t think it is reasonable to except any routing engine to understand gates or fences. Only if the ‘last mile’ is properly mapped, you can expect the routing engine to do the right thing. Marko torstai 6. lokakuuta 2016 8.01.39 UTC+3 Poutnik kirjoitti: > > Dne 06/10/2016 v 06:52 Marko Mäkelä napsal(a): > > On Wednesday, October 5, 2016 at 18:28:20 UTC+3 Poutnik wrote: >> >> I see it as logical that Avoid_unpaved avoids surface=gravel. >> > For what it is worth, my entry to OSM was via mkgmap > <http://www.mkgmap.org.uk/>, which translates maps in the proprietary > Garmin format. The Garmin routing engine seems to implement special > treating of the ‘last mile’. > The special treatment: If the closest applicable way is the only way to > reach the destination (or to leave the current location), then any access > restrictions or ‘avoid unpaved’ or similar routing preferences would be > ignored. (It might even be that oneway directions would be ignored. And > mkgmap would map access=destination, access=private, access=no in the same > way.) > So, in this case, the surface=gravel road would be chosen despite the > ‘avoid unpaved’ preference, because there is no alternative. > > > Related note: My observation from the Garmin Edge 705 was that it would > generally always select the roads that are closest to the selected starting > and destination points. I am not sure if it would avoid choosing a point on > a motorway or motorway_link, but let us give it the benefit of doubt and > assume that it would choose the parking aisle (or footpath if they are > mapped) in this case. > > > Similar approach can be obtained with Brouter profiles. > The way preference is expressed as costfactor <1.0 , 10000> > CF 10000 means the engine pretends such a way does not exist. > CF 9999 means the engine pretends such a way does not exist for routing, > but considers it geometrically for navigation hints. > CF >> 1.0, <9999 means the engine considers the ways for routing, but with > such a high penalty, that the was is used only if is the only option. > > What seems important is the internal logic of routing engine about > reaching destination. > I suppose the engines suppose the user brain is switched ON for the last > mile. > > -- > Poutnik ( The Wanderer ) > > My Brouter profiles https://github.com/poutnikl/Brouter-profiles/wiki > > -- You received this message because you are subscribed to the Google Groups "Osmand" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
