I often see inconsistent mapping styles within the same junction or between nearby junctions, and certainly different styles between distant regions of the world. These differences often occur in combinations of road and link geometry, and are particularly noticeable when turn:lanes is mapped. These choices can affect different uses of the data, from map rendering to navigation software behaviors (like GPS alignment and camera movement while driving). So I would like to ask for general opinions on each of the following choices.
Consider junctions where turn lanes continue into link ways, with solid lane markings (not considered carriageway dividers) between some lanes midway through the divergence section.
Where should the road line be drawn?
Along the geometric center of the entire carriageway (all lanes, including turn and merge lanes)
Along the geometric center of the main lanes that continue along the main route (ignoring turn, merge and parking lanes)
0voters
Where should a *_link way start when turn:lanes sections are mapped?
Following the link’s centerline as closely as possible throughout
Smoother intersection angles, allowing small offsets from the centerline
0voters
To illustrate how these practices can interact, I tried to map the same junction following different approaches. I also included the current mapping of the junction for comparison.
More examples from existing mapped junctions that illustrate other approaches would also be very helpful. I do not have a strong preference yet and am interested in understanding existing practices and their reasoning.
Where one existing lane becomes turn lane, rather than temporary extra lane appearing I would take rather centerline of entire road, without excluding turn lanes. Note that in extreme cases you may have only and solely turn lanes.
in general I would take centerline of “regular” road
(I feel that here links should start/end a bit differently but I am not annoyed enough to edit it, but I am happy about centerline)
And if you have 1km of turn lane, or after turn lane you get merge lane and then immediately new turn lane appears then I would likely take centerline of entire road, including this extra lane.
I see this very frequently on the map; in some areas, this is the most common pattern currently. In other areas, mapping turn:lanes is not a common practice, and mappers start the link at the beginning of the conversion lanes; this has been done extensively even by some organized editing teams in my region, sometimes completely ignoring the existing turn:lane mapping.
Lane mapping is not really my thing, but as a general rule, the line mapped in OSM should follow the path typically taken by a vehicle. So that rules out anything that crosses the gore, and anything that involves unnatural angles.
This method is blatantly incorrect. We wouldn’t have tagging schemes like turn:lanes=* and change:lanes=* if we take an absolutist stance on modeling vehicle movements or steering angles. While it would make little difference to a route planner, this method makes lane guidance all but impossible for turn-by-turn navigation, one of the driving factors for OSM’s improving road coverage in the last decade. If a traffic simulator needs to model vehicle movements with realistic lane change timing, it needs a more probablistic model anyways, perhaps based on the road_marking=* tagging scheme. Yes, that would be more difficult to implement than consuming the roadway geometries more literally.
