Hi @rhhs, and the rest. Thanks everyone for the thoughtful discussion, and apologies for responding late. I originally drafted a very different response, but after following the conversation it would probably be out of context now given where the discussion has evolved.
I’m thinking the main reason the wiki guidance exists is that many routing and navigation systems have historically handled roundabouts more reliably when a route traverses a distinct segment of the roundabout between entry and exit. What’s been most useful in this discussion in my perspective is the distinction between situations where a vehicle genuinely enters the circulating carriageway, even if only briefly, and situations where a driver can complete the movement without really joining it.
Our approach has always been to leave room for mapper judgment. In the more complex and less straightforward layouts, reasonable mappers may reach different conclusions. Our aim is to represent the traffic movement and road layout as accurately as possible, rather than applying a single rule mechanically.
Regarding TomTom’s MapRoulette challenges, they are intended to highlight potential cases for review, not to assume that every shared entry/exit node is incorrect. Contributors are expected to assess the local situation and determine whether separation is warranted or whether the existing mapping should remain unchanged.
Whether in the Wiki or in TomTom’s MapRoulette challenges, if the wording is perceived as too absolute, we would welcome clarifying language that emphasizes the general rule while also documenting the exceptions that this discussion has identified.
Thanks again for raising the topic. I’m happy to contribute further wherever I’d be most helpful.
Do you collect data on how many of the cases you flag for review in a Maproulette challenge result in the suggested changes?
Do you do any verification of the changes done in a Maproulette challenge, perhaps looking at a sample of the tasks and assessing retroactively whether action was warranted and then observing whether action was taken?
I’m too, with @Mateusz_Konieczny here. This model introduces non-existing carriageways & islands and adds complication to a system where a simpler representation is possible.
Yes, on that. Retroactive QA processes on challenge-led OSM edits are in place pretty much as you have described.
There is also a proactive validation step before creating the challenges, where majority of the false-positives are removed before publishing. In Maproulette terms, if the percentage of post-completion Not An Issues are relatively high compared to the Fixes, that would also hint us to look at our validation processes next time.
Somewhat offtopic regarding road geometry in LordGarySugar's specific case in NZ
The first step when remapping this roundabout should be to make it an oval like in real life rather than a perfect circle, and I will declare any mapper who changes how the roads connect without addressing the shape being inaccurate is more interested in finishing a task than in improving the map
The way they have it mapped collides with the kerb in the southeast corner and it’s more than a car width away from from where it should be
Yes we don’t always map a physically accurate depiction of the highway, but nor should we map a perfect circle where none exists simply because it’s a roundabout and editors have a “circle” button
That’s pretty neat, it would be a good idea to document this somewhere. There is currently a long list of the types of MapRoulette challenges TomTom makes on the wiki, but there doesn’t seem to be anything about what you’ve mentioned.
Sort of off topic for this thread, but if a spiky building task is flagged as “Not an issue”, is that recorded so the building will be ignored in future challenges? Is this data made available anywhere?
From the discussion so far, it seems that a single node is preferable if it enables the geometry of the way to better match the shape of the roundabout.
I would add that a single node is also preferable if lanes are marked like a point intersection. For example, in the situation pictured below, if you create two separate nodes with a short way in between, it becomes more or less impossible to determine the correct number of lanes for that short way:
since you’re posting a place next to me, well like any roundabout with multiple lanes.
The OSM model to tag em is completelly insufficient. IMO it’s better to foxus on that and not how a single line can describe it best… coz in essence it can never do good. .
A picture to visualise with highlighted area visible at that wikimedia picture :
I’ve highlighted with yellow “Double white lines where the line nearest to you is broken” that require such a mapping on overtaking rule.
Problem is that one is longer than the other one, so how do you tag the overtaking disability and how do you describe to a driver possible manouvers like the green arrow?
Using turn:lanes, changed:lanes and connectivity relation makes situations like this completly possible to tag and use on navigation.
It’s complicated but still much easier than map all lanes separately, then connect them by links (which are difficult to tag since they don’t exist on ground) and then make complicated multi-ways restriction relations to forbid impossible turns.
Whether signals should affect the choice between junction=roundabout or junction=circular is another controversial topic. Regardless, the wiki page for junction=circular contains a similar recommendation to use two nodes, and there are also unsignalised multi-lane roundabouts with “intersection-like corners”.
Some ‘roundabouts’ are more like two roads split around a common traffic island, where only the left turns actually use the inner circle. The rest of the ‘circle’ looks not round, it has angles between rounded sections. If in addition it has non-circular lineage patterns and traffic lights all around, I think it should not be mapped as a roundabout at all.
yeah, it doesn’t really matter if its actually a roundabout or an intersection that only looks like so. That’s my opinion. Can’t see where it might be faulty.
But to make things clear, the wikimedia one you linked is actually a roundabout. It’s signed properly and should the traffic lights disable, all roundabout rules apply there.
Yes, the task text (just “fix roundabout where multiple highways connect to the roundabout at the same point”) is completely inadequate and misses the nuance of all the replies above.
