Mapping turbo roundabouts

Roundabout naming… Somewhere 2021 or so the OSM mantra here became not to name the roundabout unless it has a name of it’s own. What name to give it when e.g. 4 streets each their own name used to connect of old at a crossing and modern times frenzy decided to slap a roundabout on them doing away with traffic lights on the go. At any rate my car navigator on this 4 way RB would tell me an RB is up and to take the Nth exit,

In the U.S., where relatively few roundabouts have their own names, mappers used to give the roundabout ways name tags based on the names of connecting streets. Often this resulted in a semicolon-delimited list. But Telenav and others removed most of these name tags in 2018, possibly influenced by how routers interpret these tags.

I assume @Peter_Elderson was instead referring to the name of the roundabout itself. Sometimes a street follows a boundary and therefore has a different name on either side, so I was hypothesizing that maybe the same thing can happen to a named roundabout that straddles a boundary.

I was, and it could, but AFAIK it never happened so far.

Streets/roads different names landing on the roundabout is very common, even when the through road has the same road ref and normal people would say it’s one road. To my osm-predecessors it seemed best to just leave out the names on the junction area, to simply avoid the naming issue and do away with label placement issues on this very confined map space, and I totally agree.

Most roundabouts have no official name, and it’s not customary to name them after the roads they serve. Some regions have taken to naming their precious new roundabouts, I personally think the custom wil grow because mayors and city councils like it, and I think we should have mapping for it and not wait until it’s an omission we could have prevented.
They also like to put monstruosities in the middle of roundabouts. I think they call it artwork. For the central islands, shrubbery is favourite, which means it won’t be rendered on OSM Carto, which in turn means that all kinds of substitus are being used to get some colour on the map, despite it being blatant tagging for the renderer.
You would be surprised how many village greens we have, surrounded by truck aprons, which also are not rendered. (BTW we do have true (but former) village greens, i.e. areas formerly serving as village greens and now often still with a public function, appropriate layout and named after the former village green. )

Please hold me back or I’ll be ranting all night!

I have no experience mapping junctions so apologies if this is not relevant… I remember reading somewhere about creating areas with junction=yes that share one node with each way that enters or exits the junction. The idea being that you can put a name on them, but also that routers can give simplified turn instructions such as “turn left at the junction” instead of “turn slight left then continue straight for 5m then turn slight left again”. Maybe something like that would be useful here? By looking at the area, a router would be able to identify the exits, to count and announce them correctly, and also to calculate and announce the overall direction change. For example, “take the third exit to turn left at the turbo roundabout”, instead of the more complicated directions they currently produce.

I would also suggest, if you’re going to create issues in the issue trackers of routers, that you add an example of what the routing instructions currently look like and what you would like them to look like.

Ah, right, the other other meaning of junction=*. :sweat_smile: Junction areas are a cool solution to a number of guidance-related problems. There’s probably no harm in also surrounding the roundabout with one, but no router seems to support it yet, so it would be premature to remove junction=roundabout from the ways.

Valhalla already seems to understand that this Westland roundabout is a single intersection, even though it lacks support for junction=yes areas and there isn’t one here. It just chooses to present more detail about destinations and such.

OSRM doesn’t even call it a roundabout. The ways are tagged junction=circular, which is for circular intersections in general, probably because of the traffic lights. There are many kinds of circular intersections, such as the classic American traffic circles, which can also be signalized, or “neighborhood roundabouts” that are essentially just a traffic calming island in the middle of a standard two-way stop intersection. (Exits?)

I’m not sure it would be accurate to change the Westland example to junction=roundabout roundabout=turbo just because of its shape. When a driver hears “at the roundabout” from the application, they probably expect a continuously flowing intersection.

No, it’s absolutely a roundabout, because it has the official roundabout signs and give_way signs at every entrance. It also has physical carriageway separation, drivers can not change carriageways on the roundabout, and you always arrive at an exit when following the carriageway. It’s just a biggy with three rings, where even the rings can have multiple lanes; that’s why I say separate carriageways, where usually I would say lanes.
The traffic authorities require traffic lights, because drivers at an entry point are supposed to be able to handle max 2 lanes to give way to (or pick a gap in the traffic). Most drivers actually can. This roundabout has entry points where drivers see 4 lanes to yield to. Then traffic lights are mandatory to regulate safe traffic flow.

There are many multi-lane (3 or more) non-turbo roundabouts with traffic lights for safe flow regulation. I think the sentence on the roundabout page about no traffic lights needs to be revised. The crucial requirements are: a. the roundabout sign visible form for all entries, and b. legal or explicit priority for traffic on the roundabout. You could say something like “one lane or two-lane roundabouts typically have no traffic lights. On larger or complicated roundabouts, traffic lights may be present to regulate traffic flow”.

The Westland junction as a whole has a name, Vlietpolderplein, currently tagged on all the sections of the rings. That is probably why it is mentioned in the navigation instructions. Awkward, because the driver is already on the Vlietpolderplein roundabout), then gets the instruction to enter the Vlietpolderplein (road). At that point, the driver can not make a decision, so it’s only confusing.

Then the instruction is: take the first exit to Piet Struykweg. Again, the driver cannot make a decision here; (s)he has to follow the lane onto the exit Piet Struykweg. This does not require what people what call “neem de afslag” (take the exit), which would be a manoevre. So it’s just confusing. Normally people would expect “U verlaat the rotonde” (You are leaving the roundabout), which says what happens, instead of giving an instruction to do something. The instruction even says: take the 1st exit. Usually people count the exits with regard to the whole roundabout.

This particular turbo roundabout has full guidance for left, through and right; once you have chosen the entry, you will always end up on the appropriate exit. On most turbo roundabouts, the driver will have one choice point where (s)he can take the exit (a manoevre) or continu (keep in lane) to the next exit. No choice usually means the vehicle will be forced off the roundabout at the next exit.
“Take the exit to <road name and/or destination>” or “keep in lane / continue on the roundabout” or in Dutch “Blijf op de rotonde” would be the right instruction here. At this point, no exit count should be given.

Yes - I remember mentioning this back in "Pencil tip" intersections - #9 by Richard. My use-case is junctions like this… thing where a mass of tarmac has been mapped as a mess of intersecting ways.

Fair enough. In that case, this roundabout is mistagged as junction=circular. To a router, that tag and the traffic signals must make it look a lot like a classic American traffic circle, the kind that conventional roundabouts were supposed to replace.

By design, routers sometimes issue unavoidable instructions. For example, many routers say, “Merge onto [motorway],” at the end of an on-ramp, even though there’s nowhere to go but forward. Some routers say, “At the end of the street, turn left onto [street],” as you approach an L-intersection where you can only turn left. During turn-by-turn navigation, these instructions can reassure the driver that the application is still following along, although some users prefer less chatty instructions.

There are also situations where routers should be issuing unavoidable instructions but aren’t. For example, if the upcoming intersection has an only_straight_on restriction but the driver needs to turn left at the intersection immediately past it, the router would ideally say something like, “Go past this intersection, then turn left.” Another example:

I agree, exit counts are not always very helpful in roundabouts. As I alluded to earlier, I suspect they’re often counterproductive in American roundabouts, where exits are usually assigned a direction, even when there’s only one lane. It sounds like you have enough detail to file bug reports with the major routing engines.

Hm… I’m hesitant to call it bug reports - I ask for support/adaptations for an existing but still not widespread situation. Maybe a change at the tagging end is needed to convey the necessary information to the router.

In the meantime, I’m still thinking about the give_way situation. Suppose a router/navigator does something with that information, way by way. Penalise or display or whatever.

And suppose I have a standard turbo roundabout. Then there are entry lanes crossing the outer ring to a short connection between the outer lane and the inner lane. I think if only the ways are taken into account, the short connections woud have priority over the inner ring! Because both are on the roundabout, both have equal priority because of the junction=roundabout tag, so the default priority rule would apply (in NL: traffic coming from the right has priority).

In reality of course, everybody regards this segment as an incoming way that has to yield to all traffic already on the roundabout. But suppose that we feed this osm-roundabout to a self-navigating car… am I missing something?

PS I’ll answer this myself… priority is not determined and applied at each crossing separately , but at the point where the sign is, and is then applied to all the crossing ways at once. If software had to handle the situation that two cars approach the same intersection node, and had to determine who is to continue and who is to stop, then it would matter. But that has nothing to do with routing.
So just forget it, please, my mistake!

Thanks, I will do that.

I have opened an issue for OSRM support.

I didnt think it would be manageable to open development-issues for the three routers at Arbitrarily, I chose the one with letters osm in the name.
The other two will of course follow, once we have more answers about what could be done.

I imagine most routers encounter the same issues if they would build support for turbo roundabouts.

I think a nice chain of instructions would be:

“Use the left or right lane to enter the roundabout” [depending on if both lanes lead to the desired exit]
“Take the second exit on the roundabout” [second overall exit. if the inner lane is taken, this could be the inner lane’s first exit]