Re: [sumo-user] Question about multilane roundabout modelling

[Jakob Erdmann <namdre.sumo@xxxxxxxxx>]

Hello,

generally speaking, multi-lane roundabout behavior is a mix of regulations (how are connectivity markings drawn on the road) and local customs and it's hard to give a rule that covers all localities.

Looking at the screenshot above, I'd say that the number of lanes with exit connections should be as large as the number of outgoing lanes.

In your screenshot, I would avoid having lane 0 in the roundabout target lanes 0 and 1 on the exit. Instead I would set connections between lane indices as 0 -> 0, 1 -> 1 and 2 -> 2. Lane 2 should additionally allow continuing in the roundabout. since that leaves you 4 lanes that stay within the roundabout and 6 lanes on the continuation edge I would give double connections to the highest number lanes (4 and 5) which makes them more viable for reaching an exit lane.

Unfortunately, the automatic connection generation is often not up to the task for such roundabouts and the only remedy is manual correction.

However, I strongly advise to look at the simulation in sumo-gui and discover where congestion initially forms. Typical reasons/things to fix are listed at https://sumo.dlr.de/docs/FAQ.html#the_simulation_has_lots_of_jamsdeadlocks_what_can_i_do

regards,

Jakob

Am Do., 26. März 2026 um 18:29 Uhr schrieb Gabriel Garcia Casa via sumo-user <sumo-user@xxxxxxxxxxx>:

Dear SUMO users,

For my first SUMO project, I am building a simulation model of a middle-sized European city and I have a question regarding the correct way to define lane-to-lane connections in multilane roundabouts. As seen in the image, to address this, I have been distributing connections proportionally: dividing the number of exit lanes (black and red arrow) by the number of internal circulating lanes (green arrow), so that each internal lane connects to a proportional subset of exit lanes.

However, I am not sure this is the correct approach for roundabout exit nodes. My question is whether the right definition should instead be fully permissive — i.e. each internal lane connecting to all available exit lanes regardless of proportionality — to prevent vehicles from blocking the internal carriageway when their single assigned exit lane is occupied.

Specifically:

1. At roundabout exit nodes, should connections be 1-to-N (each internal lane to all exit lanes), or is proportional distribution acceptable?

2. Does the answer differ for intermediate nodes within the roundabout (where traffic continues circulating) vs. exit nodes (where traffic leaves the roundabout)?

3. Is there a recommended bulk method to correct these connections (netconvert option or netedit batch operation) rather than editing each one manually?

For context, I am observing ~13,000 teleports per simulation run, a significant portion of which appear to be Wrong Lane type, and I suspect the 1-to-1 exit connections are a major contributing factor.

Simulation setup:
- Route generation: randomTrips + marouter (User Equilibrium) → routeSampler

- fringe-factor: 10 · max-distance: 20,000m · minimize-vehicles: 0.5 · optimize: full · min-count: 2 · threads: 12 · write-flows: number

- Simulation: step-length 0.25s · time-to-teleport 300s · time-to-teleport.highways 0s · rerouting.probability 0.2 · rerouting.adaptation-weight 0.0

- Demand: 1,164 loop detectors, flow calibrators active on edges with GEH > 5 per hourly interval (302–306 edges/interval), vehicles inserted with departPos="last" and a single-edge fallback route

Thank you very much for your time.

Best regards

_______________________________________________
sumo-user mailing list
sumo-user@xxxxxxxxxxx
To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/sumo-user