Thank you for bringing this up! The latest development version implements solution 4 ( https://github.com/eclipse/sumo/issues/12615) and you should be able to download the new binaries starting tomorrow at https://sumo.dlr.de/docs/Downloads.php#nightly_snapshots
Solution 3 is not safe from invalid phase prolongation: if there is a long jam by cars targeting the turn lane and that jam that stops on the longer thru-detetor, the thru phase will be prolonged despite nobody being able to drive. However, you can still achieve this configuration where needed by using custom detectors: https://sumo.dlr.de/docs/Simulation/Traffic_Lights.html#custom_detectors_1 The reason why actuated tls are less unaffected is, because their detectors are much closer to the stop line (i.e. 25m at 50km/h versus 100m default range of delay_based). In the real world, the delay based concept was sometimes implemented somewhat similar to solution 3. However, the design actually targets V2X communication where vehicles could unambiguously announce their desired turning movement and thus turn lane length doesn't matter. best regards, Jakob Am Di., 14. Feb. 2023 um 20:32 Uhr schrieb lcc <[email protected]>: > Hi Sumo developers, happy Valentine's day! > > > I have a question about delay-based signal’s detector placement > > Here I’m showing an intersection. The northbound has 1 lane fork into 2 > lanes, one is thru-only and one is right-turn only. (red arrow pointed) > > > The signalized lanes (thru and right-only lanes) are shorter than > delay-based detector’s default length, so the detector is extended to the > preceding segment (the 1 lane segment), and overlap each other. > > This causes a problem that, when there is a long line waiting for thru > signal phase, the right-turn detector also detects cars waiting (in the > overlapping area) and keeps the right-turn phase green for a long time, > even when there is no westbound traffic. The screenshot shows this case. > > I can think of a few solutions: > > 1. > > Manually config the detector length to be shorter. OK for one case but > not scalable. > 2. > > Change the auto-generated signal schedule so that the northbound right > turn will not be shared with westbound left turns. For example: change the > tlLayout to incoming > 3. > > Change the detector placement code, so that if two detectors are > overlapping on the same lane, only place one detector. Say, thru-detector > have higher priority. This requires a code change and definitely will have > corner cases. > 4. > > Stop the detector if there is a lane merge. I saw that there is > similar logic in actuated lights’ detector placement as well. This may not > work well if the lane fork is very close to the stop line. > 5. > > Change signal type to actuated. The actuated sensor seems to have no > such problem (AFACT) > > > Do you have a suggestion? In practice and the real world, how will sensors > be placed in this case? > > > Thanks very much and hope you have a great valentine's day! > > _______________________________________________ > sumo-user mailing list > [email protected] > To unsubscribe from this list, visit > https://www.eclipse.org/mailman/listinfo/sumo-user >
_______________________________________________ sumo-user mailing list [email protected] To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/sumo-user
