Thank you for providing example files. Here are my observations:
- your scenario is strongly affected by
https://github.com/eclipse-sumo/sumo/issues/18038 (because you have elevated road speeds of 43.89 m/s in your residential network). Updating to the nightly development build will make your scenario run ~100x faster
- Motorcycles are taking a long time to find a gap in traffic for turning left (at the junction you highlighted). This is mainly a consequence of having a lot of traffic on the main road and made worse by a very substantial fraction of left-turns which take even longer to find suitable gaps. You can increase aggressiveness / throughput by setting a higher impatience value in your vType (impatience=1 for maximum effect) but it is still not enough to prevent build-up of queues. A reduction in traffic or more effort on traffic assignment is needed:
https://sumo.dlr.de/docs/Demand/Dynamic_User_Assignment.html- Your network defines connections with attribute keepClear=False in many parts of the net (also in the highlighted region). This causes a different kind of deadlock. Once junctions are filled by spillback, cross traffic is also blocked and traffic grinds to a halt. I recommend keeping the default of keepClear=True for clusters of closely spaced junctions with heavy traffic to avoid this.
- Your scenario contains walkingareas despite not using pedestrians or crossings. This is wasteful of simulation performance. You can get rid of the walkingareas by running netconvert -s net.net.xml --walkingareas false -o net2.net.xml
regards,
Jakob