| [sumo-user] Unexpected traffic behavior after lane removal in mesoscopic simulation |
Dear SUMO users community,
I'm encountering an unexpected traffic phenomenon in a mesoscopic simulation and would greatly appreciate any insights or suggestions.
I'm observing a significant increase in segment speed after a lane is removed from a segment by modifying the network XML input via Netconvert. My expectation was a reduction in speed due to the diminished segment capacity. This counterintuitive drastic speed change is primarily localized to the segment where the lane was removed, with its upstream and downstream segments showing only minor or no speed alterations. This behavior is particularly noticeable in congested segments (i.e., simulated speed way below the speed limit), but doesn't appear under free-flow conditions.
I currently have lane-queue disabled in my simulation setup. I also noticed that sometimes in the network XML file keepClear is set to false or '0' for certain new connections, and some of their states are set to 'm' rather than 'M' by Netconvert. However, my experiments haven't shown any difference in the results when I change these settings. This suggests these particular configurations might not be the root cause of the unexpected speed increase.
Since mesoscopic models are queueing-based (with a default queue length of 100m), I wonder if segment length could be a contributing factor. However, I've observed the aforementioned speed increase after lane removal across segments varying significantly in length, from just a few meters to several hundred meters.
Has anyone experienced similar anomalies? I suspect I might be overlooking some critical configurations within the mesoscopic simulation.
Any pointers on what parameters or settings could be influencing this behavior would be highly valued.
Thank you in advance for your time and expertise.
Best,
Chao