MoST with SUMO 1.9.2

[awillecke]

Hi,

we would like to run the MoST scenario with a more recent version of SUMO (1.9.2).

We noticed that at certain locations, traffic is piling up at intersections using 1.3.1 which does not happen in 1.9.2.

First example:
With SUMO 1.3.1:

and with SUMO 1.9.2:

Second example:
With SUMO 1.3.1:

and with SUMO 1.9.2:

The output of SUMO 1.9.2 says:

Eclipse SUMO sumo Version 1.9.2
 Build features: Linux-5.12.9-arch1-1 x86_64 GNU 11.1.0 Release Proj GUI
SWIG GDAL FFmpeg OSG
 Copyright (C) 2001-2021 German Aerospace Center (DLR) and others;
https://sumo.dlr.de
....
Simulation ended at time: 50400.00
Reason: The final simulation step has been reached.
Performance:
 Duration: 3203.15s
 Real time factor: 11.2389
 UPS: 374543.147099
 UPS-Persons: 51272.532918
Vehicles:
 Inserted: 46842
 Running: 32
 Waiting: 0
Teleports: 52 (Jam: 15, Yield: 32, Wrong Lane: 5)
Emergency Stops: 2
Persons:
 Inserted: 45000
 Running: 26
 Jammed: 2560
Statistics (avg):
 RouteLength: 6839.22
 Speed: 6.92
 Duration: 6406.78
 WaitingTime: 35.84
 TimeLoss: 121.52
 DepartDelay: 0.45
Pedestrian Statistics (avg of 31421 walks):
 RouteLength: 450.72
 Duration: 385.63
 TimeLoss: 57.15
Ride Statistics (avg of 45216 rides):
 WaitingTime: 38.11
 RouteLength: 6073.44
 Duration: 589.52
 Bus: 5967
 Bike: 3455

Currently, we are unsure why this might happen and which of the behavior shown above is actually the desired one.
Do you have an idea what might be the cause of this behavior?
Also, is 1.3.1 the most recent version of SUMO capable to run this scenario?

Thank you and kind regards
Alexander

Edit: This was tested with the current master (7cee150)

[lcodeca]

Usually, when you have less traffic congestion when upgrading SUMO, it means that the previous one was more a quirk (for example 2 vehicles conflicting on the same intersection) than a feature. With all the fixes and improvements you have in 1.9.2, that's the easiest explanation.

I just want to highlight that the mobility in the scenario is not calibrated with real data, and the scenario is used mostly to play with different traffic demands and mobility compositions, starting from a solid infrastructure.

[awillecke]

Thank you for your quick response and explanation.
I am not that knowledgeable of the inner workings of SUMO but that makes sense.

So basically, using a more recent version of SUMO with changes to the driver models for MoST fine (as long as it runs the scenario without any crashes) since the traffic is synthetic anyways?

[lcodeca]

Exactly. I use the MoST infrastructure with the activity generation provided by SAGA to play with it and get the mobility I need to test my things. I always stick with the development version of SUMO, so the latest version should work fine.

All the static part was hand-fixed from a lot of open-source datasets, but I had nothing good concerning the demand calibration.

You can find an example here https://github.com/lcodeca/SUMOActivityGen/tree/master/example