Thanks for the quick reply. I will disable the meso-overtaking and try again. With respect to the random seed, thanks for the hint. I was initially taking the average of 5 runs, when I noticed the difference. Nevertheless, in the results that I have had so far, I must say random seed had almost no impact. For the original scenario, I always get the orange curve, and for the re-run the blue curve (grey is the empirical Macroscopic Fundamental Diagram). The second diagram shows the impact of random seed in multiple re-run scenarios. I will keep you posted if I find any changes.
Thanks, Sasan [image: image.png] [image: image.png] On Tue, Oct 5, 2021 at 11:29 AM Jakob Erdmann <[email protected]> wrote: > Your simulation is using option --meso-overtaking (which is stochastic) > and seems to use a different seed for both runs. This could be sufficient > cause for the deviation. > Note, that when using option --random and --seed, option --random takes > precedence to generate a random seed (based on system clock) and the given > seed is ignored (making the run not reproducible). > > Am Di., 5. Okt. 2021 um 09:50 Uhr schrieb Sasan Amini <[email protected] > >: > >> Thanks Jakob. I tried to use --default.speeddev 0 to eliminate the effect >> of speed deviation, but it actually worsened the situation. The scenario is >> not that small, but I try to make one that is manageable and send it to you. >> >> Best, >> Sasan >> >> On Mon, Oct 4, 2021 at 5:22 PM Jakob Erdmann <[email protected]> >> wrote: >> >>> If your simulation is using vehicle speed deviations (active by >>> default), I recommend that you use the latest development version which >>> supports option >>> --vehroute-output.speedfactor This reproduces the same vehicle speeds as >>> the original run and should thereby reduce discrepancy. >>> >>> There are other reasons for the discrepancy (i.e. >>> https://github.com/eclipse/sumo/issues/3024). >>> If you can provide a small scenario that shows a strong difference >>> between original run and re-run, please send it my way. >>> >>> regards, >>> Jakob >>> >>> >>> Am Mo., 4. Okt. 2021 um 17:15 Uhr schrieb Sasan Amini < >>> [email protected]>: >>> >>>> Dear all, >>>> >>>> I am trying to re-create a scenario for which I don't have a calibrated >>>> route file, and thus, I use device-rerouting as a Noniterative Stochastic >>>> Route Choice method to avoid using duaIterate and time consuming Dynamic >>>> User Equilibrium. >>>> I export the final routes by using --vehroute-output option and provide >>>> that as an input to my simulation again. Route lengths are identical, which >>>> means the problem is that when I try to re-run the scenario with the newly >>>> generated route file, I get fewer cars in my network. What I observe is >>>> that almost one-third of my 300,000 cars cannot enter the simulation due to >>>> congestion and therefore, I get a totally different traffic situation. I >>>> tried options like --vehroute-output.intended-depart, but I still get >>>> different departure times. I am running meso-simulation version 1.10.0 and >>>> do not change any other parameters of my simulation except for the input >>>> route file. Any tips that help me reproduce exactly the same scenario are >>>> appreciated. >>>> >>>> Thanks, >>>> Sasan >>>> >>>> _______________________________________________ >>>> 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 >>> >> _______________________________________________ >> 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 >
_______________________________________________ sumo-user mailing list [email protected] To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/sumo-user
