Re: [sumo-user] Simulation Random

[Jakob Erdmann <namdre.sumo@xxxxxxxxx>]

Increasing the number of runs is the best way to get meaningful averages out of a stochastic model. Fortunately, distinct runs can easiliy be run in parallel. The tool runSeeds.py makes this easy for you: https://sumo.dlr.de/docs/Tools/Misc.html#runseedspy

Am Sa., 10. Juni 2023 um 12:56 Uhr schrieb dvir n <dvirer@xxxxxxxxx>:

Dear users,

 

I am trying to examine the effect of different VSL parameter values on vehicles' average travel time.

 

For each set of values, I do a check that includes 200 simulation runs while using “--random” command. For each run, I calculate the average travel time for all the vehicles in the simulation and then calculate the total average of the 200 runs.

 

Unfortunately, I recognized that even without changing the set of values, the total average changes till 2.5% between different checks. That fact makes it impossible to compare the effect of different set of values as I don’t know if a decrease in travel time was achieved due to better parameter values choice or just due to the random. What can be the solution for that? Increasing the amount of runs to 1000 for example would require a much longer time for each check (as each run takes around 1 min)

 

Thanks

_______________________________________________
sumo-user mailing list
sumo-user@xxxxxxxxxxx
To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/sumo-user