Re: [sumo-user] Using vTypeDistribution

["Bae, Jong In" <jbae42@xxxxxxxxxx>]

Thank you Jakob for your suggestion!

I have changed my rou.file so that I'm putting all vehicle types into one distribution as shown in the attached screenshot.

However, it seems like the way I have set up my rou.file may be incorrect.

My files are loaded into simulation successfully but the simulation doesn't run (simulation ends at -0.10).

Since it's not returning any error message, I'm unable to figure out what the issue is.

I attached a screenshot of the rou.file as well as the zipped files of my experiment.

I would appreciate any guidance you can provide.

Thank you.

---

From: sumo-user <sumo-user-bounces@xxxxxxxxxxx> on behalf of Jakob Erdmann <namdre.sumo@xxxxxxxxx>
Sent: Friday, May 21, 2021 6:47 AM
To: Sumo project User discussions <sumo-user@xxxxxxxxxxx>
Subject: Re: [sumo-user] Using vTypeDistribution

 

If you keep the number of vTypeDistributions constant, the number of random calls should be constant as well regardless of how many vehicles are sampled from each distribution.

A common way to use them is to put all types into one distribution and them sample each vehicle type from that distribution.

Random depart offset and vTypeDistributions use the same RNG so you won't get any advantage by using the former without fixing the latter.

Am Fr., 21. Mai 2021 um 10:58 Uhr schrieb Bae, Jong In <jbae42@xxxxxxxxxx>:

Thank you for your explanation. 

Does this mean that every time a vehicle is deployed from one of the vTypeDistribution (each distribution only has one vehicle type in my set up), the random number generator's state changes? And if the number of vehicles change in each vTypeDistribution in the next experiment while maintaining the same lane volume (all vehicles have identical properties), the roadway condition still changes due to the randomness from the vTypeDistribution. Is this the correct way of understanding the issue?

If I am trying to create an identical roadway conditions while changing the composition of vehicle types (all of them have identical properties), should I use flow instead of vTypeDistribution?

According to the SUMO documentation, I can use --randomize-flows option to add randomness. Instead of using this option, can I use the seed value and the random-depart-offset value (defined in the .sumocfg file) to have the vehicle depart time the same?

Thank you so much!

---

From: sumo-user <sumo-user-bounces@xxxxxxxxxxx> on behalf of Jakob Erdmann <namdre.sumo@xxxxxxxxx>
Sent: Thursday, May 20, 2021 9:12 AM
To: Sumo project User discussions <sumo-user@xxxxxxxxxxx>
Subject: Re: [sumo-user] Using vTypeDistribution

 

Whenever a type is sampled from a distribution, the state of one of the random number generators changes. This can have far-reaching effects on other parts of the simulation that use the same random number generator (i.e. speedFactor distribution).

Am Do., 20. Mai 2021 um 14:08 Uhr schrieb Bae, Jong In <jbae42@xxxxxxxxxx>:

Hello,

I have some questions about how using vTypeDistribution may introduce randomness in simulation even if the seed value and the random-depart-offset value were kept the same.

My experiment is a 2-lane highway. "HVexit" vehicles merge into lane 0 to exit the highway. 

The attached screenshot shows the vTypeDistributions in my experiment. The two vehicle types (AV and HV) have the identical properties. Thus, they should behave identically.

In my next experiment, my "HVexit" volume was kept the same and all the other volumes changed. However, the volume changes were made in a way that the total volume of each approach (thru and exit) as well as the total lane volume were kept the same.

In other words, the road conditions (lane volumes) were kept the same - it's just the vehicle compositions changed but the two vehicle types behave identically.

I would expect that these two experiments would have the exact same results. The seed value and the random-depart-offset value were kept the same.

However, the number of vehicles processed (completed their routes) and the average travel time of "HVexit" were slightly different.

I was wondering if using multiple vTypeDistirbution introduced randomness in the simulation.

I know my question was lengthy. Please let me know if I can clarify my question in any way.

Thank you so much in advance for your help!

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

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

Attachment: Screen Shot 2021-05-21 at 11.15.18 PM.png
Description: Screen Shot 2021-05-21 at 11.15.18 PM.png

Attachment: test.zip
Description: test.zip