Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [sumo-user] Question regarding the influence of the ccoolness-Parameter on SUMO (using EIDM)

A review of the code makes it clear that there is no direct impact of coolness on other vehicles in the simulation.
However, the most likely path of interaction is the number of calls to the RNG (random number generator) that is triggered by any one vehicle.
At the moment the number of calls per EIDM vehicle depends on whether it intends to drive with a speed above or below the critical threshold of 3m/s
Since a large number of vehicles share the same RNG, additional calls by one vehicle affect the behavior of all other vehicles on the same RNG.

For general advice on dealing with stochastic effects, 
https://sumo.dlr.de/docs/Simulation/Randomness.html
see https://sumo.dlr.de/docs/FAQ.html#how_to_perform_repeated_simulations_with_different_results


Am Sa., 15. März 2025 um 19:09 Uhr schrieb Hanna Krett via sumo-user <sumo-user@xxxxxxxxxxx>:

Hi everyone,

I'm doing a simulation with SUMO and during a closer examination of certain parameters in the EIDM, I noticed something that I cannot logically explain.

I compared the speed profiles of two vehicles (hereafter referred to as “test vehicles”) from two nearly identical simulations. The only difference between the simulations lies in one parameter of another vehicle (hereafter referred to as the “parameter vehicle”), which is introduced into the simulation both spatially and temporally after the test vehicle. It is impossible for this vehicle, with its modified parameter, to influence other vehicles that in turn could affect the test vehicle (e.g., through the butterfly effect). The results of my comparisons were as follows:

When modifying the sigmaerror parameter of the parameter vehicle (Simulation 1: sigmaerror = 0, Simulation 2: sigmaerror = 0.1), the speed profiles of the test vehicle remain identical.
When modifying the ccoolness parameter of the parameter vehicle (Simulation 1: ccoolness = 0.99, Simulation 2: ccoolness = 0.9), not only do the test vehicles exhibit different speed profiles starting around the time the parameter vehicle is introduced, but other vehicles—independent of the test vehicle—are also affected. The simulations diverge significantly, for example, in terms of vehicle positions at the same time step. As mentioned before, it is ruled out that these changes could have resulted from the butterfly effect or direct influence.

I have now observed this phenomenon for the second time. The first time, I was unable to reproduce it because I did not immediately associate it with the ccoolness parameter. However, since I have now successfully linked it, I am fairly certain that this observation does not stem from an error on my part.
Due to the identical time series when modifying sigmaerror, I rule out differences caused by random behavior.

My research has not provided any further insights on this matter, which is why I am reaching out with the following question:
Is it possible that the ccoolness parameter generally influences the numerical development of the entire simulation, rather than just affecting changes through the vehicle’s behavior?

I would greatly appreciate any insights you may have on this!

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

Back to the top