Re: [sumo-user] SUMO simulation speed

[Jakob E. <nam...@go...>]

Hello,
The real-time speed factors can only be compared meaningfully in relation
to the simulation size (network size and number of vehicles). For the small
bologna scenario (one of the public scenarios available at
http://sumo.dlr.de/wiki/Data/Scenarios#Bologna) I'm getting a speed up of
580x real time (9k vehicles within a 3km^2 network).

For the real time factor with external control the comparison is even more
difficult since it depends very much on the usage pattern
(rerouting every vehicle during every simulation step is much slower than
just touching a few vehicles every once in a while).

Depending on your usage scenario you could see much higher real-time
factors (100x real time in some of my TraCI scenarios).
If you explain what kind of external control you want to achieve I might
hazard a guess as to the speeds you could expect.

regards,
Jakob


2016-11-27 5:03 GMT+01:00 Cathy Wu <ca...@ee...>:

> Hi everyone,
>
> I am part of a research project at UC Berkeley which requires a traffic
> microsimulator that is as fast as possible and implements significant
> external control (via TraCI).
>
> I was wondering how I can learn more about the simulation speed of SUMO as
> compared with other key simulators (Vissim, Paramics, Aimsun, etc.)? The
> following research study states that SUMO may vary from 1-2x real-time
> speed (with TraCI calls) to 50x real-time speed (without TraCI calls) for
> their small test network.
>
> Blokpoel, R. and Vreeswijk, J. (2015). *Online Micro Modelling Using
> > Proprietary Controllers and SUMO*
> > <https://link.springer.com/chapter/10.1007/978-3-319-15024-6_3>, pages
> > 35–46. Springer International Publishing, Cham.
> >
>
> In comparison, Vissim's website claims 80-160x real-time performance
> (without external control), and I believe Aimsun claims 2-3x real-time
> performance.
>
> Do you know if other simulators incur a similar performance hit when
> implementing external control? I would also like to better understand
> why/how TraCI calls cause such a significant performance hit?
>
> We are very excited to be working with SUMO, and look forward to better
> understanding its benefits from other simulators!
>
> Best wishes,
> Cathy
> ------------------------------------------------------------
> ------------------
> _______________________________________________
> sumo-user mailing list
> sum...@li...
> https://lists.sourceforge.net/lists/listinfo/sumo-user
>