Hello, 1) Again a nice idea to do improvements myself, for which I still don't have time in my current project. Maybe as some holiday hobby project. 3) I think it does not need to be configurable, especially when you use real data to calibrate the cut-off or when you just set a cut-off of 2*deviation. I would probably also keep a hard lower bound of 0.1 to prevent negative speed factors that might occur thanks to a bad distribution choice.
Best regards, Pieter On 23 April 2014 09:27, Jakob Erdmann <[email protected]> wrote: > 1) I agree that 4.5m/s is probably to high for desired deceleration and > more appropriate for a max deceleration. This is simply the historical > baggage of the Krauss car-following model which does not distinguish > between desired and max decel. Feel free to roll your own car following > model (-: > 2) The documetation at > http://sumo-sim.org/wiki/Definition_of_Vehicles,_Vehicle_Types,_and_Routes#Car-Following_Modelshas > been updated > 3) Yes. extreme values are currently possible. It would be best to > calibrate the cut-off using a real-world data set. Do you think it needs be > made configurable? > > regards, > Jakob > > > 2014-04-22 13:55 GMT+02:00 Pieter Loof <[email protected]>: > > Hello, >> >> Thanks for the quick reactions again. Some reactions from my side on this: >> >> 1) Ok, I get it. So it is better to use a realistic average value. >> However, the default values are extremely high then, since realistic >> averages for acceleration and deceleration are below 2 m/s/s. >> 2) The documentation of vehicle definitions says "The drivers reaction >> time in seconds", so that is a bit misleading then. If I understand well, >> the distance to the leading vehicle depends on both the speed and tau, so >> the precise value of tau matters for this distance. However, reactions take >> x time steps, with x = ceil(tau / step-size)? So, 1 second time steps and a >> tau of 1 means a reaction takes 1 time step and for a tau of 1.1 a reaction >> would take 2 time steps? >> 3) Hmm, a speed factor of 0.2 is very low. It would make more sense to >> cap the distribution such that any drawn value cannot differ more from the >> mean than two standard deviations. So it is currently unlikely but possible >> that a vehicle receives a speed factor of 10 such that it always wants to >> drive at it's physical maximum speed? >> >> Best regards, >> Pieter >> >> >> On 22 April 2014 12:51, Jakob Erdmann <[email protected]> wrote: >> >>> Hello, >>> 1) Using the default car-following model, vehicles tend to drive as fast >>> as possible and then brake as hard as possible (using 'decel'). This means >>> they will brake with 'decel' at the latest possible moment when approaching >>> the jam. Thus 'decel' functions more like a desired deceleration. >>> Unfortunately, there is yet not concept for discriminating between desired >>> and maximum deceleration. >>> >>> 2) Tau is the desired time gap to the follower which may be interpreted >>> as a consequence of a drivers reaction time. However, reaction time is >>> fundamentally limited by simulation step size. This why you will get a >>> warning when using tau < step-size (it may lead to collisions). Also, when >>> setting tau > step-size, drivers will still adapt their behavior during >>> every simulation step. There are plans for adding a new car following model >>> which allows reaction times above the simulation step size to be modeled ( >>> http://sumo-sim.org/trac.wsgi/ticket/1151) >>> >>> 3) A normal distribution with expectation X may return any real value. >>> However, we wish to avoid very low speedFactors so we always return values >>> of at least 0.2. (The value in the wiki documentation was wrong and is now >>> fixed). >>> >>> regards, >>> Jakob >>> >>> >>> 2014-04-22 12:03 GMT+02:00 Pieter Loof <[email protected]>: >>> >>>> Hi all, >>>> >>>> Just a few short questions to make sure that I interpret the vehicle >>>> parameters correctly: >>>> >>>> 1. 'decel' is the maximum deceleration of vehicles, so the average value >>>> will be lower. Suppose a vehicle drives towards the end of a traffic >>>> jam, >>>> how large is the deceleration it will apply compared to this maximum >>>> value? >>>> 2. 'tau' is the reaction time. Suppose we have simulation steps of 1 >>>> second >>>> and a tau of 0.8, are vehicles then able to start reacting on a >>>> fraction in >>>> between two simulation steps? >>>> 3. 'speedDev' is the deviation for drawing a desired speed factor from a >>>> normal distribution. The documentation says "The resulting values are >>>> capped at 0.1 to prevent extreme dawdling." I don't understand this >>>> sentence, can someone explain this? >>>> >>>> Best regards, >>>> Pieter >>>> >>>> ------------------------------------------------------------------------------ >>>> Start Your Social Network Today - Download eXo Platform >>>> Build your Enterprise Intranet with eXo Platform Software >>>> Java Based Open Source Intranet - Social, Extensible, Cloud Ready >>>> Get Started Now And Turn Your Intranet Into A Collaboration Platform >>>> http://p.sf.net/sfu/ExoPlatform >>>> _______________________________________________ >>>> sumo-user mailing list >>>> [email protected] >>>> https://lists.sourceforge.net/lists/listinfo/sumo-user >>>> >>> >>> >> > ------------------------------------------------------------------------------ Start Your Social Network Today - Download eXo Platform Build your Enterprise Intranet with eXo Platform Software Java Based Open Source Intranet - Social, Extensible, Cloud Ready Get Started Now And Turn Your Intranet Into A Collaboration Platform http://p.sf.net/sfu/ExoPlatform _______________________________________________ sumo-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/sumo-user
