| Re: [sumo-user] Validation of a road capacity with SUMO |
|
Hallo Jakob,
Thank you so much for taking the time to explain these issues (all the more so during the weekend). You bring several creative solutions to generate congestion situations and I will take time to study them carefully.
You made me realize that indeed we cannot create jams artificially at the front of a stream if there are no more obstacles beyond. It makes sense and this is an excellent observation! I also like your suggestion to use random speed fluctuations to create unexpected jams because that's what happens sometimes in real life.
Can we create congestion artificially in SUMO by adding an edge at the end of the network with a speed lower than the maximum speed of the vehicles? So all vehicles would slow down entering the last edge and that could increase the density at the back of the stream and create jams?
Once again thank you for your availability and as I mentioned previously during the Berlin conference in 2015, no organization is comparable to the SUMO team in terms of support and help for users.
Vielen Dank François
De : sumo-user <sumo-user-bounces@xxxxxxxxxxx> de la part de Jakob Erdmann <namdre.sumo@xxxxxxxxx>
Envoyé : dimanche 29 août 2021 08:11 À : Sumo project User discussions <sumo-user@xxxxxxxxxxx> Objet : Re: [sumo-user] Validation of a road capacity with SUMO Hello François,
There might still be some misunderstanding here so I'll try to clarify:
Speed is automatically reduced according to the current traffic density (this comes out of the vehicle interactions and is a core function of the simulation).
However, normal vehicle insertion will not reduce the speed of upstream vehicles in order to create gaps for new vehicle insertion.
Thus normal insertion cannot increase the density in a circle beyond the maximum free flow density.
And this might actually be a missing feature:
https://github.com/eclipse/sumo/issues/9031.
If you have a straight section of highway, there are no upstream vehicles at the beginning of your network and you can insert vehicles with low speeds and high densities.
However, since jams dissolve from the front, any jam you create by insertion vehicles at the start of an edge will not grow downstream. Instead you will just see the medium density outflow from a perpetual jam front.
The high densities that you seen in a jam can be achieved if you either:
- insert lots of slow vehicles at the same time all along one or more edges (rather unnatural but easily achievable with departPos="last" and departSpeed="x" for a low value of x).
- create a situation where a stream of fast vehicles approaches a section with slow vehicles and is thereby "compressed" to a high density (i.e. with an on-ramp or a bottleneck).
There is another way in which speed can suddenly drop at high densities and this is through spontaneous jam creation ("jam-out-of-nowhere").
Spontaneous jams happen if the model has random speed fluctuations and these random fluctuations can add up in a dense queue of vehicles.
Whether this can occur depends on the carFollowModel and it's parameters. The default 'Krauss' model only has spontaneous jams when increasing the 'sigma' parameter to 0.8 or higher (default is 0.5).
The IDM model has no random fluctuations and thus cannot model this at all. The new EIDM model added in 1.10 should be suitable for spontaneous jams but I haven't tried out the parameters. Maybe Dominik Salles can suggest some.
Another issue with the straight-highway fundamental diagram scenario, is that SUMO has no single input element that defines a rising level of traffic directly.
In a circle, the inserted traffic can remain and thus the density may rise steadily from a constant input of new vehicles.
In contrast, for a straight highway scenario you have to define a sequence of traffic inputs with rising levels of traffic.
regards,
Jakob
Am Fr., 27. Aug. 2021 um 15:28 Uhr schrieb François Vaudrin <francois.vaudrin.1@xxxxxxxxx>:
|