Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [sumo-user] Realistic behavior for approaching intersections

Hello,
The behavior of the Krauss model is working as intended (by the original model author). The design goal for that model was simplicism reather than realistic acceleration profiles. I'm afraid there are no model parameters that can remedy this (you can always flatten the profile by lowering accel and decel, though).

The behavior of IDM suffers from https://github.com/eclipse/sumo/issues/6080. This is fixed in the latest development version (download of updated windows-binaries available tomorrow).

regards,
Jakob


Am Do., 12. Sept. 2019 um 21:02 Uhr schrieb Martin Trat <trat@xxxxxx>:
Dear SUMO enthusiasts,

I am currently simulating trajectories approaching an intersection in
SUMO and apply various statistical and machine learning methods aiming
at inferring the traffic regulation of these trajectories (traffic
light, stop, yield, yield to right etc.) afterwards. Small network
cutouts are extracted via osmwebwizard, setting the correct traffic
regulation or edge speed limits if necessary. Using Sumo 1.2.0

Therefore, I would like to use parameters modeling such vehicle behavior
as good as possible.
Using Krauss Car Following, I notice very unrealistic accelerations /
decelerations approaching an intersection node (very high - almost each
approach leads to the default maximum deceleration of the vType, at
least for a simulation time step).
Using IDM Car Following, vehicles approaching intersections have, for my
understanding, a much more realistic character in terms of especially
acceleration / deceleration but still some unrealistic ones (using set
max deceleration of 4.5 m/s² when turning right e.g. from a priority
regulated lane onto a yielding one - driving from yielding ones onto
priority lanes it works fine with this car following model).

Parameters / settings I altered from default:
- simulation step length = 0.1 s
- junction internal links are enabled
- vType attributes passed to randomTrips generator (this is how I
generate trips and convert those to routes via Duarouter):
     - lcKeepRight="0" (to suppress lane changes after insertion with
<departlane value="best" />)


I have tried:
- varying tau as vType attribute
- varying decel and emergencyDecel as vType attributes
- varying actionsteplength to values higher than simulation step length
(I am aware that this triggers the usage of ballistic update method)
- varying the visibility attribute of connections (my thought: a vehicle
approaching notices intersection that contain connections with high
visibility values earlier and decelerates smoother - but this parameter
did not influence the behavior noticeably)
Some of these variations enable me to change some simulation
characteristics but it is hard for me to evaluate whether I alter SUMO
default behavior (which I would like to leave untouched as much as
possible) in a way that makes the whole simulation less realistic by
applying stated changes.

I would be more than thankful if you could give me some opinions, which
parameters can be "safely" varied and which Car Following Model yields
the most realistic vehicle behavior when approaching intersections, in
both cases with present lead vehicles or a "free" intersections.


Thank you very much

Best

Martin Trat

_______________________________________________
sumo-user mailing list
sumo-user@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://www.eclipse.org/mailman/listinfo/sumo-user

Back to the top