Hello,
it should be quite easy to modify the car-following model so as to force
(rear collision) accidents.
The current (default) model computes the maximum speed which still allows
braking if the leaders were to brake. This velocity is then used by the
vehicle (if the vehicle parameters and speed limit allow). By modifying the
model so as to return a larger value, the vehicle will drive in an "unsafe"
manner and will crash if the leader vehicle brakes hard.

Note, that sumo will flag the situation as a collision as soon as the
leader comes within the "minGap" of the follewer vehicle and will
immediately "teleport" to follower vehicle to a subsequent lane (if you
want to make sure that vehicles intersect closely you should set the minGap
attribute to 0).

Flanking collisions are more difficult to create. Head-on collisions are
currently not possible.
Also, sumo hardly models any physics, besides limiting acceleration and
deceleration. You need to modify these parametesr ("accel" and "decel") to
simulation different road conditions.

regards,
Jakob.


2014-08-29 11:27 GMT+02:00 Marcus Müller <[email protected]>:

> Dear SUMO Devs and MailingList Subscribers,
>
>
>
> First: A short introduction of myself. My name is Marcus Müller, I am
> currently studying
>
> Electrical Engineering in Munich, Germany. My Bachelorthesis’ Topic was the
> making of
>
> a Support Vector Machine (SVM) powered gesture Control for robots. In the
> upcoming
>
> Masterthesis I plan to stick with machine learning (ensemble methods) as
> well to solve a
>
> Car-accident related problem.
>
>
>
> For that I thought about generating some ‘artificial’ Training Data with
> SUMO first. I checked out
>
> The FAQs and Tutorials and found that accidents are not a real content of
> SUMO yet.
>
> I would like to force an accident by making one car ignoring traffic
> lights,
> speed limits or taking
>
> a forbidden u-turn, etc. I don’t need to simulate the accident itself
> (crash
> physics). It would be enough to have two cars
>
> touching each other. I then would like to go a few milliseconds in the
> past,
> starting from the moment of the crash, and vary
>
> a bit the figures which lead to the accident (e.g. relative speed, car
> angles,…). During my research I found SUMO’s Car-Following-Models
>
> and wonder now, whether it is possible to achieve such behavior (like
> ignoring traffic lights) with the help of my own cfm. Or would you suggest
>
> another approach to get this done?
>
> Furthermore, I would be interested in varying physical parameters, like
> street conditions (wet, icy). Is there already something for that or
>
> would I need to implement it. (Maybe changing deceleration could solve this
> particular problem, but I would like to know how SUMO implements
>
> Physics in general, or whether it doesn’t – is there smthg like a physical
> car model?).
>
>
>
> Many, many thanks in advance for your help and best regards,
>
>
>
> Marcus
>
>
>
>
> ------------------------------------------------------------------------------
> Slashdot TV.
> Video for Nerds.  Stuff that matters.
> http://tv.slashdot.org/
> _______________________________________________
> sumo-devel mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/sumo-devel
>
------------------------------------------------------------------------------
Slashdot TV.  
Video for Nerds.  Stuff that matters.
http://tv.slashdot.org/
_______________________________________________
sumo-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/sumo-devel

Reply via email to