Re: [sumo-devel] Acceleration Model

[Björn H. <hen...@ib...>]

Hi all,

I'm a newbie to SUMO but plan to start using it soon (in fact I already 
analysed the code to find out the v_safe computation) and I have already 
written a traffic simulator applying Krauss-model and IDM which could be 
interchanged globally during simulation run. So I would like to give some 
thoughts on adding IDM as alternate traffic model.

At Saturday 20 June 2009 16:32:10 wrote Tobias Mayer:
> a couple of days ago, Nicholas Loulloudes asked about the possibility
> of adding a new acceleration model to SUMO, namely the IDM.

> 1. At which point will the user decide which model to use?

I also strongly disrecommend the suggested command line option as it would add 
a dependency between command line options and vehicle type attributes which 
could be hard to manage.

Currently there is the suggestion to set set the model per vehicle type. This 
is nice because it's flexible but can introduce very nasty problems. Consider 
the case that a non Krauss vehicle changes the lane to the front of a 
Krauss-vehicle violating its v_safe distance. How can you be sure that the 
whole thing is still collision free? You probably have to proof collision 
freeness for all possible interactions of the models like Krauss has done for 
his model in his dissertation.

Maybe before solving those problems you should make sure that all vehicles use 
the same traffic model. I suggest to do this by a command line option to turn 
a traffic model consistency checking on or off. I would recommend to turn it 
on by default and off by a special option to be on the safe side.

> 3. Concerning IDM: While it is a very accurate model, which offers
> a higher degree of realism than Krauss' model, it does not support
> intermissions or multiple lanes. What can we do about that?

Despite this the IDM doesn't include any randomness. So if you equally 
distribute the cars at initialization you will never see any traffic jams (as 
I'm particularly interested in) because there are no fluctuations.

To overcome this I added the random deceleration of the Krauss-model also to 
the IDM as a last step. So maybe you should add the randomness parameter as a 
general vehicle type parameter.

Greetings
Björn