Re: [sumo-devel] Rerouting

[Fabian L. <fab...@tu...>]

Hello Daniel,

thanks for your response.
The problem does occur with guisim as well as with sumo.
We did a test just a few mintues ago:
- sumo.exe without having a DijkstraRouter does not crash.
- if we have just one Router per MSVehicle - without computing even one 
new route and without setting any new route, sumo crashes, "quitting on 
unknown error".
We are now wondering if "max-depart-delay" may be of interest in this 
case, as it seems to influence the timestep at which sumo quits. 
"Ignore-accidents" is set to true.

Are there any further ideas, what we should test do solve this problem?

Thanks in advance,
Jens and Fabian

Am 24.09.2010 11:18, schrieb Daniel Krajzewicz:
> Hello Jens,
>
> I suppose the problem is due to saving all routes a vehicle had for
> visualisation. Actually, in SUMO, this is only done if an according
> output which writes all routes (vehroutes actually) is used. In
> guisim, this is but enabled by default for visualisation. If you do
> not require to show the routes in guisim, you should remove the
> following line from GUINet for not caching the routes (the vehroutes
> output should still work):
>   MSCORN::setWished(MSCORN::CORN_VEH_SAVEREROUTING);
>
> sincerely,
> Daniel
>
> 2010/9/23 Jens K.<jen...@tu...>:
>>   Hello Daniel,
>>
>> beside Fabian I am the second writer on our diploma thesis.
>> You already helped us out very well, but, if you do not complain,
>> we have another question concerning rerouting.
>>
>> The Situation:
>> We have (we hope so) successfully extended the TraCI-Interface to activate
>> our trafficsigns via a traci script,
>> which actually works like the cross.py example and it works.
>> If we are running the script and load sumo.exe or sumo-gui.exe with the
>> desired config-file, the simulation
>> starts to work.
>> As we have ~8000 vehicles doing sociological calculations over a time of 90
>> minutes, the simulation both in sumo and sumo-gui gets very slow,
>> but on the pc of my colleague it crashes 2000 ticks earlier than on mine
>> (faster than Fabians).
>> So we assume, that this is not a failure of the script but depends on the
>> pc's strength (memory, cpu, e.g.).
>> The only clue we have is, that every time a vehicle wants to change its
>> route, a new routing device is build.
>> Can this lead to such a behaviour?
>> Isn't a vehicle supposed to have only *one* routing device?
>>
>> So, we know by now, everything works in general, but both sumo.exe and
>> sumo-gui.exe crash (slower PC: 2490 steps, faster PC: 4950 steps)
>> after some time and we can't think of a reason why.
>> If we do not pass trafficsign-informations (and so vehicles are not
>> rerouting in that intensity) and let sumo run, the simulation runs to end
>> with exactly the same config
>> without any problems except the low speed-performance.
>>
>> Best regards,
>> Fabian and Jens
>>