Issues with calibration. about matsim-code-examples

[kainagel]

Dear Lorena,

I can't see an obvious problem with your setup (maybe someone else can).

What is fairly obvious is that you have indeed by far not enough traffic. On the links where you have counts they are around 3600 veh/hr; at 5% you should have around 180/hr, and not the 0 to 6 you are getting (before multiplying with 20).

I would for the time being forget about counts (and even iterations; set lastIteration=0) and try to find out where that missing traffic got lost. Things to check:

• How many persons do you have in your input plans? Look at your logfile towards the beginning; it will give you the number of persons that were read.
• How many persons have car plans etc.? Assuming that you are programming Java, something like
  

  matsim-code-examples/src/main/java/org/matsim/codeexamples/population/analyzePlans/AnalyzePlans.java

  Lines 29 to 46 in ddfdd9f

  	long nCarLegs = 0 ;
  	long nCarUsingPersons = 0 ;
  	for ( Person person : pop.getPersons().values() ) {
  	boolean carUser = false ;
  	Plan plan = person.getSelectedPlan() ;
  	for ( Leg leg : TripStructureUtils.getLegs( plan ) ) {
  	if ( TransportMode.car.equals( leg.getMode() ) ) {
  	nCarLegs++ ;
  	carUser = true ;
  	}
  	}
  	if ( carUser ) nCarUsingPersons++ ;
  	}
  	System.out.println( "Number of persons =" + pop.getPersons().size() ) ;
  	System.out.println( "Number of car legs = " + nCarLegs ) ;
  	System.out.println( "Number of car legs per person = " + 1.*nCarLegs/pop.getPersons().size() ) ;
  	System.out.println( "Number of car using persons = " + nCarUsingPersons ) ;

  
  You should be able to verify approximate numbers; for example, if you have a city of 1 mio inhabitants and a car mode share of approx. 50%, then at your 5% you should have approx 25'000 car users. Or if you are modelling commuter traffic only, then maybe you have 500k employees and maybe a car mode share of approx. 60% between those, resulting in 500k * 60% * 5% = 15k car users (and 30k car trips).
• Etc.

If you cannot get it plausible at this level, then the simulation will also not be plausible further downstream.

It is also our experience that most of such problems are at this level.

Please keep us posted ...

PS: Note that the flow/storage factors (in your config file) are 0.2 und 0.3, without the "%". Corresponding to 5% they should rather be 0.05, but that should (at this point) not be the problem you are having.

[LorenaBR]

Dear,
Thank you so much for your reply.
For your information, I have around 120.000 agents in my simulation (5%), all the agents are being simulated and I have no inconsistency in my network.
As you recommended, I did only 1 iteration, and with simple plans (origin - destination) only to test, and the result continued the same in the countscompare file.

I visualized in the OTFVIS and VIA and what happens is that, at a given moment (peak hours like 06am), all the agents are stopped and remain this way suntil the end of the day. The network is very loaded and the agents stop to transit, I believe that this is why the count is so low.
I
So I’m saharing with you (and with anyone who feel confortable to help-me), a folder with the following files:
1- Config
2- Network
3-Plans
4- Events (after 20 iterations);
5 .MVI for visualization in the OFTVIS

PS: I did what you’ve recomended about the flow/storage factors, thanks!

Ps2: In the scenário base, i’m only simulating bus and car as main mode... so Walk needs to be teleported?
Do i need to change my network for suporting car and buses? Or transit vehicles use the same network as car?

Thanks in advance

https://drive.google.com/drive/folders/1qUVTypRheWwolwyBwEZeXrkR2LxnkkH6

[jiao1997]

Dear Lorena,
Have you ever solved your problem. I got a similar question with you. Could you tell me are you using the RunCadyts4CarExample.java to run your own scenario?

[whizkid-25]

Dear @LorenaBR ,

I am facing the same problem. Did you solve the problem?

[LorenaBR]

Sorry for the delay, but the main problem was the geolocalization of my network and of my agent's plans, they weren't in the same coordenate. Once I fixed this, the calibration was possible.

[LorenaBR]

solved your problem. I got a similar question with you. Could you tell me are you using the RunCadyts4CarExample.java to run your own scenario

No, i didn't use the RunCadyts4CarExample.jav to run my scenario, because after i fixed my coordinates, it was possible to calibrate "manually" my simulation.