Re: [sumo-user] Route file: repeat flow definitions in sub-time interval

[Jakob Erdmann <namdre.sumo@xxxxxxxxx>]

1) Yes. You should be able to confirm this by checking the vTypes that were assigned to the individual vehicles using tripinfo-output

Furthermore, if the time gap between vehicle insertions (implied by vehsPerHour) is not an exact divisor of the sub-interval length, you will get differences in departure time due to the way the times are re-aligned with the new sub-interval.

2) No. At least I cannot think of an easy fix for the type sampling differences. However, if type sampling has a strong influence on simulation results, I suggest you run multiple simulations and average the results to lessen the effect of sampling differences.

3) I think, the first vehicle type from the next interval is sampled on loading so it still influences other randomness even before departure.

4) This might be be due to overhead from checking multiple flow definitions for possible vehicle insertions in case the network is jammed and vehicles must delay their departure. You can check this assumption by running the simulation with option --max-depart-delay 0 and observe if the running time changes significantly. (See https://sumo.dlr.de/wiki/Simulation/VehicleInsertion)

regards,

Jakob

Am Di., 14. Mai 2019 um 16:58 Uhr schrieb Jimmy Forsman <jimmeyforsman@xxxxxxxxx>:

Hi,

I have a question regarding dividing a set of SUMO route file flow definitions into different sub-time intervals. The purpose of dividing the flow definitions is to use the same flow elements and corresponding routing (from-via-to) throughout the simulation, but with the possibility of varying the vehicles per hour inserted for each sub-time interval.

Below, I provide an example of my original route file structure and how I divide the set of flow definitions into sub-intervals.

Originally, I have a route file structure as follows (to illustrate, in the below flow elements I have intentionally left out attributes: vehsPerHour, from, to, via, departLane and departSpeed)

<trips>

<vTypeDistribution id="vehType1" >

... definitions of various vType-elemets...

</vTypeDistribution>

<flow begin="0.001" end="10200" id="5" type="vehType1" />

<flow begin="0.001" end="10200" id="7" type="vehType1" />

<flow begin="0.001" end="10200" id="9" type="vehType1" />

</trips>

Say that I run a simulation using the above structure during a simulation time interval of [0, 2001] s. I would then like to create a route file providing me the same simulation results as for the structure above, but the set of flow definitions are repeated in several sub-time intervals.

I do this as follows:

<trips>

<vTypeDistribution id="vehType1" >

... definitions of various vType-elemets...  

</vTypeDistribution>

<flow begin="0" end="1000" id="1" type="vehType1" />

<flow begin="0" end="1000" id="4" type="vehType1" />

<flow begin="0" end="1000" id="7" type="vehType1" />

<flow begin="1000" end="2000" id="10" type="vehType1" />

<flow begin="1000" end="2000" id="13" type="vehType1" />

<flow begin="1000" end="2000" id="16" type="vehType1" />

<flow begin="2000" end="2001" id="19" type="vehType1" />

<flow begin="2000" end="2001" id="22" type="vehType1" />

<flow begin="2000" end="2001" id="23" type="vehType1" />

</trips>

Note that for each sub-time interval (in my example: [0, 1000), [1000, 2000), [2000, 2001)) the flow definitions in the set (three in my example) are identical with respect to values of attributes vehsPerHour, from, to, via, departLane and departSpeed.

When running an identical simulation for [0, 2001] s using the above structure for my route file I notice the following three main differences in the simulation outcome compared to my original structure:

1. Difference in the simulation results (deduced from detector output) for the complete simulation.

I can image that the vehicle types drawn from my vehicle type distribution differ between the two ways of structuring my route file (e.g. the seed for drawing a vehicle type might be reused when going from sub-interval 1 to sub-interval 2 etc., such that the vehicle types that are inserted in the system differs between using the two route file structures).

Question 1: Do you think that is the reason? 

Question 2: Can I define my set of flow definitions in sub-time intervals such that I obtain the same simulation results as using my original structure? 

2. Difference in the simulation results for the initial interval [0, 1000) s. 

I don't see how results from the first interval would differ for the different structures since the first interval shouldn't be subject to vehicle types begin drawn differently.

Question 3: Do you have any idea of why that is the case?

3. The time running the simulation increases significantly using the second structure dividing the flow definitions into sub-intervals.

Question 4: Is this expected?

(Note that in the actual simulations I run I have set of about 600 flow definitions divided into 23 sub-intervals containing my total simulation time).

Most importantly, it is "Question 2" that I am most eager to find out about!

Best regards,

Jimmy

_______________________________________________
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