2

In my project, I am trying to make a base case, in which the market penetration is 0.0. In another word, no vehicle is using the application. I tried two ways to do this. First, I use SUMO only to run the simulation, and secondly I use SUMO, Veins and OMNeT++ to run the simulation, but in omnetpp.ini set the penetration rate to 0 by *manager.penetrationRate = 0. I assume doing so will not create any mobility module for the vehicle running in SUMO.

But what confused me is the results of the above two simulations are not the same. The congestion in the first way (SUMO only) is much heavier than it in the second way (SUMO, Veins, and OMNeT++). Can anyone help me explain this? Aren't they supposed to yield the same result, since I set the penetration rate to 0?

This is the config file in SUMO. The two simulations are based on the same configuration.

    <input>
    <net-file value="simul.net.xml"/>
    <route-files value="simul.rou.xml"/>
    <additional-files value="simul.add.xml"/>
</input>

<time>
    <begin value="0"/>
    <end value="4000"/>
    <step-length value="0.1"/>
</time>

<output>
    <tripinfo-output value="tripinfo.out.xml"/>
    <lanechange-output value="lanechange.out.xml"/>
</output>
J.G
  • 190
  • 3
  • 13
  • 1
    Are you talking about the results of a single, random simulation run each? – Christoph Sommer Apr 25 '18 at 06:20
  • @ChristophSommer The simulation is based on the same SUMO configuration, as added above. The route is pre-generated using `duarouter` tool with `--randomize-flows` flag. But it is generated before the simulation and the same route is used in both simulations. – J.G Apr 25 '18 at 14:19
  • @ChristophSommer In `omnetpp.ini` I also set the `*.manager.updateInterval = 0.1s `. Does that have something to do with the different results? – J.G Apr 25 '18 at 14:22
  • 1
    If you are looking at the results of a single simulation replication, you will need to make sure that the mobility models in SUMO consume the exact same pseudorandom numbers. Or: run the simulation enough times and use only aggregate metrics, so that random effects have no bearing on your results. – Christoph Sommer Apr 26 '18 at 15:04
  • @ChristophSommer Can you please point out where the randomness is involved? – J.G Apr 26 '18 at 16:03
  • If you are unsure about this, I would probably suggest the second option. – Christoph Sommer Apr 26 '18 at 20:30
  • 3
    SUMO should always generate the same random number sequence when started with the same inputs unless the --random option is used. Everything else I would consider a bug. – Michael Apr 27 '18 at 20:12

0 Answers0