I am using veins 5.1 with omnet 5.6.2 and I have noticed some abnormal behaviour in my simulations related to the penetration ratio manager and the number of communicating vehicles.
Actually, the road network is a very simple network composed of a two lanes bi directional road. There is so called "road 0" going from west to east, and "road 1" going from east to west. Each road is composed of two lanes. The sumo demand is symetrical. However, I have noticed that the number of communicating vehicles on each road is not symetrical, and depends on the penetration ratio : see figure below.
I have run different simulations by changing the seeds and the behaviour remains whith this bias even for 100 different runs.
Any help would be much appreciated. Sincerely yours,
The algorithm implemented in Veins 5.1 is deterministic. For every new car inserted into the mobility simulation (in your case: SUMO, I guess) it checks if ignoring it would violate the configured market penetration requirement - and, if it would not, ignores the car for the network simulation.
If your mobility simulation uses a deterministic algorithm for filling roads with cars (e.g., first try inserting as many cars as would fit on road 1, then try road 2), you will see this effect bleeding through into your network simulation, at least until enough cars have left and re-entered the mobility simulation to even out the effect.
Possible solutions would include: letting the simulation run for a longer time and ignoring everything that happened during the initial "transient period" at its start; or, changing the way cars are inserted into the mobility simulation; or, patching the algorithm to be fully random (at the cost of the exact simulated market penetration now being subject to chance as well)