MetaSeeds
As discussed with @hm-lehmberg already: I believe that there might arise problems from using only one seed throughout the simulation when varying parameters. It might happen that if a submodel is changed, the random numbers are differently distributed over the other parts. Also, if the number of pedestrians in a simulation is changed, there will be other velocities / positions assigned. For example, changing from 50 to 51 pedestrians, there will be more random numbers be used for - let's say - free-flow speeds and therefore, there will be a random number missing for the next category.
One solution seems to be to generate "metaSeeds", meaning that we generate from the main seed a set of seeds which are then used for each of the "random" places in the code, e.g. positions, distribution of groups, speeds, ....
@BZoennchen @hm-kleinmei : What do you think? Any contradictions?