-
Notifications
You must be signed in to change notification settings - Fork 179
faq 48955430
by Kai Nagel on 2016-03-01 15:34:55
by Kai Nagel on 2016-03-01 15:35:10
Yes, and it is even explicitly recommended in particular initially when gaining experience with MATSim for a large scenario. Steps to take:
- Generate a synthetic population (= initial plans) that is only a fraction of the "true" population. One can, for example, do this from the start, or take a given travel survey, or randomly sample from a full population. Say we look at a 10% sample, i.e. a factor of 0.1.
Set the
flowCapacityFactor
setting in the config file (qsim section) to that same value (i.e. 0.1). This will scale down the flow capacity of each individual link by that factor.- In principle, one should set
storageCapacityFactor
to the same value. We have better experiences with making it a bit larger, e.g. (factor)^0.75. This gives approximately storageCapacityFactor=0.18 for flowCapacityFactor=0.1, and storageCapacityFactor=0.03 for flowCapacityFactor=0.01.
To understand this, it may be best to imagine that rather than reducing flow and storage capacity, these settings make vehicles larger, and a xxxCapacityFactor of 0.01 means that a standard car is 600 metres long and has a time headway of 200 seconds. The somewhat increased storage capacity factor gives the simulation "more flexibility".
The simulation runs correspondingly faster. The mental modules scale exactly in the population size. The mobsim does not scale exactly in the population size, but the number of vehicles plays an important role.
Having said this, there are unfortunately no systematic studies of this (or at least I am not aware of them). At TU Berlin, we have normally used the samples only for debugging, and run the full productive studies then at 100%. Exceptions are the "sketch planning" papers at http://www.vsp.tu-berlin.de/publications/vspwp .
You are viewing an archive of the previous MATSim Q&A site. The real site is now at https://matsim.org/faq