search

matsim-org / matsim-code-examples

A repository containing code examples around MATSim
GNU General Public License v3.0
75 stars 173 forks source link

initial demand #279

Open stevenhz opened 4 years ago

stevenhz commented 4 years ago

Shall I ask a question about initial demand? In the book chapter 1.2 you described that "It starts with an initial demand arising from the study area population’s daily activity chains. " Does "population daily activity chain" mean that the user should provide a complete population with their daily plan? In chapter 7.3 I read that there are scripts to generate initial demand, the script call RUNpopulationgenerator. I have not run the script yet. Does it generate only population or does it generate population with their daily plan? Thank you very much!

JWJoubert commented 4 years ago

Hi @stevenhz the other MATSim users may/should chip in as well. My take on it is that the population and their plans (travel demand) are two distinct steps. First, you have to generate a synthetic population, like Sun & Erath (2015) or Joubert (2018). For that you mainly need Census data... and that is (usually) readily available.

I found the travel demand (if you want to do it properly) a lot more challenging. Here you will need something like a trip/travel diary where people reported their actual travelling patterns (activity start and end times; activity locations; mode of transport) for an entire 24-hour period. We've recently submitted a paper about this, but it is still under review.

That would be the "complete" package.

An alternative, possibly simpler approach for generating the population and travel demand in one go is if you use existing Origin-Destination (OD) matrices. The assumption is that for every trip you generate a person. So, you model trips, not really "full daily plans". Also, often OD matrices are for the morning peak only, so you have to manually create a second trip for each person to get them back home. I believe the auto-magic scripts in MATSim typically assume OD-matrices as input. I may be wrong (I have not used them).

Toying around in MATSim with fabricated data is one thing... building useful and plausible travel demand is no trivial task (at least, that's my experience in South Africa). But DO NOT underestimate the amount of learning and behavioural modelling you can do if you have carefully created (fabricated) scenarios. For research purposes, this is a critical skill (but also non-trivial).

stevenhz commented 4 years ago

Hi @JWJoubert , thanks for the clarifying! In my understanding so far, the "initial demand" is kind of output of activity based travel demand modeling where population, activity, location, time and mode are simulated. And MATSim is mainly for optimization of departure time, mode choice, etc., is that correct?

JWJoubert commented 4 years ago

And MATSim is mainly for optimization of departure time, mode choice, etc., is that correct?

Well, yes (but potentially much more). MATSim allows you rich behavioural analysis of specific scenarios. So, I will not really call it optimization. Simulation models are only as good as the scenario that you are giving it to evaluate.