Integrating Cadyts and MATSim for calibration

[MAlbertini95]

I’m trying to create a MATSim scenario for my case study. I was looking at the works regarding the integration of Cadyts with MATSim for the calibration of the travel demand, and I had some question:
-Firstly, in relation to the scaling up of the network properties, how’s made? -Secondly, how did you group all different plans from CEMDAP into a list for every agent? -Lastly, what method did you use to calculate zonal parameters and zone2zone infos to put in CEMDAP, since I’m quite new, I would like to know if thhere’s a fast method to get the infos. Hope you can help with questions, thanks for the attention, Best Regards

[kainagel]

Dear Matteo,

Your question can be understood at several levels, and I am not sure which one you are aiming for. I will try:

Concrete scenario generation

Here, I would strongly recommend to go for the simplest approach possible, and make it more complex over time if needed. The simplest approach possible is

• take network from OpenStreetMap. This is quite standardized; see, e.g., https://github.com/matsim-org/matsim-code-examples/blob/11.x/src/main/java/org/matsim/codeexamples/network/RunCreateNetworkFromOSM.java .
• generate demand in some easy way. This is unfortunately not standardized because the input data varies widely from region to region. If you don't have anything, I would still not start from CEMDAP/Cadyts, but just make something up. See, e.g., https://github.com/matsim-org/matsim-code-examples/blob/11.x/src/main/java/org/matsim/codeexamples/population/demandGeneration/RunPOnePersonPopulationGenerator.java .

I can't tell from your email if you have already undertaken steps in this direction, and if so, which ones.

Longer term plan for better scenario

The CEMDAP/Cadyts approach is one of many possible approaches; if it is useful for you depends strongly (1) on your available data, and (2) on your needs. As available data, it needs at least a commuting matrix, i.e. how many people travel from one zone to another for work in the morning. Other things, such as precise locations of facilities for activities, or traffic counts, are optional.

If this is what you need, we could provide more help, since it is something we are currently working on ourselves (in particular replacing CEMDAP by ActiTopp, which makes the whole tool chain java-only). We would, however, need to know how far you are with the concrete scenario generation described above, which tools you are using, which input data you have, etc.

General

If you want concrete help, could you please do the following:

• Make a fork of https://github.com/matsim-org/matsim-example-project .
• Program your code inside that project and push to github.
• For each question you have, there should be a runnable class with a main method. (Optimally, there would be a test case.)
• Input data should be in the repository. If you don't want to put it there, the repository needs to contain stub versions, i.e. short files with some example data. The runnable main methods should refer to these files.
• Send the community a link to that repo. Maybe you will find it awkward to make your code visible in that way. In that case, please understand that we are a community project, and this means community standards also for people who we help.

Best wishes ...

[MAlbertini95]

Dear Mr. Nagel, Ok, so I would try to explain better at which point I am with the building up of the scenario. I already have a OSM-derived network, and I have ancd i know how to generate a realistic synthetic population ( 112.500 pp) from my case study. I do have now to obtain somehow the travel plans for each person, and I don’t really know how to build it, the more feasible method I found was the one you used for the Berlin Open scenario, but if you gave suggestions I’ll be take it in consideration, it would be great. I do have a commuter matrix, but with a lower resolution than TAZ-resolution I would like to simulate. So I thought a calibration would be useful to better represent the current scenario ( I do have traffic counts ). Lastly, I know just a little of how CEMDAP works, and how to input data, but I don’t know to convert output files into readable MATSim plans (how to deal with TAZ zones and the coordinates, how to write multiple plans into one agent plans). Hope you can help and/or give advice. Thank you for your attention. Best Regards,

---

Da: kainagel notifications@github.com Inviato: Wednesday, September 11, 2019 11:55:32 AM A: matsim-org/matsim-code-examples matsim-code-examples@noreply.github.com Cc: MAlbertini95 teo.albe@hotmail.it; Author author@noreply.github.com Oggetto: Re: [matsim-org/matsim-code-examples] Integrating Cadyts and MATSim for calibration (#201)

Dear Matteo,

Your question can be understood at several levels, and I am not sure which one you are aiming for. I will try:

Concrete scenario generation

Here, I would strongly recommend to go for the simplest approach possible, and make it more complex over time if needed. The simplest approach possible is

• take network from OpenStreetMap. This is quite standardized; see, e.g., https://github.com/matsim-org/matsim-code-examples/blob/11.x/src/main/java/org/matsim/codeexamples/network/RunCreateNetworkFromOSM.java .
• generate demand in some easy way. This is unfortunately not standardized because the input data varies widely from region to region. If you don't have anything, I would still not start from CEMDAP/Cadyts, but just make something up. See, e.g., https://github.com/matsim-org/matsim-code-examples/blob/11.x/src/main/java/org/matsim/codeexamples/population/demandGeneration/RunPOnePersonPopulationGenerator.java .

I can't tell from your email if you have already undertaken steps in this direction, and if so, which ones.

Longer term plan for better scenario

The CEMDAP/Cadyts approach is one of many possible approaches; if it is useful for you depends strongly (1) on your available data, and (2) on your needs. As available data, it needs at least a commuting matrix, i.e. how many people travel from one zone to another for work in the morning. Other things, such as precise locations of facilities for activities, or traffic counts, are optional.

If this is what you need, we could provide more help, since it is something we are currently working on ourselves (in particular replacing CEMDAP by ActiTopp, which makes the whole tool chain java-only). We would, however, need to know how far you are with the concrete scenario generation described above, which tools you are using, which input data you have, etc.

General

If you want concrete help, could you please do the following:

• Make a fork of https://github.com/matsim-org/matsim-example-project .
• Program your code inside that project and push to github.
• For each question you have, there should be a runnable class with a main method. (Optimally, there would be a test case.)
• Input data should be in the repository. If you don't want to put it there, the repository needs to contain stub versions, i.e. short files with some example data. The runnable main methods should refer to these files.
• Send the community a link to that repo. Maybe you will find it awkward to make your code visible in that way. In that case, please understand that we are a community project, and this means community standards also for people who we help.

Best wishes ...

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHubhttps://github.com/matsim-org/matsim-code-examples/issues/201?email_source=notifications&email_token=AMNEGPP6W7EMUYDHEC23UZLQJC6BJA5CNFSM4IS67ZX2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD6N6HYQ#issuecomment-530310114, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AMNEGPNLUI35AHDWCT2TLDTQJC6BJANCNFSM4IS67ZXQ.

[kainagel]

Going through old emails. Don't know if this one here is still an issue. I would say that the above-mentioned RunPOnePersonPopulationGenerator would be a good starting point. There is more material in the same package, for increasingly complex cases, and then more material if you go one level up (int org.matsim.codeexamples.population). The approach that we always use for situations like yours (having a commuter matrix referring to zones) is:

• For each entry in the commuter matrix, generate one agent. Have her/him have a home activity in the home zone, a work activity in the work zone, and a home activity in the home zone. (We often call this home-work-home plans.)
• For the locations, find a coordinate randomly in the corresponding zone. The package org.matsim.codeexamples.population.demandGenerationFromShapefile describes this. Make sure that your two home locations get the same coordinate!