Improve modeling of car availability

eqasim-org / ile-de-france

An open synthetic population of Île-de-France for agent-based transport simulation

GNU General Public License v2.0

45 stars 65 forks source link

Improve modeling of car availability #107

Open sebhoerl opened 2 years ago

sebhoerl commented 2 years ago

For some scenarios (Nantes, for instance), we see that exact distance-based mode shares are hard to obtain if households in rural areas get too little car availability. The process should be improved to better represent car availability spatially to be able to work better with urban-rural continuum cases.

diallitoz commented 1 year ago

Hi @sebhoerl, Maybe we have the same problem in Lyon. We match very well the global modal shares. But, we are not able to reproduce the car modal share for long distances. parts_modales_par_distance

sebhoerl commented 1 year ago

In Nantes there was more of a problem that there was a "gap" in medium-distance trips. But sure, it may also be related to car availability.

Have you tried to play around with lambdaCostEuclideanDistance in the model? Basically, it defines how much cost makes a difference for longer distances, and depending on your PT cost model (maybe fixed price?) PT may become "unfairly" expensive
Would be interesting to map the car availability in the synpop vs. car availability from the survey and compare the two

diallitoz commented 1 year ago

Ok, I see. For this result, we have used the MATSim standard version. Indeed, I am now doing this representation. I will share the result soon.

sebhoerl commented 1 year ago

Ah ok, I think with the available parameters in standard MATSim, I think it will be difficult to model the PT trajectory ("first increasing, then reaching a tip, and decreasing again")

diallitoz commented 1 year ago

Yes, it's ! I'm trying to play with the ASC of PT and car. But, there is no major improvement.

diallitoz commented 1 year ago

Here are the car availability maps from the survey and the synpop respectively

car_map_reference

car_map_simulation_0 25

There is a slight variation in the household motorization rate in peripheral areas between the survey and the synthetic population. But I don't think that's the only reason for the decrease in car use for a long-distance trips.

diallitoz commented 1 year ago

Hello, After playing with the PT parameters waitingPt and utilityOfLineSwitch, we succeed in reproducing the modal shares of the car and the PT for long distances. reference_vs_simulation

sebhoerl commented 1 year ago

Great :)