kainagel
commented
4 years ago Depends on how you want to set this up and how you want to use it. Plain MATSim already has a randomizing component in the routing. It is in the planscalcroute config. Note that that was buggy until a few weeks ago, so you could set the randomness away from 0 but it would not pick that up but it should work with a recent weekly build.
This randomizes the ratio between time disutility and money disutility. Money disutility normally comes from distance-based cost. What you get are Pareto-optimal routes.
Practically, you should be able to do the following:
- distribute your 1000 people into 1000 different matsim persons, all with the same plan.
- make sure you have time-based costs and distance-based costs
- run this through one iteration of matsim (i.e. from iteration 0 to iteration 0) and see what you get
If you don't get different routes between the different people, put the whole thing into a git repo and I will have a look.
As stated, the above models Pareto routes. Sometimes, people are more interested in link randomness (sometimes called "de la Barra"). This you would have to do yourself in matsim, by replacing the TravelDisutility. Also not terribly difficult to do. Again, if you want help, put the whole thing into a git repo and I will have a look.
Hi,
We are in the midst of research that a part of it deals with navigation errors. For that matter, let's say I want to get from point A to point B in the shortest route. The information I have about the shortest route is inaccurate, so I may have navigation errors, so eventually, my route from A to B will be longer than the shortest route. Another person may make other navigation errors and therefore his route will be a different length. Suppose 1000 people are required to get from A to B, can I use MatSim get all route lengths? If so, what is the right way to do it?
Thanks, Oren