In my understanding, the multimodal contrib did the following:
(1) route previously teleported modes on a network
(2) simulate previously teleported modes on a network
(3) generate the network to be able to do (1) and (2)
(4) take into account slopes for walk and bike link travel times
The original motivation was that Christoph Dobler wanted to be able to do within-day rerouting for agents that were not network modes in the qsim.
For the following, let us assume that the multimodal mode is bike.
(1) has been superceded by the trip router, and in fact, most of the corresponding infrastructure in the multimodal contrib just refers to the trip router.
(2) can be emulated by
(a) generating an explicit "bike" network, where bike can either be shared with other modes, or the bike links can be completely separate
(b) making the flow capacity of these links large enough or the pce of the "bike" mode vehicle small enough that one never gets congestion.
(3) This could IMO be better done by explicit preprocessing after loading the scenario and before instantiating Controler.
(4) This might be worth rescuing since it uses values from the literature. However, the bicycle contrib also takes slopes into account, albeit by using a different paper as a basis.
Is there anybody who is still using the multimodal contrib, and who could not switch to using the core functionalities that are now there?
Just a few words on (4): For walking, I used literature that seemed quite reasonable and took also stuff like age, gender and purpose of a trip into account (however, I did not include the trip purpose back then). The model for bicycles was way simpler and more or less just to “have at least something”. Although I don’t know the bicycle contrib, I would suggest to take the slopes stuff from there.
Christoph, thanks for the info. The "bicycle" contrib in fact refers to some paper in the literature, so maybe that is the version to use for the time being.