Reproduce Lisbon example and compare results with count data #87
Description
As discussed with @juanfonsecaLS1 today, we're looking to generate route network options to support 'bike bus' interventions, with a focus on Lisbon. Juan will be working on finding profiles/cost functions that are well suited to supporting bike busses. To support this we have some good starting points:
- OD data on overall travel patterns in Lisbon
- Count datasets described in this paper https://doi.org/10.5194/isprs-archives-XLVIII-4-W1-2022-279-2022 (see repo at https://github.com/Robinlovelace/foss4g22/ )
- Count dataset with counts for different types of cyclists in the data folder here: https://github.com/bicischools/bicischools/
Breaking the problem up into pieces I think this could be tackled by some combination of:
- Just getting the example in https://github.com/Urban-Analytics-Technology-Platform/od2net/tree/main/examples/lisbon running
- Comparing the results with the 'old' way of doing things described in https://github.com/Robinlovelace/foss4g22/
- Figuring out how to change the cost function in od2net, possibly from within the R package, and possibly with upstream contributions to the R package: https://urban-analytics-technology-platform.github.io/od2net/r/
- Generating many route networks, each one based on a different set of parameters, and assessing the fit between observed and modelled flows to home-in on the cost functions that generate the most realistic route networks for cycling overall and cycling to school in particular
Heads-up @temospena on this, this should generate the 'which routing profiles work' results we need for the project. Comments/suggestions welcome here, I've suggested that Juan starts off with a test repo for experiments, and some of the code could eventually be ported upstream into this repo (if the tests go well).