search

chairemobilite / transition

Transition is a modern new approach to transit planning. It's a web application to model, simulate and plan public transit and alternative transportation.
http://transition.city
MIT License
20 stars 13 forks source link

Resync osrm lua profile with upstream #467

Open greenscientist opened 1 year ago

greenscientist commented 1 year ago

Our osrm profile lua file might have diverged from upstream. We should resync them and make sure we have the latest and greatest options.

tahini commented 1 year ago

While revisiting the lua profiles, maybe we could put them outside the current typescript code, but still accessible and packageable with chaire-lib-backend.

Also, if those files come from osrm, it should be mentioned somewhere to give the proper copyright.

greenscientist commented 1 year ago

We could also just commit some "patch" file and regen them directly from upstream. (COuld be done manually and not at each build, but we be better to have some semi automation for that)

tahini commented 1 year ago

For the record: I just had a weird walking trip from parc Tillemont to de Bordeaux street a bit north of St-Zotique, where it prefers to cross Jean-Talon street at an alley and take the alley up to St-Zotique and come back on de Bordeaux. The same trip with osrm-backend upstream (https://map.project-osrm.org/) gives the expected path. So there is something in our profile that causes this.

There are expected differences with upstream, like we allow foot on service roads at origin or destination and our speeds are more fine-grained than upstream which always uses the same walking speed (even for steps). But there is something fishy.

We should figure out why and document our own modifications in our file header how we are different from upstream, and why.

I plan to do a small zone that can be quickly prepared and test a few trips. Hypothetical sources of problem: turn duration, traffic light penalty, some speed

kaligrafy commented 1 year ago

Added the bug label, since it is quite weird

greenscientist commented 11 months ago

@kaligrafy, can you have a look at this ? The cycling and walking profile are quite different than the upstream ones. And the our walking one is crashing when running the whole north america

kaligrafy commented 11 months ago

I'll try but this will need time and tests!

greenscientist commented 11 months ago

@kaligrafy BTW, those are the profile used by the OSM routing servers. But be worth also taking a look: https://github.com/fossgis-routing-server/cbf-routing-profiles