Bike navigation rankes streets too high

[simonschaufi]

I use this app for bike navigation quite a lot but I often switch to walking navigation as the bike navigation is often not the most convenient way and ignores even official bike ways and instead sends me to big roads.

Which should be the same like:

This is the route in Maps.me:

This is the route in Komoot:

Also not optimal as it also uses only big streets instead of field ways.

[biodranik]

Did you compare it with other apps? Is OSM data correct there? It would be easier for us if you share an OSM link to the wrong route, so we can quickly reproduce the issue.

Please report all found wrong routes with examples of good ones, then it will be easier for us to fix the algorithm.

[simonschaufi]

These are the routes:

• https://www.openstreetmap.org/directions?engine=graphhopper_bicycle&route=48.9734%2C8.4036%3B48.9647%2C8.3239#map=14/48.9685/8.3616 (:star: the route I would prefer)
• https://www.openstreetmap.org/directions?engine=fossgis_osrm_bike&route=48.9734%2C8.4036%3B48.9647%2C8.3239#map=14/48.9608/8.3604

and for comparison the one from Google: https://www.google.de/maps/dir/Rastatter+Stra%C3%9Fe+50,+Karlsruhe/Epplesee,+76287+Rheinstetten/@48.9748759,8.3459009,14z/data=!3m1!4b1!4m14!4m13!1m5!1m1!1s0x479705e12e723bc9:0x5afb1a1bbcbdcc63!2m2!1d8.4036298!2d48.9734004!1m5!1m1!1s0x4797040b218ff77d:0x92ded19145cf769e!2m2!1d8.3235399!2d48.9655289!3e1

Does this help? The routes in the app are totally different from the ones from OSM.

[RedAuburn]

Definite +1 for this, an option to only route (when possible) with dedicated bike lanes would be nice too

[vng]

@endim8 If the app skips dedicated/designated bike roads - it's an obvious bug. Please, attach the examples here, will investigate.

[vng]

@simonschaufi The desired route has highway=track Not sure, but is next set of tags ok for bicycle routing?

highway=track
motor_vehicle=forestry
name=Pirschweg
smoothness=intermediate
surface=fine_gravel
tracktype=grade2

The app assumes 8kmh speed for this road, at the same time has obviously bigger speed via "big roads".

Possible solution here is to introduce more bicycle profiles like fastest, nicest/quite.

[simonschaufi]

Many tracks which are used for walking can also be easily used by bikes except some really rough ways through fields but as a biker I rarely find ways which are only useful for walking. Exceptions are usually in the fields or ways in the forest but these ways I have in the screenshot could even be used by a car even though often not allowed. So the whole algorithm is not optimized for bikers.

Possible solution here is to introduce more bicycle profiles like fastest, nicest/quite.

Yes, that would be next level and would be also useful for cars (avoid highways...)

[biodranik]

@simonschaufi it would be helpful if you can make a list of OSM tags that actually allow cycling, so we can tune our algorithms.

[simonschaufi]

@biodranik can you tell me how to do that as I'm just a user of the app and don't know much about internals of OSM?

Is there no public data available of "official bike ways" which are marked with a green arrow in Germany? I wonder how other platforms like Komoot for example do their routing.

[biodranik]

We're not bikers :) So if you help us to find answers to your good questions, it may save us a lot of time and help us to focus on development.

You can check the real roads in your country and how are they tagged in osm.org (there is a data layer there).

[simonschaufi]

I'm a developer myself and could maybe also understand the algorithm if you point me to the right file where this happens. The main question is: is the tagging in OSM bad or is your algorithm not optimal? Which tag does the road need in order to be chosen as a possible way? Where is the "switch" between pedestrians and bikers?

I have added a screenshot from Komoot to my initial post.

[biodranik]

At first, it's important to check the OSM data quality. If everything is tagged properly, and it is obvious for you that there is a mistake in the algorithm, then it's a bug. You can check routing_common/bicycle_model.cpp

[simonschaufi]

Here are some already (which I don't find in the bicycle_model.cpp but I hope that helps):

• concrete:plates
• Gravel
• Compacted (most often used in my area which is perfect for bikers)
• Asphalt
• Paved

PS: I found a special type in OSM called "Bicycle Route". This can be used and should be rated high.

[vng]

Simply increasing highway=track priority for bicycle is not a good fix, IMHO. Upcoming release makes this route, that uses designated bicycle paths and seems like its OK:

We are working on alternative routes algorithm that will allow you to select best route for the user.

[vng]

I see a problem on this route here (48.96240, 8.33978):

The app left designated bicycle highway=path and uses secondary/residential instead. Will investigate it.

[vng]

One more claim of ignoring track roads for bicycles: https://www.openstreetmap.org/directions?engine=graphhopper_bicycle&route=49.6461%2C8.6475%3B49.6624%2C8.6920#map=14/49.6540/8.6696

Let's try to increase track's weight, especially with tracktype=grade1/grade2. And it should be defined as unpaved_good (now have unpaved_bad).

[vng]

One more example: (53.2876678, 8.11889789) -> (53.2122689, 8.20143976)

[pastk]

One more claim of ignoring track roads for bicycles: https://www.openstreetmap.org/directions?engine=graphhopper_bicycle&route=49.6461%2C8.6475%3B49.6624%2C8.6920#map=14/49.6540/8.6696

Let's try to increase track's weight, especially with tracktype=grade1/grade2. And it should be defined as unpaved_good (now have unpaved_bad).

Did you adjust it? Is it fixed now?

[simonschaufi]

It is indeed better with the release of 2024.10.08. I'm closing it now and will report further issues in a new issue. Thank you very much!