Unexpected routing by bike, ignoring cycle paths

[lalebarde]

Hi,
I have tried GraphHopper to calculate a routing by bike where it should select acycle path and it does not unexpectedly.

On line link don’t show the routing but at least you have the example.

Here is a snapshot of the routing obtained off-line:

It should take the cycle path underlined in red.

As you can see in the url, it is a customed profile for EUC, but mostly as a bike with:

{
  "if": "road_class == CYCLEWAY",
  "multiply_by": 1.0
},

Other road_class are penalized.

Here is a view of the area from OSM, showing the cycle path it should have used:

My full euc.json:

{
  "priority": [
    {
      "if": "road_class == PRIMARY",
      "multiply_by": 0.5
    },
    {
      "if": "road_class == SECONDARY",
      "multiply_by": 0.7
    },
    {
      "if": "road_class == TERTIARY",
      "multiply_by": 0.8
    },
    {
      "if": "road_class == CYCLEWAY",
      "multiply_by": 1.0
    },
    {
      "if": "road_class == MOTORWAY || road_class == TRUNK || road_class == PEDESTRIAN || road_class == FOOTWAY || road_class == BRIDLEWAY",
      "multiply_by": 0.0
    },
    {
      "if": "max_speed > 80",
      "multiply_by": 0.4
    },
    {
      "if": "surface == ASPHALT || surface == CONCRETE",
      "multiply_by": 1.0
    },
    {
      "if": "surface == GRAVEL || track_type == GRADE3",
      "multiply_by": 0.6
    },
    {
      "if": "surface == DIRT || track_type == GRADE4",
      "multiply_by": 0.3
    },
    {
      "if": "track_type == GRADE5",
      "multiply_by": 0.1
    },
  ],
  "speed": [
    {
      "if": "true",
      "limit_to": "20"
    },
    {
      "if": "surface == ASPHALT || surface == CONCRETE",
      "limit_to": "25"
    },
    {
      "if": "surface == GRAVEL || track_type == GRADE3",
      "limit_to": "18"
    },
    {
      "if": "surface == DIRT || track_type == GRADE4",
      "limit_to": "10"
    },
    {
      "if": "track_type == GRADE5 || smoothness == VERY_BAD",
      "limit_to": "6"
    }
  ]
}

What could be the reason for that?

[simse]

Hi,

first of all, the term bike_network in GH does not care about real infrastructure such as cycleways but it only takes (bicycles-)routes into account.
The definition of a bicycle route in OSM can be found here:
https://wiki.openstreetmap.org/wiki/Tag:route%3Dbicycle
To get a better unterstanding, here is a overpass-query for your area of interest:

overpass turbo

Also I made a query with “real” cycleways in OSM for your area of interest:

overpass turbo

In blue you see the separate cycleway infrastucture, that is also considered as road_class == CYCLEWAY in GH. Then in green there is cycleway infrastucture that is tagged not separatly but as an attribute of a road/highway. The GH-roadclass here is not cycleway but rather the corresponding highway-type in OSM, which I tried to describe also in here:

Take `cycleway`s along roads into account for custom routing profiles

Hi, I am working on a project where I use Graphhopper with custom bicycle routing to generate isochrones (locally). To create/compare different bicycle profiles, I want to create f.e. a profile that avoids roads without cycleways. So far I played around with the following custom profiles, to get an understanding of the given “road attributes” as described in here work: { "distance_influence": 15, "priority": [ { "if": "road_class==TRUNK||road_class==PRIMARY||road_class==SECONDAR…

To achive what you want (routing on cycleway infrastucture?) you rather need customize the priority weights for road_class, however, in my understanding the not-separated cycleways (see green-colered ways in the query above) can not be taken into account in GH, so far. (?)

I simplified your example:

GraphHopper Maps | Route Planner

You can play around with the weights of
road_class == CYCLEWAY
bike_network == MISSING
to get a better unterstanding.

Disclaimer: I am quite new to GH, however know OSM pretty well. I am happy to be corrected.

[lalebarde]

Thanks for your help. So as I play already with weights for road_class, I just need to add penalty when bike_network == MISSING, isn’t it ?

It looks like you did not have GH developers feedback from your post?

EDIT:

Here is the result:

I have added the following:

    {
      "if": "road_class == CYCLEWAY || bike_network != MISSING",
      "multiply_by": 1.0
    },
    {
      "if": "bike_network == MISSING",
      "multiply_by": 0.4
    },

It takes now the strong cycle infrastructure of Bordeaux, the REVE cycle network.

[simse]

It really depends if you care more about the (touristic) bicycle routes or you just care about “real” cycling infrastucture.

For example:
Should the routing prefare a cycleway that is part of a cycling route (f.e. the cycleway along Chemin du Phare Way: 458671183 | OpenStreetMap ) over another cycleway that is not part of a cycling route (f.e. the cycleway along Avenue Marie Curie Way: 744694308 | OpenStreetMap )?

Another example:
Should the routing prefare a tertiary highway without a cycleway but which is part of bicycle route ( f.e. Rue Toussaint Catros Way: ‪Rue Toussaint Catros‬ (‪398329290‬) | OpenStreetMap ) over a “real” cycleway that is not part of a bicycle route (f.e. Avenue de la Grande Semaine Way: 755878915 | OpenStreetMap ).
Depending on the custom weights, the route will either go along one or another:

With penalty for bike_network==MISSING:

GraphHopper Maps | Route Planner

Without penalty for bike_network==MISSING:

GraphHopper Maps | Route Planner

[lalebarde]

Thanks for your deep explanations, it is not an obvious topic. My need is to maximize cycleways. Being part of a bicycle route is not by itself a criteria for me, more a consequence. So I will try your last setup.

[ratrun]

Somehow I missed this discussion. As was pointed out before we are preferring segments which are part of a bicycle route relation. If you want to avoid that I would suggest the following custom rule:

{
  "priority": [
    {
      "if": "bike_network == LOCAL",
      "limit_to": "1.1"
    }
  ]
}