Does ORS take into account the type of ’highway=’ to which a route has close connections?
For example; appropriateness of highway=tertiary|secondary for road-bike use can vary considerably. A close highway=trunk is often a good indicator of incompatible traffic with that routing profile. The cause may sometimes be counter-intuitive, i.e. characteristics of traffic may result in some spillover from faster roads, such that motorists psychologically need time to adapt to slower speeds or different conditions of smaller roads when coming from motorways or trunk roads.
The measure of distance would, in any case, be a graph measure. Spacial closeness often result in opposite traffic characteristics. Such may be the case when highway=secondary and highway=trunk|motorway|primary run in parallel. The secondary roads tend to get used for minor, local traffic only, where motorists usually have little to no intention to use the parallel motorway.
Of course, these are mere observations, and I have no data to back this up.
I am sorry, I think, I do not really understand your question. Of course ORS considers the highway tag, otherwise it could not extract a routing graph. The way categories such as tertiary may have an influence on how the graph is extracted.
When using the recommended weighting in the bike profile, there are preferences for bike lanes, in order to capture nearby parallel ways that are more suitable. However, traffic or psychological adaptation are not considered.
Could ORS consider the highway-tag for connected roads, when constructing the graph? Not meaning the highway-tag of the road considered for the route, or its parallel bike path, but the highway-tag of its junction roads.
When comparing the traffic of tertiary roads; where one of which has no connection to a motorway, and the other is an approach-road/feeder-road/Zubringer-straße; the tertiary with no such connection is significantly calmer. I suspect, adding a penalty to roads with a motorway-connection for the bicycle-profiles, would enhance the useability of the bike routes in a meaningful way.
The psychological example was just to point to the fact, that such things are easy to imagine and understand.
To put it very bluntly from a cyclists perspective, tertiary roads with connections to trunk-roads often have unbearable traffic, and I avoid them like the plague – I exaggerate, of course. I suspect, the computational effort is very controllable.
The German language has a beautiful idiom to illustrate this: über die Dörfer. Motorized traffic seems to flow torwards bigger/faster roads; bike traffic can compfortably take the small roads, one after another.
Thanks for the clarification, now I understand. ORS does not support this natively. If you are willing to do some coding and run an own instance of ORS, it could be extended by a further EncodedValue that captures information about feeder-roads. I guess the most complex part of it would be to find all joining roads while reading the OSM data. I would expect a lot of special cases in the OSM data that need to be considered.
I have to admit, I’m incapable of implementing this, but wanted to share the observation nevertheless. Looking at the OSM data, I agree, the distance of these connections is not always straightforward.
Developing a measure of distance would probably involve evaluating even more tags of the junctioned sections – such as the destination= pointing towards a trunk/motorway; or the road being part of a relation that is a motorway. highway=motorway_link might be sure sign though.