Weird routing off main road onto a _link and then back to the main road

[rickmastfan67]

Just notice this recently. For some odd reason, there is a section of highway that the OSRM routing engine is bypassing and sending drivers onto a 'ramp' and then back onto the main road.

Incorrect: http://osrm.at/9bK (32 seconds)
Correct: http://osrm.at/9bL (33 seconds)

It also seems that OSRM thinks that the '_link' has a faster speed limit then it should have. (There is no speed limit signs posted on it, hence why there isn't a maxspeed tag on it.) It seems that since the main road has two traffic light nodes on it & and the _link just one, ORSM gives the main road an extra penalty that is just enough to force routing off of it here. Pretty sure there isn't any bugs in classification here as MapQuest has no problems with this and correctly keeps people on McKnight Road. http://mapq.st/1tZEmcw And if you try to force it onto the _link, it's 12 seconds longer to use it then the main route.

On a separate note, I noticed that on the last step of the 'Incorrect' route, it tells people to 'Continue onto US 19 Truck'. The problem is, it doesn't tell people which way to turn onto 'US 19 Truck;McKnight Road'. I think that should be looked into as well, as it should tell people to turn right instead of just 'continuing' onto that road.

[DennisOSRM]

Thanks for reporting this. What's the correct speed on such links?

[rickmastfan67]

Depends on the area I would assume since some jughandles can be huge, and some really small/compact. I personally would think no more than 25 to 30 mph on the _links unless it has the maxspeed tag already added to the way.

[DennisOSRM]

thanks

[emiltin]

sounds like the issue that this PR fixes: #989

[DennisOSRM]

A fix/workaround has been pushed. Will close once data has been refreshed and the issue is (hopefully) gone.

[emiltin]

the PR https://github.com/Project-OSRM/osrm-backend/pull/989/files includes a few cuke tests that you might want to add

[DennisOSRM]

cool, let's move these over.

[DennisOSRM]

@rickmastfan67 your original issue has been fixed:

[emiltin]

want these two tests added? they provide a reality check of link speeds, although the existing tests of raw speed on way types should in principle cover much the same:

    Scenario: Car - Don't follow link when main road has a reasonable maxspeed
        Given the node map
            | a |   |   |   |   |   | e |
            |   | b |   |   |   | d |   |
            |   |   |   | c |   |   |   |

        And the ways
            | nodes | highway       | maxspeed |
            | abcde | motorway      | 60       |
            | bd    | motorway_link |          |

        When I route I should get
            | from | to | route |
            | a    | e  | abcde |

    Scenario: Car - Follow link when main road has a very low maxspeed
        Given the node map
            | a |   |   |   |   |   | e |
            |   | b |   |   |   | d |   |
            |   |   |   | c |   |   |   |

        And the ways
            | nodes | highway       | maxspeed |
            | abcde | motorway      | 30       |
            | bd    | motorway_link |          |

        When I route I should get
            | from | to | route          |
            | a    | e  | abcde,bd,abcde |

if the answer is yes, they should maybe be extended to cover trunk, primary, secondary and tertiary links.