drifty/comaps
1
0
Fork 0
mirror of https://codeberg.org/comaps/comaps synced 2025-12-24 06:53:46 +00:00
Code Issues Packages Projects Releases Wiki Activity

[traffxml] Tweak GuessDnp

Browse Source 
Signed-off-by: mvglasow <michael -at- vonglasow.com>
This commit is contained in:
mvglasow
2025-05-09 18:03:34 +03:00
parent 6e65e60c3d
commit 80a7ed503e
1 changed files with 26 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
traffxml/traff_model.cpp
View File

@@ -107,6 +107,7 @@ openlr::LinearLocationReference TraffLocation::ToLinearLocationReference(bool ba
lrp.m_functionalRoadClass = GetFrc();
if (!locationReference.m_points.empty())
{
// TODO use `distance` from TraFF reference point, if available and consistent with direct distance
locationReference.m_points.back().m_distanceToNextPoint
= GuessDnp(locationReference.m_points.back(), lrp);
}
@@ -226,19 +227,37 @@ std::optional<TrafficImpact> TraffMessage::GetTrafficImpact()
return std::nullopt;
}
// TODO tweak formula based on FRC, FOW and direct distance (lower FRC roads may have more and sharper turns)
uint32_t GuessDnp(openlr::LocationReferencePoint & p1, openlr::LocationReferencePoint & p2)
{
double doe = mercator::DistanceOnEarth(mercator::FromLatLon(p1.m_latLon),
mercator::FromLatLon(p2.m_latLon));
/*
* The tolerance factor is currently 1/0.6, or ~1.67, so that direct distance is just at the
* lower boundary for `openlr::LinearSegmentSource::FromLocationReferenceTag`. Since we use
* `openlr::LinearSegmentSource::FromCoordinatesTag`, different tolerance values apply,
* so direct distance is well within the lower boundary, where the upper boundary is ~6.67 times
* the direct distance. This should work even in mountain areas, where the shortest route from
* one valley to the next, using long-distance roads, can be up to ~3 times the direct distance.
* Acceptance boundaries for candidate paths are currently:
*
* for `openlr::LinearSegmentSource::FromLocationReferenceTag`, 0.6 to ~1.67 (i.e. 1/0.6) times
* the direct distance,
*
* for `openlr::LinearSegmentSource::FromCoordinatesTag`, 0.25 to 4 times the direct distance.
*
* A tolerance factor of 1/0.6 is the maximum for which direct distance would be accepted in all
* cases, with an upper boundary of at least ~2.78 times the direct distance. However, this may
* cause the actual distance to be overestimated and an incorrect route chosen as a result, as
* path candidates are scored based on the match between DNP and their length.
* Also, since we use `openlr::LinearSegmentSource::FromCoordinatesTag`, acceptance limits are
* much wider than that.
* In practice, the shortest route from one valley to the next in a mountain area is seldom more
* than 3 times the direct distance, based on a brief examination. This would be even within the
* limits of direct distance, hence we do not need a large correction factor for this scenario.
*
* Candidate values:
* 1.66 (1/0.6) – upper boundary for direct distance to be just within the most stringent limits
* 1.41 (2^0.5) – ratio between two sides of a square and its diagonal
* 1.3 – close to the square root of 1.66 (halfway between 1 and 1.66)
* 1.19 – close to the square root of 1.41
* 1 – direct distance unmodified
*/
return doe / 0.6f + 0.5f;
return doe * 1.19f + 0.5f;
}
/*