Relies on GetClosestPoint for the GetClosestPointUsing2dProjection method.

[francocipollone]

🎉 New feature

Related to #50 #47

Goes on top of #52

Summary

• Refactor GetClosestPointUsing2dProjection in order to make use of the kdtree3d we have.
  This reduce a lot the time it takes LaneGeoemetry::WInverse queries, and therefore ToLanePosition, ToRoadPosition and FindRoadPosition times are massively improved

Speed up

Benchmark name	cpu time[us]		speed up	map
	francocipollone/poc_line_string_kd_tree	francocipollone/use_closest_point_for_2d_projection_case
MaliputSparseGeometryTime/InterpolatePointAtCenterline	0.132	0.124	1.06
MaliputSparseGeometryTime/ToArcCoordinates/WInverse	14717	3.07	4793.81
MaliputOsmQueryTime/GetLane/0	0.023	0.023	1.00	SingleLane.osm
MaliputOsmQueryTime/GetLane/1	0.022	0.022	1.00	ArcLane.osm
MaliputOsmQueryTime/GetLane/2	0.024	0.028	0.86	ArcLaneDense.osm
MaliputOsmQueryTime/GetLane/3	0.024	0.023	1.04	Circuit.osm
MaliputOsmQueryTime/Length/0	0.003	0.003	1.00	SingleLane.osm
MaliputOsmQueryTime/Length/1	0.003	0.004	0.75	ArcLane.osm
MaliputOsmQueryTime/Length/2	0.003	0.004	0.75	ArcLaneDense.osm
MaliputOsmQueryTime/Length/3	0.003	0.003	1.00	Circuit.osm
MaliputOsmQueryTime/ToLanePosition/0	3.26	4.51	0.72	SingleLane.osm
MaliputOsmQueryTime/ToLanePosition/1	11.5	6.24	1.84	ArcLane.osm
MaliputOsmQueryTime/ToLanePosition/2	350	17.1	20.47	ArcLaneDense.osm
MaliputOsmQueryTime/ToLanePosition/3	23.8	8.85	2.69	Circuit.osm
MaliputOsmQueryTime/ToInertialPosition/0	1.22	1.41	0.87	SingleLane.osm
MaliputOsmQueryTime/ToInertialPosition/1	4.16	1.65	2.52	ArcLane.osm
MaliputOsmQueryTime/ToInertialPosition/2	129	4.5	28.67	ArcLaneDense.osm
MaliputOsmQueryTime/ToInertialPosition/3	8.45	2.45	3.45	Circuit.osm
MaliputOsmQueryTime/GetOrientation/0	1.07	1.25	0.86	SingleLane.osm
MaliputOsmQueryTime/GetOrientation/1	3.92	1.48	2.65	ArcLane.osm
MaliputOsmQueryTime/GetOrientation/2	119	4.04	29.46	ArcLaneDense.osm
MaliputOsmQueryTime/GetOrientation/3	7.92	2.31	3.43	Circuit.osm
MaliputOsmQueryTime/ToRoadPosition/0	8.97	10.4	0.86	SingleLane.osm
MaliputOsmQueryTime/ToRoadPosition/1	37.1	13.7	2.71	ArcLane.osm
MaliputOsmQueryTime/ToRoadPosition/2	1155	36.8	31.39	ArcLaneDense.osm
MaliputOsmQueryTime/ToRoadPosition/3	476	168	2.83	Circuit.osm
MaliputOsmQueryTime/FindRoadPositions/0	7.24	8.3	0.87	SingleLane.osm
MaliputOsmQueryTime/FindRoadPositions/1	24.8	10.2	2.43	ArcLane.osm
MaliputOsmQueryTime/FindRoadPositions/2	761	28.5	26.70	ArcLaneDense.osm
MaliputOsmQueryTime/FindRoadPositions/3	429	153	2.80	Circuit.osm

Checklist

• Signed all commits for DCO
• Added tests
• Added example and/or tutorial
• Updated documentation (as needed)
• Updated migration guide (as needed)
• Consider updating Python bindings (if it affects the public API)

Note to maintainers: Remember to use Squash-Merge and edit the commit message to match the pull request summary while retaining Signed-off-by messages.

[agalbachicar]

LGTM!

Wow. Impressive!
It sounds like small amounts of points still prefer the previous approach, right?

[francocipollone]

LGTM!

Wow. Impressive! It sounds like small amounts of points still prefer the previous approach, right?

Yes but only for SingleLane which each line string is only created with 2 points (start and end point).
That scalates quickly: For ArcLane (not dense one) there are about 10-20 points per linestring and the speed up is already remarkable