torchdrivesim.lanelet2#

All utilities for working with maps in Lanelet2 format. This module imports correctly and exports all the usual signatures even when the lanelet2 package is not installed, but in that case all calls to its functions raise the Lanelet2NotFound exception.

Module Contents#

Functions#

load_lanelet_map() → LaneletMap)

Load a Lanelet2 map from an OSM file on disk.

find_lanelet_directions(→ List[float])

For a given point, find local orientations of all lanelets that contain it.

find_direction(→ float)

For a given linestring and a point near it, finds the local orientation of the linestring.

pick_random_point_and_orientation(→ Tuple[float, ...)

Picks a point on the map by randomly selecting a lanelet and then randomly selecting a point

road_mesh_from_lanelet_map(→ torchdrivesim.mesh.BaseMesh)

Creates a road mesh by triangulating all lanelets in a given map.

line_segments_to_mesh(→ torchdrivesim.mesh.BaseMesh)

Converts a given collection of line segments to a mesh visualizing them.

lanelet_map_to_lane_mesh(→ torchdrivesim.mesh.BirdviewMesh)

Creates a lane marking mesh from a given map.

Attributes#

is_available

torchdrivesim.lanelet2.is_available = True[source]#
exception torchdrivesim.lanelet2.Lanelet2NotFound[source]#

Bases: ImportError

Python bindings for Lanelet2 are not installed.

exception torchdrivesim.lanelet2.LaneletError[source]#

Bases: RuntimeError

Some function related to Lanelet2 failed.

torchdrivesim.lanelet2.load_lanelet_map(map_path: str, origin: Tuple[float, float] = (0, 0)) → LaneletMap[source]#

Load a Lanelet2 map from an OSM file on disk.

Parameters:

  • map_path – local path to OSM file containing the map

  • origin – latitude and longitude of the origin to use with UTM projector

Raises:

  • Lanelet2NotFound – if lanelet2 package is not available

  • FileNotFoundError – if specified file doesn’t exist

torchdrivesim.lanelet2.find_lanelet_directions(lanelet_map: LaneletMap, x: float, y: float, tags_to_exclude: List[str] | None = None) → List[float][source]#

For a given point, find local orientations of all lanelets that contain it.

Parameters:

  • lanelet_map – the map with a lanelet layer

  • x – first coordinate of the point in the map’s coordinate frame

  • y – second coordinate of the point in the map’s coordinate frame

  • tags_to_exclude – lanelets tagged with any of those tags will be omitted as if they didn’t exist

Returns:

for each lanelet, the angle representing its local orientation in radians

Raises:

Lanelet2NotFound – if lanelet2 package is not available

torchdrivesim.lanelet2.find_direction(linestring, location3d) → float[source]#

For a given linestring and a point near it, finds the local orientation of the linestring.

Returns:

the linestring’s local orientation in radians

Raises:

  • Lanelet2NotFound – if lanelet2 package is not available

  • LaneletError – when the method fails, usually because the linestring has a weird shape

torchdrivesim.lanelet2.pick_random_point_and_orientation(lanelet_map: LaneletMap) → Tuple[float, float, float][source]#

Picks a point on the map by randomly selecting a lanelet and then randomly selecting a point along its centerline, both using random package.

Returns:

tuple (x, y, orientation), using the map’s coordinate frame and radians

Raises:

Lanelet2NotFound – if lanelet2 package is not available

torchdrivesim.lanelet2.road_mesh_from_lanelet_map(lanelet_map: LaneletMap, lanelets: List[int] | None = None) → torchdrivesim.mesh.BaseMesh[source]#

Creates a road mesh by triangulating all lanelets in a given map.

Parameters:

  • lanelet_map – map to triangulate

  • lanelets – if specified, only use lanelets with those ids

torchdrivesim.lanelet2.line_segments_to_mesh(points: torch.Tensor, line_width: float = 0.3, eps: float = 1e-06) → torchdrivesim.mesh.BaseMesh[source]#

Converts a given collection of line segments to a mesh visualizing them.

Parameters:

  • points – BxNx2x2 tensor specifying N line segments, each consisting of a pair of points

  • line_width – width of the line in meters

  • eps – small value to add for avoiding division by zero

Returns:

mesh visualizing the line, with 6*N verts and 4*N faces

torchdrivesim.lanelet2.lanelet_map_to_lane_mesh(lanelet_map: LaneletMap, left_handed: bool = False, batch_size: int = 50000) → torchdrivesim.mesh.BirdviewMesh[source]#

Creates a lane marking mesh from a given map.

Parameters:

  • lanelet_map – map to use

  • left_handed – whether the map’s coordinate system is left-handed (flips the left and right boundary designations)

  • batch_size – controls the amount of points processed in parallel