irsim.lib.algorithm.ray_casting_2d

Vectorized 2D ray casting against simulation objects.

Shared by the lidar sensors (Lidar2D and FMCWLidar2D). The high-level cast_rays() function flattens already-detected object boundaries and returns the nearest object hit by every beam. The lower-level cast_ray_segments() function performs the vectorized numerical kernel.

For a sensor origin in free space, this reproduces a GEOS difference scan to floating-point precision while avoiding the expensive overlay. FMCW exit hits also remain compatible when the origin is inside an obstacle.

Typical use:

ranges, hit_object_indices, origin, directions = cast_rays(
    lidar_geometry, detected_objects, range_max
)

Attributes

ORIGIN_EPS
SEGMENT_CHUNK_SIZE

Functions

boundary_segments(geometry)	Flatten a geometry's boundary into (start, end) segment arrays.
cast_ray_segments(→ tuple[numpy.ndarray, numpy.ndarray])	Nearest ray-segment hit distance and segment index per ray (vectorized).
cast_rays(→ tuple[numpy.ndarray, numpy.ndarray, ...)	Cast a 2D lidar geometry against already-detected objects.

Module Contents

irsim.lib.algorithm.ray_casting_2d.ORIGIN_EPS = 1e-09

irsim.lib.algorithm.ray_casting_2d.SEGMENT_CHUNK_SIZE = 1024

irsim.lib.algorithm.ray_casting_2d.boundary_segments(geometry)

Flatten a geometry's boundary into (start, end) segment arrays.

Handles every obstacle geometry a lidar may encounter: polygons (with holes), rectangles, circles (buffer polygons), linestrings, map segments, and compound MultiPolygon / GeometryCollection shapes. Points and empty geometries contribute no edges.

Polygonal parts are reduced to their ring linestrings with shapely.boundary() (which covers the outer ring and any holes); linestrings are used directly. Note shapely.boundary of a linestring returns its endpoints, not its segments, so it must not be applied to line obstacles.

参数:

geometry -- Any Shapely geometry.

返回:

(start, end) endpoint arrays, each of shape (M, 2) ((0, 2) when the geometry has no edges).

返回类型:

tuple[np.ndarray, np.ndarray]

irsim.lib.algorithm.ray_casting_2d.cast_ray_segments(origin: numpy.ndarray, directions: numpy.ndarray, seg_start: numpy.ndarray, seg_end: numpy.ndarray, max_range: float) → tuple[numpy.ndarray, numpy.ndarray]

Nearest ray-segment hit distance and segment index per ray (vectorized).

Solves, for every ray origin + t * direction against every segment seg_start + u * (seg_end - seg_start), the intersection in vectorized segment blocks and keeps the nearest valid hit (0 < t <= max_range, 0 <= u <= 1) per ray. Blocking bounds peak matrix memory by SEGMENT_CHUNK_SIZE * number_of_rays. Collinear overlaps are handled separately because their standard intersection denominator is zero.

参数:

• origin (np.ndarray) -- Shared ray origin (2,).

• directions (np.ndarray) -- Unit ray directions (N, 2).

• seg_start (np.ndarray) -- Segment start points (M, 2).

• seg_end (np.ndarray) -- Segment end points (M, 2).

• max_range (float) -- Maximum ray length; misses return this.

返回:

ranges (N,) clamped to max_range, and hit_index (N,) giving the index into the segment arrays that each ray hit (-1 on a miss).

返回类型:

tuple[np.ndarray, np.ndarray]

irsim.lib.algorithm.ray_casting_2d.cast_rays(lidar_geometry, detected_objects, max_range: float) → tuple[numpy.ndarray, numpy.ndarray, numpy.ndarray, numpy.ndarray]

Cast a 2D lidar geometry against already-detected objects.

This geometry-only operation derives ray parameters, gathers boundary segments from the supplied objects, runs the numerical kernel, and maps segment hits back to indices in detected_objects. Scene lookup remains the sensor's responsibility.

参数:

• lidar_geometry -- Max-range beams in world coordinates as a Shapely multiline geometry.

• detected_objects -- Objects selected by the sensor's scene query.

• max_range -- Maximum ray length; misses return this value.

返回:

Ranges, indices into detected_objects, origin, and directions.

返回类型:

tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray]