torchdrivesim.utils

Miscellaneous utilities, including for geometric operations on tensors.

Classes

Resolution

Functions

isin(→ torch.Tensor)	Checks whether elements of tensor x are contained in tensor y.
normalize_angle(angle)	Normalize to <-pi, pi) range by shifting by a multiple of 2*pi.
rotation_matrix(→ torch.Tensor)	Counterclockwise rotation matrix in 2D.
rotate(→ torch.Tensor)	Rotate the vector counterclockwise (from x towards y).
relative(→ Tuple[torch.Tensor, torch.Tensor])	Computes position and orientation of the target relative to origin.
transform(→ torch.Tensor)	Given points relative to a pose, produce absolute positions of the points.
is_inside_polygon(→ torch.Tensor)	Checks whether a given point is inside a given convex polygon.
merge_dicts(→ Dict)	Merges a sequence of dictionaries, giving preference to entries earlier in the sequence.
assert_equal(x, y)
line_circle_intersection(→ torch.Tensor)	Determine intersections between batched line segments and circles in 2D.

Module Contents

class torchdrivesim.utils.Resolution

Bases: tuple

width

height

torchdrivesim.utils.isin(x: torch.Tensor, y: torch.Tensor) → torch.Tensor

Checks whether elements of tensor x are contained in tensor y. This function is built-in in torch >= 1.10 and will be removed from here in the future.

Parameters:

• x – any tensor

• y – a one-dimensional tensor

Returns:

a boolean tensor with the same shape as x

torchdrivesim.utils.normalize_angle(angle)

Normalize to <-pi, pi) range by shifting by a multiple of 2*pi. Works with floats, numpy arrays, and torch tensors.

torchdrivesim.utils.rotation_matrix(theta: torch.Tensor) → torch.Tensor

Counterclockwise rotation matrix in 2D.

Parameters:

theta – tensor of shape Sx1 with rotation angle in radians

Returns:

Sx2x2 tensor with the rotation matrix.

torchdrivesim.utils.rotate(v: torch.Tensor, angle: torch.Tensor) → torch.Tensor

Rotate the vector counterclockwise (from x towards y). Works correctly in batch mode.

Parameters:

• v – tensor of shape Sx2 representing points

• angle – tensor of shape Sx1 representing rotation angle

Returns:

Sx2 tensor of rotated points

torchdrivesim.utils.relative(origin_xy: torch.Tensor, origin_psi: torch.Tensor, target_xy: torch.Tensor, target_psi: torch.Tensor) → Tuple[torch.Tensor, torch.Tensor]

Computes position and orientation of the target relative to origin. Points are represented as Sx2 tensors of coordinates and Sx1 tensors of orientations in radians.

torchdrivesim.utils.transform(points: torch.Tensor, pose: torch.Tensor) → torch.Tensor

Given points relative to a pose, produce absolute positions of the points. There can be zero or more batch dimensions.

Parameters:

• points – BxNx2 tensor

• pose – Bx3 tensor of position (x,y) and orientation (yaw angle in radians)

Returns:

Bx2 tensor of absolute positions

torchdrivesim.utils.is_inside_polygon(point: torch.Tensor, polygon: torch.Tensor) → torch.Tensor

Checks whether a given point is inside a given convex polygon. B and P can be zero or more batch dimensions, the former being batch and the latter points.

Parameters:

• point – BxPx2 tensor of x-y coordinates

• polygon – BxNx2 tensor of points specifying a convex polygon in either clockwise or counter-clockwise fashion.

Returns:

boolean tensor of shape BxP indicating whether the point is inside the polygon

torchdrivesim.utils.merge_dicts(ds: List[Dict]) → Dict

Merges a sequence of dictionaries, giving preference to entries earlier in the sequence.

torchdrivesim.utils.assert_equal(x, y)

torchdrivesim.utils.line_circle_intersection(p1: torch.Tensor, p2: torch.Tensor, circle_center: torch.Tensor, radius: torch.Tensor) → torch.Tensor

Determine intersections between batched line segments and circles in 2D.

Parameters:

• p1 (torch.Tensor) – Start points of line segments […, 2]

• p2 (torch.Tensor) – End points of line segments […, 2]

• circle_center (torch.Tensor) – Circle centers […, 2]

• radius (torch.Tensor) – Circle radii […, 1]

Returns:

Boolean tensor indicating intersections […, 1]

Return type:

torch.Tensor