rlightning.humanoid.utils.lafan_vendor.utils

rlightning.humanoid.utils.lafan_vendor.utils.angle_axis_to_quat(angle, axis)

Converts from and angle-axis representation to a quaternion representation

Parameters:

• angle – angles tensor

• axis – axis tensor

Returns:

quaternion tensor

rlightning.humanoid.utils.lafan_vendor.utils.euler_to_quat(e, order='zyx')

Converts from an euler representation to a quaternion representation

Parameters:

• e – euler tensor

• order – order of euler rotations

Returns:

quaternion tensor

rlightning.humanoid.utils.lafan_vendor.utils.extract_feet_contacts(pos, lfoot_idx, rfoot_idx, velfactor=0.02)

Extracts binary tensors of feet contacts

Parameters:

• pos – tensor of global positions of shape (Timesteps, Joints, 3)

• lfoot_idx – indices list of left foot joints

• rfoot_idx – indices list of right foot joints

• velfactor – velocity threshold to consider a joint moving or not

Returns:

binary tensors of left foot contacts and right foot contacts

rlightning.humanoid.utils.lafan_vendor.utils.interpolate_local(lcl_r_mb, lcl_q_mb, n_past, n_future)

Performs interpolation between 2 frames of an animation sequence.

The 2 frames are indirectly specified through n_past and n_future. SLERP is performed on the quaternions LERP is performed on the root’s positions.

Parameters:

• lcl_r_mb – Local/Global root positions (B, T, 1, 3)

• lcl_q_mb – Local quaternions (B, T, J, 4)

• n_past – Number of frames of past context

• n_future – Number of frames of future context

Returns:

Interpolated root and quats

rlightning.humanoid.utils.lafan_vendor.utils.length(x, axis=-1, keepdims=True)

Computes vector norm along a tensor axis(axes)

Parameters:

• x – tensor

• axis – axis(axes) along which to compute the norm

• keepdims – indicates if the dimension(s) on axis should be kept

Returns:

The length or vector of lengths.

rlightning.humanoid.utils.lafan_vendor.utils.normalize(x, axis=-1, eps=1e-08)

Normalizes a tensor over some axis (axes)

Parameters:

• x – data tensor

• axis – axis(axes) along which to compute the norm

• eps – epsilon to prevent numerical instabilities

Returns:

The normalized tensor

rlightning.humanoid.utils.lafan_vendor.utils.quat_between(x, y)

Quaternion rotations between two 3D-vector arrays

Parameters:

• x – tensor of 3D vectors

• y – tensor of 3D vetcors

Returns:

tensor of quaternions

rlightning.humanoid.utils.lafan_vendor.utils.quat_fk(lrot, lpos, parents)

Performs Forward Kinematics (FK) on local quaternions and local positions to retrieve global representations

Parameters:

• lrot – tensor of local quaternions with shape (…, Nb of joints, 4)

• lpos – tensor of local positions with shape (…, Nb of joints, 3)

• parents – list of parents indices

Returns:

tuple of tensors of global quaternion, global positions

rlightning.humanoid.utils.lafan_vendor.utils.quat_ik(grot, gpos, parents)

Performs Inverse Kinematics (IK) on global quaternions and global positions to retrieve local representations

Parameters:

• grot – tensor of global quaternions with shape (…, Nb of joints, 4)

• gpos – tensor of global positions with shape (…, Nb of joints, 3)

• parents – list of parents indices

Returns:

tuple of tensors of local quaternion, local positions

rlightning.humanoid.utils.lafan_vendor.utils.quat_inv(q)

Inverts a tensor of quaternions

Parameters:

q – quaternion tensor

Returns:

tensor of inverted quaternions

rlightning.humanoid.utils.lafan_vendor.utils.quat_mul(x, y)

Performs quaternion multiplication on arrays of quaternions

Parameters:

• x – tensor of quaternions of shape (…, Nb of joints, 4)

• y – tensor of quaternions of shape (…, Nb of joints, 4)

Returns:

The resulting quaternions

rlightning.humanoid.utils.lafan_vendor.utils.quat_mul_np(x, y)

Performs quaternion multiplication on arrays of quaternions

Parameters:

• x – tensor of quaternions of shape (…, Nb of joints, 4)

• y – tensor of quaternions of shape (…, Nb of joints, 4)

Returns:

The resulting quaternions

rlightning.humanoid.utils.lafan_vendor.utils.quat_mul_vec(q, x)

Performs multiplication of an array of 3D vectors by an array of quaternions (rotation).

Parameters:

• q – tensor of quaternions of shape (…, Nb of joints, 4)

• x – tensor of vectors of shape (…, Nb of joints, 3)

Returns:

the resulting array of rotated vectors

rlightning.humanoid.utils.lafan_vendor.utils.quat_normalize(x, eps=1e-08)

Normalizes a quaternion tensor

Parameters:

• x – data tensor

• eps – epsilon to prevent numerical instabilities

Returns:

The normalized quaternions tensor

rlightning.humanoid.utils.lafan_vendor.utils.quat_slerp(x, y, a)

Performs spherical linear interpolation (SLERP) between x and y, with proportion a

Parameters:

• x – quaternion tensor

• y – quaternion tensor

• a – indicator (between 0 and 1) of completion of the interpolation.

Returns:

tensor of interpolation results

rlightning.humanoid.utils.lafan_vendor.utils.remove_quat_discontinuities(rotations)

Removing quat discontinuities on the time dimension (removing flips)

Parameters:

rotations – Array of quaternions of shape (T, J, 4)

Returns:

The processed array without quaternion inversion.

rlightning.humanoid.utils.lafan_vendor.utils.rotate_at_frame(X, Q, parents, n_past=10)

Re-orients the animation data according to the last frame of past context.

Parameters:

• X – tensor of local positions of shape (Batchsize, Timesteps, Joints, 3)

• Q – tensor of local quaternions (Batchsize, Timesteps, Joints, 4)

• parents – list of parents’ indices

• n_past – number of frames in the past context

Returns:

The rotated positions X and quaternions Q