newton.ik.IKObjectiveRotation#

class newton.ik.IKObjectiveRotation(link_index, link_offset_rotation, target_rotations, canonicalize_quat_err=True, weight=1.0)[source]#

Bases: IKObjective

Match the world-space orientation of a link frame.

Parameters:

  • link_index (int) – Body index whose frame defines the constrained link.

  • link_offset_rotation (wp.quat) – Local rotation from the body frame to the constrained frame, stored in (x, y, z, w) order.

  • target_rotations (wp.array[wp.vec4]) – Target orientations, shape [problem_count]. Each quaternion is stored in (x, y, z, w) order.

  • canonicalize_quat_err (bool) – If True, flip equivalent quaternions so the residual follows the short rotational arc.

  • weight (float) – Scalar multiplier applied to the residual and Jacobian rows.

__init__(link_index, link_offset_rotation, target_rotations, canonicalize_quat_err=True, weight=1.0)#
compute_jacobian_analytic(body_q, joint_q, model, jacobian, joint_S_s, start_idx)#

Fill the rotation Jacobian block from the analytic motion subspace.

Parameters:

  • body_q (wp.array(dtype=wp.transformf, ndim=2)) – Batched body transforms for the evaluation rows, shape [n_batch, body_count].

  • joint_q (wp.array(dtype=wp.float32, ndim=2)) – Batched joint coordinates, shape [n_batch, joint_coord_count]. Present for interface compatibility and not used directly by this objective.

  • model (Model) – Shared articulation model.

  • jacobian (wp.array(dtype=wp.float32, ndim=3)) – Global Jacobian buffer to update, shape [n_batch, total_residual_count, joint_dof_count].

  • joint_S_s (wp.array(dtype=wp.spatial_vectorf, ndim=2)) – Batched motion-subspace columns, shape [n_batch, joint_dof_count].

  • start_idx (int) – First residual row reserved for this objective.

compute_jacobian_autodiff(tape, model, jacobian, start_idx, dq_dof)#

Fill the rotation Jacobian block using Warp autodiff.

Parameters:

  • tape (Tape) – Recorded Warp tape whose output is the global residual buffer.

  • model (Model) – Shared articulation model.

  • jacobian (wp.array(dtype=wp.float32, ndim=3)) – Global Jacobian buffer to update, shape [n_batch, total_residual_count, joint_dof_count].

  • start_idx (int) – First residual row reserved for this objective.

  • dq_dof (wp.array(dtype=wp.float32, ndim=2)) – Differentiable joint update variable for the current batch, shape [n_batch, joint_dof_count].

compute_residuals(body_q, joint_q, model, residuals, start_idx, problem_idx)#

Write weighted orientation errors into the global residual buffer.

Parameters:

  • body_q (wp.array(dtype=wp.transformf, ndim=2)) – Batched body transforms for the evaluation rows, shape [n_batch, body_count].

  • joint_q (wp.array(dtype=wp.float32, ndim=2)) – Batched joint coordinates, shape [n_batch, joint_coord_count]. Present for interface compatibility and not used directly by this objective.

  • model (Model) – Shared articulation model. Present for interface compatibility.

  • residuals (wp.array(dtype=wp.float32, ndim=2)) – Global residual buffer to update, shape [n_batch, total_residual_count].

  • start_idx (int) – First residual row reserved for this objective.

  • problem_idx (wp.array(dtype=wp.int32, ndim=1)) – Mapping from evaluation rows to base problem indices, shape [n_batch], used to fetch target_rotations.

init_buffers(model, jacobian_mode)#

Initialize caches used by analytic or autodiff Jacobian evaluation.

Parameters:

  • model (Model) – Shared articulation model.

  • jacobian_mode (IKJacobianType) – Jacobian backend selected for this objective.

residual_dim()#

Return the three rotational residual rows for this objective.

set_target_rotation(problem_idx, new_rotation)#

Update the target orientation for a single base IK problem.

Parameters:

  • problem_idx (int) – Base problem index to update.

  • new_rotation (vec4f) – Replacement target quaternion in (x, y, z, w) order.

set_target_rotations(new_rotations)#

Replace the target orientations for all base IK problems.

Parameters:

new_rotations (wp.array(dtype=wp.vec4f, ndim=1)) – Target quaternions, shape [problem_count], in (x, y, z, w) order.

supports_analytic()#

Return True because this objective has an analytic Jacobian.