Skip to content

Fix [xyz, quat] format pose ik computation for robot #149

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
yuecideng merged 2 commits into from
Feb 26, 2026
Merged

Fix [xyz, quat] format pose ik computation for robot #149

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3222078
wip
yuecideng Feb 26, 2026
21ae68d
wip
yuecideng Feb 26, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
52 changes: 19 additions & 33 deletions embodichain/lab/sim/objects/robot.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,7 @@
from embodichain.lab.sim.solvers import SolverCfg, BaseSolver
from embodichain.lab.sim.objects import Articulation
from embodichain.lab.sim.utility.tensor import to_tensor
from embodichain.utils.math import quat_from_matrix
from embodichain.utils.math import quat_from_matrix, matrix_from_quat
from embodichain.utils.string import (
is_regular_expression,
resolve_matching_names_values,
Expand Down Expand Up @@ -561,22 +561,14 @@ def compute_ik(

if pose.shape[-1] == 7 and pose.dim() == 2:
# Convert pose from (batch, 7) to (batch, 4, 4)
pose = torch.cat(
(
pose[:, :3].unsqueeze(-1), # Position
quat_from_matrix(pose[:, 3:]).unsqueeze(-1), # Quaternion
),
dim=-1,
)
pose = torch.cat(
(
pose,
torch.tensor([[0, 0, 0, 1]], device=pose.device).expand(
pose.shape[0], -1, -1
),
),
dim=1,
)
pos = pose[:, :3]
quat = pose[:, 3:]
# Convert quaternion to rotation matrix
rot = matrix_from_quat(quat)
# Build homogeneous transformation matrix efficiently
pose = torch.eye(4, device=pose.device).repeat(pose.shape[0], 1, 1)
pose[:, :3, :3] = rot
pose[:, :3, 3] = pos

base_pose = self.get_link_pose(
link_name=solver.root_link_name, env_ids=local_env_ids, to_matrix=True
Expand Down Expand Up @@ -721,23 +713,17 @@ def compute_batch_ik(

if pose.shape[-1] == 7 and pose.dim() == 3:
# Convert pose from (n_envs, n_batch, 7) to (n_envs * n_batch, 4, 4)
pose_batch = torch.reshape(-1, 7)
pose_batch = torch.cat(
(
pose_batch[:, :3].unsqueeze(-1), # Position
quat_from_matrix(pose_batch[:, 3:]).unsqueeze(-1), # Quaternion
),
dim=-1,
)
pose_batch = torch.cat(
(
pose_batch,
torch.tensor([[0, 0, 0, 1]], device=pose_batch.device).expand(
pose_batch.shape[0], -1, -1
),
),
dim=1,
pose_batch = pose.reshape(-1, 7)
pos = pose_batch[:, :3]
quat = pose_batch[:, 3:]
# Convert quaternion to rotation matrix
rot = matrix_from_quat(quat)
# Build homogeneous transformation matrix efficiently
pose_batch = torch.eye(4, device=pose.device).repeat(
pose_batch.shape[0], 1, 1
)
pose_batch[:, :3, :3] = rot
pose_batch[:, :3, 3] = pos
else:
# Convert pose from (n_envs, n_batch, 4, 4) to (n_envs * n_batch, 4, 4)
pose_batch = pose.reshape(-1, 4, 4)
Expand Down
11 changes: 11 additions & 0 deletions tests/sim/solvers/test_opw_solver.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,11 +80,22 @@ def test_ik(self, arm_name: str):
)

fk_xpos = self.robot.compute_fk(qpos=qpos_fk, name=arm_name, to_matrix=True)
fk_xpos_xyzquat = self.robot.compute_fk(
qpos=qpos_fk, name=arm_name, to_matrix=False
)

res, ik_qpos = self.robot.compute_ik(
pose=fk_xpos, joint_seed=qpos_fk, name=arm_name
)

res, ik_qpos_xyzquat = self.robot.compute_ik(
pose=fk_xpos_xyzquat, joint_seed=qpos_fk, name=arm_name
)

assert torch.allclose(
ik_qpos, ik_qpos_xyzquat, atol=1e-4, rtol=1e-4
), "IK results do not match for different pose formats"

res, ik_qpos = self.robot.compute_ik(pose=fk_xpos, name=arm_name)

if ik_qpos.dim() == 3:
Expand Down