Skip to content

[BUG FIX] Restore the carried object's pose when plan_path fails #2938

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

Open
HaozheZhang6 wants to merge 2 commits into
base: main
Choose a base branch
from
+78 −0
Open

[BUG FIX] Restore the carried object's pose when plan_path fails #2938

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ab7d961
[BUG FIX] Restore the carried object's pose when plan_path fails
HaozheZhang6 Jun 12, 2026
4af09db
Restore the carried object by its captured world pose, covering non-c…
Jun 16, 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
4 changes: 4 additions & 0 deletions genesis/utils/path_planning.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -590,6 +590,8 @@ def plan(
self._entity.set_qpos(qpos_cur, envs_idx=envs_idx, zero_velocity=False)
else:
self._entity.set_qpos(qpos_cur, zero_velocity=False)
if is_plan_with_obj:
self.update_object(ee_link_idx, obj_link_idx, _pos, _quat, envs_idx)

@chatgpt-codex-connector chatgpt-codex-connector Bot Jun 12, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Restore the original object pose for non-current starts

When callers pass a qpos_start that differs from the robot's current pose, get_exclude_geom_pairs(...) above has already moved the robot to qpos_start before _pos/_quat are captured, so this new restore applies a start-relative grasp transform after setting the robot back to qpos_cur. In that scenario a failed plan_path(with_entity=...) still teleports the attached object instead of preserving its original world pose; the same pattern was added in the RRTConnect failure branch, so the regression test only covers the qpos_start=None case.

Useful? React with 👍 / 👎.

sol = torch.zeros((num_waypoints, len(envs_idx), sol.shape[-1]), dtype=gs.tc_float, device=gs.device)
return sol, is_invalid

Expand Down Expand Up @@ -994,6 +996,8 @@ def plan(
self._entity.set_qpos(qpos_cur, envs_idx=envs_idx, zero_velocity=False)
else:
self._entity.set_qpos(qpos_cur, zero_velocity=False)
if is_plan_with_obj:
self.update_object(ee_link_idx, obj_link_idx, _pos, _quat, envs_idx)
return torch.zeros(num_waypoints, len(envs_idx), sol.shape[-1], device=gs.device), is_invalid

mask = rrt_connect_valid_mask(res_idx)
Expand Down
37 changes: 37 additions & 0 deletions tests/test_rigid_physics.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2992,6 +2992,43 @@ def test_path_planning_avoidance(backend, n_envs, show_viewer, tol):
assert_allclose(theta, 0.0, tol=5e-3)


@pytest.mark.required
@pytest.mark.parametrize("backend", [gs.cpu])
def test_plan_path_with_entity_preserves_state_on_failure(backend, tol):
# A failed plan_path(with_entity=...) must not move the attached object (#2715): the
# failure early-return restored the robot qpos but not the carried object. The goal
# below targets a pose under the floor, so the goal config is in collision and the
# single planning attempt fails deterministically regardless of the planner's RNG.
scene = gs.Scene(
sim_options=gs.options.SimOptions(dt=0.01),
rigid_options=gs.options.RigidOptions(box_box_detection=True),
show_viewer=False,
)
scene.add_entity(gs.morphs.Plane())
cube = scene.add_entity(gs.morphs.Box(size=(0.05, 0.05, 0.05), pos=(0.3, 0.1, 0.35)))
franka = scene.add_entity(gs.morphs.MJCF(file="xml/franka_emika_panda/panda.xml"))
scene.build()

hand = franka.get_link("hand")
qpos_goal = franka.inverse_kinematics(link=hand, pos=np.array([0.4, 0.0, -0.25]), quat=np.array([0, 1, 0, 0]))
qpos_goal[-2:] = 0.0

cube_pos = cube.get_pos().clone()
cube_quat = cube.get_quat().clone()
_, valid = franka.plan_path(
qpos_goal=qpos_goal,
num_waypoints=50,
max_retry=0,
max_nodes=100,
ee_link_name="hand",
with_entity=cube,
return_valid_mask=True,
)
assert not bool(valid) # the scenario under test is a failed plan
assert_allclose(cube.get_pos(), cube_pos, tol=tol)
assert_allclose(cube.get_quat(), cube_quat, tol=tol)


@pytest.mark.required
def test_all_fixed(show_viewer):
scene = gs.Scene(
Expand Down