Replaces Isaac Sim SimulationManager with IsaacLab SimulationManager by ooctipus · Pull Request #4475 · isaac-sim/IsaacLab

ooctipus · 2026-01-27T18:26:05Z

Description

Adds simulation manager that replaces Isaac sim simulation manager,
this is currently a draft, things are mostly working, but lacks clean up.

AntoineRichard

Intermediary review. I feel the code is a bit too convoluted at times. Let's sync on that.

AntoineRichard · 2026-02-03T15:15:41Z

apps/isaaclab.python.kit

Should we remove the isaacsim_4_5 apps or do we plan on supporting those still?

AntoineRichard · 2026-02-03T15:16:05Z

apps/isaacsim_4_5/isaaclab.python.kit

Should we remove the isaacsim_4_5 apps or do we plan on supporting those still?

AntoineRichard · 2026-02-03T15:16:13Z

apps/isaacsim_4_5/isaaclab.python.kit

 "isaacsim.core.cloner" = {}
 "isaacsim.core.nodes" = {}
-"isaacsim.core.simulation_manager" = {}
+# "isaacsim.core.simulation_manager" = {}  # Replaced by isaaclab.sim.simulation_manager


This can just be removed?

AntoineRichard · 2026-02-03T15:16:23Z

apps/isaaclab.python.kit

 "isaacsim.core.cloner" = {}
 "isaacsim.core.nodes" = {}
-"isaacsim.core.simulation_manager" = {}
+# "isaacsim.core.simulation_manager" = {}  # Replaced by isaaclab.sim.simulation_manager


This could be removed directly?

AntoineRichard · 2026-02-03T15:33:05Z

source/isaaclab/isaaclab/assets/asset_base.py

        )
        # register prim deletion callback
-        self._prim_deletion_callback_id = SimulationManager.register_callback(
+        self._prim_deletion_callback_id = PhysxManager.register_callback(


I think the callbacks should live in the simulation context. It's something that we could take a look at in an other PR.

AntoineRichard · 2026-02-03T16:25:05Z

source/isaaclab/isaaclab/physics/newton_manager.py

Let's not have that here. I understand that it was used for testing, but let's keep things pure PhysX for now.

AntoineRichard · 2026-02-03T16:25:14Z

source/isaaclab/isaaclab/physics/newton_manager_cfg.py

Let's not have that here

AntoineRichard · 2026-02-03T16:26:14Z

source/isaaclab/isaaclab/physics/__init__.py

+# from .newton_manager import NewtonManager
+# from .newton_manager_cfg import NewtonManagerCfg


Let's delete.

AntoineRichard · 2026-02-03T16:26:20Z

source/isaaclab/isaaclab/physics/__init__.py

+    # "NewtonManager",
+    # "NewtonManagerCfg",


AntoineRichard · 2026-02-03T16:26:44Z

source/isaaclab/isaaclab/visualizers/__init__.py

In general let's try to have only things that we can use here.

kellyguo11 · 2026-02-04T01:37:06Z

scripts/environments/state_machine/open_cabinet_sm.py

    desired_orientation[:, 1] = 1.0
    # create state machine
-    open_sm = OpenDrawerSm(env_cfg.sim.dt * env_cfg.decimation, env.unwrapped.num_envs, env.unwrapped.device)
+    open_sm = OpenDrawerSm(env_cfg.sim.physics_manager_cfg.dt * env_cfg.decimation, env.unwrapped.num_envs, env.unwrapped.device)


I think I like retrieving the attributes from env_cfg.sim.dt directly a bit more - would it make sense to still keep some common attributes in the SimulationCfg? probably also less breaking changes for user scripts

kellyguo11 · 2026-02-04T01:39:01Z

source/isaaclab/isaaclab/physics/newton_manager.py

+#
+# SPDX-License-Identifier: BSD-3-Clause
+
+"""Newton physics manager for Isaac Lab."""


is this being added here to also refactor it to match the new PhysicsManager interface? I'm not sure if we anticipate more changes for this in the dev/newton branch before merging things in.

kellyguo11 · 2026-02-04T01:39:46Z

source/isaaclab/isaaclab/physics/physx_manager.py

+#
+# SPDX-License-Identifier: BSD-3-Clause
+
+"""PhysX Manager for Isaac Lab.


I think this would likely be the OV manager, since ovphysx will probably have a different set of APIs

kellyguo11 · 2026-02-04T01:43:23Z

source/isaaclab/isaaclab/sim/interface/interface.py

+    from .simulation_context import SimulationContext
+
+
+class Interface(ABC):


this does feel maybe a bit too abstracted, perhaps we can just combine the interface idea into the SimulationContext?

we can probably go with the approach we talked about today where we design the classes with interface in mind, but flatten them out when we do the implementation.

kellyguo11 · 2026-02-04T01:43:50Z

source/isaaclab/isaaclab/sim/interface/visualizer_interface.py

+#
+# SPDX-License-Identifier: BSD-3-Clause
+
+"""Visualizer interface for SimulationContext."""


would be good for @matthewtrepte to provide feedback for the visualizers

kellyguo11 · 2026-02-04T01:44:52Z

source/isaaclab/isaaclab/sim/utils/__init__.py

 from .legacy import *  # noqa: F401, F403
 from .prims import *  # noqa: F401, F403
 from .queries import *  # noqa: F401, F403
 from .semantics import *  # noqa: F401, F403
 from .stage import *  # noqa: F401, F403
 from .transforms import *  # noqa: F401, F403
+
+
+def raise_callback_exception_if_any() -> None:


I wonder if there's a better way to handle the exceptions if we follow our own callback system? I think this was kind of a hacky way we had to do to force exceptions to be propagated from the OV callbacks

kellyguo11 · 2026-02-04T01:46:37Z

source/isaaclab/isaaclab/sim/__init__.py

+# Monkey-patch Isaac Sim's PhysxManager to use Isaac Lab's implementation
+# This ensures all code (including Isaac Sim internals) uses our manager
+try:
+    import isaacsim.core.simulation_manager as _isaacsim_sim_manager


hmm...this feels a bit scary haha, is there any way we can avoid referencing Isaac Sim's SimulationManager all together?

I think we want to avoid cases where there is isaacsim manager and isaaclab manager, since both are singeton, just like we point isaaclab stage as isaacsim stage x)

kellyguo11 · 2026-02-04T01:48:01Z

source/isaaclab/isaaclab/sim/simulation_context.py

        raise
    finally:
-        if not sim.has_gui():
-            # Stop simulation only if we aren't rendering otherwise the app will hang indefinitely
+        if not sim.get_setting("/isaaclab/has_gui"):


is the has_gui flag important for non-OV workflows as well?

kellyguo11 · 2026-02-04T01:48:44Z

source/isaaclab/isaaclab/visualizers/newton_visualizer.py

+#
+# SPDX-License-Identifier: BSD-3-Clause
+
+"""Newton OpenGL Visualizer implementation."""


let's sync with @matthewtrepte on the visualizer classes since he's been looking at integrating these with the PhysX backend.

we can probably also bring things in piece by piece so we can keep this PR somewhat smaller

…er_cfg

…e together

Mayankm96 added 30 commits December 30, 2025 11:14

adds initial code and implementation for xform class

6fd9082

wraps code inside sdf block for optimization

dcc4737

runs formatter

fcf6c1a

updates doc strings

f54ae43

fixes type-hinting

2f9afad

removes isaac sim xform class from create_prim

c1413d4

moves functions for pose to transforms module

4652ac6

adds api docs

a10e419

makes a function to abstract out convert world pose call

a1a714d

runs formatter

d347aef

adds test for different backend data types to create prim

9929597

adds test for get usd references

263054d

adds error code

53a8ddb

adds example for all transform calls

590ff75

runs formatter

8e5153f

fixes type issues and adds more tests for to_tuple

17d6936

Merge branch 'main' into fix/clear-xform

836d0c6

adds test for flaot precision in standardize

62361ac

removes benchmark test

113ac86

makes a habit of sending stage

b33eb9a

use stage attributes

c7f6b23

optimization for root parent

b406521

adds test to validate xform ops order

dc62696

adds benchmark script

083a2c3

moves xform prim view into views module

366f338

fixes type error

8c5e7c1

adds test for xform prim view

618e2f1

adds clear op

bd98dda

adds franka test

56cf54d

fixes check for non-xform xformable

cbbccdf

ooctipus added 4 commits February 3, 2026 05:57

refactor step 15: refactored physics to lab.physics

5315495

refactor step 16: refactored physics_manager

6be6ac8

refactor step 17: refactored simulation_cfg

dc20d8e

refactor step 18: refactored newton manager

4b1d3ae

ooctipus requested a review from xyao-nv as a code owner February 3, 2026 15:13

github-actions bot added the isaac-mimic Related to Isaac Mimic team label Feb 3, 2026

ooctipus added 4 commits February 3, 2026 07:13

added in dummy refactored visualizers to show structure

1a11880

refactor step 19: reorganized interface

07c0573

refactor step 20: gets simulation_context tests to pass

81c91d1

made simulation manager a lot nicer

1538f48

AntoineRichard requested changes Feb 3, 2026

View reviewed changes

ooctipus marked this pull request as draft February 3, 2026 17:06

ooctipus added 3 commits February 3, 2026 09:59

synchronized newton visualizer

aaa3d8b

refactore step 21: ensure simulation context consistency

8a11590

passing simulation context test

deb84dc

kellyguo11 reviewed Feb 4, 2026

View reviewed changes

ooctipus added 4 commits February 4, 2026 12:57

refactore step 22: moved physics_material from sim_cfg to physx_manag…

a47719a

…er_cfg

refactor step 23: refactor env_cfg simulation back style

8326029

refactor step 24: simplifies physx manager

89204d0

refactor step 25: flatten physics and visualizer interface

2fa5656

ooctipus force-pushed the feature/simulation-manager branch from 28009e4 to 2fa5656 Compare February 5, 2026 05:30

ooctipus added 9 commits February 5, 2026 06:58

small bug fix

48017c0

refactor step 26: come up with callback system to bring physics engin…

106f09c

…e together

refactor step 27: clean up physics manager interfaces

8e98259

clean up

93b2a80

fixed bug in physx_ov_visualizer

efb1e4f

refactor step 28: remove callback exception gathering

1de77d3

remove get version from simulation_context

df2b39b

pass simulation context test

751b5a0

bug fixes

a450f17

		# from .newton_manager import NewtonManager
		# from .newton_manager_cfg import NewtonManagerCfg

		from .simulation_context import SimulationContext


		class Interface(ABC):

Conversation

ooctipus commented Jan 27, 2026

Description

Uh oh!

AntoineRichard left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants