Adds animation recording for environments (#409)

# Description

This MR adds a new button to the UI which allows you to record the
transforms of the prims into an animation file. This file can then be
opened in the simulator and played back easily.

## Type of change

- New feature (non-breaking change which adds functionality)

## Screenshot

Video Tutorial:


https://github.com/isaac-orbit/orbit/assets/12863862/55ddc2b0-5217-4e21-8276-911fce36da6b

## Checklist

- [x] I have run the [`pre-commit` checks](https://pre-commit.com/) with
`./orbit.sh --format`
- [x] I have made corresponding changes to the documentation
- [x] My changes generate no new warnings
- [ ] I have added tests that prove my fix is effective or that my
feature works
- [ ] I have run all the tests with `./orbit.sh --test` and they pass
- [x] I have updated the changelog and the corresponding version in the
extension's `config/extension.toml` file
- [x] I have added my name to the `CONTRIBUTORS.md` or my name already
exists there

.gitignore

@@ -55,3 +55,4 @@ _build
 # RL-Games
 **/runs/*
 **/logs/*
+**/recordings/*

docs/source/how-to/index.rst

@@ -19,3 +19,4 @@ use Orbit. If you are new to Orbit, we recommend you start with the tutorials.
     draw_markers
     wrap_rl_env
     master_omniverse
+    record_animation

docs/source/how-to/record_animation.rst

+Recording Animations of Simulations
+===================================
+
+.. currentmodule:: omni.isaac.orbit
+
+Omniverse includes tools to record animations of physics simulations. The `Stage Recorder`_ extension
+listens to all the motion and USD property changes within a USD stage and records them to a USD file.
+This file contains the time samples of the changes, which can be played back to render the animation.
+
+The timeSampled USD file only contains the changes to the stage. It uses the same hierarchy as the original
+stage at the time of recording. This allows adding the animation to the original stage, or to a different
+stage with the same hierarchy. The timeSampled file can be directly added as a sublayer to the original stage
+to play back the animation.
+
+.. note::
+
+  Omniverse only supports playing animation or playing physics on a USD prim at the same time. If you want to
+  play back the animation of a USD prim, you need to disable the physics simulation on the prim.
+
+
+In Orbit, we directly use the `Stage Recorder`_ extension to record the animation of the physics simulation.
+This is available as a feature in the :class:`~omni.isaac.orbit.envs.ui.BaseEnvWindow` class.
+However, to record the animation of a simulation, you need to disable `Fabric`_ to allow reading and writing
+all the changes (such as motion and USD properties) to the USD stage.
+
+
+Stage Recorder Settings
+~~~~~~~~~~~~~~~~~~~~~~~
+
+Orbit integration of the `Stage Recorder`_ extension assumes certain default settings. If you want to change the
+settings, you can directly use the `Stage Recorder`_ extension in the Omniverse Create application.
+
+.. dropdown:: Settings used in base_env_window.py
+  :icon: code
+
+  .. literalinclude:: ../../../source/extensions/omni.isaac.orbit/omni/isaac/orbit/envs/ui/base_env_window.py
+    :language: python
+    :linenos:
+    :pyobject: BaseEnvWindow._toggle_recording_animation_fn
+
+
+Example Usage
+~~~~~~~~~~~~~
+
+In all environment standalone scripts, Fabric can be disabled by passing the ``--disable_fabric`` flag to the script.
+Here we run the state-machine example and record the animation of the simulation.
+
+.. code-block:: bash
+
+  ./orbit.sh -p source/standalone/environments/state_machine/lift_cube_sm.py --num_envs 8 --cpu --disable_fabric
+
+
+On running the script, the Orbit UI window opens with the button "Record Animation" in the toolbar.
+Clicking this button starts recording the animation of the simulation. On clicking the button again, the
+recording stops. The recorded animation and the original stage (with all physics disabled) are saved
+to the ``recordings`` folder in the current working directory. The files are stored in the ``usd`` format:
+
+- ``Stage.usd``: The original stage with all physics disabled
+- ``TimeSample_tk001.usd``: The timeSampled file containing the recorded animation
+
+You can open Omniverse Create application to play back the animation. On a new stage, add the ``Stage.usd``
+as a sublayer and then add the ``TimeSample_tk001.usd`` as a sublayer. You can then play the animation by
+pressing the play button.
+
+.. _Stage Recorder: https://docs.omniverse.nvidia.com/extensions/latest/ext_animation_stage-recorder.html
+.. _Fabric: https://docs.omniverse.nvidia.com/kit/docs/usdrt/latest/docs/usd_fabric_usdrt.html

source/extensions/omni.isaac.orbit/config/extension.toml

 [package]
 
 # Note: Semantic Versioning is used: https://semver.org/
-version = "0.10.26"
+version = "0.10.27"
 
 # Description
 title = "ORBIT framework for Robot Learning"

source/extensions/omni.isaac.orbit/docs/CHANGELOG.rst

 Changelog
 ---------
 
+0.10.27 (2024-02-15)
+~~~~~~~~~~~~~~~~~~~~
+
+Added
+^^^^^
+
+* Added UI feature to start and stop animation recording in the stage when running an environment.
+  To enable this feature, please pass the argument ``--disable_fabric`` to the environment script to allow
+  USD read/write operations. Be aware that this will slow down the simulation.
+
+
 0.10.26 (2024-02-26)
 ~~~~~~~~~~~~~~~~~~~~
 

source/extensions/omni.isaac.orbit/omni/isaac/orbit/app.py

@@ -653,6 +653,8 @@ class AppLauncher:
         # note: we need to always import this even with headless to make
         #   the module for orbit.envs.ui work
         enable_extension("omni.isaac.ui")
+        # enable animation recording extension
+        enable_extension("omni.kit.stagerecorder.core")
 
         # set the nucleus directory manually to the 2023.1.0 version
         # TODO: Remove this once the 2023.1.0 version is released

source/extensions/omni.isaac.orbit/omni/isaac/orbit/envs/ui/base_env_window.py

@@ -6,12 +6,18 @@
 from __future__ import annotations
 
 import asyncio
+import os
 import weakref
+from datetime import datetime
 from typing import TYPE_CHECKING
 
 import omni.isaac.ui.ui_utils as ui_utils
+import omni.kit.app
+import omni.kit.commands
 import omni.ui
+import omni.usd
 from omni.kit.window.extensions import SimpleCheckBox
+from pxr import PhysxSchema, Sdf, Usd, UsdGeom, UsdPhysics
 
 if TYPE_CHECKING:
     from ..base_env import BaseEnv
@@ -115,6 +121,19 @@ class BaseEnvWindow:
                 }
                 self.ui_window_elements["render_dropdown"] = ui_utils.dropdown_builder(**render_mode_cfg)
 
+                # create animation recording box
+                record_animate_cfg = {
+                    "label": "Record Animation",
+                    "type": "state_button",
+                    "a_text": "START",
+                    "b_text": "STOP",
+                    "tooltip": "Record the animation of the scene. Only effective if fabric is disabled.",
+                    "on_clicked_fn": lambda value: self._toggle_recording_animation_fn(value),
+                }
+                self.ui_window_elements["record_animation"] = ui_utils.state_btn_builder(**record_animate_cfg)
+                # disable the button if fabric is not enabled
+                self.ui_window_elements["record_animation"].enabled = not self.env.sim.is_fabric_enabled()
+
     def _build_viewer_frame(self):
         """Build the viewer-related control frame for the UI."""
         # create collapsable frame for viewer
@@ -215,6 +234,85 @@ class BaseEnvWindow:
     Custom callbacks for UI elements.
     """
 
+    def _toggle_recording_animation_fn(self, value: bool):
+        """Toggles the animation recording."""
+        if value:
+            # log directory to save the recording
+            if not hasattr(self, "animation_log_dir"):
+                # create a new log directory
+                log_dir = datetime.now().strftime("%Y-%m-%d_%H-%M-%S")
+                self.animation_log_dir = os.path.join(os.getcwd(), "recordings", log_dir)
+            # start the recording
+            _ = omni.kit.commands.execute(
+                "StartRecording",
+                target_paths=[("/World", True)],
+                live_mode=True,
+                use_frame_range=False,
+                start_frame=0,
+                end_frame=0,
+                use_preroll=False,
+                preroll_frame=0,
+                record_to="FILE",
+                fps=0,
+                apply_root_anim=False,
+                increment_name=True,
+                record_folder=self.animation_log_dir,
+                take_name="TimeSample",
+            )
+        else:
+            # stop the recording
+            _ = omni.kit.commands.execute("StopRecording")
+            # save the current stage
+            stage = omni.usd.get_context().get_stage()
+            source_layer = stage.GetRootLayer()
+            # output the stage to a file
+            stage_usd_path = os.path.join(self.animation_log_dir, "Stage.usd")
+            source_prim_path = "/"
+            # creates empty anon layer
+            temp_layer = Sdf.Find(stage_usd_path)
+            if temp_layer is None:
+                temp_layer = Sdf.Layer.CreateNew(stage_usd_path)
+            temp_stage = Usd.Stage.Open(temp_layer)
+            # update stage data
+            UsdGeom.SetStageUpAxis(temp_stage, UsdGeom.GetStageUpAxis(stage))
+            UsdGeom.SetStageMetersPerUnit(temp_stage, UsdGeom.GetStageMetersPerUnit(stage))
+            # copy the prim
+            Sdf.CreatePrimInLayer(temp_layer, source_prim_path)
+            Sdf.CopySpec(source_layer, source_prim_path, temp_layer, source_prim_path)
+            # set the default prim
+            temp_layer.defaultPrim = Sdf.Path(source_prim_path).name
+            # remove all physics from the stage
+            for prim in temp_stage.TraverseAll():
+                # skip if the prim is an instance
+                if prim.IsInstanceable():
+                    continue
+                # if prim has articulation then disable it
+                if prim.HasAPI(UsdPhysics.ArticulationRootAPI):
+                    prim.RemoveAPI(UsdPhysics.ArticulationRootAPI)
+                    prim.RemoveAPI(PhysxSchema.PhysxArticulationAPI)
+                # if prim has rigid body then disable it
+                if prim.HasAPI(UsdPhysics.RigidBodyAPI):
+                    prim.RemoveAPI(UsdPhysics.RigidBodyAPI)
+                    prim.RemoveAPI(PhysxSchema.PhysxRigidBodyAPI)
+                # if prim is a joint type then disable it
+                if prim.IsA(UsdPhysics.Joint):
+                    prim.GetAttribute("physics:jointEnabled").Set(False)
+            # resolve all paths relative to layer path
+            omni.usd.resolve_paths(source_layer.identifier, temp_layer.identifier)
+            # save the stage
+            temp_layer.Save()
+            # print the path to the saved stage
+            print("Recording completed.")
+            print(f"\tSaved recorded stage to    : {stage_usd_path}")
+            print(f"\tSaved recorded animation to: {os.path.join(self.animation_log_dir, 'TimeSample_tk001.usd')}")
+            print("\nTo play the animation, check the instructions in the following link:")
+            print(
+                "\thttps://docs.omniverse.nvidia.com/extensions/latest/ext_animation_stage-recorder.html#using-the-captured-timesamples"
+            )
+            print("\n")
+            # reset the log directory
+            self.animation_log_dir = None
+
     def _set_viewer_origin_type_fn(self, value: str):
         """Sets the origin of the viewport's camera. This is based on the drop-down menu in the UI."""
         # Extract the viewport camera controller from environment

source/extensions/omni.isaac.orbit/omni/isaac/orbit/sim/simulation_context.py

@@ -229,10 +229,23 @@ class SimulationContext(_SimulationContext):
     def has_gui(self) -> bool:
         """Returns whether the simulation has a GUI enabled.
 
-        The simulation has a GUI enabled either locally or livestreamed.
+        True if the simulation has a GUI enabled either locally or live-streamed.
         """
         return self._has_gui
 
+    def is_fabric_enabled(self) -> bool:
+        """Returns whether the fabric interface is enabled.
+
+        When fabric interface is enabled, USD read/write operations are disabled. Instead all applications
+        read and write the simulation state directly from the fabric interface. This reduces a lot of overhead
+        that occurs during USD read/write operations.
+
+        For more information, please check `Fabric documentation`_.
+
+        .. _Fabric documentation: https://docs.omniverse.nvidia.com/kit/docs/usdrt/latest/docs/usd_fabric_usdrt.html
+        """
+        return self._fabric_iface is not None
+
     def get_version(self) -> tuple[int, int, int]:
         """Returns the version of the simulator.
 

source/extensions/omni.isaac.orbit_tasks/omni/isaac/orbit_tasks/utils/parse_cfg.py

@@ -13,12 +13,12 @@ import inspect
 import os
 import re
 import yaml
-from typing import Any
 
+from omni.isaac.orbit.envs import RLTaskEnvCfg
 from omni.isaac.orbit.utils import update_class_from_dict, update_dict
 
 
-def load_cfg_from_registry(task_name: str, entry_point_key: str) -> dict | Any:
+def load_cfg_from_registry(task_name: str, entry_point_key: str) -> dict | RLTaskEnvCfg:
     """Load default configuration given its entry point from the gym registry.
 
     This function loads the configuration object from the gym registry for the given task name.
@@ -97,13 +97,18 @@ def load_cfg_from_registry(task_name: str, entry_point_key: str) -> dict | Any:
     return cfg
 
 
-def parse_env_cfg(task_name: str, use_gpu: bool | None = None, num_envs: int | None = None) -> dict | Any:
+def parse_env_cfg(
+    task_name: str, use_gpu: bool | None = None, num_envs: int | None = None, use_fabric: bool | None = None
+) -> dict | RLTaskEnvCfg:
     """Parse configuration for an environment and override based on inputs.
 
     Args:
         task_name: The name of the environment.
         use_gpu: Whether to use GPU/CPU pipeline. Defaults to None, in which case it is left unchanged.
         num_envs: Number of environments to create. Defaults to None, in which case it is left unchanged.
+        use_fabric: Whether to enable/disable fabric interface. If false, all read/write operations go through USD.
+            This slows down the simulation but allows seeing the changes in the USD through the USD stage.
+            Defaults to None, in which case it is left unchanged.
 
     Returns:
         The parsed configuration object. This is either a dictionary or a class object.
@@ -127,6 +132,10 @@ def parse_env_cfg(task_name: str, use_gpu: bool | None = None, num_envs: int | N
             args_cfg["sim"]["physx"]["use_gpu"] = True
             args_cfg["sim"]["device"] = "cuda:0"
 
+    # disable fabric to read/write through USD
+    if use_fabric is not None:
+        args_cfg["sim"]["use_fabric"] = use_fabric
+
     # number of environments
     if num_envs is not None:
         args_cfg["scene"]["num_envs"] = num_envs

source/standalone/environments/random_agent.py

@@ -17,6 +17,9 @@ from omni.isaac.orbit.app import AppLauncher
 # add argparse arguments
 parser = argparse.ArgumentParser(description="Random agent for Orbit environments.")
 parser.add_argument("--cpu", action="store_true", default=False, help="Use CPU pipeline.")
+parser.add_argument(
+    "--disable_fabric", action="store_true", default=False, help="Disable fabric and use USD I/O operations."
+)
 parser.add_argument("--num_envs", type=int, default=None, help="Number of environments to simulate.")
 parser.add_argument("--task", type=str, default=None, help="Name of the task.")
 # append AppLauncher cli args
@@ -45,7 +48,9 @@ from omni.isaac.orbit_tasks.utils import parse_env_cfg
 def main():
     """Random actions agent with Orbit environment."""
     # parse configuration
-    env_cfg = parse_env_cfg(args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs)
+    env_cfg = parse_env_cfg(
+        args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs, use_fabric=not args_cli.disable_fabric
+    )
     # create environment
     env = gym.make(args_cli.task, cfg=env_cfg)
 

source/standalone/environments/state_machine/lift_cube_sm.py

@@ -24,6 +24,9 @@ from omni.isaac.orbit.app import AppLauncher
 # add argparse arguments
 parser = argparse.ArgumentParser(description="Pick and lift state machine for lift environments.")
 parser.add_argument("--cpu", action="store_true", default=False, help="Use CPU pipeline.")
+parser.add_argument(
+    "--disable_fabric", action="store_true", default=False, help="Disable fabric and use USD I/O operations."
+)
 parser.add_argument("--num_envs", type=int, default=None, help="Number of environments to simulate.")
 # append AppLauncher cli args
 AppLauncher.add_app_launcher_args(parser)
@@ -221,6 +224,7 @@ class PickAndLiftSm:
                 self.des_gripper_state_wp,
                 self.offset_wp,
             ],
+            device=self.device,
         )
 
         # convert to torch
@@ -230,7 +234,10 @@ class PickAndLiftSm:
 def main():
     # parse configuration
     env_cfg: LiftEnvCfg = parse_env_cfg(
-        "Isaac-Lift-Cube-Franka-IK-Abs-v0", use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs
+        "Isaac-Lift-Cube-Franka-IK-Abs-v0",
+        use_gpu=not args_cli.cpu,
+        num_envs=args_cli.num_envs,
+        use_fabric=not args_cli.disable_fabric,
     )
     # create environment
     env = gym.make("Isaac-Lift-Cube-Franka-IK-Abs-v0", cfg=env_cfg)

source/standalone/environments/teleoperation/teleop_se3_agent.py

@@ -17,6 +17,9 @@ from omni.isaac.orbit.app import AppLauncher
 # add argparse arguments
 parser = argparse.ArgumentParser(description="Keyboard teleoperation for Orbit environments.")
 parser.add_argument("--cpu", action="store_true", default=False, help="Use CPU pipeline.")
+parser.add_argument(
+    "--disable_fabric", action="store_true", default=False, help="Disable fabric and use USD I/O operations."
+)
 parser.add_argument("--num_envs", type=int, default=1, help="Number of environments to simulate.")
 parser.add_argument("--device", type=str, default="keyboard", help="Device for interacting with environment")
 parser.add_argument("--task", type=str, default=None, help="Name of the task.")
@@ -64,7 +67,9 @@ def pre_process_actions(delta_pose: torch.Tensor, gripper_command: bool) -> torc
 def main():
     """Running keyboard teleoperation with Orbit manipulation environment."""
     # parse configuration
-    env_cfg = parse_env_cfg(args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs)
+    env_cfg = parse_env_cfg(
+        args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs, use_fabric=not args_cli.disable_fabric
+    )
     # modify configuration
     env_cfg.terminations.time_out = None
 

source/standalone/environments/zero_agent.py

@@ -17,6 +17,9 @@ from omni.isaac.orbit.app import AppLauncher
 # add argparse arguments
 parser = argparse.ArgumentParser(description="Zero agent for Orbit environments.")
 parser.add_argument("--cpu", action="store_true", default=False, help="Use CPU pipeline.")
+parser.add_argument(
+    "--disable_fabric", action="store_true", default=False, help="Disable fabric and use USD I/O operations."
+)
 parser.add_argument("--num_envs", type=int, default=None, help="Number of environments to simulate.")
 parser.add_argument("--task", type=str, default=None, help="Name of the task.")
 # append AppLauncher cli args
@@ -44,7 +47,9 @@ from omni.isaac.orbit_tasks.utils import parse_env_cfg
 def main():
     """Zero actions agent with Orbit environment."""
     # parse configuration
-    env_cfg = parse_env_cfg(args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs)
+    env_cfg = parse_env_cfg(
+        args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs, use_fabric=not args_cli.disable_fabric
+    )
     # create environment
     env = gym.make(args_cli.task, cfg=env_cfg)
 

source/standalone/workflows/rl_games/play.py

@@ -17,6 +17,9 @@ from omni.isaac.orbit.app import AppLauncher
 # add argparse arguments
 parser = argparse.ArgumentParser(description="Play a checkpoint of an RL agent from RL-Games.")
 parser.add_argument("--cpu", action="store_true", default=False, help="Use CPU pipeline.")
+parser.add_argument(
+    "--disable_fabric", action="store_true", default=False, help="Disable fabric and use USD I/O operations."
+)
 parser.add_argument("--num_envs", type=int, default=None, help="Number of environments to simulate.")
 parser.add_argument("--task", type=str, default=None, help="Name of the task.")
 parser.add_argument("--checkpoint", type=str, default=None, help="Path to model checkpoint.")
@@ -59,7 +62,9 @@ from omni.isaac.orbit_tasks.utils.wrappers.rl_games import RlGamesGpuEnv, RlGame
 def main():
     """Play with RL-Games agent."""
     # parse env configuration
-    env_cfg = parse_env_cfg(args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs)
+    env_cfg = parse_env_cfg(
+        args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs, use_fabric=not args_cli.disable_fabric
+    )
     agent_cfg = load_cfg_from_registry(args_cli.task, "rl_games_cfg_entry_point")
 
     # wrap around environment for rl-games

source/standalone/workflows/rl_games/train.py

@@ -21,6 +21,9 @@ parser.add_argument("--video", action="store_true", default=False, help="Record
 parser.add_argument("--video_length", type=int, default=200, help="Length of the recorded video (in steps).")
 parser.add_argument("--video_interval", type=int, default=2000, help="Interval between video recordings (in steps).")
 parser.add_argument("--cpu", action="store_true", default=False, help="Use CPU pipeline.")
+parser.add_argument(
+    "--disable_fabric", action="store_true", default=False, help="Disable fabric and use USD I/O operations."
+)
 parser.add_argument("--num_envs", type=int, default=None, help="Number of environments to simulate.")
 parser.add_argument("--task", type=str, default=None, help="Name of the task.")
 parser.add_argument("--seed", type=int, default=None, help="Seed used for the environment")
@@ -67,7 +70,9 @@ def main():
     args_cli_seed = args_cli.seed
 
     # parse configuration
-    env_cfg = parse_env_cfg(args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs)
+    env_cfg = parse_env_cfg(
+        args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs, use_fabric=not args_cli.disable_fabric
+    )
     agent_cfg = load_cfg_from_registry(args_cli.task, "rl_games_cfg_entry_point")
     # override from command line
     if args_cli_seed is not None:

source/standalone/workflows/robomimic/play.py

@@ -17,6 +17,9 @@ from omni.isaac.orbit.app import AppLauncher
 # add argparse arguments
 parser = argparse.ArgumentParser(description="Play policy trained using robomimic for Orbit environments.")
 parser.add_argument("--cpu", action="store_true", default=False, help="Use CPU pipeline.")
+parser.add_argument(
+    "--disable_fabric", action="store_true", default=False, help="Disable fabric and use USD I/O operations."
+)
 parser.add_argument("--task", type=str, default=None, help="Name of the task.")
 parser.add_argument("--checkpoint", type=str, default=None, help="Pytorch model checkpoint to load.")
 # append AppLauncher cli args
@@ -48,7 +51,7 @@ from omni.isaac.orbit_tasks.utils import parse_env_cfg
 def main():
     """Run a trained policy from robomimic with Orbit environment."""
     # parse configuration
-    env_cfg = parse_env_cfg(args_cli.task, use_gpu=not args_cli.cpu, num_envs=1)
+    env_cfg = parse_env_cfg(args_cli.task, use_gpu=not args_cli.cpu, num_envs=1, use_fabric=not args_cli.disable_fabric)
     # modify configuration
     env_cfg.control.control_type = "inverse_kinematics"
     env_cfg.control.inverse_kinematics.command_type = "pose_rel"

source/standalone/workflows/rsl_rl/play.py

@@ -20,6 +20,9 @@ import cli_args  # isort: skip
 # add argparse arguments
 parser = argparse.ArgumentParser(description="Train an RL agent with RSL-RL.")
 parser.add_argument("--cpu", action="store_true", default=False, help="Use CPU pipeline.")
+parser.add_argument(
+    "--disable_fabric", action="store_true", default=False, help="Disable fabric and use USD I/O operations."
+)
 parser.add_argument("--num_envs", type=int, default=None, help="Number of environments to simulate.")
 parser.add_argument("--task", type=str, default=None, help="Name of the task.")
 parser.add_argument("--seed", type=int, default=None, help="Seed used for the environment")
@@ -57,7 +60,9 @@ from omni.isaac.orbit_tasks.utils.wrappers.rsl_rl import (
 def main():
     """Play with RSL-RL agent."""
     # parse configuration
-    env_cfg = parse_env_cfg(args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs)
+    env_cfg = parse_env_cfg(
+        args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs, use_fabric=not args_cli.disable_fabric
+    )
     agent_cfg: RslRlOnPolicyRunnerCfg = cli_args.parse_rsl_rl_cfg(args_cli.task, args_cli)
 
     # create isaac environment

source/standalone/workflows/rsl_rl/train.py

@@ -25,6 +25,9 @@ parser.add_argument("--video", action="store_true", default=False, help="Record
 parser.add_argument("--video_length", type=int, default=200, help="Length of the recorded video (in steps).")
 parser.add_argument("--video_interval", type=int, default=2000, help="Interval between video recordings (in steps).")
 parser.add_argument("--cpu", action="store_true", default=False, help="Use CPU pipeline.")
+parser.add_argument(
+    "--disable_fabric", action="store_true", default=False, help="Disable fabric and use USD I/O operations."
+)
 parser.add_argument("--num_envs", type=int, default=None, help="Number of environments to simulate.")
 parser.add_argument("--task", type=str, default=None, help="Name of the task.")
 parser.add_argument("--seed", type=int, default=None, help="Seed used for the environment")
@@ -74,7 +77,9 @@ torch.backends.cudnn.benchmark = False
 def main():
     """Train with RSL-RL agent."""
     # parse configuration
-    env_cfg: RLTaskEnvCfg = parse_env_cfg(args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs)
+    env_cfg: RLTaskEnvCfg = parse_env_cfg(
+        args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs, use_fabric=not args_cli.disable_fabric
+    )
     agent_cfg: RslRlOnPolicyRunnerCfg = cli_args.parse_rsl_rl_cfg(args_cli.task, args_cli)
 
     # specify directory for logging experiments

source/standalone/workflows/sb3/play.py

@@ -18,6 +18,9 @@ from omni.isaac.orbit.app import AppLauncher
 # add argparse arguments
 parser = argparse.ArgumentParser(description="Play a checkpoint of an RL agent from Stable-Baselines3.")
 parser.add_argument("--cpu", action="store_true", default=False, help="Use CPU pipeline.")
+parser.add_argument(
+    "--disable_fabric", action="store_true", default=False, help="Disable fabric and use USD I/O operations."
+)
 parser.add_argument("--num_envs", type=int, default=None, help="Number of environments to simulate.")
 parser.add_argument("--task", type=str, default=None, help="Name of the task.")
 parser.add_argument("--checkpoint", type=str, default=None, help="Path to model checkpoint.")
@@ -56,7 +59,9 @@ from omni.isaac.orbit_tasks.utils.wrappers.sb3 import Sb3VecEnvWrapper, process_
 def main():
     """Play with stable-baselines agent."""
     # parse configuration
-    env_cfg = parse_env_cfg(args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs)
+    env_cfg = parse_env_cfg(
+        args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs, use_fabric=not args_cli.disable_fabric
+    )
     agent_cfg = load_cfg_from_registry(args_cli.task, "sb3_cfg_entry_point")
     # post-process agent configuration
     agent_cfg = process_sb3_cfg(agent_cfg)

source/standalone/workflows/sb3/train.py

@@ -27,6 +27,9 @@ parser.add_argument("--video", action="store_true", default=False, help="Record
 parser.add_argument("--video_length", type=int, default=200, help="Length of the recorded video (in steps).")
 parser.add_argument("--video_interval", type=int, default=2000, help="Interval between video recordings (in steps).")
 parser.add_argument("--cpu", action="store_true", default=False, help="Use CPU pipeline.")
+parser.add_argument(
+    "--disable_fabric", action="store_true", default=False, help="Disable fabric and use USD I/O operations."
+)
 parser.add_argument("--num_envs", type=int, default=None, help="Number of environments to simulate.")
 parser.add_argument("--task", type=str, default=None, help="Name of the task.")
 parser.add_argument("--seed", type=int, default=None, help="Seed used for the environment")
@@ -72,7 +75,9 @@ from omni.isaac.orbit_tasks.utils.wrappers.sb3 import Sb3VecEnvWrapper, process_
 def main():
     """Train with stable-baselines agent."""
     # parse configuration
-    env_cfg = parse_env_cfg(args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs)
+    env_cfg = parse_env_cfg(
+        args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs, use_fabric=not args_cli.disable_fabric
+    )
     agent_cfg = load_cfg_from_registry(args_cli.task, "sb3_cfg_entry_point")
 
     # override configuration with command line arguments

source/standalone/workflows/skrl/play.py

@@ -23,6 +23,9 @@ from omni.isaac.orbit.app import AppLauncher
 # add argparse arguments
 parser = argparse.ArgumentParser(description="Play a checkpoint of an RL agent from skrl.")
 parser.add_argument("--cpu", action="store_true", default=False, help="Use CPU pipeline.")
+parser.add_argument(
+    "--disable_fabric", action="store_true", default=False, help="Disable fabric and use USD I/O operations."
+)
 parser.add_argument("--num_envs", type=int, default=None, help="Number of environments to simulate.")
 parser.add_argument("--task", type=str, default=None, help="Name of the task.")
 parser.add_argument("--checkpoint", type=str, default=None, help="Path to model checkpoint.")
@@ -53,7 +56,9 @@ from omni.isaac.orbit_tasks.utils.wrappers.skrl import SkrlVecEnvWrapper, proces
 def main():
     """Play with skrl agent."""
     # parse env configuration
-    env_cfg = parse_env_cfg(args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs)
+    env_cfg = parse_env_cfg(
+        args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs, use_fabric=not args_cli.disable_fabric
+    )
     experiment_cfg = load_cfg_from_registry(args_cli.task, "skrl_cfg_entry_point")
 
     # create isaac environment

source/standalone/workflows/skrl/train.py

@@ -26,6 +26,9 @@ parser.add_argument("--video", action="store_true", default=False, help="Record
 parser.add_argument("--video_length", type=int, default=200, help="Length of the recorded video (in steps).")
 parser.add_argument("--video_interval", type=int, default=2000, help="Interval between video recordings (in steps).")
 parser.add_argument("--cpu", action="store_true", default=False, help="Use CPU pipeline.")
+parser.add_argument(
+    "--disable_fabric", action="store_true", default=False, help="Disable fabric and use USD I/O operations."
+)
 parser.add_argument("--num_envs", type=int, default=None, help="Number of environments to simulate.")
 parser.add_argument("--task", type=str, default=None, help="Name of the task.")
 parser.add_argument("--seed", type=int, default=None, help="Seed used for the environment")
@@ -70,7 +73,9 @@ def main():
     args_cli_seed = args_cli.seed
 
     # parse configuration
-    env_cfg = parse_env_cfg(args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs)
+    env_cfg = parse_env_cfg(
+        args_cli.task, use_gpu=not args_cli.cpu, num_envs=args_cli.num_envs, use_fabric=not args_cli.disable_fabric
+    )
     experiment_cfg = load_cfg_from_registry(args_cli.task, "skrl_cfg_entry_point")
 
     # specify directory for logging experiments