[Task Clean-up][Manager] Dexterous Part 6/10: Add the handover and camera manager counterparts

docs/source/overview/environments.rst

@@ -1152,11 +1152,15 @@ including disabling runtime perturbations used for training.
     * - Isaac-Reorient-Cube-Allegro
       - Manager Based
       - **rl_games** (PPO), **rsl_rl** (PPO), **skrl** (PPO)
-      -
+      - **physics=** ``isaacsim_physx``, ``newton_mjwarp``, ``ovphysx``
     * - Isaac-Reorient-Cube-Allegro-Direct
       - Direct
       - **rl_games** (PPO), **rsl_rl** (PPO), **skrl** (PPO)
       - **physics=** ``isaacsim_physx``, ``newton_mjwarp``, ``ovphysx``
+    * - Isaac-Reorient-Cube-Shadow
+      - Manager Based
+      - **rl_games** (PPO), **rsl_rl** (PPO), **skrl** (PPO)
+      - **physics=** ``isaacsim_physx``, ``newton_kamino``, ``newton_mjwarp``, ``ovphysx``
     * - Isaac-Reorient-Cube-Shadow-Camera-Direct
       - Direct
       - **rl_games** (PPO), **rsl_rl** (PPO)
@@ -1167,10 +1171,18 @@ including disabling runtime perturbations used for training.
       - Direct
       - **rl_games** (PPO), **rsl_rl** (PPO), **skrl** (PPO)
       - **physics=** ``isaacsim_physx``, ``newton_kamino``, ``newton_mjwarp``, ``ovphysx``
+    * - Isaac-Reorient-Cube-Shadow-OpenAI-FF
+      - Manager Based
+      - **rl_games** (PPO), **rsl_rl** (PPO), **skrl** (PPO)
+      - **physics=** ``isaacsim_physx``, ``newton_kamino``, ``newton_mjwarp``, ``ovphysx``
     * - Isaac-Reorient-Cube-Shadow-OpenAI-FF-Direct
       - Direct
       - **rl_games** (PPO), **rsl_rl** (PPO), **skrl** (PPO)
       - **physics=** ``isaacsim_physx``, ``newton_kamino``, ``newton_mjwarp``, ``ovphysx``
+    * - Isaac-Reorient-Cube-Shadow-OpenAI-LSTM
+      - Manager Based
+      - **rl_games** (PPO), **rsl_rl** (PPO)
+      - **physics=** ``isaacsim_physx``, ``newton_kamino``, ``newton_mjwarp``, ``ovphysx``
     * - Isaac-Reorient-Cube-Shadow-OpenAI-LSTM-Direct
       - Direct
       - **rl_games** (PPO), **rsl_rl** (PPO)

source/isaaclab/changelog.d/task-cleanup-dex-part08.rst

@@ -0,0 +1,6 @@
+Fixed
+^^^^^
+
+* Fixed :attr:`~isaaclab.envs.ManagerBasedRLEnv.reset_buf` not existing until the first
+  call to :meth:`~isaaclab.envs.ManagerBasedRLEnv.step`, so manager terms that run during
+  the initial reset could not read it.

source/isaaclab/isaaclab/envs/manager_based_rl_env.py

@@ -75,6 +75,7 @@ def __init__(self, cfg: ManagerBasedRLEnvCfg, render_mode: str | None = None, **
 
         # initialize the episode length buffer BEFORE loading the managers to use it in mdp functions.
         self.episode_length_buf = torch.zeros(cfg.scene.num_envs, device=cfg.sim.device, dtype=torch.long)
+        self.reset_buf = torch.zeros(cfg.scene.num_envs, device=cfg.sim.device, dtype=torch.bool)
 
         # initialize the base class to setup the scene.
         super().__init__(cfg=cfg)

source/isaaclab_assets/changelog.d/task-cleanup-dex-part11.minor.rst

@@ -0,0 +1,7 @@
+Added
+^^^^^
+
+* Added :obj:`~isaaclab_assets.robots.shadow_hand.SHADOW_ACTUATED_JOINT_NAMES_NEWTON`.
+  The Newton Shadow Hand asset numbers its finger joints one higher than the
+  PhysX one, so terms that address joints by name need the Newton ordering
+  rather than :obj:`~isaaclab_assets.robots.shadow_hand.SHADOW_ACTUATED_JOINT_NAMES`.

source/isaaclab_assets/isaaclab_assets/robots/shadow_hand.py

@@ -169,9 +169,10 @@
 )
 """Configuration of the Shadow Hand robot on the Newton (MJWarp) asset.
 
-The Newton USD renumbers the finger joints (+1) relative to :obj:`SHADOW_HAND_CFG`, but the
-names in :obj:`SHADOW_ACTUATED_JOINT_NAMES` resolve on both assets, so the two backends share
-one actuated-joint list. Gains default to the PhysX values; tasks override them as needed.
+The Newton USD renumbers the finger joints (+1) relative to :obj:`SHADOW_HAND_CFG`. Actuator
+regexes resolve on both assets, so the two backends share one actuator configuration, but terms
+that need the joints in order use :obj:`SHADOW_ACTUATED_JOINT_NAMES_NEWTON` on this asset.
+Gains default to the PhysX values; tasks override them as needed.
 """
 
 
@@ -210,3 +211,31 @@
 
 These names resolve on both the PhysX and Newton assets, so every backend shares this list.
 """
+
+SHADOW_ACTUATED_JOINT_NAMES_NEWTON: list[str] = [
+    "robot0_WRJ1",
+    "robot0_WRJ0",
+    "robot0_FFJ4",
+    "robot0_FFJ3",
+    "robot0_FFJ2",
+    "robot0_MFJ4",
+    "robot0_MFJ3",
+    "robot0_MFJ2",
+    "robot0_RFJ4",
+    "robot0_RFJ3",
+    "robot0_RFJ2",
+    "robot0_LFJ5",
+    "robot0_LFJ4",
+    "robot0_LFJ3",
+    "robot0_LFJ2",
+    "robot0_THJ4",
+    "robot0_THJ3",
+    "robot0_THJ2",
+    "robot0_THJ1",
+    "robot0_THJ0",
+]
+"""Shadow Hand actuated joint names on the Newton asset, in the same order.
+
+The Newton USD numbers the finger joints one higher than :obj:`SHADOW_ACTUATED_JOINT_NAMES`,
+so action terms that address joints by name need this list rather than the PhysX one.
+"""

source/isaaclab_tasks/changelog.d/task-cleanup-dex-part08.major.rst

@@ -0,0 +1,68 @@
+Added
+^^^^^
+
+* Added manager-based counterparts for the Shadow cube reorientation task and its
+  OpenAI FF/LSTM variants, alongside the existing Allegro manager task.
+* Added :class:`~isaaclab_tasks.core.reorient.mdp.reorient_timeout`, which restarts
+  the episode timer on every goal reach so OpenAI-variant episodes extend across
+  success streaks.
+* Added ``enable_domain_randomization`` to the Allegro and Shadow manager
+  environments.
+* Added Newton and OvPhysX presets to the manager-based reorientation environments,
+  selectable with ``physics=``.
+* Added a Direct-versus-manager value-parity test covering timing, success tolerance,
+  fall distance, and the consecutive-success cap.
+
+Changed
+^^^^^^^
+
+* **Breaking:** Changed ``Metrics/success_rate`` on the manager-based reorientation
+  tasks to a per-episode success bit drawn at
+  ``ReorientCommandCfg.success_count_threshold``, matching the Direct environments,
+  instead of a per-attempt ratio. Curves from earlier runs are not comparable.
+* **Breaking:** Changed domain randomization to default off on the Allegro and Shadow
+  manager tasks and on for the OpenAI variants, matching each task's Direct
+  counterpart. Set ``enable_domain_randomization`` to restore it.
+* **Breaking:** Changed the manager-based Allegro environment to match the Direct
+  observation, action, reset, and termination contracts, and to use the same agent
+  configurations. The observation space changes size, so existing manager checkpoints
+  must be retrained. ``rl_games_manager_ppo_cfg.yaml``, ``skrl_manager_ppo_cfg.yaml``
+  and ``AllegroCubePPORunnerCfg`` are gone; use ``rl_games_ppo_cfg.yaml``,
+  ``skrl_ppo_cfg.yaml`` and ``AllegroHandPPORunnerCfg``.
+* **Breaking:** Moved the Shadow Hand camera benchmark task to the contributed tasks
+  as ``IsaacContrib-Reorient-Cube-Shadow-Camera-Benchmark-Direct``. The released
+  ``Isaac-Reorient-Cube-Shadow-Camera-Benchmark-Direct`` identifier no longer
+  resolves; use the contributed one.
+* **Breaking:** Replaced the per-backend in-hand cube preset with one configuration
+  per hand. Newton spawned a 54 mm articulation at ``z=0.535``; both backends now
+  share the PhysX rigid body, a 60 mm cube at ``z=0.6``. Newton checkpoints must be
+  retrained, and code resolving the object through
+  :class:`~isaaclab.assets.Articulation` must use :class:`~isaaclab.assets.RigidObject`.
+* Renamed the per-robot scene constants to name what they hold: ``ROBOT_CFG`` becomes
+  ``SHADOW_HAND_ROBOT_CFG`` or ``ALLEGRO_HAND_ROBOT_CFG``, and ``OBJECT_CFG`` becomes
+  ``CUBE_CFG``.
+* Renamed the manager-based Allegro configurations after the robot rather than the object,
+  matching the Shadow counterparts: ``AllegroCubeEnvCfg`` and ``AllegroCubeSceneCfg`` become
+  ``AllegroHandManagerEnvCfg`` and ``AllegroHandManagerSceneCfg``. The
+  ``Isaac-Reorient-Cube-Allegro`` task identifier is unchanged.
+
+Removed
+^^^^^^^
+
+* Removed ``ReorientObjectEnvCfg`` and the shared reorientation observation, action,
+  and command configurations. Each manager task declares its own; derive from
+  :class:`~isaaclab.envs.ManagerBasedRLEnvCfg` directly.
+* Removed ``reorient_common``. Its constants are declared by the tasks that use them,
+  and the in-hand offset and goal-marker position are per-robot fields on the Direct
+  configurations.
+* Removed the handover ``EventCfg``, which was never wired into
+  ``HandoverEnvCfg.events``.
+* Removed ``isaaclab_tasks.core.utils``. Its helpers moved to
+  ``isaaclab_tasks.core.reorient.utils``, which the hand-over task imports.
+* Removed the ``clone_in_fabric`` settings from the reorientation scenes. The flag no
+  longer reaches the replicator, so the value had no effect.
+
+Fixed
+^^^^^
+
+* Fixed the manager-based reorientation tasks not reporting ``Metrics/success_rate``.

source/isaaclab_tasks/changelog.d/task-cleanup-dex-part11.minor.rst

@@ -0,0 +1,36 @@
+Added
+^^^^^
+
+* Added manager-based counterparts for the Shadow handover and Shadow camera
+  reorientation tasks, completing the manager coverage of the dexterous task
+  families.
+* Added a Direct-versus-manager value-parity check for the handover task,
+  alongside the reorientation one.
+
+Changed
+^^^^^^^
+
+* Changed the manager-based Shadow camera task to run on the Kit PhysX backend by
+  default, since only the Isaac RTX tiled camera renders the default modalities.
+  Select Newton with ``physics=newton_mjwarp`` for the state-only observation
+  groups.
+* Changed the handover reward to a plain reward term, moving success and
+  goal-distance bookkeeping to
+  :class:`~isaaclab_tasks.core.handover.mdp.commands.HandoverCommand`.
+* Changed the reorientation action configuration to name its term through a
+  module path, so loading a task configuration no longer imports the USD
+  bindings.
+
+Removed
+^^^^^^^
+
+* Removed the ``Isaac-Reorient-Cube-Shadow-Camera-Play`` and
+  ``Isaac-Reorient-Cube-Shadow-Camera-Direct-Play`` tasks. Use the training task
+  with ``--play`` instead; playback settings now live in
+  :meth:`~isaaclab.envs.ManagerBasedRLEnvCfg.play_mode`.
+
+Fixed
+^^^^^
+
+* Fixed the Shadow camera feature-extractor observation term ignoring its
+  declared ``feature_extractor_cfg`` parameter.

source/isaaclab_tasks/isaaclab_tasks/contrib/reorient/__init__.py

@@ -0,0 +1,6 @@
+# Copyright (c) 2022-2026, The Isaac Lab Project Developers (https://github.com/isaac-sim/IsaacLab/blob/main/CONTRIBUTORS.md).
+# All rights reserved.
+#
+# SPDX-License-Identifier: BSD-3-Clause
+
+"""Contributed variants of the in-hand reorientation tasks."""

source/isaaclab_tasks/isaaclab_tasks/contrib/reorient/config/__init__.py

@@ -0,0 +1,6 @@
+# Copyright (c) 2022-2026, The Isaac Lab Project Developers (https://github.com/isaac-sim/IsaacLab/blob/main/CONTRIBUTORS.md).
+# All rights reserved.
+#
+# SPDX-License-Identifier: BSD-3-Clause
+
+"""Configurations for the contributed reorientation environments."""

source/isaaclab_tasks/isaaclab_tasks/contrib/reorient/config/shadow_hand/__init__.py

@@ -0,0 +1,21 @@
+# Copyright (c) 2022-2026, The Isaac Lab Project Developers (https://github.com/isaac-sim/IsaacLab/blob/main/CONTRIBUTORS.md).
+# All rights reserved.
+#
+# SPDX-License-Identifier: BSD-3-Clause
+
+"""Shadow Hand rendering-throughput benchmark task."""
+
+import gymnasium as gym
+
+from isaaclab_tasks.core.reorient.config.shadow_hand import agents
+
+gym.register(
+    id="IsaacContrib-Reorient-Cube-Shadow-Camera-Benchmark-Direct",
+    entry_point="isaaclab_tasks.core.reorient.config.shadow_hand.shadow_hand_direct_camera_env:ShadowHandCameraEnv",
+    disable_env_checker=True,
+    kwargs={
+        "env_cfg_entry_point": f"{__name__}.shadow_hand_camera_benchmark_env_cfg:ShadowHandCameraBenchmarkEnvCfg",
+        "rsl_rl_cfg_entry_point": f"{agents.__name__}.rsl_rl_ppo_cfg:ShadowHandCameraFFPPORunnerCfg",
+        "rl_games_cfg_entry_point": f"{agents.__name__}:rl_games_ppo_camera_cfg.yaml",
+    },
+)

source/isaaclab_tasks/isaaclab_tasks/contrib/reorient/config/shadow_hand/shadow_hand_camera_benchmark_env_cfg.py

@@ -0,0 +1,30 @@
+# Copyright (c) 2022-2026, The Isaac Lab Project Developers (https://github.com/isaac-sim/IsaacLab/blob/main/CONTRIBUTORS.md).
+# All rights reserved.
+#
+# SPDX-License-Identifier: BSD-3-Clause
+
+"""Rendering-throughput benchmark variant of the Shadow Hand camera task."""
+
+from isaaclab.utils.configclass import configclass
+
+from isaaclab_tasks.core.reorient.config.shadow_hand.feature_extractor import FeatureExtractorCfg
+from isaaclab_tasks.core.reorient.config.shadow_hand.shadow_hand_direct_camera_env_cfg import ShadowHandCameraEnvCfg
+
+
+@configclass
+class ShadowHandCameraBenchmarkEnvCfg(ShadowHandCameraEnvCfg):
+    """Benchmark configuration with the feature extractor CNN disabled.
+
+    The tiled camera renders frames each step as normal, but the CNN forward pass is
+    bypassed — zero embeddings are returned instead. This isolates rendering throughput
+    from CNN inference overhead when profiling.
+
+    The renderer backend and camera data types can still be selected via ``presets``::
+
+        presets = newton_renderer  # benchmark with Newton renderer
+        presets = ovrtx  # benchmark with OVRTX renderer
+        presets = rgb  # benchmark RGB rendering only
+        presets = depth, newton_renderer  # benchmark depth rendering with Newton
+    """
+
+    feature_extractor: FeatureExtractorCfg = FeatureExtractorCfg(enabled=False)

source/isaaclab_tasks/isaaclab_tasks/core/handover/__init__.py

@@ -15,6 +15,16 @@
 # Register Gym environments.
 ##
 
+gym.register(
+    id="Isaac-Shadow-Handover",
+    entry_point="isaaclab.envs:ManagerBasedRLEnv",
+    disable_env_checker=True,
+    kwargs={
+        "env_cfg_entry_point": f"{__name__}.handover_manager_env_cfg:HandoverManagerEnvCfg",
+        "rsl_rl_cfg_entry_point": f"{agents.__name__}.rsl_rl_ppo_cfg:HandoverPPORunnerCfg",
+    },
+)
+
 gym.register(
     id="Isaac-Shadow-Handover-Direct",
     entry_point=f"{__name__}.handover_env:HandoverEnv",

source/isaaclab_tasks/isaaclab_tasks/core/handover/handover_common.py

@@ -15,22 +15,38 @@
 import isaaclab.sim as sim_utils
 from isaaclab.markers import VisualizationMarkersCfg
 
+from isaaclab_tasks.utils.hydra import preset
+
 from isaaclab_assets.robots.shadow_hand import (
     SHADOW_ACTUATED_JOINT_NAMES as ACTUATED_JOINT_NAMES,
 )
+from isaaclab_assets.robots.shadow_hand import (
+    SHADOW_ACTUATED_JOINT_NAMES_NEWTON,
+)
 from isaaclab_assets.robots.shadow_hand import (
     SHADOW_FINGERTIP_BODY_NAMES as FINGERTIP_BODY_NAMES,
 )
 
 __all__ = [
     "ACTUATED_JOINT_NAMES",
+    "ACTUATED_JOINT_NAMES_PRESET",
     "FINGERTIP_BODY_NAMES",
     "GOAL_MARKER_CFG",
     "GOAL_POSITION_OFFSET",
     "OBJECT_RADIUS",
 ]
 
 
+ACTUATED_JOINT_NAMES_PRESET = preset(
+    physx=ACTUATED_JOINT_NAMES,
+    isaacsim_physx=ACTUATED_JOINT_NAMES,
+    newton_mjwarp=SHADOW_ACTUATED_JOINT_NAMES_NEWTON,
+    ovphysx=ACTUATED_JOINT_NAMES,
+    default=ACTUATED_JOINT_NAMES,
+)
+"""Per-backend actuated joint names, resolved by the physics preset key."""
+
+
 OBJECT_RADIUS: float = 0.0335
 """Hand-over object sphere radius [m], also used for the goal marker."""
 

source/isaaclab_tasks/isaaclab_tasks/core/handover/handover_env.py

@@ -21,7 +21,11 @@
 from isaaclab_tasks.core.handover.handover_common import GOAL_POSITION_OFFSET
 from isaaclab_tasks.core.handover.handover_env_cfg import HandoverEnvCfg
 from isaaclab_tasks.core.handover.mdp.rewards import evaluate_handover_success, handover_reward
-from isaaclab_tasks.core.utils import EpisodeErrorRecorder, randomize_rotation, sample_joint_positions_within_limits
+from isaaclab_tasks.core.reorient.utils import (
+    EpisodeErrorRecorder,
+    randomize_rotation,
+    sample_joint_positions_within_limits,
+)
 
 
 class HandoverEnv(DirectMARLEnv):
@@ -228,8 +232,7 @@ def _get_dones(self) -> tuple[dict[str, torch.Tensor], dict[str, torch.Tensor]]:
     def _reset_idx(self, env_ids: Sequence[int] | torch.Tensor | None):
         if env_ids is None:
             env_ids = self.right_hand._ALL_INDICES
-        # Flush per-episode success (sticky binary: object ever reached the goal within threshold).
-        # 0-dim device tensor, for the same reason
+        # Flush the sticky per-episode success bit.
         self.extras.setdefault("log", {})["Metrics/success_rate"] = self._episode_succeeded[env_ids].float().mean()
         for statistic, value in self._goal_distance.reset(env_ids).items():
             self.extras["log"][f"Diagnostics/episode_min_goal_distance_{statistic}"] = value