Robot Arm: Completed Reference to Scaffold

Files changed (9) hide show

README.md +18 -0
src/isaaclab_tutorial/tasks/place_vial/config/so101/__init__.py +10 -17
src/isaaclab_tutorial/tasks/place_vial/config/so101/agents/rsl_rl_distillation_cfg.py +3 -37
src/isaaclab_tutorial/tasks/place_vial/config/so101/agents/rsl_rl_ppo_cfg.py +3 -32
src/isaaclab_tutorial/tasks/place_vial/config/so101/camera_env_cfg.py +15 -90
src/isaaclab_tutorial/tasks/place_vial/config/so101/env_cfg.py +50 -217
src/isaaclab_tutorial/tasks/place_vial/config/so101/inspection_env_cfg.py +34 -0
src/isaaclab_tutorial/tasks/place_vial/config/so101/lesson_scaffold.py +61 -0
tests/test_config.py +8 -2

README.md CHANGED Viewed

@@ -1,3 +1,21 @@
 # SO-101 vial placement
 <p align="center"><img src="media/demo.gif" alt="SO-101 workshop demonstration" width="100%"></p>

+# Isaac Lab Tutorial Learning Path
+This branch is the exercise scaffold for the completed code on `main`.
+Follow the **Train a Robot Arm** module, applying its marked regions in lesson order.
+Imports, workshop assets, the reset dataset, and supporting utilities are supplied.
+The initial scene intentionally has no robot. Lesson 02 makes inspection runnable;
+lesson 04 adds actions, lesson 05 the state teacher, and lesson 06 camera distillation.
+Run the upstream tests after completing lesson 06, not against unfinished exercises.
+Compare with the completed reference without switching branches:
+```bash
+git diff main -- src/
+```
+The reference documentation below describes the completed task and includes
+optional upstream workflows outside the course.
 # SO-101 vial placement
 <p align="center"><img src="media/demo.gif" alt="SO-101 workshop demonstration" width="100%"></p>

src/isaaclab_tutorial/tasks/place_vial/config/so101/__init__.py CHANGED Viewed

@@ -4,25 +4,18 @@ import gymnasium as gym
 _PACKAGE = "isaaclab_tutorial.tasks.place_vial.config.so101"
-gym.register(
-    id="IsaacTutorial-Place-Vial-SO101",
-    entry_point="isaaclab.envs:ManagerBasedRLEnv",
-    disable_env_checker=True,
-    kwargs={
-        "env_cfg_entry_point": f"{_PACKAGE}.env_cfg:SO101VialEnvCfg",
-        "rsl_rl_cfg_entry_point": f"{_PACKAGE}.agents.rsl_rl_ppo_cfg:SO101StatePPORunnerCfg",
-        "default_agent": "rsl_rl",
-    },
-)
 gym.register(
-    id="IsaacTutorial-Place-Vial-SO101-Camera",
     entry_point="isaaclab.envs:ManagerBasedRLEnv",
     disable_env_checker=True,
-    kwargs={
-        "env_cfg_entry_point": f"{_PACKAGE}.camera_env_cfg:SO101VialCameraEnvCfg",
-        "rsl_rl_cfg_entry_point": f"{_PACKAGE}.agents.rsl_rl_distillation_cfg:SO101CameraDistillationRunnerCfg",
-        "rsl_rl_ppo_cfg_entry_point": f"{_PACKAGE}.agents.rsl_rl_ppo_cfg:SO101CameraPPORunnerCfg",
-        "default_agent": "rsl_rl",
-    },
 )

 _PACKAGE = "isaaclab_tutorial.tasks.place_vial.config.so101"
+# BEGIN lesson-01-registration
+# Complete task registration in the matching lesson.
+# END lesson-01-registration
+# BEGIN lesson-06-camera-registration
+# Complete task registration in the matching lesson.
+# END lesson-06-camera-registration
 gym.register(
+    id="IsaacTutorial-Inspect-SO101",
     entry_point="isaaclab.envs:ManagerBasedRLEnv",
     disable_env_checker=True,
+    kwargs={"env_cfg_entry_point": f"{_PACKAGE}.inspection_env_cfg:SO101InspectionEnvCfg"},
 )

src/isaaclab_tutorial/tasks/place_vial/config/so101/agents/rsl_rl_distillation_cfg.py CHANGED Viewed

@@ -23,42 +23,8 @@ class BoundedTeacherDistillation(Distillation):
         return actions
 @configclass
 class SO101CameraDistillationRunnerCfg(RslRlDistillationRunnerCfg):
-    """DAgger-style distillation: the student acts, the loaded state teacher labels every visited state."""
-    seed = 42
-    num_steps_per_env = 32
-    max_iterations = 1600
-    save_interval = 100
-    experiment_name = "so101_vial_camera_distillation"
-    run_name = ""
-    obs_groups = {"student": ["wrist_rgb", "proprioception"], "teacher": ["teacher_state"]}
-    clip_actions = 1.0
-    student = RslRlCNNModelCfg(
-        # Compact encoder for the 64 x 48 wrist image.
-        cnn_cfg=RslRlCNNModelCfg.CNNCfg(
-            output_channels=[16, 32, 32], kernel_size=[5, 3, 3], stride=[2, 2, 2], activation="elu"
-        ),
-        hidden_dims=[256, 128],
-        activation="elu",
-        obs_normalization=True,
-        distribution_cfg=RslRlCNNModelCfg.GaussianDistributionCfg(init_std=0.05, std_type="log"),
-    )
-    # Must mirror the state actor so its checkpoint loads; the teacher always acts deterministically.
-    teacher = RslRlMLPModelCfg(
-        hidden_dims=[256, 256, 128],
-        activation="elu",
-        obs_normalization=True,
-        distribution_cfg=RslRlMLPModelCfg.GaussianDistributionCfg(init_std=0.2, std_type="log", std_range=(0.05, 0.3)),
-    )
-    algorithm = RslRlDistillationAlgorithmCfg(
-        class_name=(
-            "isaaclab_tutorial.tasks.place_vial.config.so101.agents.rsl_rl_distillation_cfg:BoundedTeacherDistillation"
-        ),
-        num_learning_epochs=4,
-        learning_rate=5.0e-4,
-        gradient_length=1,
-        max_grad_norm=1.0,
-        loss_type="mse",
-    )

         return actions
+# BEGIN lesson-06-distillation
 @configclass
 class SO101CameraDistillationRunnerCfg(RslRlDistillationRunnerCfg):
+    pass
+# END lesson-06-distillation

src/isaaclab_tutorial/tasks/place_vial/config/so101/agents/rsl_rl_ppo_cfg.py CHANGED Viewed

@@ -5,40 +5,11 @@ from isaaclab.utils.configclass import configclass
 from isaaclab_rl.rsl_rl import RslRlCNNModelCfg, RslRlMLPModelCfg, RslRlOnPolicyRunnerCfg, RslRlPpoAlgorithmCfg
 @configclass
 class SO101StatePPORunnerCfg(RslRlOnPolicyRunnerCfg):
-    """State teacher: fully observed actor, privileged critic."""
-    seed = 42
-    num_steps_per_env = 64
-    max_iterations = 800
-    save_interval = 50
-    experiment_name = "so101_vial_state"
-    run_name = ""
-    obs_groups = {"actor": ["policy"], "critic": ["critic"]}
-    clip_actions = 1.0
-    actor = RslRlMLPModelCfg(
-        hidden_dims=[256, 256, 128],
-        activation="elu",
-        obs_normalization=True,
-        # Clamp the std: with clipped actions and an entropy bonus, PPO otherwise grows it without bound.
-        distribution_cfg=RslRlMLPModelCfg.GaussianDistributionCfg(init_std=0.2, std_type="log", std_range=(0.05, 0.3)),
-    )
-    critic = RslRlMLPModelCfg(hidden_dims=[256, 256, 128], activation="elu", obs_normalization=True)
-    algorithm = RslRlPpoAlgorithmCfg(
-        value_loss_coef=1.0,
-        use_clipped_value_loss=True,
-        clip_param=0.2,
-        entropy_coef=0.005,
-        num_learning_epochs=5,
-        num_mini_batches=8,
-        learning_rate=3.0e-4,
-        schedule="adaptive",
-        gamma=0.995,
-        lam=0.95,
-        desired_kl=0.01,
-        max_grad_norm=1.0,
-    )
 @configclass

 from isaaclab_rl.rsl_rl import RslRlCNNModelCfg, RslRlMLPModelCfg, RslRlOnPolicyRunnerCfg, RslRlPpoAlgorithmCfg
+# BEGIN lesson-05-ppo
 @configclass
 class SO101StatePPORunnerCfg(RslRlOnPolicyRunnerCfg):
+    pass
+# END lesson-05-ppo
 @configclass

src/isaaclab_tutorial/tasks/place_vial/config/so101/camera_env_cfg.py CHANGED Viewed

@@ -48,98 +48,32 @@ _HORIZONTAL_APERTURE = 20.955 * 80 / 64
 WRIST_FOCAL_LENGTH_PIXELS = 80 * _FOCAL_LENGTH / _HORIZONTAL_APERTURE
 @configclass
 class SO101CameraSceneCfg(SO101SceneCfg):
-    """SO-101 scene with the rectified pinhole wrist camera and handles to the randomized visual materials."""
-    wrist_camera = CameraCfg(
-        prim_path="{ENV_REGEX_NS}/Robot/gripper/wowrobo_2MP_camera/pinhole",
-        spawn=PinholeCameraCfg(focal_length=_FOCAL_LENGTH, horizontal_aperture=_HORIZONTAL_APERTURE),
-        offset=CameraCfg.OffsetCfg(convention="opengl"),
-        data_types=["rgb"],
-        width=80,
-        height=60,
-        update_period=1.0 / 30.0,
-        update_latest_camera_pose=True,
-        renderer_cfg=_camera_renderer_cfg(),
-        background_color=(0.3, 0.3, 0.3),
-    )
-    # The Newton renderer only reads material color.
-    robot_visual = VisualMaterialCfg(prim_path="{ENV_REGEX_NS}/Robot/Looks/material_a_d_printed", spawn=None)
-    desk_visual = VisualMaterialCfg(prim_path="{ENV_REGEX_NS}/Desk/Looks/Wood", spawn=None)
-    rack_visual = VisualMaterialCfg(prim_path="{ENV_REGEX_NS}/Rack/WorkshopVisual/Looks/OmniPBR", spawn=None)
-    vial_body_visual = VisualMaterialCfg(prim_path="{ENV_REGEX_NS}/Vial/Looks/Vial_plastic", spawn=None)
-    vial_cap_visual = VisualMaterialCfg(prim_path="{ENV_REGEX_NS}/Vial/Looks/Vial_cap", spawn=None)
-    vial_label_visual = VisualMaterialCfg(prim_path="{ENV_REGEX_NS}/Vial/Looks/Label", spawn=None)
-    def __post_init__(self):
-        self.robot.spawn.hide_camera_housing = True
-        self.vial.spawn.usd_path = str(VIAL_LABELED_USD)
 @configclass
 class WristImageCfg(ObsGroup):
-    """Two-frame wrist RGB history with episode-consistent photometric (and, for sim2real, intrinsics) variation."""
-    image = ObsTerm(
-        func=mdp.wrist_rgb,
-        params={
-            "sensor_cfg": SceneEntityCfg("wrist_camera"),
-            "exposure_range": (0.75, 1.25),
-            "contrast_range": (0.85, 1.15),
-            "white_balance_range": (0.90, 1.10),
-            "brightness_range": (-0.05, 0.05),
-            "gamma_range": preset(default=(1.0, 1.0), sim2real=(0.85, 1.15)),
-            "blur_range": preset(default=(0.0, 0.0), sim2real=(0.0, 0.5)),
-            "history_length": 2,
-            "projection_size": POLICY_IMAGE_SIZE,
-            "focal_length_pixels": WRIST_FOCAL_LENGTH_PIXELS,
-            "focal_scale_range": preset(default=(1.0, 1.0), sim2real=(0.95, 1.05)),
-            "principal_point_pixels": preset(default=0.0, sim2real=1.5),
-            "radial_distortion_range": preset(default=(0.0, 0.0), sim2real=(-0.04, 0.04)),
-        },
-        noise=UniformNoiseCfg(n_min=-0.025, n_max=0.025),
-    )
-    def __post_init__(self):
-        self.enable_corruption = True
 @configclass
 class ProprioceptionCfg(ObsGroup):
-    """Proprioceptive observations available on the real robot, in the encoders' (possibly biased) coordinates."""
-    joint_pos = ObsTerm(
-        func=mdp.encoder_joint_pos,
-        params={"asset_cfg": SceneEntityCfg("robot")},
-        noise=UniformNoiseCfg(n_min=-0.01, n_max=0.01),
-    )
-    joint_vel = ObsTerm(
-        func=mdp.joint_vel,
-        params={"asset_cfg": SceneEntityCfg("robot")},
-        noise=UniformNoiseCfg(n_min=-0.02, n_max=0.02),
-    )
-    joint_target = ObsTerm(
-        func=mdp.encoder_joint_target,
-        params={"asset_cfg": SceneEntityCfg("robot")},
-        noise=UniformNoiseCfg(n_min=-0.005, n_max=0.005),
-    )
-    previous_action = ObsTerm(func=mdp.last_action)
-    def __post_init__(self):
-        self.enable_corruption = True
-        self.concatenate_terms = True
 @configclass
 class CameraObservationsCfg:
-    """Deployable student inputs, the state teacher's inputs (distillation labels), and the privileged critic."""
-    wrist_rgb: WristImageCfg = WristImageCfg()
-    proprioception: ProprioceptionCfg = ProprioceptionCfg()
-    teacher_state: PolicyStateGroupCfg = PolicyStateGroupCfg()
-    critic: CriticStateGroupCfg = CriticStateGroupCfg()
 def _color_event(materials: str, color_range) -> EventTerm:
@@ -173,17 +107,8 @@ class CameraSim2RealEventsCfg(Sim2RealEventsCfg):
     )
 @configclass
 class SO101VialCameraEnvCfg(SO101VialEnvCfg):
-    """Vial placement from wrist RGB and proprioception, distilled from the state teacher."""
-    scene: SO101CameraSceneCfg = SO101CameraSceneCfg(num_envs=1024, env_spacing=0.9, replicate_physics=True)
-    observations: CameraObservationsCfg = CameraObservationsCfg()
-    events: EventsCfg = preset(default=EventsCfg(), sim2real=CameraSim2RealEventsCfg())
-    def play_mode(self):
-        from isaaclab_tutorial.utils import evaluation
-        super().play_mode()
-        if not evaluation.EXACT_EVALUATION_ACTIVE:
-            self.scene.num_envs = min(self.scene.num_envs, 8)

 WRIST_FOCAL_LENGTH_PIXELS = 80 * _FOCAL_LENGTH / _HORIZONTAL_APERTURE
+# BEGIN lesson-06-SO101CameraSceneCfg
 @configclass
 class SO101CameraSceneCfg(SO101SceneCfg):
+    pass
+# END lesson-06-SO101CameraSceneCfg
+# BEGIN lesson-06-WristImageCfg
 @configclass
 class WristImageCfg(ObsGroup):
+    pass
+# END lesson-06-WristImageCfg
+# BEGIN lesson-06-ProprioceptionCfg
 @configclass
 class ProprioceptionCfg(ObsGroup):
+    pass
+# END lesson-06-ProprioceptionCfg
+# BEGIN lesson-06-CameraObservationsCfg
 @configclass
 class CameraObservationsCfg:
+    pass
+# END lesson-06-CameraObservationsCfg
 def _color_event(materials: str, color_range) -> EventTerm:
     )
+# BEGIN lesson-06-SO101VialCameraEnvCfg
 @configclass
 class SO101VialCameraEnvCfg(SO101VialEnvCfg):
+    pass
+# END lesson-06-SO101VialCameraEnvCfg

src/isaaclab_tutorial/tasks/place_vial/config/so101/env_cfg.py CHANGED Viewed

@@ -14,7 +14,10 @@ from pxr import Gf, Sdf, Usd, UsdGeom, UsdPhysics, UsdShade
 import isaaclab.sim as sim_utils
 from isaaclab.assets import AssetBaseCfg, RigidObjectCfg
-from isaaclab.envs import ManagerBasedRLEnvCfg
 from isaaclab.envs.mdp.actions.actions_cfg import RelativeJointPositionActionCfg
 from isaaclab.managers import EventTermCfg as EventTerm
 from isaaclab.managers import ObservationGroupCfg as ObsGroup
@@ -45,6 +48,14 @@ from isaaclab_tutorial.assets import DESK_USD, RACK_USD, RESET_DATASET, VIAL_USD
 from isaaclab_tutorial.tasks.place_vial import mdp
 from isaaclab_tutorial.tasks.place_vial.reset.curriculum import ALL_PHASES, CANONICAL_START
 JOINTS = ["shoulder_pan", "shoulder_lift", "elbow_flex", "wrist_flex", "wrist_roll", "gripper"]
 ARM_JOINTS = JOINTS[:-1]
@@ -438,77 +449,14 @@ def _jaw_support_sensor(link: str) -> ContactSensorCfg:
 class SO101SceneCfg(InteractiveSceneCfg):
     """One SO-101, one vial, one rack, and a desk; sim2real adds the gray mat and jaw/support contact sensing."""
-    robot = SO101_CFG.replace(
-        prim_path="{ENV_REGEX_NS}/Robot",
-        init_state=SO101_CFG.init_state.replace(
-            pos=preset(default=(-0.05, 0.0, 0.0), sim2real=(-0.05, 0.0, BASE_HEIGHT_CORRECTION)),
-            # +90 degrees about world Z (XYZW).
-            rot=(0.0, 0.0, 0.7071068, 0.7071068),
-            joint_pos=dict(zip(JOINTS, WORKSHOP_INITIAL_JOINT_POSITION, strict=True)),
-        ),
-        soft_joint_pos_limit_factor=0.98,
-    )
-    vial = RigidObjectCfg(
-        prim_path="{ENV_REGEX_NS}/Vial",
-        spawn=VialSpawnCfg(
-            usd_path=str(VIAL_USD),
-            rigid_props=preset(
-                default=None,
-                physx=PhysxRigidBodyCfg(solver_position_iteration_count=PHYSX_SOLVER_POSITION_ITERATIONS),
-            ),
-            activate_contact_sensors=True,
-            randomize_diameters=preset(default=False, sim2real=True),
-        ),
-        init_state=RigidObjectCfg.InitialStateCfg(
-            pos=TABLETOP_VIAL_POSITION,
-            # Horizontal vial: +90 degrees about world Y (XYZW).
-            rot=(0.0, 0.7071068, 0.0, 0.7071068),
-        ),
-    )
-    rack = RigidObjectCfg(
-        prim_path="{ENV_REGEX_NS}/Rack",
-        spawn=sim_utils.UsdFileCfg(usd_path=str(RACK_USD), func=_spawn_usd_with_contact_material),
-        init_state=RigidObjectCfg.InitialStateCfg(pos=(0.18, 0.08, 0.04)),
-    )
-    desk = AssetBaseCfg(
-        prim_path="{ENV_REGEX_NS}/Desk",
-        spawn=sim_utils.UsdFileCfg(usd_path=str(DESK_USD), func=_spawn_usd_with_contact_material),
-        init_state=AssetBaseCfg.InitialStateCfg(pos=(0.22, 0.0, 0.032), rot=(0.0, 0.0, 0.7071068, 0.7071068)),
-    )
-    support = preset(
-        default=None,
-        sim2real=RigidObjectCfg(
-            prim_path="{ENV_REGEX_NS}/Support",
-            spawn=SupportSpawnCfg(
-                size=(0.39, 0.39, 0.005),
-                rigid_props=sim_utils.UsdPhysicsRigidBodyCfg(kinematic_enabled=True),
-                collision_props=sim_utils.UsdPhysicsCollisionCfg(),
-                mass_props=sim_utils.MassCfg(mass=1.0),
-                physics_material=WORKSHOP_CONTACT_MATERIAL,
-                visual_material=sim_utils.PreviewSurfaceCfg(diffuse_color=(0.07, 0.07, 0.07)),
-            ),
-            init_state=RigidObjectCfg.InitialStateCfg(pos=(0.22, 0.02, 0.0325)),
-        ),
-    )
-    # Unfiltered: OV PhysX cannot build a filtered contact view for the gripper link in a cloned scene.
-    fixed_jaw_contact = ContactSensorCfg(prim_path="{ENV_REGEX_NS}/Robot/gripper", history_length=4)
-    moving_jaw_contact = ContactSensorCfg(
-        prim_path="{ENV_REGEX_NS}/Robot/moving_jaw_so101_v1",
-        filter_prim_paths_expr=["{ENV_REGEX_NS}/Vial"],
-        history_length=4,
-    )
-    vial_rack_contact = ContactSensorCfg(
-        prim_path="{ENV_REGEX_NS}/Vial",
-        filter_prim_paths_expr=["{ENV_REGEX_NS}/Rack"],
-        history_length=4,
-    )
-    fixed_jaw_support = preset(default=None, sim2real=_jaw_support_sensor("gripper"))
-    moving_jaw_support = preset(default=None, sim2real=_jaw_support_sensor("moving_jaw_so101_v1"))
     light = AssetBaseCfg(
         prim_path="/World/Light",
@@ -521,108 +469,39 @@ class SO101SceneCfg(InteractiveSceneCfg):
     )
 @configclass
 class ActionsCfg:
-    """Joint targets relative to the measured position, matching the real SO-101 position interface."""
-    arm_action = RelativeJointPositionActionCfg(
-        asset_name="robot",
-        joint_names=ARM_JOINTS,
-        preserve_order=True,
-        # Small offsets stall the real shoulder and elbow, so sim2real gives them larger steps.
-        scale=preset(
-            default=0.033,
-            sim2real={
-                "shoulder_pan": 0.033,
-                "shoulder_lift": 0.12,
-                "elbow_flex": 0.07,
-                "wrist_flex": 0.033,
-                "wrist_roll": 0.033,
-            },
-        ),
-        use_zero_offset=True,
-    )
-    gripper_action = RelativeJointPositionActionCfg(
-        asset_name="robot",
-        joint_names=["gripper"],
-        # Avoid opening a grasp rapidly from a small policy bias.
-        scale=0.02,
-        use_zero_offset=True,
-    )
 @configclass
 class PolicyStateGroupCfg(ObsGroup):
-    """Fully observed state actor inputs."""
-    joint_pos = ObsTerm(func=mdp.joint_pos, params={"asset_cfg": SceneEntityCfg("robot", joint_names=JOINTS)})
-    joint_vel = ObsTerm(func=mdp.joint_vel, params={"asset_cfg": SceneEntityCfg("robot", joint_names=JOINTS)})
-    joint_target = ObsTerm(func=mdp.joint_target, params={"asset_cfg": SceneEntityCfg("robot", joint_names=JOINTS)})
-    previous_action = ObsTerm(func=mdp.last_action)
-    end_effector = ObsTerm(
-        func=mdp.body_state_b,
-        params={
-            "body_asset_cfg": SceneEntityCfg("robot", body_names="gripper"),
-            "base_asset_cfg": SceneEntityCfg("robot"),
-        },
-    )
-    vial = ObsTerm(
-        func=mdp.body_state_b,
-        params={"body_asset_cfg": SceneEntityCfg("vial"), "base_asset_cfg": SceneEntityCfg("robot")},
-    )
-    vial_in_rack = ObsTerm(func=mdp.vial_in_rack_frame)
-    placement = ObsTerm(func=mdp.placement_features)
-    # Latched milestones make the once-per-episode milestone rewards Markov.
-    progress = ObsTerm(func=mdp.progress_flags)
-    def __post_init__(self):
-        self.enable_corruption = False
-        self.concatenate_terms = True
 @configclass
 class CriticStateGroupCfg(PolicyStateGroupCfg):
-    """Privileged training critic inputs."""
-    contact = ObsTerm(func=mdp.jaw_contact_flags)
 @configclass
 class ObservationsCfg:
-    policy: PolicyStateGroupCfg = PolicyStateGroupCfg()
-    critic: CriticStateGroupCfg = CriticStateGroupCfg()
 @configclass
 class EventsCfg:
-    """Task-horizon resets from the physics-validated dataset plus modest vial variation."""
-    vial_material = EventTerm(
-        func=mdp.randomize_rigid_body_material,
-        mode="startup",
-        params={
-            "asset_cfg": SceneEntityCfg("vial"),
-            "static_friction_range": preset(default=VIAL_FRICTION_RANGE, physx=PHYSX_VIAL_FRICTION_RANGE),
-            "dynamic_friction_range": preset(default=VIAL_FRICTION_RANGE, physx=PHYSX_VIAL_FRICTION_RANGE),
-            "restitution_range": (0.0, 0.02),
-            "num_buckets": 32,
-        },
-    )
-    vial_mass = EventTerm(
-        func=mdp.randomize_rigid_body_mass,
-        mode="startup",
-        params={
-            "asset_cfg": SceneEntityCfg("vial"),
-            # Newton cannot reliably infer mass from the detailed mesh.
-            "mass_distribution_params": (0.015, 0.025),
-            "operation": "abs",
-        },
-    )
-    reset_from_dataset = EventTerm(
-        func=mdp.ResetFromDataset,
-        mode="reset",
-        params={"dataset_path": str(RESET_DATASET), "sequential": False, "phase_weights": ALL_PHASES},
-    )
 def _material_event(asset_cfg: SceneEntityCfg, friction_range: tuple[float, float]) -> EventTerm:
@@ -702,43 +581,18 @@ class Sim2RealEventsCfg(EventsCfg):
         )
 @configclass
 class RewardsCfg:
-    """Sparse physical milestones, a success bonus, two dense shaping terms, and light regularization."""
-    # Sim2real aims the pad centre slightly above the cap, favouring a shallower pinch clear of the mat.
-    approach_progress = RewTerm(
-        func=mdp.approach_progress, weight=1.0, params={"height_offset": preset(default=0.0, sim2real=0.008)}
-    )
-    held_goal = RewTerm(func=mdp.held_goal, weight=0.1)
-    milestones = RewTerm(func=mdp.milestones, weight=10.0)
-    success = RewTerm(func=mdp.success, weight=200.0)
-    vial_lost = RewTerm(
-        func=mdp.is_terminated_term, weight=-50.0, params={"term_keys": ["vial_out_of_bounds", "vial_dropped"]}
-    )
-    action_rate = RewTerm(func=mdp.action_rate_l2, weight=-0.002)
-    joint_velocity = RewTerm(func=mdp.joint_vel_l2, weight=-0.0002)
-    joint_limits = RewTerm(func=mdp.joint_pos_limits, weight=-1.0)
-    jaw_support_force = preset(
-        default=None,
-        sim2real=RewTerm(func=mdp.jaw_support_force, weight=-8.0, params={"free_force": 0.25, "force_span": 2.0}),
-    )
 @configclass
 class TerminationsCfg:
-    success = DoneTerm(func=mdp.placement_success)
-    vial_dropped = DoneTerm(func=mdp.vial_dropped)
-    # The robot base frame is yawed +90 degrees, so these bounds are 0.5 m to its sides and 0.62 m ahead/behind.
-    vial_out_of_bounds = DoneTerm(
-        func=mdp.out_of_bound,
-        params={
-            "asset_cfg": SceneEntityCfg("vial"),
-            "in_bound_range": {"x": (-0.55, 0.45), "y": (-0.62, 0.62), "z": (-0.01, 1.0)},
-        },
-    )
-    unstable_robot = DoneTerm(func=mdp.abnormal_robot_state)
-    time_out = DoneTerm(func=mdp.time_out, time_out=True)
 @configclass
@@ -768,44 +622,23 @@ class PhysicsCfg(PresetCfg):
     default = newton_mjwarp
 @configclass
 class SO101VialEnvCfg(ManagerBasedRLEnvCfg):
-    """State task trained from physics-validated reset poses."""
-    scene: SO101SceneCfg = SO101SceneCfg(num_envs=4096, env_spacing=0.9, replicate_physics=True)
-    actions: ActionsCfg = ActionsCfg()
-    observations: ObservationsCfg = ObservationsCfg()
-    events: EventsCfg = preset(default=EventsCfg(), sim2real=Sim2RealEventsCfg())
-    rewards: RewardsCfg = RewardsCfg()
-    terminations: TerminationsCfg = TerminationsCfg()
     def __post_init__(self):
         self.decimation = 4
-        # The measured shoulder response is slower, so sim2real attempts get more time.
-        self.episode_length_s = preset(default=20.0, sim2real=30.0)
-        self.is_finite_horizon = False
         self.sim.dt = 1.0 / 120.0
         self.sim.render_interval = self.decimation
         self.sim.physics = PhysicsCfg()
         self.sim.default_visualizer_cfg = VisualizerCfg(eye=(0.64, 0.0, 0.36), lookat=(0.19, 0.02, 0.075))
-    def play_mode(self):
-        """Play and evaluate complete episodes from the home pose, in dataset order.
-        The nominal task pins the vial to its workshop material and mass; sim2real keeps its physics randomization.
-        """
-        from isaaclab_tutorial.utils import evaluation
-        requested_num_envs = self.scene.num_envs
-        super().play_mode()
-        if evaluation.EXACT_EVALUATION_ACTIVE:
-            self.scene.num_envs = min(requested_num_envs, evaluation.EVALUATION_EPISODES)
-        else:
-            self.scene.num_envs = min(self.scene.num_envs, 16)
-        self.events.reset_from_dataset.params["sequential"] = True
-        self.events.reset_from_dataset.params["phase_weights"] = CANONICAL_START
-        if not isinstance(self.events, Sim2RealEventsCfg):
-            self.events.vial_material.params.update(
-                static_friction_range=(0.7, 0.7), dynamic_friction_range=(0.7, 0.7), restitution_range=(0.0, 0.0)
-            )
-            self.events.vial_mass.params["mass_distribution_params"] = (0.020, 0.020)

 import isaaclab.sim as sim_utils
 from isaaclab.assets import AssetBaseCfg, RigidObjectCfg
+from isaaclab.envs import (
+    ManagerBasedRLEnvCfg,
+    VideoRecorderCfg,  # noqa: F401 -- used by the lesson 04 replacement
+)
 from isaaclab.envs.mdp.actions.actions_cfg import RelativeJointPositionActionCfg
 from isaaclab.managers import EventTermCfg as EventTerm
 from isaaclab.managers import ObservationGroupCfg as ObsGroup
 from isaaclab_tutorial.tasks.place_vial import mdp
 from isaaclab_tutorial.tasks.place_vial.reset.curriculum import ALL_PHASES, CANONICAL_START
+from .lesson_scaffold import (  # noqa: F401 -- supplied for the lesson 01-04 replacements
+    AgentObservationGroupsCfg,
+    EmptyActionsCfg,
+    InspectionObservationsCfg,
+    LessonEventsCfg,
+    LessonTerminationsCfg,
+)
 JOINTS = ["shoulder_pan", "shoulder_lift", "elbow_flex", "wrist_flex", "wrist_roll", "gripper"]
 ARM_JOINTS = JOINTS[:-1]
 class SO101SceneCfg(InteractiveSceneCfg):
     """One SO-101, one vial, one rack, and a desk; sim2real adds the gray mat and jaw/support contact sensing."""
+    # BEGIN lesson-02-robot
+    robot = None
+    # END lesson-02-robot
+    # BEGIN lesson-03-objects
+    # Add the vial, rack, desk, and contact sensors in lesson 03.
+    pass
+    # END lesson-03-objects
     light = AssetBaseCfg(
         prim_path="/World/Light",
     )
+# BEGIN lesson-04-actions
 @configclass
 class ActionsCfg:
+    pass
+# END lesson-04-actions
+# BEGIN lesson-05-policy-observations
 @configclass
 class PolicyStateGroupCfg(ObsGroup):
+    pass
+# END lesson-05-policy-observations
+# BEGIN lesson-05-critic-observations
 @configclass
 class CriticStateGroupCfg(PolicyStateGroupCfg):
+    pass
+# END lesson-05-critic-observations
+# BEGIN lesson-05-observation-groups
 @configclass
 class ObservationsCfg:
+    pass
+# END lesson-05-observation-groups
+# BEGIN lesson-05-dataset-resets
 @configclass
 class EventsCfg:
+    pass
+# END lesson-05-dataset-resets
 def _material_event(asset_cfg: SceneEntityCfg, friction_range: tuple[float, float]) -> EventTerm:
         )
+# BEGIN lesson-05-rewards
 @configclass
 class RewardsCfg:
+    pass
+# END lesson-05-rewards
+# BEGIN lesson-05-terminations
 @configclass
 class TerminationsCfg:
+    pass
+# END lesson-05-terminations
 @configclass
     default = newton_mjwarp
+# BEGIN lesson-environment
 @configclass
 class SO101VialEnvCfg(ManagerBasedRLEnvCfg):
+    """Inspection scaffold; lesson 04 adds actions and lesson 05 adds training."""
+    scene: SO101SceneCfg = SO101SceneCfg(num_envs=1, env_spacing=0.9, replicate_physics=True)
+    actions: EmptyActionsCfg = EmptyActionsCfg()
+    observations: InspectionObservationsCfg = InspectionObservationsCfg()
+    events: LessonEventsCfg = LessonEventsCfg()
+    rewards = None
+    terminations = None
     def __post_init__(self):
         self.decimation = 4
+        self.episode_length_s = 20.0
         self.sim.dt = 1.0 / 120.0
         self.sim.render_interval = self.decimation
         self.sim.physics = PhysicsCfg()
         self.sim.default_visualizer_cfg = VisualizerCfg(eye=(0.64, 0.0, 0.36), lookat=(0.19, 0.02, 0.075))
+# END lesson-environment

src/isaaclab_tutorial/tasks/place_vial/config/so101/inspection_env_cfg.py ADDED Viewed

	@@ -0,0 +1,34 @@

+"""Inspect the robot and scene before adding policy controls."""
+from isaaclab.envs import ManagerBasedRLEnvCfg, VideoRecorderCfg
+from isaaclab.utils.configclass import configclass
+from isaaclab.visualizers import VisualizerCfg
+from isaaclab_tasks.utils import preset
+from .env_cfg import PhysicsCfg, SO101SceneCfg
+from .lesson_scaffold import EmptyActionsCfg, InspectionObservationsCfg, LessonEventsCfg
+@configclass
+class SO101InspectionEnvCfg(ManagerBasedRLEnvCfg):
+    scene: SO101SceneCfg = SO101SceneCfg(num_envs=1, env_spacing=0.9, replicate_physics=True)
+    actions: EmptyActionsCfg = EmptyActionsCfg()
+    observations: InspectionObservationsCfg = InspectionObservationsCfg()
+    events: LessonEventsCfg = LessonEventsCfg()
+    rewards = None
+    terminations = None
+    def __post_init__(self):
+        self.episode_length_s = 20.0
+        self.decimation = 4
+        self.sim.dt = 1.0 / 120.0
+        self.sim.render_interval = self.decimation
+        self.sim.physics = PhysicsCfg()
+        self.sim.default_visualizer_cfg = VisualizerCfg(eye=(0.64, 0.0, 0.36), lookat=(0.19, 0.02, 0.075))
+        self.video_recorders = preset(
+            default=[],
+            record_video=[VideoRecorderCfg(
+                source="visualizer:newton", output_dir="videos/inspection", video_length=120, fps=30,
+            )],
+        )

src/isaaclab_tutorial/tasks/place_vial/config/so101/lesson_scaffold.py ADDED Viewed

	@@ -0,0 +1,61 @@

+"""Small runnable scene scaffold used before the teacher configuration is assembled."""
+# Imports are supplied for the lesson snippets, including unfinished regions.
+# ruff: noqa: F401
+from isaaclab.managers import EventTermCfg as EventTerm
+from isaaclab.managers import ObservationGroupCfg as ObsGroup
+from isaaclab.managers import ObservationTermCfg as ObsTerm
+from isaaclab.managers import SceneEntityCfg
+from isaaclab.managers import TerminationTermCfg as DoneTerm
+from isaaclab.utils.configclass import configclass
+from isaaclab_tutorial.tasks.place_vial import mdp
+@configclass
+class EmptyActionsCfg:
+    pass
+@configclass
+class JointObservationsCfg(ObsGroup):
+    joint_pos = ObsTerm(func=mdp.joint_pos, params={"asset_cfg": SceneEntityCfg("robot")})
+    joint_vel = ObsTerm(func=mdp.joint_vel, params={"asset_cfg": SceneEntityCfg("robot")})
+    def __post_init__(self):
+        self.enable_corruption = False
+        self.concatenate_terms = True
+@configclass
+class InspectionObservationsCfg:
+    policy: JointObservationsCfg = JointObservationsCfg()
+# BEGIN lesson-04-AgentObservationsCfg
+@configclass
+class AgentObservationsCfg(JointObservationsCfg):
+    pass
+# END lesson-04-AgentObservationsCfg
+@configclass
+class AgentObservationGroupsCfg:
+    policy: AgentObservationsCfg = AgentObservationsCfg()
+@configclass
+class LessonEventsCfg:
+    reset_scene = EventTerm(func=mdp.reset_scene_to_default, mode="reset", params={"reset_joint_targets": True})
+    # BEGIN lesson-03-vial-mass
+    pass
+    # END lesson-03-vial-mass
+# BEGIN lesson-04-LessonTerminationsCfg
+@configclass
+class LessonTerminationsCfg:
+    pass
+# END lesson-04-LessonTerminationsCfg

tests/test_config.py CHANGED Viewed

@@ -42,8 +42,14 @@ def _camera(*presets):
 def test_tasks_are_registered_with_rsl_rl_entry_points():
     ids = sorted(task for task in gym.registry if task.startswith("IsaacTutorial-"))
-    assert ids == ["IsaacTutorial-Place-Vial-SO101", "IsaacTutorial-Place-Vial-SO101-Camera"]
-    state, camera = (gym.spec(task).kwargs for task in ids)
     assert state["env_cfg_entry_point"].endswith("env_cfg:SO101VialEnvCfg")
     assert state["rsl_rl_cfg_entry_point"].endswith("rsl_rl_ppo_cfg:SO101StatePPORunnerCfg")
     assert camera["env_cfg_entry_point"].endswith("camera_env_cfg:SO101VialCameraEnvCfg")

 def test_tasks_are_registered_with_rsl_rl_entry_points():
     ids = sorted(task for task in gym.registry if task.startswith("IsaacTutorial-"))
+    assert ids == [
+        "IsaacTutorial-Inspect-SO101",
+        "IsaacTutorial-Place-Vial-SO101",
+        "IsaacTutorial-Place-Vial-SO101-Camera",
+    ]
+    inspection, state, camera = (gym.spec(task).kwargs for task in ids)
+    assert inspection["env_cfg_entry_point"].endswith("inspection_env_cfg:SO101InspectionEnvCfg")
+    assert "rsl_rl_cfg_entry_point" not in inspection
     assert state["env_cfg_entry_point"].endswith("env_cfg:SO101VialEnvCfg")
     assert state["rsl_rl_cfg_entry_point"].endswith("rsl_rl_ppo_cfg:SO101StatePPORunnerCfg")
     assert camera["env_cfg_entry_point"].endswith("camera_env_cfg:SO101VialCameraEnvCfg")