Example Environments#

Isaac Lab Arena ships a catalog of ready-to-run environments under isaaclab_arena_environments/. Environments can be provided in two ways:

Python registered environments: small compositions of the building blocks introduced in Concept Overview — Scene, Embodiment, and Task — wrapped in an ExampleEnvironmentBase subclass and registered with the global EnvironmentRegistry. The registered Task ID is passed as the positional example_environment argument to scripts such as isaaclab_arena/evaluation/policy_runner.py.
Environment graph YAML specs: linked graph specs that describe the same scene, embodiment, task, objects, and relations declaratively. These are passed with --env_graph_spec_yaml and can be generated from prompts by the Agentic Environment Generation and Policy Evaluation workflow.

The metadata below follows the same structure as the Key Specifications tables in the Imitation Learning and Reinforcement Learning workflow guides.

Agentically Generated Graph Specs #

The Robolab examples below are linked environment graph YAMLs generated from natural-language prompts and checked in under isaaclab_arena_environments/robolab/. They are consumed with --env_graph_spec_yaml instead of the positional example_environment name.

bagel_plate_banana_bowl#

Environment YAML: isaaclab_arena_environments/robolab/bagel_plate_banana_bowl_linked.yaml

Task Description: Pick up the banana and place it on the plate.

Property	Value
Source Type	Environment graph YAML
Generation Prompt	`Droid picks up the banana from the maple table and places it on the plate. Include two bagels and a bowl on the table.`
Tags	Agentic generation, Robolab, table-top manipulation
Skills	Reach, Grasp, Pick & place
Embodiment	`droid_abs_joint_pos` (node id: `droid`)
Scene	`maple_table_robolab` background
Objects	Pick: `banana`; Destination: `plate`; Distractors: `bagel_1`, `bagel_2`, `bowl`
Task Class	`PickAndPlaceTask`
Object Placement	Initial relations: all objects `On(maple_table_robolab)`; success: `banana On(plate)`
CLI Args	`--env_graph_spec_yaml isaaclab_arena_environments/robolab/bagel_plate_banana_bowl_linked.yaml`

bin_mug_mustard_marker_bowl#

Environment YAML: isaaclab_arena_environments/robolab/bin_mug_marker_bowl_linked.yaml

Task Description: Pick up the bowl and place it in the grey bin, with a mustard bottle, dry-erase marker, and mug as additional tabletop objects.

Property	Value
Source Type	Environment graph YAML
Generation Prompt	`Droid picks up the bowl from the maple table and places it in the grey bin. Using maple table background. Other objects on the table as distractors: mustard, dry erase marker, mug.`
Tags	Agentic generation, Robolab, table-top manipulation
Skills	Reach, Grasp, Pick & place
Embodiment	`droid_abs_joint_pos` (node id: `droid`)
Scene	`maple_table_robolab` background
Objects	Pick: `bowl`; Destination: `grey_bin`; Distractors: `mustard`, `dry_erase_marker`, `mug`
Task Class	`PickAndPlaceTask`
Object Placement	Initial relations: all objects `On(maple_table_robolab)`; success: `bowl On(grey_bin)`
CLI Args	`--env_graph_spec_yaml isaaclab_arena_environments/robolab/bin_mug_marker_bowl_linked.yaml`

butter_raisin_box_grey_bin#

Environment YAML: isaaclab_arena_environments/robolab/butter_raisin_box_grey_bin_linked.yaml

Task Description: Pick up the red raisin box and place it into the grey bin.

Property	Value
Source Type	Environment graph YAML
Generation Prompt	`Droid picks up the red raisin box from the maple table and places it into the grey bin. Using maple table background. Other objects on the table as distractors: butter.`
Tags	Agentic generation, Robolab, table-top manipulation
Skills	Reach, Grasp, Pick & place
Embodiment	`droid_abs_joint_pos`
Scene	`maple_table_robolab` background
Objects	Pick: `raisin_box`; Destination: `grey_bin`; Distractor: `butter`
Task Class	`PickAndPlaceTask`
Object Placement	Initial relations: objects `On(maple_table_robolab)`; rotation markers on `raisin_box` and `butter`; success: `raisin_box On(grey_bin)`
CLI Args	`--env_graph_spec_yaml isaaclab_arena_environments/robolab/butter_raisin_box_grey_bin_linked.yaml`

mustard_raisin_box#

Environment YAML: isaaclab_arena_environments/robolab/mustard_raisin_box_linked.yaml

Task Description: Pick up the mustard bottle and place it on the raisin box.

Property	Value
Source Type	Environment graph YAML
Generation Prompt	`Droid picks up the mustard bottle from the maple table and places it on the raisin box.`
Tags	Agentic generation, Robolab, table-top manipulation
Skills	Reach, Grasp, Pick & place
Embodiment	`droid_abs_joint_pos` (node id: `droid`)
Scene	`maple_table_robolab` background
Objects	Pick: `mustard_bottle`; Destination: `raisin_box`
Task Class	`PickAndPlaceTask`
Object Placement	Initial relations: objects `On(maple_table_robolab)`; rotation marker on `raisin_box`; success: `mustard_bottle On(raisin_box)`
CLI Args	`--env_graph_spec_yaml isaaclab_arena_environments/robolab/mustard_raisin_box_linked.yaml`

Pick & Place #

kitchen_pick_and_place#

Task ID: kitchen_pick_and_place

Class: KitchenPickAndPlaceEnvironment (isaaclab_arena_environments/kitchen_pick_and_place_environment.py)

Task Description: Pick an object off the kitchen counter top and place it inside a kitchen cabinet. Supports a single object via --object or a heterogeneous --object_set spawning a different object per environment.

Property	Value
Tags	Table-top manipulation
Skills	Reach, Grasp, Pick & place
Embodiment	`franka_ik` (default), configurable via `--embodiment`
Scene	`kitchen` background, counter top reference (anchor), cabinet destination
Objects	Configurable via `--object` / `--object_set` (e.g. `tomato_soup_can`, `cracker_box`)
Task Class	`PickAndPlaceTask`
Object Placement	Relations: `On(table_top)`, `AtPosition(x=0.4, y=0.0)`
CLI Args	`--object`, `--object_set`, `--embodiment`, `--teleop_device`

pick_and_place_maple_table#

Task ID: pick_and_place_maple_table

Class: PickAndPlaceMapleTableEnvironment (isaaclab_arena_environments/pick_and_place_maple_table_environment.py)

Task Description: Tabletop pick-and-place on the maple Robolab table; used as the introductory First Arena Environment walkthrough.

Property	Value
Tags	Table-top manipulation
Skills	Reach, Grasp, Pick & place
Embodiment	`droid_abs_joint_pos` (default), configurable via `--embodiment`
Scene	`maple_table_robolab` background, dome `light` (configurable HDR / intensity)
Objects	Pick: `rubiks_cube_hot3d_robolab` (default); Destination: `bowl_ycb_robolab` (default); plus optional `--additional_table_objects`
Task Class	`PickAndPlaceTask` (episode_length_s = 20)
Object Placement	Relations: `On(table)`, `PositionLimits(x=0.55..0.70, y=-0.4..-0.1)`
CLI Args	`--pick_up_object`, `--destination_location`, `--additional_table_objects`, `--embodiment`, `--teleop_device`, `--hdr`, `--light_intensity`

galileo_pick_and_place#

Task ID: galileo_pick_and_place

Class: GalileoPickAndPlaceEnvironment (isaaclab_arena_environments/galileo_pick_and_place_environment.py)

Task Description: Pick up an object in the Galileo lab environment and place it into a small bin on the shelf.

Property	Value
Tags	Table-top manipulation, lab scene
Skills	Reach, Grasp, Pick & place
Embodiment	`gr1_pink` (default), configurable via `--embodiment`
Scene	`galileo` lab background; bin lid (`small_bin_grid_01/lid`) used as destination reference
Objects	`power_drill` (default), configurable via `--object`
Task Class	`PickAndPlaceTask`
CLI Args	`--object`, `--embodiment`, `--teleop_device`

galileo_g1_locomanip_pick_and_place#

Task ID: galileo_g1_locomanip_pick_and_place

Class: GalileoG1LocomanipPickAndPlaceEnvironment (isaaclab_arena_environments/galileo_g1_locomanip_pick_and_place_environment.py)

Task Description: The G1 humanoid navigates the lab, squats, and picks an object off a shelf to place it into a bin on a table to its right. Featured in the G1 Loco-Manipulation Box Pick and Place Task workflow.

Property	Value
Tags	Room-scale loco-manipulation
Skills	Squat, Turn, Walk, Pick, Place
Embodiment	`g1_wbc_pink` (default; whole-body controller w/ navigation P-controller in Mimic)
Scene	`galileo_locomanip` background
Objects	Pick: `brown_box` (default); Destination: `blue_sorting_bin` (default)
Task Class	`PickAndPlaceTask` with `G1PickAndPlaceMimicEnvCfg` injected via `mimic_env_cfg_factory` (episode_length_s = 30, force / velocity success thresholds)
Interop	Isaac Lab Mimic (legacy `locomanip_pick_and_place_D0` datagen for the brown-box → blue-bin pair)
CLI Args	`--object`, `--destination`, `--embodiment`, `--teleop_device`, `--task_description`

Articulated Object Manipulation #

gr1_open_microwave#

Task ID: gr1_open_microwave

Class: Gr1OpenMicrowaveEnvironment (isaaclab_arena_environments/gr1_open_microwave_environment.py)

Task Description: The GR1T2 humanoid reaches with its upper body to open a microwave door. Featured in the GR1 Open Microwave Door Task workflow.

Property	Value
Tags	Table-top manipulation, articulated objects
Skills	Reach, Open door
Embodiment	`gr1_pink` (default) or `gr1_joint` via `--embodiment`
Scene	`kitchen` background, `microwave` placed on the packing table
Objects	`microwave` (articulated); optional `--object` placed in front of the microwave
Task Class	`OpenDoorTask` (openness_threshold = 0.8, reset_openness = 0.2, episode_length_s = 5)
CLI Args	`--object`, `--teleop_device`, `--embodiment`

gr1_turn_stand_mixer_knob#

Task ID: gr1_turn_stand_mixer_knob

Class: Gr1TurnStandMixerKnobEnvironment (isaaclab_arena_environments/gr1_turn_stand_mixer_knob_environment.py)

Task Description: GR1 humanoid turns the dial on a stand mixer to a target level.

Property	Value
Tags	Table-top manipulation, articulated objects
Skills	Reach, Grasp knob, Turn
Embodiment	`gr1_pink` (default) or `gr1_joint` via `--embodiment`
Scene	`kitchen` background, `stand_mixer` on the packing table
Objects	`stand_mixer` (articulated); optional `--object` placed in front
Task Class	`TurnKnobTask`
CLI Args	`--object`, `--target_level` (default 4), `--reset_level` (default -1), `--embodiment`, `--teleop_device`

press_button#

Task ID: press_button

Class: PressButtonEnvironment (isaaclab_arena_environments/press_button_environment.py)

Task Description: Press the button on a coffee machine.

Property	Value
Tags	Table-top manipulation, articulated objects
Skills	Reach, Press
Embodiment	`franka_ik` (default) via `--embodiment`
Scene	`packing_table` background, `coffee_machine` placed on top
Objects	`coffee_machine` (articulated)
Task Class	`PressButtonTask` (reset_pressedness = 0.8)
CLI Args	`--embodiment`, `--teleop_device`

Sorting #

tabletop_sort_cubes#

Task ID: tabletop_sort_cubes

Class: TableTopSortCubesEnvironment (isaaclab_arena_environments/sorting_environment.py)

Task Description: Sort two cubes into two color-matching containers on a table.

Property	Value
Tags	Table-top manipulation, multi-object
Skills	Pick, Place, Sort
Embodiment	`franka_ik` (only supported value)
Scene	`table` background (configurable), `light`
Objects	`--objects` (default `red_cube green_cube`); `--destinations` (default `red_container green_container`); exactly 2 of each required
Task Class	`SortMultiObjectTask` (custom success force_threshold = 0.1)
CLI Args	`--objects`, `--destinations`, `--background`, `--embodiment`, `--teleop_device`

Assembly #

peg_insert#

Task ID: peg_insert

Class: PegInsertEnvironment (isaaclab_arena_environments/tabletop_peginsert_environment.py)

Task Description: Assemble a peg into a hole on a tabletop.

Property	Value
Tags	Tabletop, contact-rich assembly
Skills	Reach, Grasp, Insert
Embodiment	`franka_ik` (default; assembly high-PD config) via `--embodiment`
Scene	`table` background (configurable), dome `light`
Objects	Pick: `peg` (default); Destination: `hole` (default)
Task Class	`AssemblyTask` (min_separation = 0.1, randomized x/y/yaw pose range)
CLI Args	`--object`, `--destination_object`, `--background`, `--embodiment`, `--teleop_device`

gear_mesh#

Task ID: gear_mesh

Class: GearMeshEnvironment (isaaclab_arena_environments/tabletop_gearmesh_environment.py)

Task Description: Pick a medium gear and mesh it onto a gear base, with small and large reference gears already mounted.

Property	Value
Tags	Tabletop, contact-rich assembly
Skills	Reach, Grasp, Mesh
Embodiment	`franka_ik` (default; assembly high-PD config) via `--embodiment`
Scene	`table` background (configurable), dome `light`
Objects	`gear_base`, `medium_gear` (held), `small_gear` and `large_gear` (auxiliary)
Task Class	`AssemblyTask` (held-fixed-and-auxiliary randomization, min_separation = 0.18)
CLI Args	`--background`, `--embodiment`, `--teleop_device`

tabletop_place_upright#

Task ID: tabletop_place_upright

Class: TableTopPlaceUprightEnvironment (isaaclab_arena_environments/tabletop_place_upright_environment.py)

Task Description: Pick a tipped-over mug on the table and place it upright.

Property	Value
Tags	Table-top manipulation, re-orientation
Skills	Reach, Grasp, Re-orient, Place
Embodiment	`agibot` (only supported value; `ArmMode.LEFT`)
Scene	`table` background (configurable), `ground_plane`, `light`
Objects	`mug` (default) via `--object`
Task Class	`PlaceUprightTask` (custom event `randomize_mug_positions`)
CLI Args	`--object`, `--background`, `--embodiment`, `--teleop_device`

Goal-Pose / Lift (RL)#

cube_goal_pose#

Task ID: cube_goal_pose

Class: CubeGoalPoseEnvironment (isaaclab_arena_environments/cube_goal_pose_environment.py)

Task Description: Reach a target 6-DoF pose with a cube (goal-conditioned manipulation).

Property	Value
Tags	Table-top manipulation, goal-conditioned
Skills	Reach, Grasp, Re-orient
Embodiment	`franka_ik` (default) via `--embodiment`
Scene	`table` background (configurable), `light`
Objects	`dex_cube` (default) via `--object`
Task Class	`GoalPoseTask` (target_z_range = [0.2, 1.0], target_orientation_xyzw = yaw 90°, tolerance = 0.2 rad)
CLI Args	`--object`, `--background`, `--embodiment`, `--teleop_device`

lift_object#

Task ID: lift_object

Class: LiftObjectEnvironment (isaaclab_arena_environments/lift_object_environment.py)

Task Description: Reinforcement-learning task in which the Franka Panda learns to grasp and lift an object to a commanded target position. Featured in the Franka Lift Object Task workflow.

Property	Value
Tags	Table-top manipulation, RL training
Skills	Reach, Grasp, Lift
Embodiment	`franka_joint_pos` (default; joint-position control yields better RL success than IK) via `--embodiment`
Scene	`table` background, `ground_plane`, `light`
Objects	`dex_cube` (default) via `--object`
Task Class	`LiftObjectTaskRL` (minimum_height_to_lift = 0.04, episode_length_s = 5)
Training Method	Reinforcement Learning (RSL-RL PPO; `rl_policy_cfg` = `base_rsl_rl_policy:RLPolicyCfg`)
CLI Args	`--object`, `--embodiment`, `--teleop_device`, `--rl_training_mode`

dexsuite_lift#

Task ID: dexsuite_lift

Class: DexsuiteLiftEnvironment (isaaclab_arena_environments/dexsuite_lift_environment.py)

Task Description: Evaluation wrapper around the Isaac Lab Isaac-Dexsuite-Kuka-Allegro-Lift-v0 MDP. The Kuka arm with an Allegro dexterous hand lifts a procedurally generated cuboid to a commanded target position. Featured in the Dexsuite Kuka Allegro Lift Task (Newton) workflow.

Property	Value
Tags	Dexterous manipulation, contact-rich, RL evaluation
Skills	Reach, Grasp, Lift (multi-finger)
Embodiment	`kuka_allegro` (fixed)
Scene	`procedural_table` background, `ground_plane`, `light`
Objects	`procedural_cube` (randomized initial pose with a wide `PoseRange`)
Task Class	`DexsuiteLiftTask` (object_pose command, position-only, resampled every 2–3 s)
Training Method	Pre-trained in Isaac Lab via `DexsuiteKukaAllegroPPORunnerCfg` (RSL-RL PPO)
Physics Backend	PhysX (default) or Newton (`--presets newton`)
CLI Args	(none environment-specific; uses common ``ArenaEnvBuilder`` flags)

Sandbox #

gr1_table_multi_object_no_collision#

Task ID: gr1_table_multi_object_no_collision

Class: GR1TableMultiObjectNoCollisionEnvironment (isaaclab_arena_environments/gr1_table_multi_object_no_collision_environment.py)

Task Description: Sandbox scene for testing the relation solver: an office table with multiple objects placed via On(table) plus the built-in no-overlap solver. No success task — useful for policy_runner smoke tests with zero_action or any policy.

Property	Value
Tags	Sandbox, multi-object placement
Skills	(none — no task)
Embodiment	`gr1_joint` (default) via `--embodiment`
Scene	`ground_plane`, `office_table`, `light`, table-top anchor
Objects	Default set: `cracker_box`, `sugar_box`, `tomato_soup_can`, `dex_cube`, `power_drill`, `red_container` (override via `--objects`)
Task Class	`NoTask` (optional time-out termination via `--episode_length_s`)
CLI Args	`--objects`, `--embodiment`, `--teleop_device`, `--episode_length_s`

Sequential / Composite Tasks #

put_item_in_fridge_and_close_door#

Task ID: put_item_in_fridge_and_close_door

Class: GR1PutAndCloseDoorEnvironment (isaaclab_arena_environments/gr1_put_and_close_door_environment.py)

Task Description: GR1 humanoid sequentially picks an object, places it on the refrigerator shelf, then closes the refrigerator door. Featured in the GR1 Sequential Pick & Place and Close Door Task workflow.

Property	Value
Tags	Sequential manipulation, articulated objects
Skills	Pick, Place, Close door
Embodiment	`gr1_pink` (default) via `--embodiment`
Scene	`lightwheel_robocasa_kitchen` background (`--kitchen_style` selectable), `light`, kitchen counter anchor
Objects	Pick: `ranch_dressing_hope_robolab` (default), or `--object_set` for heterogeneous spawning; Destination: refrigerator shelf reference; Container: `refrigerator` (articulated)
Task Class	`PutAndCloseDoorTask` (sequential: `PickAndPlaceTask` → `CloseDoorTask`, episode_length_s = 10)
Interop	Isaac Lab Mimic (`put_and_close_door_task_D0` datagen)
CLI Args	`--object`, `--object_set`, `--kitchen_style`, `--embodiment`, `--teleop_device`

franka_put_and_close_door#

Task ID: franka_put_and_close_door

Class: FrankaPutAndCloseDoorEnvironment (isaaclab_arena_environments/franka_put_and_close_door_environment.py)

Task Description: Sequential pick-and-place of an object into a microwave, followed by closing the microwave door.

Property	Value
Tags	Sequential manipulation, articulated objects
Skills	Pick, Place, Close door
Embodiment	`franka_ik` (default) via `--embodiment`
Scene	`kitchen` background, `microwave` (articulated, starts open)
Objects	Pick: `dex_cube` (default) via `--object`; Container: `microwave`
Task Class	`FrankaPutAndCloseDoorTask` (sequential: `PickAndPlaceTask` → `CloseDoorTask`)
CLI Args	`--object`, `--embodiment`, `--teleop_device`