Robot and articulation configuration

Robot and articulation configuration#

A jointed robot is represented as an articulation in Isaac Lab. This guide covers reusing an existing robot configuration and authoring a new ArticulationCfg. The ArticulationCfg is a configuration object that defines the properties of an Articulation in Isaac Lab.

Note

While we only cover the creation of an ArticulationCfg in this guide, the process is similar for creating any other asset configuration object.

Reusing a robot configuration#

Maintained robot configurations live in source/isaaclab_assets/isaaclab_assets/robots. Import a configuration from isaaclab_assets and copy it before changing its spawn properties, initial state, or actuators. Keep project-specific configurations in a Python module in your own project; they do not need to be added to Isaac Lab.

For example, robot_cfg = CARTPOLE_CFG.copy() creates an independent configuration. Use robot_cfg.replace(prim_path="{ENV_REGEX_NS}/Robot") when adding it to an InteractiveSceneCfg. The physics backend is selected separately from this robot configuration, as explained below.

We will use the Cartpole example to demonstrate how to create an ArticulationCfg. The Cartpole is a simple robot that consists of a cart with a pole attached to it. The cart is free to move along a rail, and the pole is free to rotate about the cart. The file for this configuration example is source/isaaclab_assets/isaaclab_assets/robots/cartpole.py.

Choosing shared and backend-specific settings#

Use the shared ArticulationCfg for the robot’s initial state and actuators. The simulation selects the physics backend separately.

For common USD properties, use classes from isaaclab.sim.schemas, such as RigidBodyBaseCfg and ArticulationRootBaseCfg. Use isaaclab_physx.sim.schemas.Physx*Cfg for PhysX tuning and isaaclab_newton.sim.schemas.Newton*Cfg / Mujoco*Cfg for Newton and MJWarp-specific settings. The Schema Configuration Classes guide explains the available classes and their USD namespaces; Schema Fragments shows how to author both backends’ attributes in one spawn configuration.

The Cartpole below uses the compatibility names RigidBodyPropertiesCfg and ArticulationRootPropertiesCfg. Its PhysX solver iterations and sleep thresholds do not configure Newton’s solver. When adapting it to Newton, retain the shared initial-state and actuator configuration and configure the Newton solver separately. For example, to override Newton’s self-collision setting:

from isaaclab_assets import CARTPOLE_CFG
from isaaclab_newton.sim.schemas import NewtonArticulationRootPropertiesCfg

robot_cfg = CARTPOLE_CFG.copy()
robot_cfg.spawn.articulation_props = NewtonArticulationRootPropertiesCfg(self_collision_enabled=False)

See Tune MJWarp for solver settings, Prepare an Asset for Newton with MJWarp for asset tuning, and Multi-backend assets for converted USD variants.

Defining the spawn configuration#

As explained in Spawning prims into the scene tutorials, the spawn configuration defines the properties of the assets to be spawned. This spawning may happen procedurally, or through an existing asset file (e.g. USD or URDF). In this example, we will spawn the Cartpole from a USD file.

When spawning an asset from a USD file, we define its UsdFileCfg. This configuration object takes in the following parameters:

usd_path: The USD file path to spawn from
rigid_props: The properties of the articulation’s rigid-body links
articulation_props: The properties of the articulation root

The last two parameters are optional. If not specified, they are kept at their default values in the USD file.

spawn=sim_utils.UsdFileCfg(
    usd_path=f"{ISAACLAB_NUCLEUS_DIR}/Robots/Classic/Cartpole/cartpole.usd",
    rigid_props=[
        sim_utils.UsdPhysicsRigidBodyCfg(rigid_body_enabled=True),
        PhysxRigidBodyCfg(
            max_linear_velocity=1000.0,
            max_angular_velocity=1000.0,
            max_depenetration_velocity=100.0,
            enable_gyroscopic_forces=True,
        ),
    ],
    articulation_props=[
        PhysxArticulationCfg(
            enabled_self_collisions=False,
            solver_position_iteration_count=4,
            solver_velocity_iteration_count=0,
            sleep_threshold=0.005,
            stabilization_threshold=0.001,
        ),
        NewtonArticulationCfg(self_collision_enabled=False),
    ],
),

To import articulation from a URDF file instead of a USD file, you can replace the UsdFileCfg with a UrdfFileCfg. For more details, please check the API documentation.

Defining the initial state#

Every asset requires defining their initial or default state in the simulation through its configuration. This configuration is stored into the asset’s default state buffers that can be accessed when the asset’s state needs to be reset.

Note

The initial state of an asset is defined w.r.t. its local environment frame. This then needs to be transformed into the global simulation frame when resetting the asset’s state. For more details, please check the Interacting with an articulation tutorial.

For an articulation, the InitialStateCfg object defines the initial state of the root of the articulation and the initial state of all its joints. In this example, we will spawn the Cartpole at the origin of the XY plane at a Z height of 2.0 meters. Meanwhile, the joint positions and velocities are set to 0.0.

init_state=ArticulationCfg.InitialStateCfg(
    pos=(0.0, 0.0, 2.0), joint_pos={"slider_to_cart": 0.0, "cart_to_pole": 0.0}
),

Defining the actuator configuration#

Actuators are a crucial component of an articulation. Through this configuration, it is possible to define the type of actuator model to use. We can use the internal actuator model provided by the physics engine (i.e. the implicit actuator model), or use a custom actuator model which is governed by a user-defined system of equations (i.e. the explicit actuator model). For more details on actuators, see Actuators.

The cartpole’s articulation has two actuators, one corresponding to its each joint: cart_to_pole and slider_to_cart. We use two different actuator models for these actuators as an example. However, since they are both using the same actuator model, it is possible to combine them into a single actuator model.

Note

Newton resolves the target mode of joints configured with ImplicitActuatorCfg before solver construction: stiffness-only selects position mode, damping-only velocity mode, both gains combined position/velocity mode, and zero gains effort mode. A gain of None retains the imported USD value; explicit actuator configurations use effort mode. Zero-gain USD drives therefore need no placeholder solely for a configured actuator. See Joint drives on each physics backend for when ensure_drives_exist remains useful.

ActuatorCfg velocity/effort limits considerations#

Use the following fields in an actuator configuration. They select joints and are resolved when the articulation is constructed; the canonical runtime values live on ArticulationData. See Joint and actuator property ownership for the ownership model and runtime mutation paths.

Limit configuration#
Field	Implicit actuator	Explicit actuator
`joint_effort_limit`	Writes the solver drive effort limit.	Writes the solver effort limit; defaults high to avoid a second model clip.
`actuator_effort_limit`	Not supported.	Clips actuator-model output.
`joint_velocity_limit`	Requests a solver velocity constraint.	Requests a solver velocity constraint.
`actuator_velocity_limit`	Creates the soft velocity-limit snapshot; it is not a solver request.	Describes the actuator rated speed; speed-dependent models use it in their torque curve.
`effort_limit`	Deprecated alias for `joint_effort_limit`.	Deprecated alias for `actuator_effort_limit`.
`velocity_limit`	Deprecated alias for `actuator_velocity_limit`.	Deprecated alias for `actuator_velocity_limit`.

Solver velocity enforcement is backend-dependent. joint_velocity_limit records the requested joint state but is not a backend-independent safety clamp; see Validate actuators and limits.

USD vs. ActuatorCfg discrepancy resolution#

USD having default value and the fact that ActuatorCfg can be specified with None, or a overriding value can sometime be confusing what exactly gets written into simulation. The resolution follows these simple rules,per joint and per property:

Resolution Rules for USD vs. ActuatorCfg#
Condition	ActuatorCfg Value	Applied
No override provided	Not Specified	USD Value
Override provided	User’s ActuatorCfg	Same as ActuatorCfg

Digging into USD can sometime be unconvinent, to help clarify what exact value is written, we designed a flag actuator_value_resolution_debug_print, to help user figure out what exact value gets used in simulation.

Whenever an actuator parameter is overridden in the user’s ActuatorCfg (or left unspecified), we compare it to the value read from the USD definition and record any differences. For each joint and each property, if unmatching value is found, we log the resolution:

USD Value The default limit or gain parsed from the USD asset.

ActuatorCfg Value The user-provided override (or “Not Specified” if none was given).

Applied The final value actually used for simulation: if the user didn’t override it, this matches the USD value; otherwise it reflects the user’s setting.

This resolution info is emitted as a warning table only when discrepancies exist. Here’s an example of what you’ll see:

+----------------+------------------------+---------------------+----+-------------+--------------------+----------+
|     Group      |      Property          |         Name        | ID |  USD Value  | ActuatorCfg Value  | Applied  |
+----------------+------------------------+---------------------+----+-------------+--------------------+----------+
| panda_shoulder | joint_velocity_limit   |    panda_joint1     |  0 |    2.17e+00 |   Not Specified    | 2.17e+00 |
|                |                        |    panda_joint2     |  1 |    2.17e+00 |   Not Specified    | 2.17e+00 |
|                |                        |    panda_joint3     |  2 |    2.17e+00 |   Not Specified    | 2.17e+00 |
|                |                        |    panda_joint4     |  3 |    2.17e+00 |   Not Specified    | 2.17e+00 |
|                |     stiffness          |    panda_joint1     |  0 |    2.29e+04 |      8.00e+01      | 8.00e+01 |
|                |                        |    panda_joint2     |  1 |    2.29e+04 |      8.00e+01      | 8.00e+01 |
|                |                        |    panda_joint3     |  2 |    2.29e+04 |      8.00e+01      | 8.00e+01 |
|                |                        |    panda_joint4     |  3 |    2.29e+04 |      8.00e+01      | 8.00e+01 |
|                |      damping           |    panda_joint1     |  0 |    4.58e+03 |      4.00e+00      | 4.00e+00 |
|                |                        |    panda_joint2     |  1 |    4.58e+03 |      4.00e+00      | 4.00e+00 |
|                |                        |    panda_joint3     |  2 |    4.58e+03 |      4.00e+00      | 4.00e+00 |
|                |                        |    panda_joint4     |  3 |    4.58e+03 |      4.00e+00      | 4.00e+00 |
|                |      armature          |    panda_joint1     |  0 |    0.00e+00 |   Not Specified    | 0.00e+00 |
|                |                        |    panda_joint2     |  1 |    0.00e+00 |   Not Specified    | 0.00e+00 |
|                |                        |    panda_joint3     |  2 |    0.00e+00 |   Not Specified    | 0.00e+00 |
|                |                        |    panda_joint4     |  3 |    0.00e+00 |   Not Specified    | 0.00e+00 |
| panda_forearm  | joint_velocity_limit   |    panda_joint5     |  4 |    2.61e+00 |   Not Specified    | 2.61e+00 |
|                |                        |    panda_joint6     |  5 |    2.61e+00 |   Not Specified    | 2.61e+00 |
|                |                        |    panda_joint7     |  6 |    2.61e+00 |   Not Specified    | 2.61e+00 |
|                |     stiffness          |    panda_joint5     |  4 |    2.29e+04 |      8.00e+01      | 8.00e+01 |
|                |                        |    panda_joint6     |  5 |    2.29e+04 |      8.00e+01      | 8.00e+01 |
|                |                        |    panda_joint7     |  6 |    2.29e+04 |      8.00e+01      | 8.00e+01 |
|                |      damping           |    panda_joint5     |  4 |    4.58e+03 |      4.00e+00      | 4.00e+00 |
|                |                        |    panda_joint6     |  5 |    4.58e+03 |      4.00e+00      | 4.00e+00 |
|                |                        |    panda_joint7     |  6 |    4.58e+03 |      4.00e+00      | 4.00e+00 |
|                |      armature          |    panda_joint5     |  4 |    0.00e+00 |   Not Specified    | 0.00e+00 |
|                |                        |    panda_joint6     |  5 |    0.00e+00 |   Not Specified    | 0.00e+00 |
|                |                        |    panda_joint7     |  6 |    0.00e+00 |   Not Specified    | 0.00e+00 |
|                |      friction          |    panda_joint5     |  4 |    0.00e+00 |   Not Specified    | 0.00e+00 |
|                |                        |    panda_joint6     |  5 |    0.00e+00 |   Not Specified    | 0.00e+00 |
|                |                        |    panda_joint7     |  6 |    0.00e+00 |   Not Specified    | 0.00e+00 |
|  panda_hand    | joint_velocity_limit   | panda_finger_joint1 |  7 |    2.00e-01 |   Not Specified    | 2.00e-01 |
|                |                        | panda_finger_joint2 |  8 |    2.00e-01 |   Not Specified    | 2.00e-01 |
|                |     stiffness          | panda_finger_joint1 |  7 |    1.00e+06 |      2.00e+03      | 2.00e+03 |
|                |                        | panda_finger_joint2 |  8 |    1.00e+06 |      2.00e+03      | 2.00e+03 |
|                |      armature          | panda_finger_joint1 |  7 |    0.00e+00 |   Not Specified    | 0.00e+00 |
|                |                        | panda_finger_joint2 |  8 |    0.00e+00 |   Not Specified    | 0.00e+00 |
|                |      friction          | panda_finger_joint1 |  7 |    0.00e+00 |   Not Specified    | 0.00e+00 |
|                |                        | panda_finger_joint2 |  8 |    0.00e+00 |   Not Specified    | 0.00e+00 |
+----------------+------------------------+---------------------+----+-------------+--------------------+----------+

To keep the cleaniness of logging, actuator_value_resolution_debug_print default to False, remember to turn it on when wishes.

Example: configure and run two robots#

The runnable example scripts/tutorials/01_assets/add_new_robot.py contrasts a minimal Jetbot configuration with a more detailed Dofbot configuration. Start with an imported USD asset (see Importing a New Asset) and define its spawn properties and actuators. Jetbot retains the joint gains authored in the USD by setting stiffness and damping to None. Both fields must be specified, even when using these USD defaults:

JETBOT_CONFIG = ArticulationCfg(
    spawn=sim_utils.UsdFileCfg(usd_path=f"{ISAAC_NUCLEUS_DIR}/Robots/NVIDIA/Jetbot/jetbot.usd"),
    actuators={"wheel_acts": ImplicitActuatorCfg(joint_names_expr=[".*"], damping=None, stiffness=None)},
)

Dofbot additionally sets initial joint positions, groups joints by name, and specifies actuator gains and limits. Its solver iterations and maximum depenetration velocity are PhysX-specific; use Choosing shared and backend-specific settings when adapting these properties to Newton. The keys in init_state.joint_pos identify USD joints, not actuator groups. Joint names can be matched with regular expressions; for example, .* selects all joints.

The example adds both configurations to an InteractiveSceneCfg, assigns each robot a path under every environment, and constructs the scene. Its loop resets root and joint states, sets joint targets, writes commands, steps physics, and updates the scene buffers. See Using the Interactive Scene for scene construction and Interacting with an articulation for the reset and control loop.

Complete runnable example

# 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

import argparse

from isaaclab.app import AppLauncher

# add argparse arguments
parser = argparse.ArgumentParser(
    description="This script demonstrates adding a custom robot to an Isaac Lab environment."
)
parser.add_argument("--num_envs", type=int, default=1, help="Number of environments to spawn.")
# append AppLauncher cli args
AppLauncher.add_app_launcher_args(parser)
# parse the arguments
args_cli = parser.parse_args()

# launch omniverse app
app_launcher = AppLauncher(args_cli)
simulation_app = app_launcher.app

import numpy as np
import torch

import isaaclab.sim as sim_utils
from isaaclab.actuators import ImplicitActuatorCfg
from isaaclab.assets import AssetBaseCfg
from isaaclab.assets.articulation import ArticulationCfg
from isaaclab.scene import InteractiveScene, InteractiveSceneCfg
from isaaclab.utils.assets import ISAAC_NUCLEUS_DIR

JETBOT_CONFIG = ArticulationCfg(
    spawn=sim_utils.UsdFileCfg(usd_path=f"{ISAAC_NUCLEUS_DIR}/Robots/NVIDIA/Jetbot/jetbot.usd"),
    actuators={"wheel_acts": ImplicitActuatorCfg(joint_names_expr=[".*"], damping=None, stiffness=None)},
)

DOFBOT_CONFIG = ArticulationCfg(
    spawn=sim_utils.UsdFileCfg(
        usd_path=f"{ISAAC_NUCLEUS_DIR}/Robots/Yahboom/Dofbot/dofbot.usd",
        rigid_props=sim_utils.RigidBodyPropertiesCfg(
            disable_gravity=False,
            max_depenetration_velocity=5.0,
        ),
        articulation_props=sim_utils.ArticulationRootPropertiesCfg(
            enabled_self_collisions=True, solver_position_iteration_count=8, solver_velocity_iteration_count=0
        ),
    ),
    init_state=ArticulationCfg.InitialStateCfg(
        joint_pos={
            "joint1": 0.0,
            "joint2": 0.0,
            "joint3": 0.0,
            "joint4": 0.0,
        },
        pos=(0.25, -0.25, 0.0),
    ),
    actuators={
        "front_joints": ImplicitActuatorCfg(
            joint_names_expr=["joint[1-2]"],
            joint_effort_limit=100.0,
            joint_velocity_limit=100.0,
            stiffness=10000.0,
            damping=100.0,
        ),
        "joint3_act": ImplicitActuatorCfg(
            joint_names_expr=["joint3"],
            joint_effort_limit=100.0,
            joint_velocity_limit=100.0,
            stiffness=10000.0,
            damping=100.0,
        ),
        "joint4_act": ImplicitActuatorCfg(
            joint_names_expr=["joint4"],
            joint_effort_limit=100.0,
            joint_velocity_limit=100.0,
            stiffness=10000.0,
            damping=100.0,
        ),
    },
)


class NewRobotsSceneCfg(InteractiveSceneCfg):
    """Designs the scene."""

    # Ground-plane
    ground = AssetBaseCfg(prim_path="/World/defaultGroundPlane", spawn=sim_utils.GroundPlaneCfg())

    # lights
    dome_light = AssetBaseCfg(
        prim_path="/World/Light", spawn=sim_utils.DomeLightCfg(intensity=3000.0, color=(0.75, 0.75, 0.75))
    )

    # robot
    Jetbot = JETBOT_CONFIG.replace(prim_path="{ENV_REGEX_NS}/Jetbot")
    Dofbot = DOFBOT_CONFIG.replace(prim_path="{ENV_REGEX_NS}/Dofbot")


def run_simulator(sim: sim_utils.SimulationContext, scene: InteractiveScene):
    sim_dt = sim.get_physics_dt()
    sim_time = 0.0
    count = 0

    # wheel-velocity templates allocated once on the simulation device; the joint
    # target setters dispatch to GPU Warp kernels and reject CPU tensors.
    straight_action = torch.tensor([[10.0, 10.0]], device=sim.device).repeat(scene.num_envs, 1)
    turn_action = torch.tensor([[5.0, -5.0]], device=sim.device).repeat(scene.num_envs, 1)

    while simulation_app.is_running():
        # reset
        if count % 500 == 0:
            # reset counters
            count = 0
            # reset the scene entities to their initial positions offset by the environment origins
            root_jetbot_pose = scene["Jetbot"].data.default_root_pose.torch.clone()
            root_jetbot_pose[:, :3] += scene.env_origins
            root_dofbot_pose = scene["Dofbot"].data.default_root_pose.torch.clone()
            root_dofbot_pose[:, :3] += scene.env_origins

            # copy the default root state to the sim for the jetbot's orientation and velocity
            scene["Jetbot"].write_root_pose_to_sim_index(root_pose=root_jetbot_pose)
            root_jetbot_vel = scene["Jetbot"].data.default_root_vel.torch.clone()
            scene["Jetbot"].write_root_velocity_to_sim_index(root_velocity=root_jetbot_vel)
            scene["Dofbot"].write_root_pose_to_sim_index(root_pose=root_dofbot_pose)
            root_dofbot_vel = scene["Dofbot"].data.default_root_vel.torch.clone()
            scene["Dofbot"].write_root_velocity_to_sim_index(root_velocity=root_dofbot_vel)

            # copy the default joint states to the sim
            joint_pos, joint_vel = (
                scene["Jetbot"].data.default_joint_pos.torch.clone(),
                scene["Jetbot"].data.default_joint_vel.torch.clone(),
            )
            scene["Jetbot"].write_joint_position_to_sim_index(position=joint_pos)
            scene["Jetbot"].write_joint_velocity_to_sim_index(velocity=joint_vel)
            joint_pos, joint_vel = (
                scene["Dofbot"].data.default_joint_pos.torch.clone(),
                scene["Dofbot"].data.default_joint_vel.torch.clone(),
            )
            scene["Dofbot"].write_joint_position_to_sim_index(position=joint_pos)
            scene["Dofbot"].write_joint_velocity_to_sim_index(velocity=joint_vel)
            # clear internal buffers
            scene.reset()
            print("[INFO]: Resetting Jetbot and Dofbot state...")

        # drive around
        if count % 100 < 75:
            # Drive straight by setting equal wheel velocities
            action = straight_action
        else:
            # Turn by applying different velocities
            action = turn_action

        scene["Jetbot"].set_joint_velocity_target_index(target=action)

        # wave
        wave_action = scene["Dofbot"].data.default_joint_pos.torch.clone()
        wave_action[:, 0:4] = 0.25 * np.sin(2 * np.pi * 0.5 * sim_time)
        scene["Dofbot"].set_joint_position_target_index(target=wave_action)

        scene.write_data_to_sim()
        sim.step()
        sim_time += sim_dt
        count += 1
        scene.update(sim_dt)


def main():
    """Main function."""
    # Initialize the simulation context
    sim_cfg = sim_utils.SimulationCfg(device=args_cli.device)
    sim = sim_utils.SimulationContext(sim_cfg)
    sim.set_camera_view([3.5, 0.0, 3.2], [0.0, 0.0, 0.5])
    # Design scene
    scene_cfg = NewRobotsSceneCfg(num_envs=args_cli.num_envs, env_spacing=2.0)
    scene = InteractiveScene(scene_cfg)
    # Play the simulator
    sim.reset()
    # Now we are ready!
    print("[INFO]: Setup complete...")
    # Run the simulator
    run_simulator(sim, scene)


if __name__ == "__main__":
    main()
    simulation_app.close()

Run the example in the Isaac Sim viewport:

uv run isaaclab -p scripts/tutorials/01_assets/add_new_robot.py --viz kit

This example uses PhysX physics and requires Isaac Sim. The Dofbot gripper is not actuated in this example, so a warning about unconfigured joints is expected. Stop the example with Ctrl+C.

Jetbot and Dofbot running in the example scene.

Robot and articulation configuration

Contents

Robot and articulation configuration#

Reusing a robot configuration#

Choosing shared and backend-specific settings#

Defining the spawn configuration#

Defining the initial state#

Defining the actuator configuration#

ActuatorCfg velocity/effort limits considerations#

USD vs. ActuatorCfg discrepancy resolution#

Example: configure and run two robots#