Skip to navigation
References

Release Notes

View as Markdown

The release notes are also available in the Isaac Lab GitHub repository. The entries below reproduce the published release text in reStructuredText format, excluding externally hosted media.

v3.0.0-beta2.patch1

View this release on GitHub.

This is a small patch release on top of the previous Isaac Lab 3.0.0 Beta 2 release, including an update to support Isaac Sim 6.0.1, which includes fixes and improvements for NuRec workflows. See the Isaac Sim 6.0.1 release notes. Additionally, some fixes were introduced for Isaac Lab dependencies and docker image to better support compatibility with Isaac Sim.

What’s Changed

  • Bump h5py version to >=3.16.0 by @peterd-NV in PR #6266.
  • Updates Isaac Sim to version 6.0.1 by @kellyguo11 in PR #6277.
  • [Fix] Cherry-pick Isaac Sim 6.0 streaming crash fix by @hujc7 in PR #6295.

Full Changelog.

v3.0.0-beta2

View this release on GitHub.

Isaac Lab 3.0 Beta 2 is a stabilization and enablement release for the Isaac Lab 3.0 beta development. It builds on v3.0.0-beta with additional features and improvements on Newton support (VBD, solver coupling, Kamino, rough terrain, sensors), multi-backend physics, simplified training and installation commands, kit-less workflows, visualizers, rendering, teleoperation, learning exports, installation, CI, and documentation.

This release is compatible with the latest release of Isaac Sim 6.0.

This is a beta release. The develop branch is still under active development and may continue to receive breaking changes, error-message changes, or performance tuning before the final 3.0 release. Please check out the develop branch for the latest development updates.

Highlights

Multi-Backend Stabilization

The multi-backend architecture introduced in the first 3.0 beta has been hardened across PhysX, Newton, OVPhysX, Isaac RTX, OVRTX, and kit-less execution paths. This release improves backend selection, scene-data routing, clone-plan handling, sensor reset behavior, runtime compatibility checks, and error reporting when unsupported physics, renderer, or visualizer combinations are requested.

Newton Physics & Kit-Less Workflows

Newton and OVPhysX support have been expanded and stabilized for larger kit-less training and visualization workflows:

  • Adds Newton ray-caster, frame-transformer, IMU/PVA, contact, joint-wrench, deformable, and VBD-coupling support across the relevant packages.
  • Adds Newton rough-terrain locomotion presets for quadrupeds and bipeds, plus additional MJWarp presets for manipulation, navigation, and Spot environments.
  • Adds documentation for Newton installation, supported features, MJWarp, Kamino, VBD, solver comparison, Warp environment migration, and the Newton manager abstraction.
  • Fixes heterogeneous replication, clone-plan source resolution, stale sensor data after resets, Newton visualizer updates on PhysX simulations, and Newton camera/site frame poses.
  • Adds initial experimental support for OvPhysX (kit-less version of PhysX).

Rendering, Cameras & PPISP

Rendering received broad updates across Isaac RTX, OVRTX, and Newton Warp:

  • Adds isaaclab_ppisp, a renderer-backend-agnostic physically plausible image signal processing pipeline for converting HDR scene-linear output to LDR camera images.
  • Adds HDR render outputs and internal PPISP composition to Isaac RTX, OVRTX, and Newton Warp renderers when CameraCfg.isp_cfg is configured.
  • Moves the tiled-camera implementation into Camera and improves tiled-camera support in Kit and Newton visualizers.
  • Fixes OVRTX multi-GPU operation, cloned-environment camera output, temp-file handling, missing-runtime errors, and package resolution for optional PPISP installs.
  • Adds rendering correctness, image comparison, visualizer integration, and performance tests.

Visualizers & Scene Data

The visualizer system now has stronger backend-agnostic scene-data support:

  • Adds a scene-data backend hook for active InteractiveScene access.
  • Improves scene-data providers for PhysX, Newton, and OVPhysX so visualizers can discover transforms, contacts, and sensors consistently.
  • Adds tiled camera views to Kit and Newton visualizers.
  • Fixes visualizer startup ordering, Kit visualizer crashes, Newton visualizer updates on PhysX scenes, and camera/video recording behavior when multiple visualizers are active.

Teleoperation, Mimic & Policy Deployment

Teleoperation and imitation-learning workflows were updated for the 3.0 stack:

  • Adds Isaac Capture MCAP record/replay support and a non-interactive teleop_replay_agent.py for CI replay.
  • Defers optional isaacteleop imports so task configs can be parsed without the Isaac Capture runtime installed.
  • Improves CloudXR defaults, XR performance guidance, native keyboard/gamepad/SpaceMouse mappings, and RMPFlow teleoperation behavior.
  • Updates Mimic installation to include isaaclab_teleop and removes the obsolete nvidia-srl-usd-to-urdf dependency.
  • Adds LEAPP policy export documentation and fixes RSL-RL recurrent policy export, projected-gravity export, checkpoint selection, and policy-deployment guides.

Installation, Docker & CI

This release simplifies setup while adding significant infrastructure hardening:

  • Adds experimental root-level uv run train and uv run play workflows so training and playback can be launched from a fresh checkout without manually creating or activating an environment.
  • Simplifies install extras to isaacsim and all, replacing older per-submodule extras such as tasks, rl, and assets.
  • Pins the released Isaac Sim Docker image to 6.0.0 and keeps Python pinned to 3.12.
  • Updates PyTorch to 2.10.0 with CUDA 12.8 wheels on x86_64/Windows and CUDA 13.0 wheels for aarch64.
  • Adds install CI, wheel build automation, package-scoped test runners, Docker volume and non-root-user fixes, and published image tags.
  • Improves aarch64, DGX Spark, ARM Docker, Windows, uv, and pip installation documentation.
  • Pins third-party GitHub Actions to commit SHAs and improves CI cleanup, cancellation, sharding, diagnostics, and docs-only behavior.

New & Expanded Capabilities

Core

  • Adds ProxyArray and asset/sensor-level property caching for faster repeated data access.
  • Adds scene-data provider APIs and docs for backend-neutral transform access.
  • Adds schema-configuration documentation and backend-neutral deformable APIs.
  • Adds GPU transform reads, performance optimizations for asset writes, and startup-time profiling tools.

Physics & Rendering Backends

  • Adds OVPhysX articulation, rigid object collection, contact sensor, frame view, scene-data backend, and runtime install diagnostics.
  • Adds HDR camera output and PPISP integration, deformable API migration, ray-caster fixes, and multi-GPU Fabric frame-view support.
  • Adds Newton deformable support, rigid-deformable coupling, ray-caster support, solver configuration exports, contact accessors, and updated Newton package pins.

Learning & Environments

  • Adds unified train and play console entry points that dispatch by --rl_library across rsl_rl, rl_games, skrl, sb3, and rlinf.
  • Refactors per-library train/play scripts under scripts/reinforcement_learning/ around a shared dispatch helper, reducing duplicated launcher and parser logic.
  • Adds typed preset selectors such as physics=physx, physics=newton_mjwarp, and renderer=... to unified training and play scripts. For example: train --rl_library rl_games --task=Isaac-Cartpole-RGB-Camera-Direct-v0 physics=newton_mjwarp renderer=newton_renderer presets=rgb.
  • Adds validation for unsupported backend/renderer/visualizer combinations with clearer errors.
  • Adds public helpers for enumerating task presets.
  • Expands Newton presets for locomotion, manipulation, navigation, and Warp-first experimental environments.
  • Adds success-rate metric logging across RL tasks.

Documentation

  • Reworks installation, quickstart, ecosystem, Docker, CloudXR, Teleop, visualizer, renderer, physical-backend, migration, and environment documentation.
  • Adds policy-deployment guides for reach policies and LEAPP exports.
  • Adds profiling, proxy-array, camera-output, video-recording, tiled-camera, schema-cfg, and physical-backend documentation.
  • Removes remaining --headless references in favor of --viz / --visualizer.

Breaking Changes

Preset CLI

isaaclab_tasks.utils.fold_preset_tokens has been removed. setup_preset_cli() now returns physics=, renderer=, and presets= tokens verbatim, and register_task() parses them directly. Scripts should assign the returned remainder to sys.argv unchanged.

Teleoperation Replay

The transitional isaaclab_teleop.automation package, including XcrReplayConfig and start_xcr_replay, has been removed. Replay workflows now use Isaac Capture MCAP replay through scripts/environments/teleoperation/teleop_replay_agent.py.

The legacy lazy teleop_devices accessor on GR1T2 pick-place environment configs has also been removed. Use env_cfg.isaac_teleop or construct legacy OpenXR devices explicitly.

Deformable APIs

Generic deformable object and material APIs have moved from isaaclab_physx into backend-neutral isaaclab.assets and isaaclab.sim imports. The old generic PhysX deformable aliases are deprecated in favor of PhysX-specific names.

OVRTX Renderer Configuration

OVRTXRendererCfg.use_cloning has been renamed to use_ovrtx_cloning, and temp_usd_suffix has been removed. Temporary USD output now uses a fixed filename under temp_usd_dir when enabled.

Camera Aliases

TiledCamera has been folded into Camera. Existing tiled-camera aliases remain as compatibility surface where available, but new code should use Camera and CameraCfg.

Notable Fixes

Core Runtime

  • Fixes SimulationContext reset ordering so initial visualizers are created before timeline event pumping can invalidate new PhysX tensor views.
  • Guards environment, asset, sensor, and callback destructors during Python shutdown.
  • Fixes memory leaks when closing manager-based, direct RL, and direct MARL environments.
  • Fixes safe callable resolution for lambda strings by validating expressions and evaluating them without Python builtins.
  • Fixes local and remote asset dependency retrieval for MDL materials and textures.

Backends & Sensors

  • Fixes stale sensor data after environment resets for PhysX and Newton contact, IMU, PVA, and joint-wrench sensors.
  • Fixes PhysX scene-data rigid-body discovery when joint prims share names or carry rigid-body API metadata.
  • Fixes Newton and OVRTX heterogeneous replication, clone-plan path resolution, and camera transform handling.
  • Fixes OVPhysX optional runtime pins and missing-runtime install errors.
  • Fixes OVRTX package resolution so PPISP is only required when camera ISP is configured.

Training & Deployment

  • Fixes rl_games and SB3 checkpoint fallback when best/final checkpoints are not yet written.
  • Fixes skrl JAX multihost imports, RSL-RL recurrent LEAPP export, and LEAPP projected-gravity export.
  • Fixes AutoMate PhysX sizing, assembly ID validation, Numba dependency removal, and downstream argument forwarding.
  • Fixes RLinf optional dependencies for DGX Spark and aarch64.
  • Fixes native teleoperation mappings for Franka reach tasks and RMPFlow quaternion handling.

Updated Dependencies

  • Isaac Sim 6.0.0 Docker image.
  • Python >=3.12,<3.13.
  • PyTorch 2.10.0, TorchAudio 2.10.0, TorchVision 0.25.0.
  • CUDA 12.8 PyTorch wheels for Linux x86_64, Windows, and AMD64 markers; CUDA 13.0 wheels for Linux aarch64/arm64 markers.
  • Newton package pins updated along the 1.2.1 line.
  • OVPhysX optional runtime pinned to ovphysx==0.4.13.
  • OVRTX optional runtime updated along the 0.3.x line.
  • pyglet>=2.1.6,<3 added for Newton GL video recording.
  • isaaclab_ppisp added as the shared camera ISP extension, loaded by renderers only when camera ISP is configured.

Known Limitations

  • This remains a beta release; the release branch may still receive breaking changes before the final Isaac Lab 3.0 release. We anticipate the branch will stabilize towards a final Isaac Lab 3.0 release in the coming 1-2 months.
  • Newton remains under active development. Some features, task presets, and surface gripper workflows remain backend-specific or unsupported.
  • Surface grippers require CPU simulation with physics=isaacsim_physx, including the UR10 Long/Short Suction and Galbot Right Arm Suction stacking tasks and the Galbot Mimic variants. Pass --device cpu for teleoperation; zero and random agents preserve the tasks’ CPU defaults when --device is omitted.
  • Pink IK dependencies are provided by the standard installation only on Linux x86_64 and aarch64. The Windows uv/pip installation does not provide Pinocchio, so Pink IK tasks cannot run with that installation.
  • OVRTX and OVPhysX require their optional runtime wheels. Isaac Lab now reports clearer install guidance when those wheels are missing, but those backends are still optional installs.

Migration Notes

  • Replace --headless usage with --viz / --visualizer or omit visualizers for fully headless runs.
  • Replace presets=... only where broad preset broadcast is intended; prefer typed selectors such as physics=... and renderer=... when selecting backend-specific behavior.
  • Remove fold_preset_tokens(...) wrappers from custom scripts that integrate with the preset CLI.
  • Move deformable imports to backend-neutral isaaclab.assets and isaaclab.sim APIs where possible.
  • Migrate tiled-camera code toward Camera / CameraCfg.
  • Use ProxyArray.torch or ProxyArray.warp explicitly when crossing between Torch and Warp code.
  • Use Isaac Capture MCAP replay and teleop_replay_agent.py for replay automation instead of the removed XCR automation helpers.

v3.0.0-beta

View this release on GitHub.

Isaac Lab 3.0 Beta is the next major release of Isaac Lab, built on Isaac Sim 6.0 and introducing a ground-up architectural overhaul. This release brings multi-backend physics, a pluggable renderer system, Warp-native data pipelines, and a kit-less installation mode — enabling faster, more flexible robot learning research.

This is a beta release. The develop branch is under active development and may experience breaking changes, error messages, or performance regressions in some use cases.

Highlights

Multi-Backend Physics Architecture

Isaac Lab 3.0 introduces a factory-based multi-backend architecture that separates simulation backend-specific code from the core API. Asset and sensor classes (e.g., Articulation, RigidObject, and ContactSensor) are now backed by abstract base classes, with backend-specific implementations in dedicated extension packages:

  • isaaclab_physx — Full PhysX backend (default), including deformable objects, surface grippers, contact sensors, IMU, and frame transformers.
  • isaaclab_newton — New Newton physics backend powered by MuJoCo-Warp, supporting MJWarp, XPBD, and Featherstone solvers with CUDA-graph acceleration.

Your existing imports from isaaclab.assets and isaaclab.sensors continue to work — the factory automatically dispatches to the active backend at runtime.

Newton Physics Backend

The new isaaclab_newton extension enables running Isaac Lab environments without Isaac Sim (kit-less mode). Newton support includes:

  • Articulations, rigid objects, and rigid object collections.
  • Contact sensors.
  • MuJoCo-Warp solver with configurable integrators (implicitfast, euler) and contact models (pyramidal, elliptic).
  • CUDA graph support for high-throughput stepping.
  • Newton-compatible presets for 20+ environments (locomotion, manipulation, classic control).

Pluggable Renderer System

A new BaseRenderer abstraction supports configuration-owned implementations managed by RenderContext:

BackendRequires Isaac Sim?Best For
Isaac RTXYesFull sensor fidelity, photorealistic rendering
OVRTXNo (+ isaaclab_ov[ovrtx])Kit-less RTX pipeline
Newton WarpNo (kit-less)Fast training without RTX

Pluggable Visualizer System

Isaac Lab 3.0 introduces a new pluggable visualizer framework (isaaclab_visualizers) with four interchangeable backends, all decoupled from the physics engine and renderer:

VisualizerBest ForKey Features
Omniverse (Kit)High-fidelity, Isaac Sim integrationUSD stage, visual markers, live training plots
NewtonFast iteration, low overheadOpenGL rendering, physics debug markers (joints, contacts, COM), interactive camera controls
RerunRemote viewing & replayWeb viewer, timeline scrubbing, .rrd recording/export
ViserWeb-based sharing & recordingBrowser-based via Newton Warp renderer, public share URLs, .viser recording

Launch from the CLI with the new --viz flag (replaces the deprecated --headless flag).

Multiple visualizers at once:

python train.py --task Isaac-Cartpole-v0 --viz kit,newton,rerun

Single visualizer:

python train.py --task Isaac-Cartpole-v0 --viz newton

Web-based Viser visualizer:

python train.py --task Isaac-Cartpole-v0 --viz viser

Headless (no visualizers):

python train.py --task Isaac-Cartpole-v0

Warp-Native Data Pipeline

All .data.* properties on asset and sensor classes now return wp.array (NVIDIA Warp) instead of torch.Tensor. Internal state buffers use structured warp types (wp.vec3f, wp.quatf, wp.transformf, wp.spatial_vectorf), and fused GPU kernels replace Python-level loops for state extraction, velocity transforms, and data write-back. Convert back to torch with wp.to_torch() when needed.

Preset System for Multi-Backend Environments

A new PresetCfg pattern allows a single environment config to declare named variants for different backends. Select a preset at launch:

Run with Newton backend:

python train.py task=Isaac-Ant-v0 presets=newton

Run with default PhysX backend:

python train.py task=Isaac-Ant-v0

Lazy Exporting & Resolvable Strings

All packages now use lazy exporting with .pyi stubs, so importing a top-level module (e.g., import isaaclab.sensors) no longer eagerly pulls in heavyweight dependencies. Config fields like class_type store references as resolvable strings that are resolved only after SimulationApp is initialized, enabling automatic physics-backend selection.

Isaac Capture Integration

XR teleoperation is now powered by Isaac Capture via the new isaaclab_teleop extension, replacing the previous isaaclab.devices.openxr stack. The new system provides a unified teleoperation interface with pipeline-based retargeting configuration.

Breaking Changes

Quaternion Format: WXYZ → XYZW

All quaternions throughout Isaac Lab now use XYZW ordering to align with Warp, PhysX, and Newton conventions. Hard-coded quaternion values must be updated. A quaternion finder tool is provided to help locate and fix quaternions.

Data Properties Return wp.array

Data Properties Return wp.array

All .data.* properties now return wp.array. Wrap with wp.to_torch() for PyTorch compatibility. An automated migration tool is available at scripts/tools/wrap_warp_to_torch.py.

Write Method Index/Mask Split

All asset write methods have been split into _index and _mask variants. The previous write_*_to_sim(data, env_ids) pattern is replaced by:

  • write_*_to_sim_index(data, env_ids) — sparse indexed data.
  • write_*_to_sim_mask(data, env_mask) — full data with boolean mask.

Deprecated APIs

  • root_physx_view → use root_view.
  • object_* naming on RigidObjectCollection → use body_*.
  • --headless CLI argument → use --visualizer / --viz.
  • OpenXRDevice / OpenXRDeviceCfg → use IsaacTeleopDevice / IsaacTeleopCfg.
  • Sensor pose_w, pos_w, and quat_w properties → use FrameTransformer.

URDF & MJCF Importers Updated

Both importers have been rewritten for Isaac Sim 6.0. Several configuration settings have been removed, renamed, or replaced. See the migration guide for details.

Updated Dependencies

  • Python 3.12.
  • PyTorch 2.10.0+cu128.
  • NumPy 2.3.1.
  • Isaac Sim 6.0.

Known Limitations

  • Ubuntu only — The develop branch is currently available on Ubuntu. Windows support and Isaac Lab pip wheels will be available soon.
  • Newton backend is in active development — Not all environments have Newton presets yet. Some features (e.g., deformable objects, surface grippers, material randomization) are PhysX-only.
  • Performance regressions may be observed in some use cases as the multi-backend architecture stabilizes.
  • Breaking changes may still occur on the develop branch before the final 3.0 release.

Migration Guide

For a comprehensive guide on migrating from Isaac Lab 2.x to 3.0, including code examples, API rename tables, and automated tooling, see the migration guide.

v2.3.2

What’s Changed

This release focuses on stability, infrastructure improvements, workflow refinements, and incremental feature expansions, along with some significant new features, including Multirotor and thruster support for drones, Multi-mesh RayCaster, Visual-based tactile sensor, Haply device integration, and new OpenArm environments. It includes improvements to training workflows, teleoperation and Mimic pipelines, Ray integration, simulation utilities, and developer tooling, along with a large number of robustness and quality-of-life fixes.

This will be our final release on the current main branch as we shift our development focus towards the develop branch. We anticipate large restructuring changes to happen on develop. While we hope to continue taking in contributions from the community, we will focus more time on our development towards Isaac Lab 3.0. For existing PRs, please re-target the target branch to develop to stay up-to-date with the latest changes.

New Features

Core & Simulation

Learning & Environments

Mimic & Teleoperation

Improvements

Core & Simulation

Learning & Environments

Infrastructure

Bug Fixes

Core & Simulation

Learning & Environments

Infrastructure & Tooling

Documentation

Migration Guide

External Force and Torque Application - Wrench Composers

The set_external_force_and_torque() method on articulations, rigid bodies, and rigid body collections has been deprecated in favor of a new composable wrench system.

Related PR: https://github.com/isaac-sim/IsaacLab/pull/3287

New Features:

  • Permanent Wrench Composer: Applies forces/torques that persist across simulation steps until explicitly changed
  • Instantaneous Wrench Composer: Applies forces/torques for a single simulation step, then automatically resets
  • Composability: Multiple forces and torques can now be added together on the same body
  • Mixed Frame Support: Seamlessly compose local and global frame wrenches

Migration Guide:

Old API (Deprecated):

# Old method - overwrites previous forces
asset.set_external_force_and_torque(
    forces=torch.ones(1, 1, 3),
    torques=torch.ones(1, 1, 3),
    body_ids=[0],
    env_ids=[0],
    is_global=False,
)

New API:

# Set initial permanent forces (replaces previous)
asset.permanent_wrench_composer.set_forces_and_torques(
    forces=torch.ones(1, 1, 3),
    env_ids=[0],
    body_ids=[0],
)
# Compose additional forces on the same body
asset.permanent_wrench_composer.add_forces_and_torques(
    forces=torch.ones(1, 1, 3),
    env_ids=[0],
    body_ids=[0],
    is_global=True,  # Mix local and global frames
)
# Add torques independently
asset.permanent_wrench_composer.add_forces_and_torques(
    torques=torch.ones(1, 1, 3),
    env_ids=[0],
    body_ids=[0],
)
# Apply forces and torques together with custom application points
asset.permanent_wrench_composer.add_forces_and_torques(
    forces=torch.ones(1, 1, 3),
    torques=torch.ones(1, 1, 3),
    positions=torch.ones(1, 1, 3),
    env_ids=[0],
    body_ids=[0],
)

Instantaneous Wrenches (New):

# Apply forces for a single simulation step only
asset.instantaneous_wrench_composer.add_forces_and_torques(
    forces=torch.ones(1, 1, 3),
    env_ids=[0],
    body_ids=[0],
)
# Multiple instantaneous wrenches compose automatically
asset.instantaneous_wrench_composer.add_forces_and_torques(
    forces=torch.ones(1, 2, 3),  # Add more forces
    env_ids=[0],
    body_ids=[0, 1],
)
# These are automatically reset after write_data_to_sim()

Key Differences:

  • set_forces_and_torques() replaces existing wrenches
  • add_forces_and_torques() composes with existing wrenches
  • Permanent and instantaneous wrenches compose automatically
  • Instantaneous wrenches auto-clear after each simulation step

Use Cases:

  • Drones: Compose thrust forces with aerodynamic drag and wind disturbances
  • Boats: Apply buoyancy forces with wave-induced motions

Formatting and Linting - Migration to Ruff

The project has migrated from multiple tools (flake8 for linting, black for formatting, isort for import sorting) to a unified toolchain using ruff for all formatting and linting tasks.

Related PRs: https://github.com/isaac-sim/IsaacLab/pull/4329, https://github.com/isaac-sim/IsaacLab/pull/4377, https://github.com/isaac-sim/IsaacLab/pull/4335, https://github.com/isaac-sim/IsaacLab/pull/4376

Why:

  • Faster performance (10-100x speedup)
  • Unified configuration in pyproject.toml
  • More consistent formatting and linting rules
  • Simplified developer workflow

Migration Steps:

  1. Update configuration files:
# Copy the updated configuration from the main branch
# Files to update: pyproject.toml, .pre-commit-config.yaml
  1. Apply new formatting:
./isaaclab.sh --format
  1. Resolve merge conflicts: If you encounter merge conflicts after updating, they likely originate from formatting differences. After copying the new configuration files, rerun the formatting command and commit the changes.

Pre-commit hooks will automatically run ruff on staged files. Ensure your code is formatted before committing to avoid CI failures.

USD Utilities - Unified isaaclab.sim.utils Module

USD Utilities - Unified isaaclab.sim.utils Module

Isaac Lab now provides its own comprehensive USD utility module (isaaclab.sim.utils) instead of relying on scattered utilities from Isaac Sim’s isaacsim.core.utils packages.

Related PR: https://github.com/isaac-sim/IsaacLab/pull/4286

Why:

  • Better Organization: All USD operations grouped into logical submodules (stage, prims, queries, transforms, semantics)
  • Type Hints: Full type annotations for better IDE support and code safety
  • Version Compatibility: Handles differences between Isaac Sim versions automatically

Old API (Isaac Sim utilities):

import isaac.core.utils.stage as stage_utils
import isaac.core.utils.prims as prim_utils
# Stage operations
stage_utils.create_new_stage()
current_stage = stage_utils.get_current_stage()
# Prim operations
prim_utils.create_prim("/World/Cube", "Cube")
prim_utils.delete_prim("/World/OldObject")

New API (Isaac Lab utilities):

import isaaclab.sim as sim_utils
# Stage operations
sim_utils.create_new_stage()
current_stage = sim_utils.get_current_stage()
# Prim operations
sim_utils.create_prim("/World/Cube", "Cube", attributes={"size": 1.0})
sim_utils.delete_prim("/World/OldObject")

Legacy Support:

For backward compatibility, legacy functions are still available in isaaclab.sim.utils.legacy, but it’s recommended to migrate to the new APIs or use USD directly.

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v2.3.1…v2.3.2

v2.3.1

What’s Changed

This is a small patch release with a few critical fixes that impacted user workflows.

Key fixes include:

  • The behavior of termination logging has changed in the manager-based workflow, where get_done_term now returns the current step value instead of the last episode value.
  • Additionally, a breaking change in the URDF importer was introduced in Isaac Sim 5.1, where the merge joints flag is no longer supported. We have now introduced a patch in the importer to return the behavior. Moving forward, we plan to deprecate this flag in favor of preserving asset definitions from URDFs directly without performing additional processing during the import process.

Bug Fixes

Documentation

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v2.3.0…v2.3.1

v2.3.0

What’s Changed

The Isaac Lab 2.3.0 release, built on Isaac Sim 5.1, delivers enhancements across dexterous manipulation, teleoperation, and learning workflows. It introduces new dexterous environments with advanced training capabilities, expands surface gripper and teleoperation support for a wider range of robots and devices, and integrates SkillGen with the Mimic imitation learning pipeline to enable GPU-accelerated motion planning and skill-based data generation with cuRobo integration.

Key highlights of this release include:

  • Dexterous RL (DexSuite): Introduction of two new dexterous manipulation environments using the Kuka arm and Allegro hand setup, with addition of support for Automatic Domain Randomization (ADR) and PBT (Population-Based Training).
  • Surface gripper updates: Surface gripper has been extended to support Manager-based workflows, including the addition of SurfaceGripperAction and SurfaceGripperActionCfg, along with several new environments demonstrating teleoperation examples with surface grippers and the RMPFlow controller. New robots and variations are introduced, including Franka and UR10 with robotiq grippers and suction cups, and Galbot and Agibot robots.
  • Mimic - SkillGen: SkillGen support has been added for the Mimic Imitation Learning pipeline, introducing cuRobo integration, integrating GPU motion planning with skill-segmented data generation. Note that cuRobo has proprietary licensing terms, please review the cuRobo license carefully before use.
  • Mimic - Locomanipulation: Added a new G1 humanoid environment combining RL-based locomotion with IK-based manipulation. A full robot navigation stack is integrated to augment demonstrations with randomization of tabletop pick/place locations, destination and ground obstacles. By segmenting tasks into pick-navigate-place phases, this method enables generation of large-scale loco-manipulation datasets from manipulation-only demonstrations.
  • Teleoperation: Upper body inverse kinematics controller is improved by adding a null space posture task that helps enable waist movement on humanoid tasks while regularizing redundant degrees-of-freedom to a preferred upright posture. Additionally, support for Vive and Manus Glove are introduced, providing more options for teleoperation devices.

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v2.2.1…v2.3.0

Isaac Sim 5.1 Updates

  • Introduced support for DGX Spark, including multi-architecture Docker images with support for ARM platforms.
  • PhysX now offers a new joint parameter tuning tutorial for robotic grippers, along with a new feature for solving articulation collision contacts last to improve on gripper penetration issues, especially for cases with sub-optimally tuned joints.
  • Surface grippers has been optimized for better performance. Although support continues to be CPU-only, performance has improved by several orders of magnitude compared to previous releases.
  • Windows 10 support ended on October 14, 2025. Microsoft will no longer provide free security, feature, or technical updates for Windows 10. As a result, we will be dropping support for Windows 10 in future releases of Isaac Sim and Lab to ensure the security and functionality of our software.

New Features

Core

Mimic and Teleoperation

Environments

Infrastructure

Improvements

Core and Infrastructure

Mimic and Teleoperation

Documentation

Core

Mimic and Teleoperation

Infrastructure

Bug Fixes

Core

Mimic and Teleoperation

Infrastructure

Breaking Changes

Known Issues

  • The ROS2 docker image is not currently expected to work due to the update to Python 3.11. We are actively working on a fix to resolve this.
  • We have received reports of performance regressions in the previous Isaac Sim release for both physics and rendering workflows. We are still working on addressing some of these, but have also found some workarounds. For viewport regressions, Omniverse settings can be set by adding --kit_args="--/app/usdrt/hierarchy/partialGpuUpdate=1 --/rtx/post/dlss/execMode=0 --/app/runLoops/main/rateLimitEnabled=false --/app/runLoops/main/manualModeEnabled=true --enable omni.kit.loop-isaac". Additionally, Isaac Sim 5.0 introduced new actuator models for PhysX, including drive model and friction model improvements. These improvements also introduced a small performance regression. We have observed up to ~20% slowdown in some state-based environments.

v2.2.1

Overview

This is a minor patch release with some improvements and bug fixes.

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v2.2.0…v2.2.1

New Features

  • Adds contact point location reporting to ContactSensor by @jtigue-bdai
  • Adds environments actions/observations descriptors for export by @AntoineRichard
  • Adds RSL-RL symmetry example for cartpole and ANYmal locomotion by @Mayankm96

Improvements

Core API

  • Enhances Pink IK controller with null-space posture control and improvements by @michaellin6
  • Adds periodic logging when checking USD path on Nucleus server by @matthewtrepte
  • Disallows string value written in sb3_ppo_cfg.yaml from being evaluated in process_sb3_cfg by @ooctipus

Infrastructure

  • Application Settings
    • Disables rate limit for headless and headless rendering app by @matthewtrepte, @kellyguo11
    • Disables rtx.indirrectDiffuse.enabled in render preset balanced and performance modes by @matthewtrepte
    • Sets profiler backend to NVTX by default by @soowanpNV, @rwiltz
  • Dependencies
    • Adds hf-xet license by @hhansen-bdai
    • Fixes new typing-inspection dependency license by @kellyguo11
  • Testing & Benchmarking
    • Adds basic validation tests for scale-based randomization ranges by @louislelay
    • Adds SensorBase tests by @jtigue-bdai
  • Repository Utilities
    • Adds improved readout from install_deps.py by @hhansen-bdai
    • Fixes isaaclab.sh to detect isaacsim_version accurately 4.5 or >= 5.0 by @ooctipus
    • Disables verbose printing in conftest.py by @ooctipus
    • Updates pytest flags for isaacsim integration testing by @ben-johnston-nv
    • Updates CodeOwners to be more fine-grained by @pascal-roth
    • Fixes minor issues in CI by @nv-apoddubny

Bug Fixes

Core API

  • Asset Interfaces
    • Fixes setting friction coefficients into PhysX in the articulation classes by @ossamaAhmed
    • Sets joint_friction_coeff only for selected physx_env_ids by @ashwinvkNV
  • Manager Interfaces
    • Fixes observation space Dict for non-concatenated groups only keeping the last term by @CSCSX
  • MDP Terms
    • Fixes termination term effort limit check logic by @moribots
    • Broadcasts environment ids inside mdp.randomize_rigid_body_com by @Foruck
    • Fixes IndexError in reset_joints_by_scale and reset_joints_by_offset by @Creampelt
    • Fixes terrain_out_of_bounds to return tensor instead of bool by @fan-ziqi

Infrastructure

  • Fixes distributed training hanging issue by @kellyguo11
  • Disables generation of internal template when detecting isaaclab install via pip by @ooctipus
  • Fixes typo in isaaclab.bat by @ooctipus
  • Updates app pathing for user-provided rendering preset mode by @matthewtrepte

Documentation

  • Adds documentation for Newton integration by @mpgussert
  • Adapts FAQ section in docs with Isaac Sim open-sourcing by @Mayankm96
  • Changes checkpoint path in rsl-rl to an absolute path in documentation by @fan-ziqi
  • Fixes MuJoCo link in docs by @fan-ziqi
  • Adds client version direction to XR document by @lotusl-code
  • Fixes broken link in doc by @kellyguo11
  • Fixes typo in list_envs.py script path by @fbeltrao
  • Fixes Franka blueprint env ID in docs by @louislelay

Breaking Changes

  • Improves termination manager logging to report aggregated percentage of environments done due to each term by @ooctipus

v2.2.0

Overview

Isaac Lab 2.2 brings major upgrades across simulation capabilities, tooling, and developer experience. It expands support for advanced physics features, new environments, and improved testing and documentation workflows. This release includes full compatibility with Isaac Sim 5.0 as well as backwards compatibility with Isaac Sim 4.5.

Key highlights of this release include:

  • Enhanced Physics Support: Updated joint friction modeling using the latest PhysX APIs, added support for spatial tendons, and improved surface gripper interactions.
  • New Environments for Imitation Learning: Introduction of two new GR1 mimic environments, with domain randomization and visual robustness evaluation, and improved pick-and-place tasks.
  • New Contact-Rich Manipulation Tasks: Integration of FORGE and AutoMate tasks for learning fine-grained contact interactions in simulation.
  • Teleoperation Improvements: Teleoperation tools have been enhanced with configurable parameters and CloudXR runtime updates, including head tracking and hand tracking.
  • Performance & Usability Improvements: Includes support for Stage in Memory and Cloning in Fabric for faster scene creation, new OVD recorder for large-scene GPU-based animation recording, and FSD (Fabric Scene Delegate) for improved rendering speed.
  • Improved Documentation: The documentation has been extended and updated to cover new features, resolve common issues, and streamline setup, including updates to teleop system requirements, VS Code integration, and Python environment management.

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v2.1.1…v2.2.0

Isaac Sim 5.0 Updates

  • Fixes rendering issues on Blackwell GPUs that previously resulted in overly noisy renders

  • Updates Python version from 3.10 to 3.11

  • Updates PyTorch version to torch 2.7.0+cu128, which will include Blackwell support

  • Drops official support for Ubuntu 20.04, we now officially support Ubuntu 22.04 and 24.04 Linux platforms

  • Isaac Sim 5.0 no longer sets /app/player/useFixedTimeStepping=False by default. We now do this in Isaac Lab.

  • isaaclab.sim.spawners.PhysicsMaterialCfg.improve_patch_friction is now removed. The simulation will always behave as if this attribute is set to true.

  • Native Livestreaming support has been removed. LIVESTREAM=1 can now be used for WebRTC streaming over public networks and LIVESTREAM=2 for private and local networks with WebRTC streaming.

  • Some assets in Isaac Sim have been reworked and restructured. Notably, the following asset paths were updated:

    • Robots/Ant/ant_instanceable.usd —> Robots/IsaacSim/Ant/ant_instanceable.usd
    • Robots/Humanoid/humanoid_instanceable.usd —> Robots/IsaacSim/Humanoid/humanoid_instanceable.usd
    • Robots/ANYbotics/anymal_instanceable.usd —> Robots/ANYbotics/anymal_c/anymal_c.usd
    • Robots/ANYbotics/anymal_c.usd —> Robots/ANYbotics/anymal_c/anymal_c.usd
    • Robots/Franka/franka.usd —> Robots/FrankaRobotics/FrankaPanda/franka.usd
    • Robots/AllegroHand/allegro_hand_instanceable.usd —> Robots/WonikRobotics/AllegroHand/allegro_hand_instanceable.usd
    • Robots/Crazyflie/cf2x.usd —> Robots/Bitcraze/Crazyflie/cf2x.usd
    • Robots/RethinkRobotics/sawyer_instanceable.usd —> Robots/RethinkRobotics/Sawyer/sawyer_instanceable.usd
    • Robots/ShadowHand/shadow_hand_instanceable.usd —> Robots/ShadowRobot/ShadowHand/shadow_hand_instanceable.usd

New Features

  • Adds FORGE tasks for contact-rich manipulation with force sensing to IsaacLab by @noseworm in #2968
  • Adds two new GR1 environments for IsaacLab Mimic by @peterd-NV
  • Adds stack environment, scripts for Cosmos, and visual robustness evaluation by @shauryadNv
  • Updates Joint Friction Parameters to Isaac Sim 5.0 PhysX APIs by @ossamaAhmed
  • Adds support for spatial tendons by @ossamaAhmed
  • Adds support and example for SurfaceGrippers by @AntoineRichard
  • Adds support for stage in memory by @matthewtrepte
  • Adds OVD animation recording feature by @matthewtrepte

Improvements

  • Enables FSD for faster rendering by @nv-mm
  • Sets rtx.indirectDiffuse.enabled to True for performance & balanced rendering presets by @matthewtrepte
  • Changes runner for post-merge pipeline on self-hosted runners by @nv-apoddubny
  • Fixes and improvements for CI pipeline by @nv-apoddubny
  • Adds flaky annotation for tests by @kellyguo11
  • Updates Mimic test cases to pytest format by @peterd-NV
  • Updates cosmos test files to use pytest by @shauryadNv
  • Updates onnx and protobuf version due to vulnerabilities by @kellyguo11
  • Updates minimum skrl version to 1.4.3 by @Toni-SM
  • Updates to Isaac Sim 5.0 by @kellyguo11
  • Updates docker CloudXR runtime version by @lotusl-code
  • Removes xr rendering mode by @rwiltz
  • Migrates OpenXRDevice from isaacsim.xr.openxr to omni.xr.kitxr by @rwiltz
  • Implements teleop config parameters and device factory by @rwiltz
  • Updates pick place env to use steering wheel asset by @peterd-NV
  • Adds a CLI argument to set epochs for Robomimic training script by @peterd-NV

Bug Fixes

  • Fixes operational space unit test to avoid pi rotation error by @ooctipus
  • Fixes GLIBC errors with importing torch before AppLauncher by @kellyguo11
  • Fixes rendering preset by @matthewtrepte in cc0dab6cd50778507efc3c9c2d74a28919ab2092
  • Fixes callbacks with stage in memory and organize environment tests by @matthewtrepte
  • Fixes XR and external camera bug with async rendering by @rwiltz
  • Disables selection for rl_games when marl is selected for template generator by @ooctipus
  • Adds check for .gitignore when generating template by @kellyguo11
  • Fixes camera obs errors in stack instance randomize envs by @peterd-NV
  • Fixes parsing for play envs by @matthewtrepte
  • Fixes issues with consecutive python exe calls in isaaclab.bat by @kellyguo11
  • Fixes spacemouse add callback function by @peterd-NV
  • Fixes humanoid training with new velocity_limit_sim by @AntoineRichard

Documentation

  • Adds note to mimic cosmos pipeline doc for eval by @shauryadNv
  • Updates teleop docs for 2.2 release by @rwiltz
  • Fixes outdated dofbot path in tutorial scripts by @mpgussert
  • Updates docs for VS Code IntelliSense setup and JAX installation by @Toni-SM
  • Updates Jax doc to overwrite version < 0.6.0 for torch by @kellyguo11
  • Adds docs for fabric cloning & stage in memory by @matthewtrepte
  • Updates driver requirements to point to our official technical docs by @mpgussert
  • Adds warning for ovd recording warning logs spam by @matthewtrepte
  • Adds documentation to specify HOVER version and known GLIBCXX error by @kellyguo11
  • Updates teleop system requirements doc by @lotusl-code
  • Add network requirements to cloudxr teleop doc by @lotusl-code

v2.1.1

Overview

This release has been in development over the past few months and includes a significant number of updates, enhancements, and new features across the entire codebase. Given the volume of changes, we’ve grouped them into relevant categories to improve readability. This version is compatible with NVIDIA Isaac Sim 4.5.

We appreciate the community’s patience and contributions in ensuring quality and stability throughout. We’re aiming for more frequent patch releases moving forward to improve the developer experience.

Note: This minor release does not include a Docker image or pip package.

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v2.1.0…v2.1.1

New Features

  • Asset Interfaces
    • Adds position argument to set external forces and torques at different locations on the rigid body by @AntoineRichard
    • Adds body_incoming_joint_wrench_b to ArticulationData field by @jtigue-bdai
    • Allows selecting articulation root prim explicitly by @lgulich
  • Sensor Interfaces
    • Draws connection lines for FrameTransformer visualization by @Mayankm96
    • Uses visualization marker for connecting lines inside FrameTransformer by @bikcrum
  • MDP Terms
    • Adds body_pose_w and body_projected_gravity_b observations by @jtigue-bdai
    • Adds joint effort observation by @jtigue-bdai
    • Adds CoM randomization term to manager-based events by @shendredm
    • Adds time-based mdp (observation) functions by @TheIndoorDad
    • Adds curriculum mdp term to modify any environment parameters by @ooctipus
  • New Example Tasks
    • Adds assembly tasks from the Automate project by @yijieg
    • Adds digit locomotion examples by @lgulich

Improvements

Core API

  • Actuator Interfaces
    • Fixes implicit actuator limits configs for assets by @ooctipus
    • Updates actuator configs for Franka arm by @reeceomahoney
  • Asset Interfaces
    • Optimizes getters of data inside asset classes by @Mayankm96
    • Adds method to set the visibility of the Asset’s prims by @Mayankm96
  • Sensor Interfaces
    • Updates to ray caster ray alignment and customizable drift sampling by @jsmith-bdai
    • Extends ContactSensorData by force_matrix_w_history attribute by @bikcrum
    • Adds IMU projected_gravity_b and optimizations by @jtigue-bdai
  • Manager Interfaces
    • Adds serialization to observation and action managers by @jsmith-bdai
    • Adds concatenation dimension to ObservationManager by @pascal-roth
    • Supports composite observation space with min/max by @ooctipus
    • Changes counter update in CommandManager by @pascal-roth
    • Integrates NoiseModel to manager-based workflows by @ozhanozen
    • Updates NoiseModelWithAdditiveBias to apply per-feature bias by @ozhanozen
    • Fixes isaaclab.scene.reset_to to accept None by @ooctipus
    • Resets step reward buffer properly by @bikcrum
  • Terrain Generation
    • Custom TerrainGenerator support by @pascal-roth
    • Adds terrain border options by @pascal-roth
    • Platform height independent of object height by @jtigue-bdai
    • Adds noise to MeshRepeatedObjectsTerrain by @jtigue-bdai
  • Simulation
    • Raises exceptions from SimContext init callbacks
    • Applies semantic_tags to ground by @KumoLiu
    • Sets enable_stabilization to false by default by @AntoineRichard
    • Fixes deprecation for pxr.Semantics by @kellyguo11
  • Math Utilities
    • Improves euler_xyz_from_quat by @ShaoshuSu
    • Optimizes yaw_quat by @hapatel-bdai
    • Changes quat_apply and quat_apply_inverse by @jtigue-bdai
    • Changes quat_box_minus by @jtigue-bdai
    • Adds quat_box_plus and rigid_body_twist_transform by @jtigue-bdai
    • Adds math tests for transforms by @jtigue-bdai
  • General Utilities
    • Simplifies buffer validation for CircularBuffer by @Mayankm96
    • Modifies update_class_from_dict() by @ozhanozen
    • Allows slicing from list values in dicts by @LinghengMeng @kellyguo11

Tasks API

  • Adds support for module:task and gymnasium >=1.0 by @kellyguo11
  • Adds RL library error hints by @Toni-SM
  • Enables hydra for play.py scripts by @ooctipus
  • Fixes ray metric reporting and hangs by @ozhanozen
  • Adds gradient clipping for distillation (RSL-RL) by @alessandroassirelli98
  • GRU-based RNNs ONNX export in RSL RL by @WT-MM
  • Adds wandb support in rl_games by @ooctipus
  • Optimizes SB3 wrapper by @araffin
  • Enables SB3 checkpoint loading by @ooctipus
  • Pre-processes SB3 env image obs-space for CNN pipeline by @ooctipus

Infrastructure

  • Dependencies
    • Updates torch to 2.7.0 with CUDA 12.8 by @kellyguo11
    • Updates gymnasium to 1.2.0 by @kellyguo11
    • Fixes numpy version to <2 by @ooctipus
    • Adds license file for OSS by @kellyguo11
    • Sets robomimic to v0.4.0 by @masoudmoghani
    • Upgrades pillow for Kit 107.3.1 by @ooctipus
    • Removes protobuf upper pin by @kwlzn
  • Docker
    • Uses --gpus instead of Nvidia runtime by @yanziz-nvidia
    • Adds docker name suffix parameter by @zoemcc
    • Adds bash history support in docker by @AntoineRichard
  • Testing & Benchmarking
    • Switches unittest to pytest by @kellyguo11 @pascal-roth
    • Adds training benchmark unit tests by @matthewtrepte
    • Fixes env and IK test failures by @kellyguo11
  • Repository Utilities
    • Adds URDF to USD batch conversion script by @hapatel-bdai
    • Adds repository citation link by @kellyguo11
    • Adds pip install warning for internal templates by @ooctipus

Bug Fixes

Core API

  • Actuator Interfaces
    • Fixes DCMotor clipping for negative power by @jtigue-bdai
  • Asset Interfaces
    • Fixes inconsistent data reads for body/link/com by @ooctipus
  • Sensor Interfaces
    • Fixes pose update in Camera and TiledCamera by @pascal-roth
    • Fixes CPU fallback in camera.py by @renaudponcelet
    • Fixes camera intrinsics logic by @jtigue-bdai
  • Manager Interfaces
    • Fixes ObservationManager buffer overwrite by @patrickhaoy
    • Fixes term check in event manager by @miguelalonsojr
    • Fixes Modifiers and history buffer bug by @ZiwenZhuang
    • Fixes re-init check in ManagerBase by @Mayankm96
    • Fixes CPU collision filtering by @kellyguo11
    • Fixes imports in InteractiveScene/LiveVisualizer by @Mayankm96
    • Fixes image plot import in Live Visualizer by @pascal-roth
  • MDP Terms
    • Fixes CoM randomization shape mismatch by @shendredm
    • Fixes visual prim handling in texture randomization by @KumoLiu
    • Resets joint targets in reset_scene_to_default by @wghou
    • Fixes joint limit terminations by @GiulioRomualdi
    • Fixes joint reset scope in SceneEntityCfg by @ooctipus
  • Math Utilities
    • Fixes quat_inv() implementation by @ozhanozen

Tasks API

  • Fixes LSTM to ONNX export by @jtigue-bdai

Example Tasks

  • Removes contact termination redundancy by @louislelay
  • Fixes memory leak in SDF by @leondavi
  • Changes randomization to events in Digit envs by @fan-ziqi

Documentation

  • Adds Isaac Sim version section to README by @kellyguo11
  • Adds physics performance guide by @kellyguo11
  • Adds jetbot tutorial to walkthrough docs by @mpgussert
  • Changes quickstart install to conda by @mpgussert
  • Fixes typo in library docs by @norbertcygiert
  • Updates docs for conda, fabric, inference by @kellyguo11
  • Adds license/contributing updates with DCO by @kellyguo11
  • Updates pytest docs and help by @louislelay
  • Adds actuator reference docs by @AntoineRichard
  • Updates multi-GPU PyTorch setup docs by @Alex-Omar-Nvidia
  • Removes deprecated env var in docs by @Kyu3224

v2.1.0

Overview

This release introduces the official support for teleoperation using the Apple Vision Pro for collecting high-quality and dexterous hand data, including the addition of bi-manual teleoperation and imitation learning workflows through Isaac Lab Mimic.

We have also introduced new randomization methods for USD attributes, including the randomization of scale, color, and textures. In this release, we updated RSL RL to v2.3.1, which introduces many additional features including distributed training, student-teacher distillation, and recurrent student-teacher distillation.

Additionally, we revamped the Extension Template to include an automatic template generator tool from within the Isaac Lab repo. The extension template is a powerful method for users to develop new projects in user-hosted repos, allowing for isolation from the core Isaac Lab repo and changes. The previous IsaacLabExtensionTemplate repo showed a limited example pertaining only to the Manager-based workflow and RSL RL. In the new template generator, users can choose from any supported workflow and RL library, along with the desired RL algorithm. We will be deprecating the standalone IsaacLabExtensionTemplate in the near future.

NVIDIA has also released HOVER as an independent repo, hosting a neural whole body controller for humanoids built on top of Isaac Lab. HOVER includes sim-to-real workflows for deployment on the Unitree H1 robot, which we have also added a tutorial guide for the deployment process in the Isaac Lab documentation.

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v2.0.2…v2.1.0

New Features

  • Adds new external project / internal task template generator by @Toni-SM
  • Adds dummy agents to the external task template generator by @louislelay
  • Adds USD-level randomization mode to event manager by @Mayankm96
  • Adds texture and scale randomization event terms by @hapatel-bdai
  • Adds replicator event for randomizing colors by @Mayankm96
  • Adds interactive demo script for H1 locomotion by @kellyguo11
  • Adds blueprint environment for Franka stacking mimic by @chengronglai
  • Adds action clipping to rsl-rl wrapper by @Mayankm96
  • Adds Gymnasium spaces showcase tasks by @Toni-SM
  • Add configs and adapt exporter for RSL-RL distillation by @ClemensSchwarke
  • Adds support for head pose for Open XR device by @rwiltz
  • Adds handtracking joints and retargetting pipeline by @rwiltz
  • Adds documentation for openxr device and retargeters by @rwiltz
  • Adds tutorial for training & validating HOVER policy using Isaac Lab by @pulkitg01
  • Adds rendering mode presets by @matthewtrepte
  • Adds GR1 scene with Pink IK + Groot Mimic data generation and training by @ashwinvkNV
  • Adds absolute pose franka cube stacking environment for mimic by @rwiltz
  • Enables CloudXR OpenXR runtime container by @jaczhangnv
  • Adds a quick start guide for quick installation and introduction by @mpgussert

Improvements

  • Clarifies the default parameters in ArticulationData by @Mayankm96
  • Removes storage of meshes inside the TerrainImporter class by @Mayankm96
  • Adds more details about state in InteractiveScene by @Mayankm96
  • Mounts scripts to docker container by @Mayankm96
  • Initializes manager term classes only when sim starts by @Mayankm96
  • Updates to latest RSL-RL v2.3.0 release by @Mayankm96
  • Skips dependency installation for directories with no extension.toml by @jsmith-bdai
  • Clarifies layer instructions in animation docs by @tylerlum
  • Lowers the default number of environments for camera envs by @kellyguo11
  • Updates Rendering Mode guide in documentation by @matthewtrepte
  • Adds task instruction UI support for mimic by @chengronglai
  • Adds ExplicitAction class to track argument usage in AppLauncher by @nv-mhaselton
  • Allows physics reset during simulation by @oahmednv
  • Updates mimic to support multi-eef (DexMimicGen) data generation by @nvcyc

Bug Fixes

  • Fixes default effort limit behavior for implicit actuators by @jtigue-bdai
  • Fixes docstrings inconsistencies the code by @Bardreamaster
  • Fixes missing stage recorder extension for animation recorder by @kellyguo11
  • Fixes ground height in factory environment by @louislelay
  • Removes double definition of render settings by @pascal-roth
  • Fixes device settings in env tutorials by @Mayankm96
  • Changes default ground color back to dark grey by @Mayankm96
  • Initializes extras dict before loading managers by @kousheekc
  • Fixes typos in development.rst by @vi3itor
  • Fixes SE gamepad omniverse subscription API by @PinkPanther-ny
  • Fixes modify_action_space in RslRlVecEnvWrapper by @felipemohr
  • Fixes distributed setup in benchmarking scripts by @kellyguo11
  • Fixes typo RF_FOOT to RH_FOOT in tutorials by @likecanyon
  • Checks if success term exists before recording in RecorderManager by @peterd-NV
  • Unsubscribes from debug vis handle when timeline is stopped by @jsmith-bdai
  • Fixes wait time in play.py by using env.step_dt by @tylerlum
  • Fixes 50 series installation instruction to include torchvision by @kellyguo11
  • Fixes importing MotionViewer from external scripts by @T-K-233
  • Resets cuda device after each app.update call by @kellyguo11
  • Fixes resume flag in rsl-rl cli args by @Mayankm96

v2.0.2

Overview

This patch release focuses on improving actuator configuration and fixing key bugs while reverting unintended behavioral changes from v2.0.1. We strongly recommend switching to this new version if you’re migrating from a pre-2.0 release of Isaac Lab.

Key Changes:

  • Actuator Limit Handling: Introduced isaaclab.actuators.ActuatorBaseCfg.velocity_limit_sim and isaaclab.actuators.ActuatorBaseCfg.effort_limit_sim to clearly distinguish simulation solver limits from actuator model constraints. Reverted implicit actuator velocity limits to pre-v2.0 behavior
  • Simulation configuration update: Removed isaaclab.sim.SimulationCfg.disable_contact_processing flag to simplify behavior
  • Rendering configuration update: Reverted to pre-2.0 configuration to improve the quality of the render product
  • Tiled camera fixes: Fixed motion vector processing and added a hotfix for retrieving semantic images from the ~isaaclab.sensors.TiledCamera
  • WebRTC Support: Added IP specification for live-streaming

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v2.0.1…v2.0.2

New Features

  • Adds isaaclab.actuators.ActuatorBaseCfg.velocity_limit_sim and isaaclab.actuators.ActuatorBaseCfg.effort_limit_sim to actuator.
  • Adds WebRTC livestreaming support with IP specification.

Improvements

  • Adds guidelines and examples for code contribution
  • Separates joint state setters inside Articulation class
  • Implements deterministic evaluation for skrl’s multi-agent algorithms
  • Adds new extensions to pyproject.toml
  • Updates docs on Isaac Sim binary installation path and VSCode integration
  • Removes remaining deprecation warning in RigidObject deprecation
  • Adds security and show&tell notes to documentation
  • Updates docs for segmentation and 50 series GPUs
  • Adds workaround for semantic segmentation issue with tiled camera

Bug Fixes

  • Fixes offset from object obs for Franka stacking env when using parallel envs
  • Adds scene update to ManagerBasedEnv, DirectRLEnv, and MARL envs initialization
  • Loads actuator networks in eval() mode to prevent gradients
  • Fixes instructions on importing ANYmal URDF in docs
  • Fixes setting of root velocities in the event term ~isaaclab.mdp.reset_root_state_from_terrain
  • Fixes activate_contact_sensors when using ~isaaclab.sim.MultiUsdFileCfg
  • Fixes misalignment in motion vectors from ~isaaclab.sim.TiledCamera
  • Sets default tensor device to CPU for Camera rot buffer

Breaking Changes

  • Reverts the setting of joint velocity limits for implicit actuators
  • Removes disable_contact_processing flag from SimulationContext
  • Reverts to old render settings in kit experience files

Migration Guide

We strongly recommend reviewing the details to fully understand the change in behavior, as it may impact the deployment of learned policies. Please open an issue on GitHub if you face any problems.

Introduction of simulation’s effort and velocity limits parameters in ActuatorBaseCfg

We have introduced the configuration variables isaaclab.actuators.ActuatorBaseCfg.velocity_limit_sim and isaaclab.actuators.ActuatorBaseCfg.effort_limit_sim to the isaaclab.actuators.ActuatorBaseCfg to allow users to set the simulation joint velocity and effort limits through the actuator configuration class.

Previously, we were overusing the attributes isaaclab.actuators.ActuatorBaseCfg.velocity_limit and isaaclab.actuators.ActuatorBaseCfg.effort_limit inside the actuator configuration. A series of changes in-between led to a regression from v1.4.0 to v2.0.1 release of IsaacLab. To make this clearer to understand, we note the change in their behavior in a tabular form:

Actuator TypeAttributev1.4.0 Behaviorv2.0.1 Behavior
Implicit Implicit Explicit Explicitvelocity_limit effort_limit velocity_limit effort_limitIgnored, not set into simulation Set into simulation Used by actuator models (e.g., DC Motor), not set into simulation Used by actuator models, not set into simulationSet into simulation Set into simulation Used by actuator models (e.g., DC Motor), set into simulation Used by actuator models, set into simulation

Setting the limits from the configuration into the simulation directly affects the behavior of the underlying physics engine solver. This impact is particularly noticeable when velocity limits are too restrictive, especially in joints with high stiffness, where it becomes easier to reach these limits. As a result, the change in behavior caused previously trained policies to not function correctly in IsaacLab v2.0.1.

Consequently, we have reverted back to the prior behavior and added velocity_limit_sim and effort_limit_sim attributes to make it clear that setting those parameters means changing solver’s configuration. The new behavior is as follows:

AttributeImplicit ActuatorExplicit Actuator
velocity_limit effort_limit velocity_limit_sim effort_limit_simIgnored, not set into simulation Set into simulation (same as effort_limit_sim) Set into simulation Set into simulation (same as effort_limit)Used by the model (e.g., DC Motor), not set into simulation Used by the models, not set into simulation Set into simulation Set into simulation

Users are advised to use the xxx_sim flag if they want to directly modify the solver limits.

Removal of disable_contact_processing flag in SimulationCfg

Removal of disable_contact_processing flag in SimulationCfg

We have now removed the disable_contact_processing flag from the isaaclab.sim.SimulationCfg to not have the user worry about these intricacies of the simulator. The flag is always True by default unless a contact sensor is created (which will internally set this flag to False).

Previously, the flag disable_contact_processing led to confusion about its behavior. As the name suggests, the flag controls the contact reporting from the underlying physics engine, PhysX. Disabling this flag (note the double negation) means that PhysX collects the contact information from its solver and allows reporting them to the user. Enabling this flag means this operation is not performed and the overhead of it is avoided.

Many of our examples (for instance, the locomotion environments) were setting this flag to True which meant the contacts should not get reported. However, this issue was not noticed earlier since GPU simulation bypasses this flag, and only CPU simulation gets affected. Running the same examples on CPU device led to different behaviors because of this reason.

Existing users, who currently set this flag themselves, should receive a deprecated warning mentioning the removal of this flag and the switch to the new default behavior.

Switch to older rendering settings to improve render quality

With the IsaacLab 2.0.0 release, we switched to new render settings aimed at improving tiled-rendering performance, but at the cost of reduced rendering quality. This change particularly affected dome lighting in the scene, which is the default in many of our examples.

As reported by several users, this change negatively impacted render quality, even in cases where it wasn’t necessary (such as when recording videos of the simulation). In response to this feedback, we have reverted to the previous render settings by default to restore the quality users expected.

For users looking to trade render quality for speed, we will provide guidelines in the future.

v2.0.1

Overview

This release contains a small set of fixes and improvements.

The main change was to maintain combability with the updated library name for RSL RL, which breaks the previous installation methods for Isaac Lab. This release provides the necessary fixes and updates in Isaac Lab to accommodate for the name change and maintain compatibility with installation for RSL RL.

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v2.0.0…v2.0.1

Improvements

  • Switches to RSL-RL install from PyPI by @Mayankm96
  • Updates the script path in the document by @fan-ziqi
  • Disables extension auto-reload when saving files by @kellyguo11
  • Updates documentation for v2.0.1 installation by @kellyguo11

Bug Fixes

  • Fixes timestamp of com and link buffers when writing articulation pose to sim by @Jackkert
  • Fixes incorrect local documentation preview path in xdg-open command by @louislelay
  • Fixes no matching distribution found for rsl-rl (unavailable) by @samibouziri
  • Fixes reset of sensor drift inside the RayCaster sensor by @zoctipus

v2.0.0

Overview

Isaac Lab 2.0 brings some exciting new features, including a new addition to the Imitation Learning workflow with the Isaac Lab Mimic extension.

Isaac Lab Mimic provides the ability to automatically generate additional trajectories based on just a few human collected demonstrations, allowing for larger training datasets with less human effort. This work is based on the MimicGen work for Scalable Robot Learning using Human Demonstrations.

Additionally, we introduced a new set of AMP tasks based on Adversarial Motion Priors, training humanoid robots to walk, run, and dance.

Along with Isaac Lab 2.0, Isaac Sim 4.5 brings several new and breaking changes, including a full refactor of the Isaac Sim extensions, an improved URDF importer, an update to the PyTorch dependency to version 2.5.1, and many fixes for tiled rendering that now supports multiple tiled cameras at different resolutions.

To follow the refactoring in Isaac Sim, we made similar refactoring and restructuring changes to Isaac Lab. These breaking changes will no longer be compatible with previous Isaac Sim versions.

Please make sure to update to Isaac Sim 4.5 when using the Isaac Lab 2.0 release.

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v1.4.1…v2.0.0

Highlights from the Isaac Sim 4.5 release

  • Support for multiple TiledCamera instances and varying resolutions
  • Improved rendering performance by up to 1.2x
  • Faster startup time through optimizations in the Cloner class that improves startup time by 30%
  • Enhanced OmniPVD for debugging physics simulation, enabling capturing reinforcement learning simulation
  • Physics simulation performance optimizations improving throughput of up to 70%
  • Physics support for dedicated cylinder and cone geometry designed for robot wheels that is fully GPU accelerated
  • A new physics GPU filtering mechanism allowing co-location of reinforcement learning environments at the origin with minimal performance loss for scenes with limited collider counts
  • Improvements in simulation stability for mimic joints at high joint gains

New Features

  • Adds humanoid AMP tasks for direct workflow by @Toni-SM
  • Adds Isaac Lab Mimic based on MimicGen data generation for Imitation Learning by @peterd-NV @nvcyc @ashwinvkNV @karsten-nvidia
  • Adds consolidated demo script for showcasing recording and mimic dataset generation in real-time in one simulation script by @nvcyc
  • Adds Franka stacking environment for GR00T mimic by @peterd-NV @nvcyc
  • Adds option to filter collisions and real-time playback by @kellyguo11

Improvements

  • Adds a tutorial for policy inference in a prebuilt USD scene by @oahmednv
  • Adds unit tests for multi-tiled cameras by @matthewtrepte
  • Updates render setting defaults for better quality by @kellyguo11
  • Adds a flag to wait for texture loading completion when reset by @oahmednv
  • Adds pre-trained checkpoints and tools for generating and uploading checkpoints by @nv-cupright
  • Adds new denoiser optimization flags for rendering by @kellyguo11
  • Updates torch to 2.5.1 by @kellyguo11

Bug Fixes

  • Fixes external force buffers to set to zero when no forces/torques are applied by @matthewtrepte
  • Fixes RSL-RL package name in setup.py according to PyPI installation by @samibouziri

Breaking Changes

  • Updates the URDF and MJCF importers for Isaac Sim 4.5 by @Dhoeller19
  • Renames Isaac Lab extensions and folders by @kellyguo11
  • Restructures extension folders and removes old imitation learning scripts by @kellyguo11
  • Renames default conda and venv Python environment from isaaclab to env_isaaclab by @Toni-SM

We have identified a breaking feature for semantic segmentation and instance segmentation when using Camera and TiledCamera with instanceable assets. Since the Isaac Sim 4.5 / Isaac Lab 2.0 release, semantic and instance segmentation outputs only render the first tile correctly and produces blank outputs for the remaining tiles. We will be introducing a workaround for this fix to remove scene instancing if semantic segmentation or instance segmentation is required for Camera and TiledCamera until we receive a proper fix from Omniverse as part of the next Isaac Sim release.

Migration Guide

Renaming of Isaac Sim Extensions

Previously, Isaac Sim extensions have been following the convention of omni.isaac.*, such as omni.isaac.core. In Isaac Sim 4.5, Isaac Sim extensions have been renamed to use the prefix isaacsim, replacing omni.isaac. In addition, many extensions have been renamed and split into multiple extensions to prepare for a more modular framework that can be customized by users through the use of app templates.

Notably, the following commonly used Isaac Sim extensions in Isaac Lab are renamed as follow:

  • omni.isaac.cloner —> isaacsim.core.cloner
  • omni.isaac.core.prims —> isaacsim.core.prims
  • omni.isaac.core.simulation_context —> isaacsim.core.api.simulation_context
  • omni.isaac.core.utils —> isaacsim.core.utils
  • omni.isaac.core.world —> isaacsim.core.api.world
  • omni.isaac.kit.SimulationApp —> isaacsim.SimulationApp
  • omni.isaac.ui —> isaacsim.gui.components

Renaming of the URDF and MJCF Importers

Starting from Isaac Sim 4.5, the URDF and MJCF importers have been renamed to be more consistent with the other extensions in Isaac Sim. The importers are available on isaac-sim GitHub as open source projects.

Due to the extension name change, the Python module names have also been changed:

  • URDF Importer: isaacsim.asset.importer.urdf (previously omni.importer.urdf)
  • MJCF Importer: isaacsim.asset.importer.mjcf (previously omni.importer.mjcf)

From the Isaac Sim UI, both URDF and MJCF importers can now be accessed directly from the File > Import menu when selecting a corresponding .urdf or .xml file in the file browser.

Changes in URDF Importer

Isaac Sim 4.5 brings some updates to the URDF Importer, with a fresh UI to allow for better configurations when importing robots from URDF. As a result, the Isaac Lab URDF Converter has also been updated to reflect these changes. The isaaclab.sim.converters.UrdfConverterCfg includes some new settings, such as ~isaaclab.sim.converters.JointDriveCfg.PDGainsCfg and ~isaaclab.sim.converters.JointDriveCfg.NaturalFrequencyGainsCfg classes for configuring the gains of the drives.

One breaking change to note is that the isaaclab.sim.converters.UrdfConverterCfg.JointDriveCfg.gains attribute must be of class type ~isaaclab.sim.converters.JointDriveCfg.PDGainsCfg or ~isaaclab.sim.converters.JointDriveCfg.NaturalFrequencyGainsCfg.

The stiffness of the ~isaaclab.sim.converters.JointDriveCfg.PDGainsCfg must be specified, as such:

joint_drive=sim_utils.UrdfConverterCfg.JointDriveCfg(
    gains=sim_utils.UrdfConverterCfg.JointDriveCfg.PDGainsCfg(stiffness=None, damping=None)
)

The isaaclab.sim.converters.JointDriveCfg.NaturalFrequencyGainsCfg.natural_frequency attribute must be specified for ~isaaclab.sim.converters.JointDriveCfg.NaturalFrequencyGainsCfg.

Renaming of Isaac Lab Extensions and Folders

Corresponding to Isaac Sim 4.5 changes, we have also made some updates to the Isaac Lab directories and extensions. All extensions that were previously under source/extensions are now under the source/ directory directly. The source/apps and source/standalone folders have been moved to the root directory and are now called apps/ and scripts/.

Isaac Lab extensions have been renamed to:

  • omni.isaac.lab —> isaaclab
  • omni.isaac.lab_assets —> isaaclab_assets
  • omni.isaac.lab_tasks —> isaaclab_tasks

In addition, we have split up the previous source/standalone/workflows directory into scripts/imitation_learning and scripts/reinforcement_learning directories. The RSL RL, Stable-Baselines, RL_Games, SKRL, and Ray directories are under scripts/reinforcement_learning, while Robomimic and the new Isaac Lab Mimic directories are under scripts/imitation_learning.

To assist with the renaming of Isaac Lab extensions in your project, we have provided a simple script that will traverse through the source and docs directories in your local Isaac Lab project and replace any instance of the renamed directories and imports. Please use the script at your own risk as it will overwrite source files directly.

Restructuring of Isaac Lab Extensions

With the introduction of isaaclab_mimic, designed for supporting data generation workflows for imitation learning, we have also split out the previous wrappers folder under isaaclab_tasks to its own module, named isaaclab_rl. This new extension will contain reinforcement learning specific wrappers for the various RL libraries supported by Isaac Lab.

The new isaaclab_mimic extension will also replace the previous imitation learning scripts under the robomimic folder. We have removed the old scripts for data collection and dataset preparation in favor of the new mimic workflow. For users who prefer to use the previous scripts, they will be available in previous release branches.

Additionally, we have also restructured the isaaclab_assets extension to be split into robots and sensors subdirectories. This allows for clearer separation between the pre-defined configurations provided in the extension.

As an example, the following import:

from omni.isaac.lab_assets.anymal import ANYMAL_C_CFG

should be replaced with:

from isaaclab_assets.robots.anymal import ANYMAL_C_CFG

v1.4.1

Overview

This release contains a set of improvements and bug fixes.

Most importantly, we reverted one of the changes from the previous release to ensure the training throughput performance remains the same.

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v1.4.0…v1.4.1

This is the final release compatible with Isaac Sim 4.2. The next release will target Isaac Sim 4.5, which introduces breaking changes that will make Isaac Lab incompatible with earlier versions of Isaac Sim.

New Features

  • Adds documentation and demo script for IMU sensor by @mpgussert

Improvements

  • Removes deprecation for root_state_w properties and setters by @jtigue-bdai
  • Fixes MARL workflows for recording videos during training/inferencing by @Rishi-V
  • Adds body tracking option to ViewerCfg by @KyleM73
  • Fixes the joint_parameter_lookup type in RemotizedPDActuatorCfg to support list format by @fan-ziqi
  • Updates pip installation documentation to clarify options by @steple
  • Fixes docstrings in Articulation Data that report wrong return dimension by @zoctipus
  • Fixes documentation error for PD Actuator by @kellyguo11
  • Clarifies ray documentation and fixes minor issues by @garylvov
  • Updates code snippets in documentation to reference scripts by @mpgussert
  • Adds dict conversion test for ActuatorBase configs by @mschweig

Bug Fixes

  • Fixes JointAction not preserving order when using all joints by @T-K-233
  • Fixes event term for pushing root by setting velocity by @Mayankm96
  • Fixes error in Articulation where default_joint_stiffness and default_joint_damping are not correctly set for implicit actuator by @zoctipus
  • Fixes action reset of pre_trained_policy_action in navigation environment by @nicolaloi
  • Fixes rigid object’s root com velocities timestamp check by @ori-gadot
  • Adds interval resampling on event manager’s reset call by @Mayankm96
  • Corrects calculation of target height adjustment based on sensor data by @fan-ziqi
  • Fixes infinite loop in repeated_objects_terrain method by @nicolaloi
  • Fixes issue where the indices were not created correctly for articulation setters by @AntoineRichard

v1.4.0

Overview

Due to a great amount of amazing updates, we are putting out one more Isaac Lab release based off of Isaac Sim 4.2. This release contains many great new additions and bug fixes, including several new environments, distributed training and hyperparameter support with Ray, new live plot feature for Manager-based environments, and more.

We will now spend more focus on the next Isaac Lab release geared towards the new Isaac Sim 4.5 release coming soon. The upcoming release will contain breaking changes in both Isaac Lab and Isaac Sim and breaks backwards compatibility, but will come with many great fixes and improvements.

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v1.3.0…v1.4.0

New Features

  • Adds Factory contact-rich manipulation tasks to IsaacLab by @noseworm
  • Adds a Franka stacking ManagerBasedRLEnv by @peterd-NV
  • Adds recorder manager in manager-based environments by @nvcyc
  • Adds Ray Workflow: Multiple Run Support, Distributed Hyperparameter Tuning, and Consistent Setup Across Local/Cloud by @glvov-bdai
  • Adds OperationSpaceController to docs and tests and implement corresponding action/action_cfg classes by @ozhanozen
  • Adds null-space control option within OperationSpaceController by @ozhanozen
  • Adds observation term history support to Observation Manager by @jtigue-bdai
  • Adds live plots to managers by @pascal-roth

Improvements

  • Adds documentation and example scripts for sensors by @mpgussert
  • Removes duplicated TerminationsCfg code in G1 and H1 RoughEnvCfg by @fan-ziqi
  • Adds option to change the clipping behavior for all Cameras and unifies the default by @pascal-roth
  • Adds check that no articulation root API is applied on rigid bodies by @lgulich
  • Adds RayCaster rough terrain base height to reward by @Andy-xiong6
  • Adds position threshold check for state transitions by @DorsaRoh
  • Adds clip range for JointAction by @fan-ziqi

Bug Fixes

  • Fixes noise_model initialized in direct_marl_env by @NoneJou072
  • Fixes entry_point and kwargs in isaaclab_tasks README by @fan-ziqi
  • Fixes syntax for checking if pre-commit is installed in isaaclab.sh by @louislelay
  • Corrects fisheye camera projection types in spawner configuration by @command-z-z
  • Fixes actuator velocity limits propagation down the articulation root_physx_view by @jtigue-bdai
  • Computes Jacobian in the root frame inside the DifferentialInverseKinematicsAction class by @zoctipus
  • Adds transform for mesh_prim of ray caster sensor by @clearsky-mio
  • Fixes configclass dict conversion for torch tensors by @lgulich
  • Fixes error in apply_actions method in NonHolonomicAction action term. by @KyleM73
  • Fixes outdated sensor data after reset by @kellyguo11
  • Fixes order of logging metrics and sampling commands in command manager by @Mayankm96

Breaking Changes

  • Refactors pose and velocities to link frame and COM frame APIs by @jtigue-bdai

v1.3.0

Overview

This release will be a final release based on Isaac Sim 4.2 before the transition to Isaac Sim 4.5, which will likely contain breaking changes and no longer backwards compatible with Isaac Sim 4.2 and earlier. In this release, we introduce many features, improvements, and bug fixes, including IMU sensors, support for various types of gymnasium spaces, manager-based perception environments, and more.

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v1.2.0…v1.3.0

New Features

  • Adds IMU sensor by @pascal-roth
  • Add Camera Benchmark Tool and Allow Correct Unprojection of distance_to_camera depth image by @glvov-bdai
  • Creates Manager Based Cartpole Vision Example Environments by @glvov-bdai
  • Adds image extracted features observation term and cartpole examples for it by @glvov-bdai
  • Supports other gymnasium spaces in Direct workflow by @Toni-SM
  • Adds configuration classes for spawning different assets at prim paths by @Mayankm96
  • Adds a rigid body collection class by @Dhoeller19
  • Adds option to scale/translate/rotate meshes in the mesh_converter by @pascal-roth
  • Adds event term to randomize gains of explicit actuators by @MoreTore
  • Adds Isaac Lab Reference Architecture documentation by @OOmotuyi

Improvements

  • Expands functionality of FrameTransformer to allow multi-body transforms by @jsmith-bdai
  • Inverts SE-2 keyboard device actions (Z, X) for yaw command by @riccardorancan
  • Disables backward pass compilation of warp kernels by @Mayankm96
  • Replaces TensorDict with native dictionary by @Toni-SM
  • Improves omni.isaac.lab_tasks loading time by @Toni-SM
  • Caches PhysX view’s joint paths when processing fixed articulation tendons by @Toni-SM
  • Replaces hardcoded module paths with __name__ dunder by @Mayankm96
  • Expands observation term scaling to support list of floats by @pascal-roth
  • Removes extension startup messages from the Simulation App by @Mayankm96
  • Adds a render config to the simulation and tiledCamera limitations to the docs by @kellyguo11
  • Adds Kit command line argument support by @kellyguo11
  • Modifies workflow scripts to generate random seed when seed=-1 by @kellyguo11
  • Adds benchmark script to measure robot loading by @Mayankm96
  • Switches from carb to omni.log for logging by @Mayankm96
  • Excludes cache files from vscode explorer by @Divelix
  • Adds versioning to the docs by @sheikh-nv
  • Adds better error message for invalid actuator parameters by @lgulich
  • Updates tested docker and apptainer versions for cluster deployment by @pascal-roth
  • Removes ml_archive as a dependency of omni.isaac.lab extension by @fan-ziqi
  • Adds a validity check for configclasses by @Dhoeller19
  • Ensures mesh name is compatible with USD convention in mesh converter by @fan-ziqi
  • Adds sanity check for the term type inside the command manager by @command-z-z
  • Allows configclass to_dict operation to handle a list of configclasses by @jtigue-bdai

Bug Fixes

  • Disables replicate physics for deformable teddy lift environment by @Mayankm96
  • Fixes Jacobian joint indices for floating base articulations by @lorenwel
  • Fixes setting the seed from CLI for RSL-RL by @kaixi287
  • Fixes camera MDP term name and reprojection docstrings by @Mayankm96
  • Fixes deprecation notice for using pxr.Semantics by @Mayankm96
  • Fixes scaling of default ground plane by @kellyguo11
  • Fixes Isaac Sim executable on pip installation by @Toni-SM
  • Passes device from CLI args to simulation config in standalone scripts by @Mayankm96
  • Fixes the event for randomizing rigid body material by @pascal-roth
  • Fixes the ray_caster_camera tutorial script when saving the data by @mpgussert
  • Fixes running the docker container when the DISPLAY env variable is not defined by @GiulioRomualdi
  • Fixes default joint pos when setting joint limits by @kellyguo11
  • Fixes device propagation for noise and adds noise tests by @jtigue-bdai
  • Removes additional sbatch and fixes default profile in cluster deployment by @pascal-roth
  • Fixes the checkpoint loading error in RSL-RL training script by @bearpaw
  • Fixes pytorch broadcasting issue in EMAJointPositionToLimitsAction by @bearpaw
  • Fixes body IDs selection when computing feet_slide reward for locomotion-velocity task by @dtc103
  • Fixes broken URLs in markdown files by @DorsaRoh
  • Fixes net_arch in sb3_ppo_cfg.yaml for Isaac-Lift-Cube-Franka-v0 task by @LinghengMeng

v1.2.0

Overview

We leverage the new release of Isaac Sim, 4.2.0, and bring RTX-based tiled rendering, support for multi-agent environments, and introduce many bug fixes and improvements.

Additionally, we have published an example for generating rewards using an LLM based on Eureka, available here: https://github.com/isaac-sim/IsaacLabEureka

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v1.1.0…v1.2.0

New Features

  • Adds RTX-based tiled rendering. This improves the overall rendering speed and quality.
  • Adds the direct workflow perceptive Shadowhand Cube Repose environment Isaac-Repose-Cube-Shadow-Vision-Direct-v0 by @kellyguo11.
  • Adds support for multi-agent environments with the Direct workflow, with support for MAPPO and IPPO in SKRL by @Toni-SM
  • Adds the direct workflow multi-agent environments Isaac-Cart-Double-Pendulum-Direct-v0 and Isaac-Shadow-Hand-Over-Direct-v0 by @Toni-SM
  • Adds throughput benchmarking scripts for the different learning workflows by @kellyguo11
  • Adds results for the benchmarks in the documentation here for different types of hardware by @kellyguo11
  • Adds the direct workflow Allegro hand environment by @kellyguo11
  • Adds video recording to the play scripts in RL workflows by @j3soon
  • Adds comparison tables for the supported RL libraries here by @kellyguo11
  • Add APIs for deformable asset by @masoudmoghani
  • Adds support for MJCF converter by @qqqwan
  • Adds a function to define camera configs through intrinsic matrix by @pascal-roth
  • Adds configurable modifiers to observation manager by @jtigue-bdai
  • Adds the Hydra configuration system for RL training by @Dhoeller19

Improvements

  • Uses PhysX accelerations for rigid body acceleration data by @Mayankm96
  • Adds documentation on the frames for asset data by @Mayankm96
  • Renames Unitree configs in locomotion tasks to match properly by @Mayankm96
  • Adds option to set the height of the border in the TerrainGenerator by @pascal-roth
  • Adds a cli arg to run_all_tests.py for testing a selected extension by @jsmith-bdai
  • Decouples rigid object and articulation asset classes by @Mayankm96
  • Adds performance optimizations for domain randomization by @kellyguo11
  • Allows having hybrid dimensional terms inside an observation group by @Mayankm96
  • Adds a flag to preserve joint order inside JointActionCfg action term by @xav-nal
  • Adds the ability to resume training from a checkpoint with rl_games by @sizsJEon
  • Adds windows configuration to VS code tasks by @johnBuffer
  • Adapts A and D button bindings in the keyboard device by @zoctipus
  • Uses torch.einsum for quat_rotate and quat_rotate_inverse operations by @dxyy1
  • Expands on articulation test for multiple instances and devices by @jsmith-bdai
  • Adds setting of environment seed at initialization by @Mayankm96
  • Disables default viewport when headless but cameras are enabled by @kellyguo11
  • Simplifies the return type for parse_env_cfg method by @Mayankm96
  • Simplifies the if-elses inside the event manager apply method by @Mayankm96

Bug Fixes

  • Fixes rendering frame delays. Rendered images now faithfully represent the latest state of the physics scene. We added the flag rerender_on_reset in the environment configs to toggle an additional render step when a reset happens. When activated, the images/observation always represent the latest state of the environment, but this also reduces performance.
  • Fixes wrap_to_pi function in math utilities by @Mayankm96
  • Fixes setting of pose when spawning a mesh by @masoudmoghani
  • Fixes caching of the terrain using the terrain generator by @Mayankm96
  • Fixes running train scripts when rsl_rl is not installed by @Dhoeller19
  • Adds flag to recompute inertia when randomizing the mass of a rigid body by @Mayankm96
  • Fixes support for classmethod when defining a configclass by @Mayankm96
  • Fixes Sb3VecEnvWrapper to clear buffer on reset by @EricJin2002
  • Fixes venv and conda pip installation on windows by @kellyguo11
  • Sets native livestream extensions to Isaac Sim 4.1-4.0 defaults by @jtigue-bdai
  • Defaults the gym video recorder fps to match episode decimation by @ozhanozen
  • Fixes the event manager’s apply method by @kellyguo11
  • Updates camera docs with world units and introduces new test for intrinsics by @pascal-roth
  • Adds the ability to resume training from a checkpoint with rl_games by @sizsJEon

Breaking Changes

  • Simplifies device setting in SimulationCfg and AppLauncher by @Dhoeller19
  • Fixes conflict in teleop-device command line argument in scripts by @Dhoeller19
  • Converts container.sh into Python utilities by @hhansen-bdai
  • Drops support for TiledCamera for Isaac Sim 4.1

Migration Guide

Setting the simulation device into the simulation context

Previously, changing the simulation device to CPU required users to set other simulation parameters (such as disabling GPU physics and GPU pipelines). This made setting up the device appear complex. We now simplify the checks for device directly inside the simulation context, so users only need to specify the device through the configuration object.

Before:

sim_utils.SimulationCfg(device="cpu", use_gpu_pipeline=False, dt=0.01, physx=sim_utils.PhysxCfg(use_gpu=False))

Now:

sim_utils.SimulationCfg(device="cpu", dt=0.01, physx=sim_utils.PhysxCfg())

Setting the simulation device from CLI

Previously, users could specify the device through the command line argument --device_id. However, this made it ambiguous when users wanted to set the device to CPU. Thus, instead of the device ID, users need to specify the device explicitly through the argument --device. The valid options for the device name are:

  • “cpu”: runs simulation on CPU
  • “cuda”: runs simulation on GPU with device ID at default index
  • “cuda:N”: runs simulation on GPU with device ID at N. For instance, “cuda:0” will use device at index “0”.

Due to the above change, the command line interaction with some of the scripts has changed.

Before:

uv run python source/standalone/workflows/sb3/train.py --task Isaac-Cartpole --cpu

Now:

uv run python source/standalone/workflows/sb3/train.py --task Isaac-Cartpole --device cpu

Renaming of teleoperation device CLI in standalone scripts

Since --device is now an argument provided by the AppLauncher, it conflicted with the command-line argument used for specifying the teleoperation-device in some of the standalone scripts. Thus, to fix this conflict, the teleoperation-device now needs to be specified through --teleop_device argument.

Before:

uv run python source/standalone/environments/teleoperation/teleop_se3_agent.py --task IsaacContrib-Lift-Cube-Franka-IK-Rel --num_envs 1 --device keyboard

Now:

uv run python source/standalone/environments/teleoperation/teleop_se3_agent.py --task IsaacContrib-Lift-Cube-Franka-IK-Rel --num_envs 1 --teleop_device keyboard

Using Python-version of container utility script

The prior container.sh became quite complex as it had many different use cases in one script. For instance, building a docker image for “base” or “ros2”, as well as cluster deployment. As more users wanted to have the flexibility to overlay their own docker settings, maintaining this bash script became cumbersome. Hence, we migrated its features into a Python script in this release. Additionally, we split the cluster-related utilities into their own script inside the docker/cluster directory.

We still maintain backward compatibility for container.sh. Internally, it calls the Python script container.py. We request users to use the Python script directly.

Before:

./docker/container.sh start

Now:

./docker/container.py start

Using separate directories for logging videos in RL workflows

Previously, users could record videos during the RL training by specifying the --video flag to the train.py script. The videos would be saved inside the videos directory in the corresponding log directory of the run.

Since many users requested to also be able to record videos while inferencing the policy, recording videos have also been added to the play.py script. Since changing the prefix of the video file names is not possible, the videos from the train and play scripts are saved inside the videos/train and videos/play directories, respectively.

Drops support for the tiled camera with Isaac Sim 4.1

Various fixes have been made to the tiled camera rendering pipeline in Isaac Sim 4.2. This made supporting the tiled camera with Isaac Sim 4.1 difficult. Hence, for the best experience, we advice switching to Isaac Sim 4.2 with this release of Isaac Lab.

v1.1.0

Overview

With the release of Isaac Sim 4.0 and 4.1, support for Isaac Sim 2023.1.1 has been discontinued. We strongly encourage all users to upgrade to Isaac Sim 4.1 to take advantage of the latest features and improvements. For detailed information on this upgrade, please refer to the release notes available here.

Besides the above, the Isaac Lab release brings new features and improvements, as detailed below. We thank all our contributors for their continued support.

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v1.0.0…v1.1.0

New Features

  • Adds distributed multi-GPU learning support for skrl by @Toni-SM
  • Updates skrl integration to support training/evaluation using JAX by @Toni-SM
  • Adds lidar pattern for raycaster sensor by @pascal-roth
  • Adds support for PBS job scheduler-based clusters by @shafeef901
  • Adds APIs for spawning deformable meshes by @Mayankm96

Improvements

  • Changes documentation color to the green theme by @Mayankm96
  • Fixes sphinx tabs to make them work in dark theme by @Mayankm96
  • Fixes VSCode settings to work with pip installation of Isaac Sim by @Mayankm96
  • Fixes isaaclab scripts to deal with Isaac Sim pip installation by @Mayankm96
  • Optimizes interactive scene for homogeneous cloning by @kellyguo11
  • Improves docker X11 forwarding documentation by @j3soon

Bug Fixes

  • Reads gravity direction from simulation inside RigidObjectData by @Mayankm96
  • Fixes reference count in asset instances due to circular references by @Mayankm96
  • Fixes issue with asset deinitialization due to torch > 2.1 by @Mayankm96
  • Fixes the rendering logic regression in environments by @Dhoeller19
  • Fixes the check for action-space inside Stable-Baselines3 wrapper by @Mayankm96
  • Fixes warning message in Articulation config processing by @locoxsoco
  • Fixes action term in the reach environment by @masoudmoghani
  • Fixes training UR10 reach with RL_GAMES and SKRL by @sudhirpratapyadav
  • Adds event manager call to simple manage-based env by @Mayankm96

Breaking Changes

  • Drops official support for Isaac Sim 2023.1.1
  • Removes the use of body view inside the asset classes by @Mayankm96
  • Renames SimulationCfg.substeps to SimulationCfg.render_interval by @Dhoeller19

Migration Guide

Renaming of SimulationCfg.substeps

Renaming of SimulationCfg.substeps

Previously, the users set both omni.isaac.lab.sim.SimulationCfg.dt and omni.isaac.lab.sim.SimulationCfg.substeps, which marked the physics insulation time-step and sub-steps, respectively. It was unclear whether sub-steps meant the number of integration steps inside the physics time-step dt or the number of physics steps inside a rendering step.

Since in the code base, the attribute was used as the latter, it has been renamed to render_interval for clarity.

Removal of Deprecated Attributes

As notified in previous releases, we removed the classes and attributes marked as deprecated. These are as follows:

  • The mdp.add_body_mass method in the events. Please use the mdp.randomize_rigid_body_mass instead.

  • The classes managers.RandomizationManager and managers.RandomizationTermCfg. Please use the managers.EventManager and managers.EventTermCfg classes instead.

  • The following properties in omni.isaac.lab.sensors.FrameTransformerData:

    • target_rot_source —> target_quat_w
    • target_rot_w —> target_quat_source
    • source_rot_w —> source_quat_w

  • The attribute body_physx_view from the omni.isaac.lab.assets.Articulation and omni.isaac.lab.assets.RigidObject classes. These caused confusion when used with the articulation view since the body names did not follow the same ordering.

v1.0.0

Overview

Welcome to the first official release of Isaac Lab!

Building upon the foundation of the Orbit framework, we have integrated the RL environment designing workflow from OmniIsaacGymEnvs. This allows users to choose a suitable task-design approach for their applications.

While we maintain backward compatibility with Isaac Sim 2023.1.1, we highly recommend using Isaac Lab with Isaac Sim 4.0.0 version for the latest features and improvements.

Full Changelog: https://github.com/isaac-sim/IsaacLab/compare/v0.3.1…v1.0.0

New Features

  • Integrated CI/CD pipeline, which is triggered on pull requests and publishes the results publicly
  • Extended support for Windows OS platforms
  • Added tiled render based Camera sensor implementation. This provides optimized RGB-D rendering throughputs of up to 10k frames per second.
  • Added support for multi-GPU and multi-node training for the RL-Games library
  • Integrated APIs for environment designing (direct workflow) without relying on managers
  • Added implementation of delayed PD actuator model
  • Added various new learning environments:
    • Cartpole balancing using images
    • Shadow hand cube reorientation
    • Boston Dynamics Spot locomotion
    • Unitree H1 and G1 locomotion
    • ANYmal-C navigation
    • Quadcopter target reaching

Improvements

  • Reduced start-up time for scripts (inherited from Isaac Sim 4.0 improvements)
  • Added lazy buffer implementation for rigid object and articulation data. Instead of updating all the quantities at every step call, the lazy buffers are updated only when the user queries them
  • Added SKRL support to more environments