NVIDIA Releases Developer Previews Of Isaac Sim 5.0 And Isaac Lab 2.2 On GitHub
In Brief
NVIDIA has released developer previews of Isaac Sim 5.0 and Isaac Lab 2.2 on GitHub, introducing advanced simulation, synthetic data generation, and improved robotics AI training tools to enhance the realism, efficiency, and accuracy of robot development in virtual environments.
Technology firm NVIDIA has introduced developer previews of two robotics-focused platforms—Isaac Sim and Isaac Lab—offering early-stage access to tools for simulating, training, and testing AI-powered robots in physics-accurate virtual environments. These frameworks, now publicly available on GitHub, are intended to support robotics development through advanced simulation workflows and synthetic data generation.
Isaac Sim, built on the NVIDIA Omniverse platform, functions as a reference application for creating and evaluating intelligent robotic systems in realistic digital settings. The 5.0 version includes an important update: the open sourcing of Isaac Sim-specific extensions. A dedicated GitHub repository now allows developers to access and utilize these components, although core elements of the Omniverse Kit remain proprietary, and external contributions are currently not accepted. This public development branch will be regularly maintained with incremental updates and bug fixes. Full binary releases aligned with NVIDIA’s quality assurance standards will be made available upon the official 5.0 launch.
The update introduces several new capabilities designed to enhance synthetic data workflows for robotics AI development. This includes MobilityGen, an extension that enables the creation of diverse, physics-informed datasets for perception models—such as occupancy maps, velocity fields, poses, and imagery. A new tutorial facilitates automated grasp data generation, simulating grasp attempts and capturing performance metrics to support model training and evaluation.
An additional writer for Omniverse Replicator has been optimized for compatibility with NVIDIA Cosmos Transfer input. This enhancement simplifies the export of synthetic datasets and supports both standalone and Script Editor-driven workflows, allowing seamless integration with existing Isaac Sim pipelines.
Expanded functionality also includes new workflows that enable synthetic data generation across larger and more complex physical environments. Incident Simulation introduces a variety of real-world scenario triggers, while Caption Generation provides automated annotations to enrich datasets used for vision-based AI. These new tools, along with refinements to existing Actor and Object Simulation processes, are designed to improve the efficiency and fidelity of training data production for robotics and AI applications.
Isaac Sim 5.0 brings forward enhancements aimed at improving the efficiency and realism of robotics simulation, particularly through updated robot modeling standards and streamlined import processes. A newly introduced robot schema establishes a unified format for defining robots using OpenUSD, which has now been implemented across NVIDIA’s growing library of robotic assets, including recently added models. Robot importers have been updated to align with this schema, and a newly developed Robot Import Wizard provides a guided, step-by-step interface to simplify the rigging and integration of robots into simulation environments.
The latest schema also includes support for an updated joint friction model specified within the OpenUSD framework. This model incorporates detailed actuator and friction parameters, created in collaboration with Hexagon Robotics and maxon, to better replicate the mechanical behavior of physical systems. By capturing realistic joint and motor dynamics, this feature helps narrow the gap between simulated and real-world performance, making virtual testing environments more reliable for application development.
Isaac Sim 5.0 also advances the platform’s sensor simulation capabilities, allowing for more accurate definition and evaluation of sensor models. A new OmniSensor USD schema enables native specification of RTX-based sensors within the USD format. This is accompanied by the introduction of a depth sensor model that delivers realistic stereo depth simulation, including disparity artifacts to reflect actual sensor limitations. The sensor asset library has also been extended with additional models to support a broader range of simulation needs, improving control and fidelity across perception system development workflows.
Key Innovations In Isaac Lab
Isaac Lab, NVIDIA’s open-source framework designed specifically for training and assessing robotic learning models, has received updates in version 2.2. One key addition is support for GR00T N1 benchmarking, which facilitates a closed-loop process involving data generation, post-training procedures, and evaluation of NVIDIA’s Isaac GR00T N open foundation models. This functionality can be applied to structured tasks such as nut pouring and pipe sorting, offering a consistent methodology for performance measurement.
The update also enhances synthetic motion generation, enabling more reliable data collection in robot manipulation tasks, particularly when using NVIDIA Isaac GR00T-Mimic’s bi-manual training environment. Integration with Omniverse Fabric further improves Isaac Lab’s efficiency, resulting in reduced loading times and more optimized execution of physics simulations and sensor data acquisition. A newly introduced tensorized suction cup gripper expands the toolkit available for simulation, providing a more adaptable and realistic approach to modeling gripping behavior—an essential feature for accurate reinforcement learning in robotic manipulation scenarios.
Disclaimer
In line with the Trust Project guidelines, please note that the information provided on this page is not intended to be and should not be interpreted as legal, tax, investment, financial, or any other form of advice. It is important to only invest what you can afford to lose and to seek independent financial advice if you have any doubts. For further information, we suggest referring to the terms and conditions as well as the help and support pages provided by the issuer or advertiser. MetaversePost is committed to accurate, unbiased reporting, but market conditions are subject to change without notice.
About The Author
Alisa, a dedicated journalist at the MPost, specializes in cryptocurrency, zero-knowledge proofs, investments, and the expansive realm of Web3. With a keen eye for emerging trends and technologies, she delivers comprehensive coverage to inform and engage readers in the ever-evolving landscape of digital finance.
More articles
Alisa, a dedicated journalist at the MPost, specializes in cryptocurrency, zero-knowledge proofs, investments, and the expansive realm of Web3. With a keen eye for emerging trends and technologies, she delivers comprehensive coverage to inform and engage readers in the ever-evolving landscape of digital finance.
