Robotics startups are boosting their capabilities with the help of NVIDIA’s Isaac Sim, leveraging cloud-based technologies offered by Amazon Web Services (AWS).
Prominent names such as Field AI, Vention and Cobot are leveraging these tools to push the boundaries of technology, enabling innovations in areas ranging from industrial processes to collaborative robots.
Field AI focuses on developing fundamental robotic brains capable of autonomously managing complex industrial workflows. Vention simplifies the development of robotic tasks by providing pre-trained skills, while Cobot unveiled Proxie, an AI-driven collaborative robot designed to seamlessly navigate dynamic environments alongside humans.
The common core of these advancements is NVIDIA’s Isaac Sim, a reference application powered by NVIDIA Omniverse for simulating and testing AI-enabled robots in hyper-realistic virtual environments. The platform is now enhanced with new cloud-based performance enhancements.
AWS and NVIDIA GPU: a powerful combo for Isaac Sim’s innovation
Unveiled at the AWS re:Invent event, NVIDIA announced the deployment of Isaac Sim on Amazon Elastic Cloud Computing (EC2) G6e instances, accelerated by NVIDIA’s latest L40S GPUs. These instances deliver twice the compute performance of previous generations and provide increased flexibility as developers tackle increasingly complex robotics use cases.
Coupled with NVIDIA OSMO – a cloud-native orchestration platform – developers can streamline and scale the complex workflows involved in robotics development, whether using AWS cloud infrastructure or on-premises systems.
This unified framework of high-performance hardware and software transforms robotics innovation for teams large and small, enabling the development of “physical AI.”
Physical AI, an emerging concept, describes intelligent models trained to understand and interact with the physical environment, thereby driving the advancement of autonomous machines such as humanoid robots, autonomous vehicles, and industrial systems.
The role of simulation in the development of robotics
Robotic infrastructure relies on robust training data to ensure the accuracy of AI physical models. Collecting such datasets from real-world environments is often prohibitively expensive and logistically challenging. Virtual simulations, however, offer a practical solution, accelerating the training of AI models while optimizing processes before real-world deployment.
Simulation plays a central role in the validation of robotic systems, facility design testing, and the development of computer vision AI models.
Amazon EC2 G6e instances, supported by NVIDIA L40S GPUs play a crucial role in this workflow, with enhanced data generation, simulation, and model optimization capabilities. These solutions enable developers to improve operational efficiency while minimizing costly errors during the design and implementation phases.
By integrating NVIDIA OSMO into cloud workflows, teams can synchronize their robot development projects, leveraging solutions like Isaac Sim for cutting-edge collaboration. Developers can generate synthetic data, create 3D assets, and streamline workflows using NVIDIA Omniverse Replicator, a framework that integrates generative AI tools.
Generative AI meets robotics training
The ability to generate synthetic data has become the cornerstone of the accelerated development of robotics.
NVIDIA’s solutions use generative AI to create data-driven workflows, encompassing tools such as USD Code NIM for Python scripting and 3D asset manipulation, Edify HDRi for generating virtual environments, and Edify 3D to create 3D object files ready for editing from text or image prompts.
Rendered.ai, for example, uses Omniverse Replicator to create synthetic computer vision datasets, serving industries ranging from manufacturing and agriculture to surveillance and security. Similarly, Tata Consultancy Services uses synthetic data pipelines to power its Mobility AI suite, which focuses on automotive use cases ranging from fault detection to hazard avoidance.
This technological leap reduces tedious manual steps, enables scalable data production and promotes the development of robotic systems capable of meeting increasingly sophisticated challenges.
Real-world use cases leveraging Isaac Sim
A growing number of startups and companies are capitalizing on Isaac Sim’s simulation and synthetic data capabilities to improve their robotics projects.
For example, Aescape leverages Isaac Sim to refine sensors in robots used to deliver precision massages. Similarly, Cobot used Isaac Sim to optimize its logistics-focused collaborative robot, Proxie, suitable for dynamic industries such as manufacturing, healthcare and warehousing.
Other notable adopters of Isaac Sim include:
- Field AI: Uses Isaac Lab, an open source robotics learning platform, to test robotic systems in unstructured environments in industries such as construction, manufacturing and mining.
- Agreement: Uses Isaac Sim to develop robotic cell capabilities for small and medium manufacturers.
- Swiss thousand: Integrates Isaac Lab and Isaac Sim for robotics learning, enhancing the capabilities of wheeled quadruped robots in navigating factory and warehouse tasks.
- Standard bots: Simulates and validates the performance of its R01 manufacturing robot in real conditions.
- Software server: Collaborates with food producer Pfeifer & Langen to develop agricultural robots optimized for vertical installations.
- Cohesive robotics: Integrates Isaac Sim into its Argus OS software, which powers robotic work cells for high-variation manufacturing systems.
The repeatability and control of virtual learning environments significantly reduces hardware testing cycles, while ensuring that deployment-ready solutions emerge more quickly.
By combining NVIDIA Isaac Sim with the AWS cloud ecosystem, developers around the world are advancing robotics innovation. This integration of high-performance simulation, generative AI tools, and scalable infrastructure ensures that robotics can tackle increasingly complex scenarios in diverse environments.
Robotic systems are expected to evolve rapidly, with fewer limitations in testing, validation and optimization. As a result, experts anticipate an acceleration in physical applications of AI.
(Image credit: Nvidia)
See also: Figure 02: A leap forward in humanoid robotics
Want to learn more about IoT from industry leaders? Check IoT technology fair taking place in Amsterdam, California and London. The full event is co-located with other premier events including Cybersecurity and Cloud Expo, AI and Big Data Exhibition, Intelligent Automation Conference, Advanced Computing ExpoAnd Digital Transformation Week.
Check out more upcoming enterprise technology events and webinars from TechForge here.