NVIDIA Isaac Sim

Isaac Sim is an open-source framework for simulating and testing robots in physics-based virtual environments and generating synthetic data. It converts data from various sources, such as CAD, URDF, and real-world captures, into USD, then sets up materials, physics, robots, and sensors to build a scene. It generates synthetic training data by randomly varying lighting, color, position, and more, and can augment data with Cosmos models. Perception and navigation stacks are trained in simulation, and the whole system is evaluated with SIL/HIL testing.

• Convert CAD, URDF, and real-world data to USD to build scenes
• Physics simulation of rigid bodies, vehicle dynamics, articulated bodies, and SDF colliders
• Generate synthetic data by randomizing lighting, reflections, color, and position
• Software/hardware-in-the-loop (SIL/HIL) testing
• ROS integration through the Isaac ROS/ROS2 bridge
• OnShape, URDF, MJCF, and ShapeNet importers
• Run on cloud GPUs in one click with Brev

• One of 8 AI tools in the Embodied AI & Robotics field.
• Connected to 2 other tools — #5 in its field.