NVIDIA PhysX is a powerful, open-source multi-physics SDK that provides scalable simulation and modeling capabilities for robotics and autonomous vehicle applications. It is designed to accurately replicate the physical behavior of objects and systems, which is foundational in building physical AI for powering the next generation of autonomous systems. PhysX runs on various platforms, from mobile CPUs to high-end GPUs, and supports end-to-end reinforcement learning via NVIDIA Isaac™ Lab.

Features

• Rigid Body Dynamics: Analyze multi-body dynamics under external forces like gravity, with scalable simulation for both CPU and GPU.
• Scene Query: Conduct spatial queries in simulated environments using raycast, overlap, and sweep functions with customizable filtering.
• Joints: Utilize a suite of common built-in joint types and support custom joints through a flexible callback mechanism.
• Reduced Coordinate Articulations: Provide linear-time, guaranteed joint-error-free simulation of a tree of rigid bodies.
• Vehicle Dynamics: Study the movement of multi-body interactions under external forces, with industry-proven scalable simulation.
• Character Controllers: Navigate a simulated world with support for both static and dynamically simulated bodies.
• Soft Body Dynamics: Finite Element Method (FEM) simulation of soft bodies for accurate and efficient models of elastic deformable bodies.
• SDF Colliders: Simulate non-convex shapes like gears and cams without convex decomposition using Signed Distance Field-based collision representation.
• Position-Based Dynamics: Flexible framework for simulating liquids, granular materials, cloth, rigid bodies, deformable bodies, and more.
• Custom Geometry: Utilize built-in geometries and a flexible callback mechanism for custom geometry types.
• Fracture & Destruction: Includes Blast, a destruction and fracture library designed for performance, scalability, and flexibility.
• Smoke & Fire: Flow enables realistic combustible fluid, smoke, and fire simulations as part of the PhysX SDK.

Benefits

• Open-Source: Fully open-source, enabling free use and sharing of work with other developers, and contributing innovative developments back to the community.
• Unified Model: FEM soft body, cloth, particles, and fluid simulation with two-way coupled interactions under a unified solver framework.
• Scalable: Runs on various platforms, from mobile CPUs to high-end GPUs, with a GPU API supporting end-to-end reinforcement learning.
• Accurate: Enhanced simulation stability with robust collision detection, stacking, and joints, while supporting momentum conservation and gyroscopic forces.