Newshot Motphys Releases Physics Simulation Engine MotrixSim 1.0, Core Developed in Rust
IT Home, November 15 news: Motphys's next-generation physics simulation engine, MotrixSim 1.0, was officially released yesterday.
According to the introduction, MotrixSim is a high-performance physics simulation engine designed specifically for multibody dynamics and robot simulation, supporting a wide range of application scenarios, including robot control, reinforcement learning, industrial simulation, and more. Its main features are as follows:
- Physics simulation: Supports complete physics simulation capabilities such as rigid body dynamics and collision detection
- Generalized coordinate modeling: Uses a generalized coordinate system to support modeling of complex multibody systems
- Brand-new solver: Uses a self-developed constraint model and solver, providing efficient and stable multibody dynamics computation
- High-performance computing: The CPU version is developed in Rust, providing excellent performance and memory safety
- Python API: A concise and easy-to-use Python interface, facilitating rapid development and prototyping
- Robot support: Specially optimized robot simulation capabilities, highly compatible with the MJCF model format
According to the official statement, MotrixSim replicates real-world physical parameters 1:1 into the simulation environment, combining multimodal sensors with high-fidelity rendering to create a parallel world for machines that can both "see clearly" and "feel accurately."
According to the introduction, through its original CPU / GPU heterogeneous architecture, leveraging the advantages of high-performance communication and unified memory, as well as a unified constraint chain that breaks through million-level concurrency, it achieves dual amplification of training scale and throughput while ensuring physical accuracy.
Paper data shows that under a 4096-concurrency environment, it runs millions of physics steps per second, allowing end-to-end training of the G1 humanoid robot to be completed in only 10.2 minutes (traditional solutions take 130.8 minutes).
By integrating visual (seeing colors and textures), geometric (measuring shapes, distances, and depth), and tactile (sensing pressure, friction, and hardness) perception, it completes a closed loop of environmental observation and behavioral feedback, solving the challenge of "robot cognitive blind spots."
In addition, unlike the traditional asset-heavy approach that relies heavily on high-end graphics cards, the core of MotrixSim is developed in the Rust language, balancing high performance and memory safety. At the same time, the engine is deeply compatible with diverse computing ecosystems such as Kunpeng, Iluvatar CoreX, Moore Threads, 英伟达, AMD, and Apple.
IT Home attaches the official documentation as follows:
- https://motrixsim.motphys.com/
Advertising statement: External redirect links contained in the article (including but not limited to hyperlinks, QR codes, passwords, and other forms) are used to convey more information and save selection time; the results are for reference only. All articles on IT Home containing external links include this statement.
Related Articles
- Classic physics engine PhysX comes to an end: 英伟达 RTX 50 series graphics cards no longer support 32-bit CUDA applications
- Understanding physics engines through "predicting" the World Cup