On Thursday, the Genesis AI physics model was introduced, showcasing its ability to simulate four-dimensional (4D) worlds. This innovative model represents a significant leap in generative artificial intelligence (AI), merging various capabilities to enhance scenarios for general-purpose robotics and physical AI applications. Researchers declare that Genesis operates with remarkable efficiency, outpacing conventional GPU-accelerated systems by up to 80 times. Moreover, the open-source model can be accessed through the Python Package Index (PyPI), though users must also install PyTorch.
Genesis AI Physics Model Can Simulate Dynamic Worlds for Robotics Training
Lead researcher Zhou Xian shared details about Genesis in a post on X (formerly Twitter), noting that its development followed a two-year extensive collaborative research effort involving more than 20 labs. Genesis stands out for its integration of several physics solvers into a cohesive framework.
Genesis supports simulating various types of physical phenomena. We developed from scratch a unified physics engine that integrates various SOTA physics solvers (MPM, SPH, FEM, Rigid Body, PBD, etc.), supporting simulation of a wide range of materials: rigid body, articulated… pic.twitter.com/PqhIWULKgp
— Zhou Xian (@zhou_xian_) December 18, 2024
Completely constructed in Python, Genesis features a generative agent framework supported by a universal physics engine. Currently, the team has made the core physics engine and the simulation platform open-source, with plans to release the generative framework at a later date.
The potential of this AI system is considerable, according to the research team’s assertions. It purportedly achieves speeds that are 10 to 80 times greater than platforms such as Isaac Gym and MJX, both of which depend on GPU acceleration. In specific instances, the engine is claimed to reach simulation speeds that are 430,000 times faster than real-time, with the capability to train a robotic locomotion policy in just 26 seconds using a single Nvidia RTX4090 GPU.
When examining the key features of the Genesis AI physics engine, it becomes evident that both the frontend and backend have been seamlessly developed in Python. The system is also accessible through an application programming interface (API). While achieving significantly faster simulation speeds, it purportedly maintains high levels of simulation accuracy and fidelity. Additionally, the unified framework allows for the integration of various physics solvers to produce a diverse array of physical phenomena and materials, with the physics engine also supporting ray-tracing rendering.
