BEIJING, Nov. 14, 2026 /PRNewswire/ -- If the global expansion of Chinese-made robots defined the hardware boom in the first half of the embodied AI race, the second half will be won by those that can bring down the high cost of real-world trial and error and make robot training more universal and accessible. In this underlying physics engine race long dominated by Western players, a Chinese team is now bringing a potential game-changer to the field.

Recently, Chinese company Motphys was invited to attend CoRL 2026, where it showcased its fully in-house embodied intelligence simulation engine—MotrixSim 1.0. In extreme benchmark tests against leading international physics engines, MotrixSim 1.0 demonstrates extraordinary capabilities: overwhelming established open-source engines in the contact and grasping test and vastly surpassing mainstream GPU-accelerated simulation platforms in training efficiency.

In the past, simulation was primarily used for error-proof verification at the end of the design phase. But in the era of embodied AI, AI requires massive amounts of interaction experience.

"Simulation once served to validate how machines perform. Now, it is where they learn how to perform," said Zhang Heng, co-founder and CTO of Motphys. Simulation has evolved from a "post-design validation tool" into a "physical experience production system". To power this system, Motphys builds its physics engine entirely in-house, developing the core from the ground up in Rust rather than building on existing open-source engines. This equips MotrixSim 1.0 with strong memory safety and computation power, enabling it to compete with established international players at the underlying architecture level.

True technical strength is proven in extreme tests.

In a "devil-level" disturbance-rejection grasping test where a Franka arm grasps objects of different shapes, MotrixSim 1.0 is the only engine to achieve a 100% pass rate at both 2 ms and 10 ms simulation step sizes. In the microscopic critical friction point test, MotrixSim exhibits hold-and-slip behavior near the critical friction consistent with Coulomb's law, achieving 0mm slip distance.

In terms of compute and training efficiency. Through its unique pipeline supporting CPU/GPU parallel computation, MotrixSim can achieve a physics simulation throughput of millions of physics steps per second across 4,096 parallel environments. In a controlled comparison conducted on the G1 whole-body tracking task, under the same task environment configuration and the same RL algorithm, MotrixSim reaches the current tracking score in about 10.2 minutes of end-to-end time, while the Isaac Sim-based solution takes 130.8 minutes—a gap of about 12.8×.

Furthermore, for the Strandbeest complex closed-loop mechanism with 92 linkages, when traditional engines fail to maintain the stability of the mechanism at a large time step of 5ms, MotrixSim solves closed-loop constraints together with the multibody dynamics in generalized coordinates, so the mechanism maintains its assembly relationships during motion.

Beyond extreme physical conditions and compute, MotrixSim 1.0 has achieved breakthroughs in multimodal perception and developer experience: LiDAR simulation throughput reaches 124 million rays per second; visual synthesis maintains stable throughput even when resolution surges 16 times. In terms of developer experience and industrial deployment, MotrixSim 1.0 streamlines development with high-speed solving in a single line of code (98.42% success rate) and a "zero-conversion + AI auto-modeling" workflow. Forge is also launched as the industry's first Blender-based extension tool, enabling AI Agents to automatically generate physics scenes.

Zhang Heng emphasized that the core proposition of Physical AI is to transform expensive and slow "real-world physical experience" into digital learning resources that can be generated at scale.

Motphys's relevant research has been accepted to top international conferences such as RSS and CoRL, and the company has built a simulation and real-robot training center for embodied AI in China. Currently, the world's leading physics engines primarily include Google's MuJoCo and NVIDIA's Isaac Sim.

View original content:https://www.prnewswire.com/news-releases/chinese-team-drops-a-game-changer-motphys-launches-motrixsim-1-0--sparking-industry-buzz-302939513.html

SOURCE Motphys