Train Virtually. Evolve in Reality
MATRiX
An Open-World Simulation and Training Platform for Embodied Intelligence
MATRiX
An Open-World Simulation and Training Platform for Embodied Intelligence
See Real. Move Real.
Built on Unreal Engine, MATRiX delivers photorealistic environments combined with accurate physics and multi-modal perception, bringing simulation training closer to real-world conditions.
500 Hz Precision Physics Simulation
Powered by the MuJoCo physics engine, MATRiX simulates robot motion, collisions, and contact interactions at a 500 Hz control frequency, providing a reliable physical foundation for motion training and algorithm validation.
Photorealistic Visual Simulation
Urban Roads
Waterfront Environments
Warehouse Spaces
Stair Terrain
Start Fast. Train Freely.
With a fully graphical interface and visual configuration, MATRiX lets users launch simulation training quickly—no ROS installation or coding required.
Build with Ease, Create on Demand
Download map DLCs as needed, build custom environments through drag-and-drop tools, AI-assisted generation, and 3DGS real-world scene import. Dynamically add obstacles, vehicles, pedestrians, and other elements to create more diverse training conditions.
Sense More. Switch in Real Time.
Native support for multiple sensor types continuously provides the simulation data required for perception and algorithm development.
Multi-Sensor Simulation
Supports RGB, infrared, depth, fisheye, and panoramic cameras, as well as LiDAR including Mid360, IMU, GPS, and odometry. Sensors can be switched in real time to support different perception tasks.
Distributed Multi-Robot Simulation
Native support for Linux and Windows enables large-scale, concurrent multi-robot simulation through distributed deployment, supporting formation control, reinforcement learning, navigation, and control training.
Built-In Motion Control
Built-in motion control cores including xgb, xg2, xgw, and zsl-1 use Linux shared memory to emulate hardware communication links, with one-click switching between SIL and HIL.
Fast Sim-to-Real Deployment
Deploy simulation algorithms and motion control models directly to physical robots such as L1 and M1, bridging virtual training with real-world deployment and shortening the Sim-to-Real cycle.
Start Your Simulation Journey
Get Started in Three Steps
01 Get the MATRiX runtime package. 02 Start the platform, then select and download the required map DLCs. 03 Select a robot and scenario, configure the sensors, and start simulating.
View Quick Start Guide