WIRobotics Begins Building a Physical AI Development Ecosystem: The First Technology Release...

• Additional core ALLEX technologies to be released sequentially

• Expanding the Physical AI development ecosystem through high-fidelity Sim-to-Real validation

• Building an open Physical AI development ecosystem for researchers and robotics developers

WIRobotics, a global robotics company (Co-CEOs: YounBaek Lee and Yong-Jae Kim), announced the launch of its technology disclosure roadmap for building a Physical AI development ecosystem.

WIRobotics plans to gradually release the core technologies developed throughout its humanoid robot program to researchers and developers. As the first technology release, the company has unveiled the simulation model of its humanoid robot, ALLEX, together with the results of its Sim-to-Real (Simulation-to-Real) validation.

Across the global robotics industry, Physical AI—which enables robots to perceive, reason, and act in the physical world—is gaining attention as one of the next-generation core technologies. In particular, development approaches based on simulation environments that accurately reflect the characteristics of real robots are rapidly expanding in the humanoid robotics field.

In response to this trend, WIRobotics plans to go beyond robot development by building an open development ecosystem where external researchers and developers can actively participate. The goal is to provide an environment in which a wide range of algorithms and AI models can be researched and validated through simulation, even before physical robots become commercially available.

The newly released ALLEX simulation model focuses on minimizing the Sim-to-Real Gap between the simulation environment and the physical robot. In particular, the model precisely reproduces the characteristics of the real ALLEX, including its high backdrivability and force transparency. This demonstrates the robot’s highly differentiated Real-to-Sim characteristics, where the behavior of the physical robot closely matches that of the ideal simulation model. As a result, researchers and developers can conduct ALLEX-based robot control, learning, and synthetic data generation research without requiring access to physical hardware. The released model is available in MJCF format for MuJoCo, USD format for Isaac Sim, and URDF format for ROS.

출처: aithority