Humanoid Robot Startup Open Sources Training Software

Robot Era, a humanoid robot startup, is open sourcing its humanoid robot training software, Humanoid-Gym, to improve the accessibility of humanoid robot technology and accelerate the uptake of humanoid designs.

Humanoid-Gym, which leverages Nvidia’s Isaac Gym to train humanoid robots in locomotion skills, will be made available on GitHub.

The Beijing-based company pitches its reinforcement learning software as providing engineers with a practical, easy to use platform to build on to “avoid reinventing the wheel.”

In a video demonstrating the software,  the company's full-size humanoid robot, XBot-L, is shown performing actions first in simulation and then in the real world. 

XBot-L is shown walking across uneven surfaces, climbing an incline, and walking across snow.

The company’s platform simplifies the process of translating a simulation avatar to a working prototype, known as sim-to-real policy transfer.

The system uses “specific reward functions and domain randomization signals” to train robots and the team say it can “significantly streamline the development process, reducing technical difficulty and bolstering engineering productivity.”

The framework also introduced several variables for the assessment of reinforcement learning results, including but not limited to speed tracking and the smoothness of robots.

Related:System Enables Human-to-Humanoid Robot Operation

By open sourcing its platform, Robot Era said it hopes to “foster knowledge-sharing and innovative exchange in the area of humanoid robotics.”