LightOriginsHQ introduced the Light-O1-Preview model for robot management

Team LightOriginsHQ has released Light-O1-Preview — a specialized model for generating sequences of actions for humanoid robots based on text instructions. The development is focused on automating the control of complex physical systems through natural language.

The model generates movements for a skeleton consisting of 22 joints at a frequency of 20 frames per second. The system is classified as a reasoning text-to-action model, capable of interpreting complex commands and converting them into precise physical trajectories. In preliminary tests, the solution received a rating of 17 stars with a training or data processing time of 2 hours.

The software is distributed under the open Apache 2.0 license, which allows the use of model weights in commercial and research projects without restrictions. The source code and documentation are available in the official repository on the Hugging Face platform.

The emergence of such tools simplifies the programming of movements for anthropomorphic robots, eliminating the need for manual coding for each joint. The use of a 22-joint model allows covering the basic kinematics of the human body, which is a standard for most modern research platforms in the field of robotics.