Google rolls out new Gemini model that can run on robots locally
Google DeepMind released a new language model on Tuesday with the name Gemini Robotics On-Device that can perform tasks locally on robots without requiring an internet connection.
Building on the previous Gemini Robotics model of the company that was released in March, Gemini Robotics On-Device can control the movements of a robot. Developers can control and refine the model to meet different needs with the help of natural language prompts.
In Benchmarks, Google claims that the model is performing at a level close to the cloud-based Gemini-Robotics model. The company says that it is performing better than other models on the devices in general benchmarks, although it did not mention those models.
In a demo, the company showed robots that run this local model, things like exhausting bags and folding clothes. Google says that while the model is trained for Aloha RobotsIt adjusted it later to work on A double Franka fr3 robot And The Apollo Humanoid Robot from Apptronik.
Google claims that the BI-arm Franka FR3 was successful in tackling scenarios and objects that it had never “seen”, such as doing Assembly on an industrial belt.
Google DeepMind also comes out A Gemini Robotics SDK. The company said that developers can show robots 50 to 100 demonstrations of tasks to train them on new tasks using these models on the Mujoco Physics Simulator.
Other AI model developers also immerse their toes in robotics. Nvidia is building a platform to create foundation models for humanoids; Hugging face is not only the development of open models and data sets for robotics, it actually also works on robots; And Mirae-by asset-supported Korean startup RLWRLD is working on making fundamental models for robots.
