Google DeepMind has introduced Gemini Robotics On-Device, a cutting-edge AI model engineered for robots. This innovative system allows robots to operate independently of an internet connection, allowing them to perform tasks without relying on cloud connectivity. The model is specifically designed for bi-arm robots, providing them with advanced capabilities in communication, comprehension, and action. Its low-latency performance makes it perfect for immediate real-time operations. This enables robots to comprehend and respond to human commands like humans do. During internal evaluations, the AI model successfully executed various tasks, including folding clothing, opening and closing bags, object recognition, and industrial assembly work. The model is capable of interpreting instructions given in natural language and taking the corresponding actions. The system has been tested across diverse robotic platforms such as ALOHA, Franka Emika FR3, and Apptronik Apollo humanoid robots. Presently, access to this system is restricted to select developers under Google’s Trusted Tester program, giving them the opportunity to experiment with the Gemini Robotics SDK and the MuJoCo physics simulator.
Gemini Robotics On-Device: Google’s Offline AI Revolutionizes Robot Capabilities
By The Chhattisgarh1 Min Read