Google’s DeepMind has introduced Gemini Robotics On-Device, a powerful offline large language model (LLM) tailored for robotic platforms, launched on June 24. Designed for latency-sensitive and low-connectivity environments, it allows robots to work autonomously without cloud dependence, enhancing response times, data privacy, and reliability. Gemini On-Device excels at complex tasks, including delicate operations like folding clothes and mechanical assembly, while achieving performance similar to cloud models. Originally developed for ALOHA robots, its compatibility extends to advanced systems like Franka FR3 and Apollo humanoid robots, demonstrating capabilities such as voice-command execution. Additionally, Google has released an SDK for developers, enabling the deployment of this AI across compatible machines, which fosters innovation in robotic applications. As the AI race in robotics intensifies, competitors like NVIDIA and Hugging Face are also advancing in this field, marking significant progress in autonomous systems.
Source link
