Google Introduces Gemini AI Models for Robotics Use
Google DeepMind Unveils New Robotics AI Models
Google DeepMind has recently introduced two innovative models in the field of artificial intelligence tailored specifically for robotics: Gemini Robotics and Gemini Robotics-ER. These models are designed to offer enhanced capabilities in robot control and spatial comprehension, making them pivotal in advancing robotic technology.
Key Features of Gemini Robotics
Gemini Robotics is an advanced model that integrates vision, language, and action. It adds a significant new output modality, allowing it to control robots directly through physical actions. This capability is designed to improve how robots interact with their environments, enabling them to undertake a broader range of tasks in real-world scenarios.
- Vision-Language-Action Integration: This model combines visual perception with linguistic understanding, allowing robots to understand commands and respond accordingly.
- Enhanced Interaction: The aim is to create robots that can adapt and react to their surroundings more efficiently.
Focusing on Spatial Understanding
The second model, Gemini Robotics-ER, is dedicated to refining spatial understanding. This model provides robotic systems with advanced reasoning abilities, allowing for better navigation and interaction with their environment.
- Utilization of Embodied Reasoning: This feature enables robots to process and understand spatial data in a way that enhances their operation.
- Integration with Existing Systems: The Gemini Robotics-ER model is built to work alongside existing low-level controllers, facilitating smoother adaptation and enhancing overall functionality.
Training and Collaborations
Google DeepMind has highlighted that the Gemini Robotics model was trained using data from the bi-arm robotic platform known as ALOHA 2. This training aids the model in effectively controlling robotic systems like the Franka arms.
- Complex Embodiments: The model is adaptable to various robotic structures, including human-like robots such as the Apollo robot developed by Apptronik.
- Partnership with Apptronik: Google is collaborating with Apptronik to push the boundaries of humanoid robot development using insights from Gemini 2.0.
Investment in Humanoid Robotics
In February 2025, Apptronik secured $350 million in a funding round led by B Capital and Capital Factory, with significant participation from Google. This investment is aimed at advancing the development of humanoid robots, specifically the Apollo model.
- Funding Utilization: The funds will enhance Apollo’s capabilities, expand its applications in various industries, including logistics and manufacturing, and increase production to meet rising demand for humanoid robots.
Future Directions
Google DeepMind has announced its plans to work closely with select testers to fine-tune the future of the Gemini Robotics-ER model. With ongoing collaborations and investments, the robotics sector is poised for significant advancements facilitated by these cutting-edge AI models.
- Strategic Partnerships: Continued collaboration with industry leaders reflects the emphasis on real-world applicability and practical integration of advanced robotic technologies.
Through these developments, Google DeepMind is taking crucial steps towards revolutionizing the robotics industry, enabling smarter and more efficient robotic systems that can seamlessly operate in various environments.
