On August 2, Google DeepMind introduced Gemini Robotics 2, the latest version of its visual-language-action model. This new model enhances full-body control, advanced dexterity, and multi-robot collaboration, marking a significant advancement from its predecessor, which only managed upper-body tasks.
Gemini Robotics 2 allows the Apptronik Apollo 2 humanoid robot to autonomously execute complex tasks, such as placing a watering can into a designated bucket, requiring coordinated decision-making and balance. While there is room for improvement in movement speed, this development represents a crucial step towards handling more intricate real-world tasks.
The model also introduces multi-robot collaboration, enabling different robots to communicate and work together on complex workflows. With the ASIMOV-Agentic benchmark, Gemini Robotics 2 excels in safety compliance and human proximity detection, ensuring safe operation. The Gemini Robotics On-Device 2 is designed for applications requiring offline functionality, adapting to new robotic forms with minimal data and time.
Editor's Note
The introduction of Gemini Robotics 2 by Google DeepMind signifies a pivotal moment in robotics, particularly in enhancing the capabilities of humanoid robots. This advancement not only improves task execution but also emphasizes safety and adaptability in various environments, which is crucial for industries like manufacturing and logistics. As robots become more integrated into daily life, their ability to perform complex tasks autonomously will reshape operational efficiencies and human-robot interactions.
Leave a comment