Google Cloud opens Gemini Robotics ER 2 in public preview
Google Cloud opened public-preview access to Gemini Robotics ER 2, which helps robots interpret surroundings and plan tasks. The offering brings an already introduced model into an enterprise cloud workflow. Separate local systems execute movement. The announcement describes a delivery channel and company-presented examples rather than a universally tested robot configuration.
Artificial Intelligence··Evening
A new enterprise delivery channel
Google Cloud announced an enterprise public preview for its robotics reasoning model, Gemini Robotics Embodied Reasoning 2. The underlying model was previously presented by Google DeepMind. The new offering concerns cloud delivery with identity, access and enterprise security controls. Teams building robots can connect high-level task planning to Google’s cloud infrastructure, while the preview label still distinguishes the offering from a universally validated production configuration.[1]
Planning and movement have separate roles
The model accepts video, audio and text, interprets a robot’s surroundings and breaks work into steps. Developers can expose vision-language-action models, navigation systems or their own movement controllers as tools. The component carrying out physical motion therefore has a separate role from the reasoning model. Google also describes monitoring progress and reassessing unsuccessful steps during a longer task.[1]
From robot data to simulation
Cloud Storage and BigQuery provide the proposed storage and analysis workflow. Google names MuJoCo Warp and NVIDIA Isaac Sim among simulation tools, alongside examples involving Boston Dynamics’ Spot, DeepReach and Arm. These are company-presented applications. Customers still need tools, perception inputs and movement controls suited to their robot, with latency-sensitive navigation workloads able to run locally.[1]
