Gemini Robotics-ER 1.6: Powering real-world robotics tasks through enhanced embodied reasoning

Google introduces Gemini Robotics-ER 1.6, an AI model for embodied reasoning that enhances robots' understanding of physical environments.

• •Spatial reasoning through pointing enables object detection, counting, and spatial relationship definition for robotic manipulation
• •Multi-view success detection allows robots to understand multiple camera streams and determine task completion in dynamic environments
• •Instrument reading capability interprets analog gauges and digital readouts through visual reasoning and code execution
• •Agentic vision combines visual reasoning with intermediate steps like zooming and pointing for accurate readings
• •Improved safety features ensure physical constraint adherence with 6-10% improvement over baseline models on safety perception