Robotic Instrument Reading
Robotic instrument reading is the use of embodied visual reasoning to interpret physical instruments such as pressure gauges, sight glasses, thermometers, level indicators, and digital readouts.
Key points
- Google DeepMind presents instrument reading as a new Gemini Robotics-ER 1.6 capability discovered through collaboration with Boston Dynamics [src-039].
- Industrial facilities contain instruments that require constant monitoring, and robots such as Boston Dynamics Spot can visit those instruments and capture images [src-039].
- Instrument reading requires perceiving needles, liquid levels, boundaries, tick marks, text labels, units, and multiple needles, then combining those elements into a reading [src-039].
- Sight glasses add perspective distortion, so the model must estimate fill levels while accounting for camera viewpoint [src-039].
- Gemini Robotics-ER 1.6 uses Agentic Vision with pointing and code execution to zoom into gauges, estimate proportions and intervals, and interpret the reading with world knowledge [src-039].
Related entities
Related concepts
Source references
- [src-039] Laura Graesser and Peng Xu — “Gemini Robotics-ER 1.6: Powering real-world robotics tasks through enhanced embodied reasoning” (2026-04-14)