StorySoftware
Spot gets Google DeepMind's Gemini Robotics-ER 1.6 model for reading analog gauges and spotting spills during inspections
Boston Dynamics added Google DeepMind's Gemini Robotics-ER 1.6 embodied reasoning model to Spot, enabling it to read analog gauges and sight glasses, spot spills and debris, and confirm grasps using multiple camera views. The goal is fully autonomous inspection rounds; several thousand Spots are already deployed commercially.
- Gemini Robotics-ER 1.6: high-level embodied reasoning model connecting how robots understand the world to human comprehension.
- Capabilities: read analog gauges, detect spills/debris, assess successful grasps via multi-camera analysis, reason about safety.
- Vision-only processing (touch sensing training data insufficient); model reasons semantifically, e.g., avoiding cup placement on table edge.
- Thousands of Spots already deployed; autonomous inspection rounds reduce manual labor and documentation time.