Boston Dynamics has integrated Google DeepMind's Gemini Robotics-ER 1.6 embodied reasoning model into its Spot quadruped robot, targeting industrial inspection use cases. The upgrade enables Spot to autonomously detect hazardous debris, read complex gauges, and reason about tasks in unstructured environments. The model is currently vision-only due to limited tactile training data on the internet, and a new 'success detection' feature uses multiple camera angles to confirm object grasps. Customers deploying the new capabilities must share operational data with Boston Dynamics to help improve future models. The partnership also aims to apply lessons learned from Spot's commercial deployments to other platforms, including the humanoid Atlas robot.

7m read timeFrom spectrum.ieee.org
Post cover image
Table of contents
Understanding Robot UnderstandingReal-World Robots That Are Useful

Sort: