Physical Safety Constraints for Robots

Physical Safety Constraints for Robots

Physical safety constraints for robots are limits on what a robot should perceive, select, handle, move, or avoid when acting in the real world.

Key points

  • Google DeepMind says safety is integrated into every level of its embodied reasoning models and describes Gemini Robotics-ER 1.6 as its safest robotics model to date [src-039].
  • The model shows stronger compliance with Gemini safety policies on adversarial spatial reasoning tasks than previous generations [src-039].
  • Physical constraints include gripper limits, material restrictions, liquid handling rules, and weight limits such as not picking up objects heavier than 20kg [src-039].
  • Gemini Robotics-ER 1.6 makes safer decisions through spatial outputs like pointing when asked which objects can be manipulated under those constraints [src-039].
  • Google DeepMind also evaluated injury-risk perception in text and video scenarios derived from real-life injury reports; Gemini Robotics-ER models outperformed Gemini 3.0 Flash by 6 percent in text and 10 percent in video on those tasks [src-039].
  • Back to Engineering adds a practical builder constraint: physical robot projects expose safety and reliability issues through wiring, power, motors, calibration, and edge/cloud trade-offs before any high-level AI policy layer is involved [src-076].

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)
  • [src-076] Back to Engineering (iulia) – physical AI, robotics, and data science cluster (41 videos, 2018-12-16 to 2026-05-10)