Robotic Success Detection

Robotic success detection is the ability of a robot or embodied AI system to decide whether a physical task has been completed successfully and whether to retry or advance.

Key points

• Google DeepMind calls success detection a cornerstone of autonomy because a robot needs to know when a task is finished, not only how to begin it ^[src-039].
• Success detection lets an agent decide between retrying a failed attempt and progressing to the next stage of a plan ^[src-039].
• Physical success detection requires perception plus reasoning under occlusion, poor lighting, ambiguous instructions, and changing scenes ^[src-039].
• Modern robot setups often use multiple camera views, such as overhead and wrist-mounted feeds, so the system must integrate viewpoints into a coherent scene ^[src-039].
• Gemini Robotics-ER 1.6 improves multi-view reasoning and can use multiple camera streams to determine whether tasks such as placing an object into a holder are complete ^[src-039].

Related entities

• Gemini Robotics-ER
• Google DeepMind

Related concepts

• Embodied Reasoning
• Agentic AI
• Agentic Vision
• Physical Safety Constraints for Robots

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)