You Don't Need To Stay in The Loop: An Agentic Robotics Loop for Robot-Policy Improvement
This work provides a robust and recoverable system for robot-policy improvement, benefiting robotics researchers and developers by reducing manual oversight and increasing reliability in policy development.
This paper introduces AgenticRobotics, an agentic architecture for robot-policy improvement that addresses the high failure rate of robotic tools by measuring and recording tool quality at every call and expiring it when artifacts change. The system achieves false-promotion control of 0.001 per run hardened, catches six of six artifact-tampering classes, and ensures anytime-valid decisions and zero lost/duplicate effects under kill injection.
Coding agents such as Claude Code and Codex close the software loop: a main agent manages the loop, subagents analyze and execute, tools do the work. We port this architecture to robot-policy improvement, where one difference dominates the design: robotic tools---trained policies, training pipelines, data collection---fail routinely, so a tool's quality must be measured, recorded at every call, and expired when the artifact behind it changes. AgenticRobotics is a backend-independent control plane in which an LLM controller drives disposable workers through durable train--evaluate--improve transactions: an immutable objective, controller-owned measurement, commit-keyed crash recovery, an evidence-graded skill library, and a tool registry with a standardized, recorded call surface. The title is an operational claim, not a selection claim: the operator can leave because promotion is evidence-gated, state is recoverable, and capability quality is derived from records---not because the loop picks better checkpoints than a human; on the one lineage we measured, it does not. The gates measurably buy false-promotion control (0.001 per run hardened versus 0.005--0.021 shipped), anytime-valid decisions under optional stopping, zero lost or duplicate effects under kill injection, and six of six artifact-tampering classes caught by a signed verifier.