Learning Robust Execution in Robotic Manipulation with Agentic Reinforcement Learning

2026-07-15

AI Summary

This work addresses the instability and task failure in robotic manipulation caused by environmental uncertainty, long-horizon execution, and error accumulation. The authors propose an agent-based reinforcement learning framework that dynamically selects execution modes via a high-level policy and triggers a recovery mechanism upon detecting performance degradation, guiding the system back to a viable state. A key innovation is the introduction of a runtime execution quality metric, enabling the agent to make high-level recovery decisions—rather than directly learning low-level actions—by integrating execution mode switching, historical state analysis, and a vision-language-action (VLA) model. Evaluated on the LIBERO benchmark, the approach improves success rates by 13.7% under standard conditions and by as much as 39.2% in perturbed environments, demonstrating substantially enhanced robustness.

Abstract

Robotic manipulation poses fundamental challenges due to uncertainty, long-horizon execution, and compounding errors, which can easily destabilize execution and lead to task failure. Although recent vision-language-action (VLA) models exhibit strong generalization, they typically lack explicit mechanisms to assess execution stability and to recover when execution deviates from its nominal behavior. In this paper, we propose: (1) two complementary metrics to assess execution quality at runtime, and (2) an agentic reinforcement learning framework that learns to restore effective execution through high-level decision-making rather than directly learning low-level actions. In this framework, an agentic policy reasons over recent execution history and selects among a small set of execution modes to regulate the execution process. Under execution degradation, it triggers appropriate recovery mechanisms to restore the robot to previously visited nominal states, enabling the task to continue. We evaluate the proposed method on the LIBERO benchmark, achieving up to a 13.7% improvement in success rate under standard settings and up to a 39.2% improvement under disturbance settings, demonstrating substantially enhanced execution robustness.

Problem

• robotic manipulation
• execution robustness
• uncertainty
• compounding errors
• task failure

Innovation

• agentic reinforcement learning
• execution robustness
• recovery mechanism
• vision-language-action models
• robotic manipulation