APEX: Adaptive Policy Execution for Precise Manipulation

2026-06-15

AI Summary

This work addresses the execution discrepancy in imitation learning caused by the lack of dynamic awareness in high-level policies and low-level controllers. To bridge this gap, the authors propose a plug-and-play adaptive policy execution framework that operates between a pretrained policy and a black-box controller without requiring modifications to either component. The framework online reconstructs dynamically feasible reference trajectories and integrates low-level state feedback for real-time adjustments. Theoretical analysis guarantees convergence of the execution error. Experimental results demonstrate that the method reduces trajectory tracking error by 41.2% in demonstration replay and improves task success rates by 4.8 to 25.8 percentage points across four categories of vision-based and vision-language-action policies.

Abstract

Modern imitation learning methods, including visuomotor and Vision-Language-Action (VLA) policies, typically output high-level action references that are executed by low-level controllers. However, the absence of higher-order reference signals, together with the policy's lack of awareness of the underlying low-level control dynamics during training, inevitably induces an execution gap. As a result, realized actions deviate systematically from policy-commanded ones, with a critical impact on precision-sensitive manipulation. Prior work either modifies the policy architecture or the low-level controller, both requiring intrusive changes to the pretrained policy or packaged controller. This raises a natural question: when the policy and controller are both treated as inaccessible black boxes, can we bridge the execution gap? We propose Adaptive Policy Execution (APEX), a plug-and-play framework inserted between the policy and the controller that reconstructs a dynamically feasible reference from policy outputs and adapts at test-time according to low-level state feedback, with a provable convergence guarantee. Extensive empirical studies show that APEX reduces controller-induced tracking error by 41.2% on demonstration replay and improves manipulation success by 4.8--25.8 percentage points across four visuomotor and VLA policy classes.

Problem

• execution gap
• imitation learning
• visuomotor policies
• Vision-Language-Action
• precise manipulation

Innovation

• Adaptive Policy Execution
• imitation learning
• execution gap
• plug-and-play framework
• visuomotor control