APEX: Adaptive Policy Execution for Precise Manipulation
2026-06-15 • Robotics
Robotics
AI summaryⓘ
The authors address a problem where robots, guided by high-level commands from a policy, often don't execute actions as intended because the policy doesn't fully account for the robot's low-level movement details. Instead of changing the policy or the robot's controller, which can be complicated, they propose a method called Adaptive Policy Execution (APEX) that sits between the two. APEX adjusts the commands on the fly using feedback from the robot's actual movements to make them more accurate. Their tests show that APEX significantly reduces errors and improves task success in various robot control scenarios.
Imitation LearningVisuomotor PoliciesVision-Language-Action (VLA) PoliciesLow-level ControllersExecution GapAdaptive Policy Execution (APEX)Test-time AdaptationState FeedbackTracking ErrorManipulation Success
Authors
Mengfei Zhao, Chenxi Jiang, Tuo An, Jindou Jia, Jianfei Yang
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.