EventSync: Event-Driven Memory Updates for History-Dependent Manipulation
Automatically prepared brief
EventSync is a hierarchical framework that uses event-driven memory updates to improve history-dependent robotic manipulation by coordinating task reasoning and action execution.
The research question addresses the challenge of history-dependent robotic manipulation, where the current observation is insufficient to determine the next action due to the need for task context and interaction history.
The proposed approach, EventSync, is a hierarchical framework that uses a vision-language model to generate subgoals and a vision-language-action model to execute them. It includes a lightweight Event Monitor to assess subgoal completion based on visual observations and end-effector trajectories.
The evaluation settings involve testing EventSync on the RoboMME benchmark, where it achieved 64.5% overall success, surpassing the best baseline by 20 percentage points. Additional experiments on four real-robot tasks validated the framework's effectiveness.
About this brief
This article summarizes a public preprint abstract. Its findings have not been independently reproduced for this publication. Consult the paper for methods, evaluation scope and limitations. Automated briefs may contain errors; the original source takes precedence.