Timing-aware robot queries improve policy learning and success
TimelyDAgger: Timing-Aware Expert Querying for VLA Policy Improvement
Robotics
Summary
Robots learn better when they know the right time to ask a human expert for help. The authors found that not only detecting when help is needed but also choosing when to take over affects how well the robot learns. They developed a method called TimelyDAgger that watches what the robot’s vision and action system is doing and adjusts when it asks for expert help. This method leads to better robot performance after training while keeping the amount of expert help the same. The authors also introduced ways to measure how well this timing helps robot learning.
What this means in practice
- •For robotics developers: Improve robot control software by automatically deciding when to ask for expert help to enhance learning and autonomy.
- •For automation engineers: Optimize timing of human intervention in automated systems to maximize learning from expert demonstrations without increasing supervision cost.
Authors
Zhixuan Zhao, Peiyan Li, Enhao Zhang, Yueran Tao, Hao Wang, Chenghao Yue, Lei Lv, Wentao Zhao, Jiahao Chen, Xin Liu, Kangyao Huang, Yu Luo, Huaping Liu
Abstract
DAgger improves robot policies by aggregating expert supervision from states visited during policy execution. Robot-gated DAgger automates expert queries, allowing the robot to decide when to request expert takeover. While existing gates emphasize detecting the need for assistance, takeover timing also shapes the content of these demonstrations and their value for policy learning. We propose TimelyDAgger, combining Bridge-PCA monitoring of internal vision-language-action (VLA) features with Feedback-guided Threshold Adaptation based on expert behavior to improve takeover timing. We introduce an evaluation framework linking failure detection, takeover timing, and policy improvement, including Target-Aligned Supervision Ratio (TASR) for assessing supervision quality without retraining. Experiments show that takeover timing affects policy learning, with TimelyDAgger achieving competitive failure detection and higher post-training success in most evaluated settings under matched expert-action budgets.