ReGIL: Retrieval-Guided Imitation Learning from a Single Demonstration
2026-06-08 • Robotics
Robotics
AI summaryⓘ
The authors developed ReGIL, a method to teach robots how to perform tasks by using just one example demonstration. Instead of relying on lots of trial-and-error, their approach treats the demonstration like a memory that the robot can look back on during learning. It helps the robot know how well it is doing at each step by comparing its current actions to parts of the demonstration. Tests showed that ReGIL works better and faster than previous methods, even on real robots with some starting position changes. This means a single example can be reused in smart ways to help robots learn efficiently.
robot manipulationimitation learningdeep neural networkssingle demonstrationreinforcement learningtemporal alignmentpolicy improvementrobot training efficiencyMeta-World benchmarkrobot exploration
Authors
Yuying Zhang, Francesco Verdoja, Wenyan Yang, Ville Kyrki
Abstract
Learning robot manipulation policies with deep neural networks from a single demonstration remains highly challenging, as even small deviations from the demonstrated trajectory can quickly compound into failure, while collecting substantial online interaction data is costly. We propose ReGIL, a retrieval-guided imitation learning framework that treats a single demonstration as an external memory. ReGIL repeatedly queries this static memory throughout training to simultaneously guide exploration, generate the regularization buffer, and construct rewards. Specifically, it computes rewards through local temporal alignment between the current trajectory and the retrieved segment, providing step-wise and informative feedback for policy improvement. We evaluate ReGIL on robotic manipulation tasks from the LIBERO and Meta-World benchmarks under the single demonstration setting. ReGIL outperforms prior baselines in both success rate and training efficiency. In real-robot experiments, using only one demonstration and less than one hour of online training, ReGIL achieves over 75% success rate across three manipulation tasks with randomness in both initial robot pose and target position. These results demonstrate that leveraging the single demonstration as reusable memory can provide more than static supervision for efficient robot learning. More details can be found on our website: https://regil2026.github.io/