ReGIL: Retrieval-Guided Imitation Learning from a Single Demonstration

2026-06-08

AI Summary

This work addresses the challenge of efficiently learning robust robotic manipulation policies from a single demonstration, mitigating task failure due to minor trajectory deviations and reducing reliance on costly online interaction data. The authors propose ReGIL, a novel framework that models the single demonstration as a dynamic, queryable external memory. During training, ReGIL leverages retrieval-guided exploration, constructs a regularized replay buffer, and generates fine-grained rewards. Its key innovation lies in combining local temporal alignment to compute step-level rewards, enabling a synergistic mechanism of retrieval-guided reward shaping and policy regularization that transcends the limitations of traditional imitation learning’s dependence on static supervisory signals. Evaluated on LIBERO and Meta-World benchmarks, ReGIL significantly outperforms existing methods; real-robot experiments achieve over 75% success rate across three randomized initial conditions with only a single demonstration and less than one hour of training.

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/

Problem

• single demonstration
• robot manipulation
• imitation learning
• online interaction
• trajectory deviation

Innovation

• Retrieval-Guided Imitation Learning
• Single Demonstration
• External Memory
• Temporal Alignment
• Robotic Manipulation