aiqbench analysis
This paper proposes two data augmentation methods for imitation learning in bimanual robot manipulation: reversing demonstrations between paired tasks with opposite objectives, and resampling trajectories at varied temporal strides. The reported gains are modest and tied to specific benchmarks (16 paired tasks, 25 and 50 demonstrations per task), and the authors explicitly treat reversal as an approximation because contact-rich manipulation is not truly reversible. For ML readers, the useful takeaway is that augmentation may reduce expensive demonstration collection in robot learning, though the results need replication across other tasks and robots before they generalize.
RESTA: Time-Scaling Augmentation for Data-Efficient Robot Imitation Learning
arXiv:2610.18163v1 Announce Type: new Abstract: Robot manipulation policies trained through imitation learning commonly rely on costly demonstrations recorded in one task direction and at one temporal resolution. We propose Reversal and Stride-based Trajectory Augmentation (RESTA), which exploits time-scaling data augmentation to improve data efficiency through trajectory reversal and temporal resampling. Trajectory-reversal augmentation reverses and relabels demonstrations between paired tasks with opposite objectives, reusing recorded rollouts without additional environment interaction. Because contact-rich manipulation is not exactly reversible, we treat reversal as an approximate paired-task prior rather than a physical symmetry. Temporal-stride augmentation varies observation and action sampling strides and conditions the policy on the resulting offsets, enabling fixed or randomized temporal schedules. Experiments across 16 paired bimanual manipulation tasks show that trajectory reversal is particularly effective in the low-data regime and that temporal-stride augmentation improves success under randomized temporal schedules. With 25 and 50 independently collected demonstrations per task, RESTA achieves 78.4% and 85.2% average success, respectively, matching or exceeding baselines trained with twice as many demonstrations (77.3% and 85.4%). These results show that RESTA can improve data efficiency in robot imitation learning.
Abstract:Robot manipulation policies trained through imitation learning commonly rely on costly demonstrations recorded in one task direction and at one temporal resolution. We propose Reversal and Stride-based Trajectory Augmentation (RESTA), which exploits time-scaling data augmentation to improve data efficiency through trajectory reversal and temporal resampling. Trajectory-reversal augmentation reverses and relabels demonstrations between paired tasks with opposite objectives, reusing recorded rollouts without additional environment interaction. Because contact-rich manipulation is not exactly reversible, we treat reversal as an approximate paired-task prior rather than a physical symmetry. Temporal-stride augmentation varies observation and action sampling strides and conditions the policy on the resulting offsets, enabling fixed or randomized temporal schedules. Experiments across 16 paired bimanual manipulation tasks show that trajectory reversal is particularly effective in the low-data regime and that temporal-stride augmentation improves success under randomized temporal schedules. With 25 and 50 independently collected demonstrations per task, RESTA achieves 78.4% and 85.2% average success, respectively, matching or exceeding baselines trained with twice as many demonstrations (77.3% and 85.4%). These results show that RESTA can improve data efficiency in robot imitation learning.
Robotics (cs.RO)
arXiv:2610.18163 [cs.RO]
(or arXiv:2610.18163v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2610.18163 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 270 words
Related coverage
Vision-Language-Action Jump-Starting for Reinforcement Learning Robotic Agents
arXiv · Jun 12, 2026
Build a Physical AI model factory with NVIDIA Cosmos 3 on SageMaker HyperPod
AWS · Sep 4, 2026
The State of Simulation for Physical AI: An Overview
Hugging Face · Jul 21, 2026
NVIDIA and Japan Bring Full-Stack AI and Robotics to Every Industry
NVIDIA · Jul 15, 2026