Reproduce “RESTA: Time-Scaling Augmentation for Data-Efficient Robot Imitation Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yunpeng Jiang, Jianshu Hu, Han Fang, Zhiyuan Guan, Paul Weng, Yutong Ban
RESTA reuses demonstration data for paired robot tasks and trains policies with varied temporal spacing. It reverses observations and relabels targets to create offline examples, while conditioning policies on observation and action offsets. In simulation-only RoboTwin tasks, mean success reaches 78.4% with 25 independently collected demonstrations per direction, versus 67.5% for DP3 with 25 and 77.3% with 50.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.