Reproduce “Variable-Speed Teaching-Playback as Real-World Data Augmentation for 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
Nozomu Masuya, Hiroshi Sato, Koki Yamane, Takuya Kusume, Sho Sakaino, Toshiaki Tsuji
The authors augment speed-specific imitation learning by replaying two fixed-speed human demonstrations at different speeds on a real robot, recording contact responses, and training on successful command–response pairs. In pick-and-place, this achieved 74/140 successes (53%) versus 42/140 (30%) for speed-adjusted duplication, and 70/80 (88%) versus 25/80 (31%) at 4.5–12 seconds. Results were not uniformly better, and playback lacks continuous feedback.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.