Reproduce “Imitation Learning via Discrete and Rhythmic Primitive Composition” 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
Rafid Saif, Auke Ijspeert, Hsiu-Chin Lin
Neural Dynamic Composition (NDC) learns to compose unsegmented robot demonstrations into ordered sequences of reusable skills from a predefined library. It predicts discrete and rhythmic primitive choices and parameters using differentiable parallel combinations and staged training. In simulation, it reports 92% overall task success versus 83% for a stronger DMP baseline, but assumes a fixed primitive count and library.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.