Reproduce “Learning human behaviors from motion capture by adversarial imitation” 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
Josh Merel, Yuval Tassa, Dhruva TB, Sriram Srinivasan, Jay Lemmon, Ziyu Wang, Greg Wayne, Nicolas Heess
The authors present adversarial imitation learning for training humanoid controllers from motion-capture examples, then reuse the learned skills through a higher-level controller. A discriminator rewards matching selected movement features, which may omit actions; a context signal selects among behaviors. In simulation, the humanoid demonstrates walking and getting up, but behavior quality lacks an objective score and demonstrated reuse remains limited.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.