Reproduce “DAPPT: Domain-Aware Policies from Procedural Trajectories for Sim-to-Real Dexterous Tool Use” 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
Beining Han, Jia Deng
DAPPT trains dexterous robot policies to use tools across tasks involving materials that are difficult to simulate. It conditions a policy on parameters defining procedural tool-motion paths, trains across settings in rigid-body simulation, and selects settings using task metrics. In Genesis sand-scooping evaluation, the best setting yielded 13.9 g versus 3.64 g on average; this is a simulation result, not a real-world gain.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.