Reproduce “Learning to Pick Apples: Targeted Domain Randomization via Real-to-Sim System Identification” 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
Abhinav Jain, Connor Brooks, Stefan Lee, Cindy Grimm
The authors measure pulls on nine fruiting systems and tune a deformable simulator to the force data. They train policies with randomized parameters sampled around identified values. On 10,000 unseen simulated systems, the policy reaches 90% success after 5.1M environment steps on average, versus 15.8M for wide randomization, and detaches apples in 5/5 preliminary trials on proxies, not orchard trees.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.