Reproduce “I3L: Iterative Interactive Imitation Learning for Whole-Body Mobile Manipulation” 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
Wensi Ai, Christopher Luey, Karthik Dharmarajan, Evans Han, Yunfan Jiang, Stef Ren, Jianwen Xie, Jiajun Wu and 2 more
I3L improves whole-body mobile manipulation policies by having an operator take over at failures and using targeted corrections to retrain the policy over successive rounds. Across eight evaluated tasks, under the tested protocol, 100 correction trajectories outperform 100 added demonstrations, improving average full-task success from 25.6% to 93.9%. Final policies reach at least 90% success on each task.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.