Reproduce “Learning Modular Neural Network Policies for Multi-Task and Multi-Robot Transfer” 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
Coline Devin, Abhishek Gupta, Trevor Darrell, Pieter Abbeel, Sergey Levine
The method factors a robotic policy into task and robot modules, training them across different robot-task pairings so they can be recombined for an unseen pairing. In MuJoCo simulations, a four-link arm successfully pushed a block using a task module it had not trained with. Another held-out task required further learning, and the experiments do not establish real-robot transfer.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.