Reproduce “Energy-Based Policy Orchestration” 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
Weirui Ye, Yen-Jen Wang, Elvis Hsieh, Zhiyang Dou, Shibo Zhao, Pieter Abbeel, Yilun Du, Leslie Pack Kaelbling and 1 more
EPO lets a robot choose at runtime a frozen low-level policy and its goal, while leaving the planner’s target unchanged. Policy-specific models estimate feasibility and the probability of tracking that target, combining them to select a policy-goal pair. In simulated standing manipulation, EPO-M4 lowers normalized tracking error to 0.420 versus MoE-M4’s 0.461, with weighted survival of 97.7% versus 98.3%.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.