Reproduce “Physics-informed reinforcement learning via probabilistic co-adjustment functions” 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
Nat Wannawas, A. Aldo Faisal
The authors introduce CKA, which adjusts a simplified physics model’s predictions with a learned scale and bias. Gaussian-process priors let predictions default to the physics model where observations are absent. In an OpenSim arm simulation, after trial 9, CKA’s reaching-error summary was 5.59 cm versus 11.23 cm for Dyna using the unadjusted model; this does not demonstrate physical-human control.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.