Reproduce “K-PWM: Control-Oriented Structured Bilinear World Models under Partial Observation” 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

Santosh Mohan Rajkumar, Sriram Narayanan, Samuel E. Otto, Debdipta Goswami

K-PWM models hidden robot state with a compact, action-dependent latent dynamics model, inferring a probability distribution from noisy, partial observations and planning with its predictions. On Half Cheetah, its 12-second observation RMSE is 0.048 ± 0.0060, versus 0.14 ± 0.018 for MTS3. It also tracks circular references on four simulated robots; these results do not establish hardware performance.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.