Reproduce “DreamLander: A 6-DOF World Model based Reinforcement Learning Planetary Lander” 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
George Nehma, Patrick Quinn, Madhur Tiwari, Camilo Riano-Rios
DreamLander uses imagined futures to learn steering for a simulated lander. In 100 simulations targeting the origin, 92 met the criteria of landing within 2 m and speed at most 0.3 m/s, though the paper’s stated 98% conflicts with that count. For a shifted target, 81 of 100 succeeded; these results apply only to the tested simulator.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.