Reproduce “MILER: Semantic Mid-Level Representation for Sim-to-Real Reinforcement Learning in Unstructured Autonomous Driving” 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
Thomas Steinecker, Denis Trescher, Alexander Bienemann, Thorsten Luettel, Mirko Maehlisch
MILER trains a reinforcement-learning policy in an abstract simulator using semantic bird’s-eye-view maps shared with real-world perception. During deployment, the policy controls a virtual vehicle, while controllers make the real vehicle follow its trajectory. MuCAR-4 completed three 3.0 km rounds without intervention and reached 33.6 km/h; vehicle B stopped in its final round, limiting the evidence.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.