Reproduce “From Prediction Uncertainty to Conformalized Distance Fields for Safe Motion Planning” 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
Jaeuk Shin, Yoonseok Ra, Insoon Yang
The paper introduces Functional Conformal Prediction-Model Predictive Control (FCP-MPC), a framework for safe motion planning that quantifies prediction uncertainty on distance fields. It achieves real-time robust safety guarantees, showing improved scalability and goal reachability in dense, dynamic environments compared to prior methods.
Key insights
- Prediction errors in distance fields are not arbitrary; they are spatially structured, approximately time-invariant, and low-rank, allowing for a compact functional representation.
- Conformalizing the entire spatial field of prediction errors, rather than scalar scores, yields a continuous, distribution-free safety certificate applicable uniformly over the workspace, resolving spatial incoherence issues.
- The low-rank and approximately time-invariant properties of residual fields enable an offline-online decomposition, making online computation largely insensitive to the number of obstacles and suitable for real-time deployment.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.