Reproduce “DriftingPlanner: Drifting Generative Models for One-Step Trajectory Planning in End-to-End 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
Ziming Liu, Yizhan Qu, Leichen Wang, Hao Sun
DriftingPlanner trains a one-step trajectory generator with an attraction–repulsion field, using multiple noise samples for each scene’s single expert path without a coordinate-regression objective. At inference, it decodes paired noise inputs and their negations in one batch and averages the outputs. In a shared-backbone comparison, it scores 90.19 PDMS versus 90.32 for ten-step flow matching, with lower measured latency.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.