FlowDrive: moderated flow matching with data balancing for trajectory planning
FlowDrive: moderated flow matching with data balancing for trajectory planning
This paper introduces FlowDrive, a learning-based trajectory planner that addresses data imbalance in autonomous driving by reweighting training samples according to trajectory patterns. The method employs conditional rectified flow to map noise directly to trajectory distributions and incorporates moderated guidance to enhance diversity while maintaining scene consistency. Evaluated on nuPlan and interPlan benchmarks, FlowDrive achieves state-of-the-art performance among learning-based planners and approaches rule-based methods. A variant with additional post-processing (FlowDrive*) further improves overall results across benchmark splits. The implementation is available publicly.
The authors present FlowDrive, a flow-matching trajectory planner that uses data balancing and moderated in-the-loop guidance to improve performance on long-tailed driving scenarios; results are reported on nuPlan and interPlan benchmarks with code released.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- code: Available Artifact link Open source S4
- paper: not_reported
- dataset: not_reported
- weights: not_reported
- code: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
eighting by trajectory pattern an effective approach. We then present FlowDrive, a flow-matching trajectory planner that learns a conditional rectified flow to map noise directly to trajectory distributions with few flow-matching steps. We further introduce moderated, in-the-loop guidance that injects small perturbation between flow steps to systematically increase trajectory diversity while remaining scene-consisten
Open source S3
t. On nuPlan and the interaction-focused interPlan benchmarks, FlowDrive achieves state-of-the-art results among learning-based planners and approaches methods with rule-based refinements. After adding moderated guidance and light post-processing (FlowDrive*), it achieves overall state-of-the-art performance across nearly all benchmark splits. Our code is available at https://github.com/einsteinguang/flow_drive_plann
Open source S4
Source：arXiv Robotics — research abstracts · arxiv.org