README.md

We propose a lightweight, self-supervised approach that enhances the synergy between online mapping and trajectory prediction, providing interpretability on when and where map uncertainty is beneficial. Additionally, we introduce a covariance-based uncertainty modeling method that better aligns with road geometry. Extensive ablation studies show that our method outperforms previous integration strategies, achieving up to 23.6% improvement in mapless trajectory prediction on the nuScenes dataset.

Project Page

https://ethan-zheng136.github.io/Dev-Unc/

Overview

We first estimate map elements online by encoding multi-view images into a common BEV feature space to regress map element vertices. Each vertex’s uncertainty is modeled using our proposed Covariance-based Uncertainty method, which leverages 2D Gaussian distribution templates. This uncertainty information, along with the original map vertices, is then passed to the downstream trajectory prediction module, which operates in two parallel streams: one that incorporates uncertainty and one that does not. Finally, the proposed Proprioceptive Scenario Gating (MLP network) dynamically adapts the optimal trajectory prediction based on the initial future trajectories prediction from these two streams.

Our results

Our demo video

Click the cover image to watch the HD demo on YouTube.

Getting Started

• Environment Setup
• Prepare Dataset
• Mapping Train and Eval
• Merge Map and Trajectory Dataset
• Trajectory Train and Eval
• Visualization

Checkpoints

Our trajectory prediction checkpoints are here.

Dataset

All the trajectory prediction data（for MapTR, StreamMapNet, MapTRv2 and MapTRv2 CL）can be generated using our future checkpoints, with a total size of approximately 600GB. Dataset Structure is as follows:

DelvingUncPrediction
├── trj_data/
│   ├── maptr/
│   |   ├── train/
│   |   |   ├── data/
│   |   |   |   ├── scene-{scene_id}.pkl
│   |   ├── val/
│   ├── maptrv2/
│   ├── maptrv2_CL/
│   ├── stream/

Catalog

• Code release
  • MapTR
  • MapTRv2
  • StreamMapNet
  • HiVT
  • DenseTNT
• Visualization Code
• Untested version released + Instructions
• Initialization

License

This repository is licensed under Apache 2.0.

Citation

If you find this project useful, feel free to cite our work!

@article{zhang2025delving,
  title={Delving into Mapping Uncertainty for Mapless Trajectory Prediction},
  author={Zhang, Zongzheng and Qiu, Xuchong and Zhang, Boran and Zheng, Guantian and Gu, Xunjiang and Chi, Guoxuan and Gao, Huan-ang and Wang, Leichen and Liu, Ziming and Li, Xinrun and others},
  journal={arXiv preprint arXiv:2507.18498},
  year={2025}
}