lanelet2anchors 0.1.0

Easy anchor generation for autonomous driving based on Lanelet2 maps

lanelet2anchors

This package enables the generation of diverse map-based anchor paths for a given lanelet or vehicle pose. It was developed as part of our paper Lanelet2 for nuScenes: Enabling Spatial Semantic Relationships and Diverse Map-based Anchor Paths. For details check our project page.

Installation

Use Python >= 3.8 and pip to install

pip install lanelet2anchors

Usage

Check the notebooks in the documentation for details and examples.

Generate anchors for given vehicle

from lanelet2anchors import AnchorGenerator

lanelet_map = AnchorGenerator(
    osm_file=osm_file, origin_latitude=49.00345654351, origin_longitude=8.42427590707
)

anchors = lanelet_map.create_anchors_for_vehicle(
    vehicle_pose=X,
    anchor_length=100,
    num_anchors: 5,
)

Generate anchors for a given Lanelet

from lanelet2anchors import AnchorGenerator

lanelet_map = AnchorGenerator(
    osm_file=osm_file, origin_latitude=49.00345654351, origin_longitude=8.42427590707
)

anchors = lanelet_map.create_anchors_for_lanelet(
    lanelet_id=lanelet_map.lanelet_ids[0], anchor_length=100
)

Citation

If you use this code for scientific research, please consider citing

@InProceedings{naumannHertleinLanelet2NuScenes,
    author    = {Naumann, Alexander and Hertlein, Felix and Grimm, Daniel and Zipfl, Maximilian and Thoma, Steffen and Rettinger, Achim and Halilaj, Lavdim and Luettin, Juergen and Schmid, Stefan and Caesar, Holger},
    title     = {Lanelet2 for nuScenes: Enabling Spatial Semantic Relationships and Diverse Map-Based Anchor Paths},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
    month     = {June},
    year      = {2023},
    pages     = {3247-3256}
}

License

This project is licensed under CC-BY-NC-SA. Development by FZI Forschungszentrum Informatik and all rights reserved by Robert Bosch GmbH.