gisbi-kim / lt-mapper

A Modular Framework for LiDAR-based Lifelong Mapping

LT-mapper (ICRA 22)

What is LT-mapper?

• A Modular Framework for LiDAR-based Lifelong Mapping
  • Paper
  • Video

Why LT-mapper?

• For LiDAR-based long-term mapping, three challenges exist.

  

Features

• TBA

How to use?

Prerequisites

• TBA

build

• TBA

Tutorial and examples

0. Single-session Data Generation

• Using the saver provided with SC-LIO-SAM (also in SC-A-LOAM or FAST_LIO_SLAM), a user should generate the set of sesssion data (i.e., keyframe point cloud scans, keyframe scan context descriptors (SCDs), and an initial pose-graph text file) for each session.

1. LT-SLAM

• Tutorial video

• command

    # change the paths in ltslam/config/params.yaml
    roslaunch ltslam run.launch
  

2. LT-removert and LT-map

• ltremovert directory runs removert (for single-session cleaning) and change detection between two sessions.
• command

    TBA
  

• Outputs

  

LT-mapper ParkingLot dataset

• LT-mapper ParkingLot dataset contains six sequences during three days for the same spatial site, but different initial poses.
• LT-SLAM automatically aligns them in a shared coordinate.

  

• Dataset Download Link. A sequence is replayable using MulRan File Player.
  • For the details of use, see this tutorial video (TBA).

Citation

@INPROCEEDINGS { kim2021ltmapper,
    AUTHOR = { Giseop Kim, Ayoung Kim },
    TITLE = { {LT-mapper: A Modular Framework for LiDAR-based Lifelong Mapping} },
    BOOKTITLE = { Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) },
    YEAR = { 2022 },
}

Contact

• Maintained by Giseop Kim and please contact the author via [e-mail]