About

In this repository we provide the prototype implementation of the MCL GLOSM/ELOFF filter proposed in our paper in [1].

Related repositories:

• SLOPY: Point-to-plane odometry implemented for validating the approach;
• DF-VO: Odometry source used for comparing SLOPY for the Kitti demo;
• 3D Kitti Mapper: Maps the Kitti trajectories and export them as PLY. Also provide the tools for exporting the mapped trajectories to digital surface map raster files.

If having trouble for associating the date/drive to their respective sequences, we provide the mappings under scripts/demos/kitti/sequence_utils.py.

Setup

For running the demos, we suggest the user to prepare an Anaconda environment using the environment.yml file.

Kitti demo

For evaluating our method, we implemented a script for running the Kitti sequences under scripts/demos/kitti/demo_kitti.py.

Setting the configuration file

In the configuration file under cfg/demo.yaml, you will find the parameters for the particle filter, SLOPY, GLOSM, ELOFF and the output video demo. Even though not necessary to change (our experiments were executed using those), you can adjust the parameters as you would like there.

The parameters are detailed in the file.

Running the pipeline

Run the command for executing the SLOPY+GLOSM+ELOFF pipeline with the required arguments:

python3 scripts/demos/kitti/demo_kitti.py \
    --dataset ${DATASET_DIR} \
    --sequence_id ${SEQUENCE} \
    --dsm_path ${DSM_PATH}

The required arguments are:

• --dataset, the absolute path to the Kitti dataset extracted in the date-drive format as expected by the PyKitti toolkit;
• --sequence_id, the id of the sequence (00, 01,..., 10) to be tested;
• --dsm_path, the absolute path to the DSM of the sequence.

Other arguments that are optional are detailed if you run python3 scripts/demos/kitti/demo_kitti.py -h.

Citation

To be included.