mirror of
https://github.com/Mobile-Robotics-W20-Team-9/polex.git
synced 2025-09-07 13:23:13 +00:00
63 lines
3.3 KiB
Markdown
63 lines
3.3 KiB
Markdown
# Long-Term Urban Vehicle Localization Using Pole Landmarks Extracted from 3-D Lidar Scans
|
|
|
|
This repository contains the Python code that accompanies our paper ["Long-Term Urban Vehicle Localization Using Pole Landmarks Extracted from 3-D Lidar Scans"](http://ais.informatik.uni-freiburg.de/publications/papers/schaefer19ecmr.pdf) submitted to the [European Conference on Mobile Robots](https://www.ecmr2019.eu/).
|
|
The implementation allows to
|
|
|
|
* extract the parameters of pole-like objects from 3-D lidar scans,
|
|
* create a global reference map of pole landmarks,
|
|
* localize a vehicle online based on the reference map and live lidar measurements,
|
|
* replicate the experiments on the [NCLT dataset](http://robots.engin.umich.edu/nclt/) and on the [KITTI dataset](http://www.cvlibs.net/datasets/kitti/) described in the paper.
|
|
|
|
It provides the following three software modules.
|
|
|
|
### [Pole extractor](poles/poles.py)
|
|
|
|
This module takes odometry and 3-D lidar scans accumulated over a short trajectory segment as input, searches for pole-like objects in the data, and outputs the parameters of the corresponding pole estimates.
|
|
|
|
|
|
<br/>*Pole extraction from NCLT lidar data.*
|
|
|
|
|
|
<br/>*Pole extraction from KITTI lidar data.*
|
|
|
|
### Mapping module
|
|
|
|
Given a set of possibly overlapping local landmark maps generated by the pole extractor, this module resolves all ambiguities and creates a global reference map of pole landmarks.
|
|
|
|
|
|
<br/>*NCLT landmark map.*
|
|
|
|
|
|
<br/>*KITTI landmark map with vehicle trajectory.*
|
|
|
|
### Localization module
|
|
|
|
On the basis of the global map, live odometry measurements, and pole landmark estimates, this module computes an estimate of the current vehicle pose using a particle filter.
|
|
|
|
|
|
<br/>*Particle filter localization on NCLT.
|
|
The red dots denote the particles, the blue dots denote the reference landmarks, ad the black crosses visualize the online landmarks.*
|
|
|
|
## Running the code
|
|
|
|
First of all, please make sure you are running Python 2.7.
|
|
|
|
While the pole extractor is represented by its own Python module [poles.py](poles/poles.py), the mapping and localization module are implemented separately for NCLT [(ncltpoles.py)](poles/ncltpoles.py) and KITTI [(kittipoles.py)](poles/kittipoles.py) due to the different representations of the datasets.
|
|
For closer information about the workings of the implementation, please read the paper and follow the source code documentation.
|
|
|
|
In order to run the scripts with the experiments on NCLT [(ncltpoles.py)](poles/ncltpoles.py) and KITTI [(kittipoles.py)](poles/kittipoles.py), please install the package manager `pip` via
|
|
|
|
```bash
|
|
sudo apt install python-pip python-tk
|
|
```
|
|
|
|
and use it to install the following Python packages:
|
|
|
|
```bash
|
|
pip install numpy matplotlib open3d-python progressbar pyquaternion transforms3d scipy scikit-image networkx psutil
|
|
```
|
|
|
|
Then, please check out the [ray tracing repository](https://github.com/acschaefer/ray_tracing) and build it.
|
|
|
|
With these prerequisites, you are ready to run the experiments and the different modules.
|