A simple 2D graph-slam system based on ICP in Python3. For Intel dataset, first 2000 datapoints would look like this:
where the green dots are localization after optimization and blue dots are the map.
- scipy
- numpy
- sklearn
- unittest
- matplotlib (for plotting only)
- g2o-python
pip3 install scipy numpy matplotlib scikit-learn g2o-python unittest
The datasets used in this SLAM demo are from http://ais.informatik.uni-freiburg.de/slamevaluation/datasets.php. Note, some datasets contain 360° lidar.
- ACES Building (Austin): aces
- Intel Research Lab (Seattle): intel
Install the required packages first and then run python3 setup.py install under the root directory
Download the data from http://ais.informatik.uni-freiburg.de/slamevaluation/datasets.php
Run
python3 utils/read_data.py intel
under utils to convert the data to the .npy dataset. The dafault value argument is aces.
Run
python3 graph_slam.py intel 0.1
to use icp for 2D SLAM. intel is the dataset name, 0.1 is the threshold for adding an edge to the graph.