CMAX++ : Leveraging Experience in Planning and Execution using Inaccurate Models

BibTeX Citation

@article{DBLP:journals/corr/abs-2009-09942,
  author    = {Anirudh Vemula and
               J. Andrew Bagnell and
               Maxim Likhachev},
  title     = {{CMAX++} : Leveraging Experience in Planning and Execution using Inaccurate
               Models},
  journal   = {CoRR},
  volume    = {abs/2009.09942},
  year      = {2020},
  url       = {https://arxiv.org/abs/2009.09942},
  archivePrefix = {arXiv},
  eprint    = {2009.09942},
  timestamp = {Wed, 23 Sep 2020 15:51:46 +0200},
  biburl    = {https://dblp.org/rec/journals/corr/abs-2009-09942.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}

Dependencies

To install all the required dependencies, run the following command

pip install -r requirements.txt

Path issues

The project uses global path everywhere and requires that the project folder be placed at a specific path. Run the following commands from your home folder,

mkdir -p workspaces/szeth_ws/src/

and then copy the project folder into the `szeth_ws/src' folder

Reproducing Experiments

3D Mobile Robot Navigation

To reproduce the results for 3D mobile robot navigation experiment in the paper, run the following command inside the src/ folder

python -m szeth.experiment.experiment_car_racing --agent <agent>

where <agent> should be one of {'cmax', 'cmaxpp', 'adaptive_cmaxpp'}. This saves the results in the data/ folder

To reproduce the bar graph, run the following command in project root,

python scripts/plot_car_racing_bar.py

The plot will be saved inside plot/ folder.

To reproduce the sensitivity experiments reported in the appendix run the following script inside the src/ folder

./alpha_expts.sh

7D Pick-and-Place

To reproduce the results for 7D pick-and-place experiment in the paper, run the following command inside src/ folder

python -m szeth.experiment.experiment_pr2_7d_approximate --agent <agent>

where <agent> should be one of {'cmax', 'cmaxpp', 'adaptive_cmaxpp', 'model', 'knn', 'qlearning'} and model refers to the NN residual model learning approach. This saves the results in the data/ folder

Contributors

The repository is maintained and developed by Anirudh Vemula from the Search based Planning Lab (SBPL) at CMU.