MB-RRL2

a library for model-based relational reinforcement learning

mbrrl2 is a fork of libPRADA

Main features of this fork

• New object-oriented interface of the relational rule learner
• Task-sensitive learning as presented in [1]
• Incremental rule learning
• Several bugfixes

Requirements

The new rule learner interface requires boost. If you don't want to use it set RULELEARNER2 = 0 in the make-config file in the root directory.

Installation

Please type make in the root directory. Then, try out the demos in mbrrl2/test.

New demos are

• test/incremental (showing how to use incremental rule learning)
• test/tasksensitive_example (showing a the task-sensitive cost function presented in [2])

To understand the main features of this library, please read the guide: doc/guide.pdf

When using this library...

... please cite the following paper:

[1] Sebastian Höfer, Tobias Lang, Oliver Brock
Extracting Kinematic Background Knowledge from Interactions Using Task-Sensitive Relational Learning. 
Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), 
2014, pp. 4342-4347.

Bibtex:

@inproceedings{hoefer_extracting_2014,
  Title = {Extracting Kinematic Background Knowledge from Interactions Using Task-Sensitive Relational Learning},
  Author = {Sebastian Höfer and Tobias Lang and Oliver Brock},
  Booktitle = {Proceedings of the IEEE International Conference on Robotics and Automation (ICRA)},
  Pages = {4342-4347},
  Year = {2014},
  Doi = {10.1109/ICRA.2014.6907491},
  Location = {Hong Kong, China},
  Url = {http://www.robotics.tu-berlin.de/fileadmin/fg170/Publikationen_pdf/hoefer_extracting_2014.pdf},
}

If you use the planner PRADA, please cite:

[2] Planning with Noisy Probabilistic Relational Rules
Journal of Artificial Intelligence Research
2010, Volume 39, pages 1-49  

Bibtex:

@article{lang-toussaint-10jair,
  author = {Tobias Lang and Marc Toussaint},
  title = {Planning with Noisy Probabilistic Relational Rules},
  journal = {Journal of Artificial Intelligence Research},
  year = {2010},
  volume = {39},
  pages = {1-49},
  pdfurl = {http://www.jair.org/media/3093/live-3093-5172-jair.pdf},
}

If you use the relational rule learner, please cite the original publication presenting this learner:

[3]  Pasula, Hanna M., Luke S. Zettlemoyer, and Leslie Pack Kaelbling. 
Learning symbolic models of stochastic domains. 
Journal of Artificial Intelligence Research 29 (2007): 309-352.

Bibtex:

@article{pasula2007learning,
  title={Learning symbolic models of stochastic domains},
  author={Pasula, Hanna M and Zettlemoyer, Luke S and Kaelbling, Leslie Pack},
  journal={Journal of Artificial Intelligence Research},
  volume={29},
  pages={309--352},
  year={2007}
}

GPL licence statement:

mbrrl2 is a fork of libPRADA by Sebastian Höfer science@sebastianhoefer.de.

libPRADA Copyright 2008-2012 Tobias Lang email: tobias.lang@fu-berlin.de

This file is part of mbrrl2/libPRADA.

mbrrl2/libPRADA is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

mbrrl2 is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with mbrrl2. If not, see http://www.gnu.org/licenses/