Warning: This code has not been touched since 2019; however, someone has requested to use this work as a benchmark (so it is being made available). You can use this code as you wish as long as you cite my PhD and/or the related publication.
PhD: Harman, H. (2020). Symbolic artificial intelligence techniques to facilitate proactive robot assistance (Doctoral dissertation, Ghent University).
Publication: Harman, H. and Simoens, P. (2021). Learning Symbolic Action Definitions from Unlabelled Image Pairs. In Proceedings of the 4th International Conference on Advances in Artificial Intelligence (ICAAI '20). Association for Computing Machinery, New York, NY, USA, 72–78. https://doi.org/10.1145/3441417.3441419
Initial workshop paper: Harman, H., & Simoens, P. (2020). Generating symbolic action definitions from pairs of images: Applied to solving towers of Hanoi. In the Workshop on Plan, Activity, and Intent Recognition (PAIR) at 34th AAAI conference on Artificial Intelligence, AAAI2020.
Note: I think you may need to download FastDownward to run this code. You will also need to edit the horrible hardcoded absolute file paths, e.g., in source/problem/fast_downward.py
python3 opencv: pip install --user opencv-python
sudo apt-get install python3-pip sudo python3 -m pip install opencv-python
http://www.fast-downward.org/ObtainingAndRunningFastDownward (Helmert, M. (2006). The fast downward planning system. Journal of Artificial Intelligence Research, 26, 191-246.)