This repository contains all code for the following evaluations presented in the paper:
Note: the project was implemented and tested on Python 3.9.16
Set a virtual environment to run the code:
cd level_crossing
python -m venv env
source env/bin/activate
pip install --upgrade pip
pip install -r requirements.txt
To run a DFS and get the state space, use the main_dfs_single.py
:
python main_dfs_single.py --n 50 --k 1 --m 1
To run the DRL algorithm and get the time that takes the algorithm to reach a policy, use the main_mask.py
:
python main_mask.py --n 50 --k 1 --m 1 --total_timesteps 100000 --repeat 5
Note: the project requires Conda
Create a virtual environment and activate it:
cd sokoban
conda create --name bp-sokoban python=3.7 --file requirements.txt
conda activate bp-sokoban
IMPORTANT: The evaluation requires a lot of RAM. We recommend using a machine with at least 64GB of RAM.
- map name - the name of the map to run on. one of the following:
- map_6_8_3
- map_12_11_1
- map_13_12_1
- map_9_10_2
- map_7_7_2
- map_7_7_3
- map_11_9_2
- map_6_6_3
- map_6_7_3
- map_8_8_2
- map_8_2
- map_10_9_2
- map_9_7_2_
- map_9_7_2
- map_10_7_2
- map_8_9_2
- map_8_7_2
- map_9_8_2
- map_9_9_2
- map_11_8_2
- single/multiple requirements - whether to run the single ("0") or multiple ("1") liveness requirements scenario.
An example of running the code on the map_6_8_3 map with the single liveness requirement scenario is as follows:
python main_sokoban.py "map_6_8_3" "0"
Note: the project was implemented and tested on Python 3.9.16
Set a virtual environment to run the code:
cd dwyer_patterns
python -m venv env
source env/bin/activate
pip install --upgrade pip
pip install -r requirements.txt
To run a check of a single pattern run the check.py
:
python check.py --pattern 8