This repository implements stable_baselines training on the environment gym_novel_gridworlds
conda create --name <env> --file requirements.txt
conda activate <env>
python train.py -E <name-of-env> -N <novelty-name> -D <novelty-difficulty> -N1 <novelty-arg1> -N2 <novelty-arg2> -I <timestep-to-inject-novelty> -T <number-of-tests> -M <number-of-models-to-save>
python train.py -N breakincrease -N1 stick
python train.py -N remapaction
python train.py -N firewall
python plot_results.py
PDDL files for all the environments are found here
CSV files from the planning architecture used for evaluations in the paper are found here