RL_BipedalWalker

RL BipedalWalker

The core CureFuzz is in the folder ./rl-baselines3-zoo/fuzz/cure_fuzz.py.

The RL algorithm is in the folder ./rl-baselines3-zoo.

The RL model we evaluate is borrowed from these awesome repositories: https://github.com/DLR-RM/rl-baselines3-zoo, https://github.com/DLR-RM/rl-trained-agents, which are under MIT license.

---

Setting up environment:

Run the following:

# Setup environment
conda create -n RLWalk python=3.6.3
conda env update --name RLWalk --file environment_RLWalk.yml
conda activate RLWalk
cp ./gym/setup.py ./
pip install -e .
cp ./stable_baselines3/setup.py ./
pip install -e .

# Download trained models
cd ./rl-baselines3-zoo
git clone https://github.com/DLR-RM/rl-trained-agents

---

Fuzz testing:

Check the default path of the model is correct in ./enjoy_cure.py.

Run python enjoy_cure.py --alg tqc --env BipedalWalkerHardcore-v3 --folder rl-trained-agents/ --guide --no-render --n-timesteps 300 to start fuzz testing.