manila95 / metadrive_clean

MetaDrive-0.2.6.0 that is compatible with newest gym version.

fsrl_metadrive

We provide nine different safe driving environments in metadrive, named as: SafeMetaDrive-{Road}{Vehicle}-v0, where {Road} includes three different levels for self-driving cars: easy, medium, hard, {Vehicle} includes four different levels of surrounding traffic: sparse, mean, dense.

1. Set up the environments:

# set up metadrive
git clone --recursive https://github.com/HenryLHH/metadrive_clean.git
cd metadrive_clean/metadrive_clean
pip3 install -e .

# set up data collector
cd ../
pip3 install -e .

2. MetaDrive Safe Environment import

import gym
import metadrive.fsrl_metadrive

env = gym.make("SafeMetaDrive-easysparse-v0")

3. MetaDrive data collection for fsrl

3.1. Run collecting the dataset:

bash run_collect.sh [your_env_name]

e.g.

bash run_collect.sh SafeMetaDrive-harddense-v0

3.2 Visualize the collected dataset:

bash run_visualize.sh [your_env_name]

e.g.

bash run_collect.sh SafeMetaDrive-harddense-v0

The cost-reward plot will be saved in figs/ folder.