Construction of 3D digital twin environment for vehicle using Unity ML-Agents and implementation of self-driving technology using deep reinforcement learning(DQN)
- Unity: 2021.3
- ML-Agents: 2.2.1 (release 19)
- Python: 3.7
- ML-Agents Python package: 0.28.0
- Torch: 1.7.1+cu110
-
Anaconda3: Download and installation instructions here: https://www.anaconda.com/products/individual
-
Nvidia driver: Download by setting it to your GPU specifications. https://www.nvidia.com/Download/index.aspx?lang=kr
-
cuDNN, CUDA, Python version: Check the version here: https://www.tensorflow.org/install/source_windows
-
CUDA: Download here:
-
cuDNN: Download here:
https://developer.nvidia.com/rdp/cudnn-archive
After downloading, extract the file and move it to
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.2
-
-
Create a Virtual Environment in Anaconda3:
conda create -n your_virtual_env_name
conda activate your_virtual_env_name
pip install tensorflow-gpu
python -m pip install -q mlagents==0.28.0
pip3 install torch~=1.7.1 -f https://download.pytorch.org/whl/torch_stable.html
-
Open Visual Studio Code(Set up virtual environment)
Ctrl + Shift + P (Python select interpreter)
-
Open python scripts(Car_gym.py, DQN_220523.py)
-
Run DQN_220523.py script