tested on ubuntu 18.04LTS
- install torcs
cd vtorcs-RL-color sudo apt-get install libglib2.0-dev libgl1-mesa-dev libglu1-mesa-dev freeglut3-dev ibplib-dev libopenal-dev libalut-dev libxi-dev libxmu-dev libxrender-dev libxrandr-dev libpng-dev ./configure sudo make sudo make install sudo make datainstall
- install xautomation
for key automation
sudo apt-get update -y sudo apt-get install -y xautomation
first need to change configureation of game environment
# start torcs env
sudo torcs -vision
- go to race configuration and select racing track and set the driver
scr_server 1
race -> practice -> configure race -> select racing track -> select driver
- go to display setup, select
64x64
RGB observation and16
color depthoptions -> Display
start training!
python train.py
DAgger_example
├── README.md
├── train.py
│ train agent with DAgger algorithm
│
├── agent.py
│ neural network architecture of learner
│
├── snakeoil3_gym.py
│ Torcs client communicate with Torcs game server
│
├── gym_torcs.py
│ make Torcs client gym environment
│
├── autostart.sh
│ key automation
│
├── vtorcs-RL-color
│ Torcs source, copied from https://github.com/ugo-nama-kun/gym_torcs
│
├── images
sampled images