K_Road

A 2-D vehicle simulator for autonomous vehicle research.

pip install -r requirements.txt

Directory Structure:

• command_line_tools/ -- a library for parsing JSON-like configs from the command line
• coordinated_optimizer/ -- a library of distributed direct policy search methods
• data_logging/ -- a library for logging experimental data
• factored_gym/ -- a library for implementing AI gyms as separate components (process, reward function, observation function, terminator, etc)
• hpc/ -- scripts and utilities for running K_Road on HPC systems
• k_road/ -- the K_Road simulator
• scenario/ -- example scenarios for K_Road
• trainer/ -- Various RL algorithm implementations for use with RLLib and KRoad

Setting up an environment

# create env
conda create --name k_road

# install dependencies
conda install numpy scipy
pip install pygame pymunk

# Optional: 
conda install nb_conda_kernels

# with GPU:
# first, install CUDA drivers: http://www.nvidia.com/getcuda
sudo nvidia-xconfig
# then reboot
glxinfo

conda install tensorflow-gpu

# without GPU:
conda install tensorflow

pip install stable_baselines