Skip to content

Repository files navigation

AD4RL

ICRA2024

AD4RL: Autonomous Driving based on Dataset-driven Deep Reinforcement Learning

[Webpage] [Dataset]

1. FLOW Framework

See https://flow-project.github.io/ for Detail information of this framework

Installation

A. Anaconda with Python 3

  1. Install prequisites: sudo apt-get install libgl1-mesa-glx libegl1-mesa libxrandr2 libxss1 libxcursor1 libxcomposite1 libasound2 libxi6 libxtst6
  2. Download the Anaconda installation file for Linux in Anaconda, and unzip the file
  3. Install Anaconda bash ~/Downloads/Anaconda3-2023.03-1-Linux-x86_64.sh

NOTE: we recommend you to running conda init 'yes'.

B. FLOW installation

Following the below scripts in your terminal.

# Download FLOW github repo'.
git clone https://github.com/flow-project/flow.git
cd flow

# Create a conda env and install the FLOW
conda env create -f environment.yml
conda activate flow
python setup.py develop

# install flow on previoulsy created environment 
pip install -e .
B-1. SUMO installation

Install driving simulator (SUMO)

bash scripts/setup_sumo_ubuntu1804.sh
which sumo
sumo --version
sumo-gui

Testing the connection between FLOW and SUMO

conda activate flow
python examples/simulate.py ring
B-2. Pytorch installation

Install torch: conda install pytorch torchvision cudatoolkit=10.2 -c pytorch

NOTE: Should install at least 1.6.0 version of pytorch (Recommend torch = 1.11.0 & cudatoolkit=10.2).
Check the Pytorch Documents.

B-3. Ray RLlib installation

Install Ray: pip install -U ray==0.8.7

NOTE: Should install at least 0.8.6 version of Ray. (Recommend 0.8.7).

B-4. Python Library installation

2. AD4RL

Repository

Clone this library: git clone (Will be updated)

Dependencies Update

sh ./requirements/env_requirements.sh

Driving Scenarios

We provide the three driving scenarios as following table.

  • Click Driving Scenario, You can check the illustrative image about driving scenario
  • Click exp_config, You can check the code about driving scenario
Driving Scenario exp_config
Cut-in UnifiedRing
Lane Reduction MA_4BL
Highway MA_5LC

Dataset

You can access the AD4RL googledrive by clicking the title name of driving scenario.

Cut-in Lane Reduction Highway
cutin-expert lanereduction-expert highway-expert
cutin-medium lanereduction-medium highway-medium
cutin-random lanereduction-random highway-random
cutin-expert-medium lanereduction-expert-medium highway-expert-meidum
cutin-expert-random lanereduction-expert-random highway-expert-random
cutin-humanlike lanereduction-humanlike highway-humanlike

Train

python [algorithm] [exp_config] --dataset [dataset]
  • [algorithm]: main_BC.py, main_BCQ.py, main_DDPGBC.py, main_EDAC.py, main_PLAS.py
  • [exp_config]: UnifiedRing, MA_4BL, MA_5LC
  • [dataset]: See the above table (e.g., cutin-expert, highway-NGSIM)

About

ICRA 2024

Resources

Stars

18 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages