trafficRL

Adaptive Traffic Control System Using Reinforcment Learning

Run

Setup

1. Install SUMO: http://sumo.dlr.de/wiki/Downloads

2. Install MongoDB: Follow the official instructions at:

   https://docs.mongodb.org/manual/tutorial/install-mongodb-on-ubuntu/ (for ubuntu) https://docs.mongodb.com/manual/tutorial/install-mongodb-on-windows/ (for windows)

   Check mongo is installed by starting the server from the command line:

   For Ubuntu: mongod

   For Windows: C:\Program Files\MongoDB\Server\3.4\bin\mongod.exe

3. Clone the project using git and cd into it:

    git clone https://github.com/codemerlin19/trafficRL.git
   
    cd trafficRL
   

4. (Optionally) Install Virtual Env

    pip3 install virtualenv
    virtualenv .venv
    source .venv/bin/activate
   

5. Install requirements

    pip3 install -r requirements.txt
   

6. Run project

    python runner.py
   

Usage: runner.py [options]

    -h, --help          show this help message and exit
    --nogui             run the commandline version of sumo
    -C NUM, --cars=NUM  specify the number of cars generated for simulation
    --bracket=BRACKET   specify the number with which to partition the range of
                            queue length/cumulative delay
    --learning=NUM      specify learning method (1 = Q-Learning, 2 = SARSA)
    --state=NUM         specify traffic state representation to be used (1 =
                            Queue Length, 2 = Cumulative Delay)
    --phasing=NUM       specify phasing scheme (1 = Fixed Phasing, 2 = Variable
                            Phasing)
    --action=NUM        specify action selection method (1 = epsilon greedy, 2 =
                            softmax)

NOTE: Ensure that SUMO_HOME path is defined in environment variables

Uses

SUMO

Last tested with Eclipse SUMO Version 1.6.0.

• Traci

Mongo

Currently it expects mongo on localhost on the default port without any authentication.

Development

Linting

For auto fixes use autopep8:

python -m pip install autopep8
autopep8 --in-place --aggressive  --exclude "__pycache__,.venv"  ./*.py