The code uses CityFlow simulation software with several utility functions added. Therefore one needs to download and install my personal fork of CityFlow: https://github.com/mbkorecki/CityFlow. To run the simulation, change directory to the src folder and call:

python3 traffic_sim.py

This would run the simulation with the default arguments. If you want to specify your own arguments here is what you have at your disposal with example uses:

--sim_config "../4x4/1.config"

The path to the cityflow config file for the simulation you want to run

--num_episodes 1

The number of episodes you want your learning algorithm to learn for

--num_sim_steps 1800

Number of simulation steps you want the simulation to run for in a single episode

--agents_type "learning"

The type of agents you want to run, for now available options are: learning or analytical

--update_freq 10

How often the reinforcement learning agent updates its q-network

--batch_size 64

The size of the mini-batch used to train the deep-q-network

--lr 5e-4

The learning rate

Example call would look like this:

python3 traffic_sim.py --agents_type 'analytical' --sim_config '../4x4/1.config' --num_episodes 1 --num_sim_steps 1800

The docs folder contains documentation build files which can be run with eg. make html.