This repository is showing the method to optimise Traffic Signal Control in X Crosswalk with Deep Reinforcement Learning.
The tested installation environment is Ubuntu(ver =< 20.04).
As a premise, to implement this simulation, instration of SUMO(Simulaton of Urban Mobility) is eccentially required following below:
SUMO INSTRATION HERE: https://www.eclipse.org/sumo/.
After getting the instration done, run the commands in your terminal
export SUMO_HOME=/usr/share/sumo
cd $SUMO_HOME
Then you are now able to see the SUMO system in your machine.
My CUDA version is 11.0. So to the best of my knowledge, TensorFlow2.4+CUDA==11.0 is reckoned as the best couple.
In terms of the optimal combination between NVIDIA cuDNN and python packages, you can dive deep into the following web site https://pytorch.org/get-started/previous-versions/.
After completion of setting CUDA properly, I would recommend you to run this command firstly,
conda install pytorch==1.7.1 torchvision==0.8.2 torchaudio==0.7.2 cudatoolkit=11.0 -c pytorch
Next My Python environment version is python == 3.7.11 which works well on my machine(you might be able to change the version but it does noeed to be corresponding to your CUDA Version)
To install all of the above, it should be enough to run
pip install -r requirements.txt
from the base directory of the package.
To see if what kind of intersection you will work with is
The method I use to optimise the TSC is straightforwardlly Double Deep Q-Network
The diagram of the amount of pedestrian on this intersection is
python main.py
After traning this model, you can see the output result in models. like below:
Thank you.