The code in this project simulates an four way traffic intersection and allows a neural network to train off of data produced. In main.py the user can fail over the traffic light to run from the neural network.
The project requires python3.5+ with tensorflow 1.7. I highly recommend using the gpu version of tensorflow called tensorflow-gpu. It requires several additional pieces of software from NVIDIA and the installation procedure found here.
Once you have finished the NVIDIA installation and have python 3.5+ installed you can enter the following in your terminal:
python3 -m pip install numpy pandas tensorflow-gpu==1.7.0
To train the model you can run train_failure_model.py this will create a training_log.csv that will be used to train and the model will be saved FailureModel/saved_model/model
python3 train_failure_model.py
Now that you have created training data, and trained the model it's time to run the simulation. To do this run the main.py file.
python3 main.py
Once main.py has been run you can graph the results of the simulation by running primary_vs_failover.py
python3 primary_vs_failover.py