Deep Learning for Traffic Prediction with an Application to Traffic Lights Optimization.

Abstract

This work proposes the use of deep neural networks for the prediction of traffic variables for measuring traffic congestion. Deep neural networks are used in this work in order to determine how much time each vehicle spends in traffic, considering a certain amount of vehicles in the traffic network and traffic light configurations. A genetic algorithm is also implemented that finds an optimal traffic light configuration. With the implementation of a deep neural network for the simulation of traffic instead of using a simulation software, the computation time of the fitness function in the genetic algorithm improved considerably, with a decrease of precision of less than 10%. Genetic algorithms are used in order to show how useful deep neural networks models can be when dealing with vehicular flow slowdown.