Predicting of this phenomenon. Swarms intelligence paradigm specially

Predicting traffic congestion has been the subject of several research works offering different models. These havebeen developed on the basis of general prediction methods and theories. They were then modified in order to adaptthem to the stochastic and non-linear nature of this phenomenon. These techniques include a non-exhaustive list ofthe following: statistics, time series, the machine learning techniques, Bayesian networks, genetic algorithms, fuzzylogic and hybrid approaches. The most widely used model in the literature to address this phenomenon is the ARMA(autoregressive and moving average). Another model proposes the decomposition of network flows as scalingcoefficient and applies ARIMA to predict traffic congestion. The Markov and Gray models also predict trafficcongestion1. Genetic algorithms were also used for predicting network traffic. A comparative study aims to determinethe optimal method in the prediction of short-term traffic congestion between three methods is given in2,3. Anotherapproach proposes a model based on recurrent neural networks in association with Boltzmann machines4, aims topredict the propagation and the dissipation of this phenomenon. Swarms intelligence paradigm specially the ant-typehas largely contributed to solving complex problems. Ant colonies algorithms are inspired by the behavior of ants orother species forming a super-organism, which are a family of Doriogo optimization meta-heuristic5. Several workshave been proposed in this research topic, see Ando et al6, Kurihara et al7, Jiang et al8, Kurihara9. In this paper, wepropose a system based on the semi-macroscopic traffic modeling. The main objective of this work is to develop aflexible simulation of traffic on an urban transport network and to estimate and to predict travel times. These traveltimes are the costs considered for the road sections which are the input of a robust algorithm to find the optimum pathaccording travel time criteria from an origin to a destination of the network. The structure of this work is organized asfollows. Section 2 provides a brief semi-macroscopic simulation model that can work together or substitute for realtimedata collector. In Section 3, we present the research approaches in the field of the robust shortest paths in dynamicgraphs. The system architecture and related software components are explained in Section 4. The conclusion of thiswork is given in section 5.