The purpose of this thesis, conducted in collaboration with the Phoenix-ISI company is to develop a model for the traffic light control problem in urban network. The objective is to determine a signal timing plan which minimizes the total delay to which vehicles are subjected during their passage through the network. The research presented here is divided into two main parts. The first part is to develop a traffic flow model based on network traffic data (inflow, flow directional distribution) to represent accurately the flow of different traffic flows in the studied network. The second part consists in solving a MIP (Mixed Integer Problem) issued from the simulation model in the first part. We propose an algorithm based on the Bender's decomposition method. Basically, such an algorithm should solve a huge linear programming problem to generate Bender's cuts in each iteration. We prove that, for our MIP we can do it much faster by an iterative method (“inverse simulation” procedure) which exploits a special structure of the matrix of constraints. We present finally a series of numerical experiments proving the efficiency of the proposed algorithm.