Modeling traffic flow dynamics using cellular automata allow us to run large network traffic simulations with only comparatively low computational efforts. The objective of this work is to search global optimization techniques in order to optimize the parameters asociated to a two lanes emergent microscopic traffic model based on a cellular automata. The variables taken into account are: the car density ratio between the two lanes, the braking probabilities of each lane, and the mean and variance of the velocity distributions of each lane. This type of problem is known in the literature as a nonlinear global optimization problem with restrictions. Thus, a robust optimization algorithm able to calculate the global optimums in a reasonable number of function evaluations and to guarantee an adequate answer for any arbitrary set of flow data, is required. The proposed technique is a multistart type algorithm that combine a stochastic exploration of the domain and a heuristic calculation of a descent direction, in order to avoid stopping the algorithm at a local optimum. The efficiency of this algorithm is determined by the following criteria: (i) The number of the objective function evaluations, (ii) the execution time, and (iii) the quality of the final result. Among the advantages of the algorithm are the simultaneous localization of all possibles global optimums of the objective function and the low number of function evaluations with compared to other method reported in the literature.