Parameter estimation in a trip distribution model by random perturbation of a descent method

We consider the problem of the estimation of some parameters involved in a trip distribution model issued from Transportation Planning. The estimators of the maximum likelihood of the model are the global minima of a non-convex function. The numerical method must prevent convergence to local minima and we apply a new algorithm of global optimization involving random perturbations of the gradient method. Numerical experiments involving real data show that the method is efficient for calculation.