Optimizing the Likelihood with Sequential Monte–Carlo methods

In this paper, a Sequential Monte–Carlo (SMC) method is studied to deal with nonlinear multivariate optimization problems arising from Maximum Likelihood (ML) estimation approaches. In this context, gradient–like methods are not efficient being the computational cost burdensome. Moreover, SMC provide an appealing way of introducing prior information in the estimation of parameters in general state–space models. Relying on SMC methods, the optimization algorithm is completely exposed and its pseudocode delivered for completely general ML problems. For the sake of completeness, the algorithm has been adapted and simulated in a GNSS application, where the use of prior information is of great interest.