One of the major tasks performed in the MPD experiment is to classify events into centrality classes. The centrality procedure allows us to estimate the initial geometry in heavy ion collisions using the relation between the observables and the impact parameter. In this work, we present results on the centrality determination using the Bayesian approach. The performance of the proposed method has been tested on simulated data for Xe+W and Xe+Xe collisions at Ekin =2.5 AGeV with MPD-FXT