• Markov chains
• Inverse problems
• Identification
• Differential Evolution
• DEMC
• AUV
• MCMC
• Monte Carlo
• Parameters

In automation engineering, the need of mathematical models for the dynamic systems to be controlled is ubiquitous; mathematical models are embedded in the control strategy, in state observers, in fault detection systems and trajectory planners; it is thus important to have numerical methods which would provide reliable models, because the model degree of accuracy affects the whole automation system performances. This thesis deals with parametric identification. The aim of the work is to define and test an identification procedure to provide the model parameters and quantify the uncertainty associated with them, given a model structure and a set of measurements. The general case of nonlinear MIMO systems is considered. In order to achieve the goal, the identification issue is addressed as a Bayesian inference problem and it is solved by employing a Markov Chain Monte Carlo (MCMC) sampling algorithm conveniently adapted and implemented for this purpose. In this thesis the method is tested for the identification of an autonomous underwater vehicle (AUV) dynamic model. Data for the identification experiment are simulated from the model of the vehicle V-Fides produced by Whitehead Sistemi Subacquei (WASS, Livorno).