Systems Classification and Control - Monte Carlo Approach

摘要

Usual approach in system identification and control is to represent all conditional probability density functions (c.p.d.f.) as a functions over the state space. Latest trend in "computational statistic" is to represent c.p.d.f. as a set of random samples. Using large number of samples an alternative representation of c.p.d.f. is obtained. In [1], Bootstrap filter is described for updating these samples for discrete time system. From this samples the estimates of moments like mean and covariances can be obtained. In this way any system nonlinearity and nonnormality of the noise can be handled. So we are able using this approach to estimate the correct structure, states and parameters of the system - all based on real data obtained from the system. In this article it is discussed how to use the results of system classification for the synthesis of optimal control.