Continuous-time system model reduction by identification via Markov parameters

An attractive and novel algorithm for Markov parameter estimation in continuous-time single-input-single-output systems from input-output data is presented. The attractive features are: (a) the model is general; (b) the estimation is linear and can be made free from bias which may arise due to truncation effects of the Markov series; (c) the Markov parameters are useful in ascertaining system order; and (d) model reduction via Markov parameters is possible in the light of the existing techniques. Certain problems typically associated with Markov parameter estimation have been solved by introducing features of data band compression, Markov-Poisson parameterization, pole placement based Markov sequence finitization, etc. leading to an efficient algorithm which provides results where techniques available now may be used for model reduction