Identification and Control of Random-parameter, Discrete Systems
Description:... In the time-domain, state-space description of linear dynamic systems, the system is represented by the imput and transition matrices, which determine how the control input and present state affect the next state. If one or more elements of the matrices are unknown or randomly varying parameters, the system is a random-parameter system. The present work is a study of discrete systems with random parameters that are correlated in time. The correlation of the random parameters permits parameter identification, or realtime learning of the random parameters, by processing the system output sequence. The parameter correlation renders the system description nonlinear in the general case. The identification technique is developed from a linearized model of the system and represents an extension of the filtering procedure of Kalman. The model of the system is then adjusted with each learning calculation. Conditions are given for which the adaptive learning procedure is optimal, with examples and interpretation. (Author).
Show description