Search

In this work, recursive identification algorithms are developed for Hammerstein systems under the conditions considerably weaker than those in the existing literature. For example, orders of linear subsystems may be unknown and no specific conditions are imposed on their moving average part. The recursive algorithms for estimating both linear and nonlinear parts are based on stochastic approximation and kernel functions. Almost sure convergence and strong convergence rates are derived for all estimates. In addition, the asymptotic normality of the estimates for the nonlinear part is also established. The nonlinearity considered in the paper is more general than those discussed in the previous papers. A numerical example verifies the theoretical analysis with simulation results. [ABSTRACT FROM PUBLISHER]

System identification extracts information from a system's operational data to derive a representative model for the system so that a decision can be made with desired accuracy and reliability. When resources are limited, especially for networked systems sharing data and communication power and bandwidth, identification must consider complexity as a critical limitation. Focusing on optimal resource allocation under a given reliability requirement, this paper studies identification complexity and its relations to decision making. Dynamic resource assignments are investigated. Algorithms are developed and their convergence properties are established, including strong convergence, almost sure convergence rate, and asymptotic normality. By a suitable design of resource updating step sizes, the algorithms are shown to achieve the CR lower bound asymptotically, and hence are asymptotically efficient. Illustrative examples demonstrate significant advantages of our real-time and individualized resource allocation methodologies over population-based worst-case strategies. [ABSTRACT FROM PUBLISHER]

This technical note presents new results on probabilistic characterization of identification errors in their relationships to data sizes and accuracy requirements. Employing the moderate deviations principle, this technical note shows that if the identification accuracy progressively increases with a suitable rate, the probability of an estimate going outside the precision bounds decays exponentially with the data size. The precise rate of the decaying probability is obtained. System identification under regular, binary, and quantized observations are considered. Impact of unmodeled dynamics is also investigated. [ABSTRACT FROM AUTHOR]