Experimental Design and Identifiability for Nonlinear Systems

摘要

Experiment design for the identification of nonlinear systems based on the criterion that the accuracy of the parameter estimates should be maximized is discussed. In open-loop operation the optimum input sequence for a completely unknown nonlinear system is derived using results from information theory. The optimum input is an independent sequence with a Gaussian distribution for a power constraint on the input and an independent sequence with a uniform distribution for an amplitude constraint on the input. The results differ from the optimum input sequences for linear systems and it is shown that identifiability problems can arise if optimum sequences for linear systems are used as inputs for nonlinear systems. Practical guidelines for the design of experiments in closed-loop operation are given.