ABSTRACT
System identification is the process of determining a dynamical model for an unknown plant that can be used for feedback control purposes. The state estimation problem involves determining the unknown internal state of a dynamical plant. System identification provides one technique for estimating the states. The area of system identification has received much attention over the past two decades. It is now a mature field, and many powerful methods are at the disposal of control engineers. On-line system identification methods used to date are mostly based on recursive methods such as least squares (Ren and Kumar 1994). However, most of these techniques are for models that are linear in the parameters. In order to relax the linearity in the parameters assumption, NN are being widely employed for system identification since these networks learn complex mappings from a set of examples. Due to their approximation properties as well as the inherent adaptation features of these networks, NN present a potentially appealing alternative to modeling of nonlinear systems. Furthermore, from a practical perspective, the massive parallelism and fast adaptability of neural network implementations provide additional incentives for further investigation.
