System Identification
Lennart Ljung A · Wiley Encyclopedia of Electrical and Electronics Engineering · 1998
Abstract System identification is the term used in the automatic control field for estimating dynamical models of systems, based on measurements of the system's input and output signals. The models are typically difference or differential equations relating the measured signals, and possibly some auxiliary states. The models can be constructed from the signals only (black‐box models) or from the signals in combination with prior physical insights into mechanisms in the system (gray‐box models). The basic technique is to compare the model's (predicted) output with the system's measured output for the same input signal. Any free parameters in the model are then adjusted to make the fit between the outputs as good as possible. This general description includes advanced statistical procedures such as the maximum likelihood method.
Cited by 3,778 works. doi:10.1002/047134608x.w1046
Authors
References
- Classification and Regression Trees.
- A new look at the statistical model identification
- Generalized Linear Models
- System Identification: Theory for the User
- Approximation by superpositions of a sigmoidal function
- Nearest neighbor pattern classification
- System identification—A survey
- Identification and control of dynamical systems using neural networks
- Modeling by shortest data description
- Partial least-squares regression: a tutorial