Identification by a combined smoothing nonlinear programming algorithm

✍ Scribed by A.B. Cox; A.E. Bryson Jr.

• Publisher: Elsevier Science
• Year: 1980
• Tongue: English
• Weight: 440 KB
• Volume: 16
• Category: Article
• ISSN: 0005-1098
• DOI: 10.1016/0005-1098(80)90010-2

✦ Synopsis

An offline algorithm is developed for identification of parameters of linear, stationary, discrete, dynamic systems with known control inputs and subjected to process and measurement noise with known statistics. Results of the algorithm include estimates of the parameters and smoothed estimates of the state and process noise sequences. The problem is stated as the minimization of a quadratic performance index. This minimization problem is then converted to a nonlinear programming problem for determining the optimum parameter estimates. The new algorithm is shown to be cost competitive with the currently popular filtering-sensitivity function method. A third order example with simulated data is presented for comparison.