Basic Research on Adaptive Model Algorithmic Control
This report has yet to be scanned by Contrails staff
Report Number: AFWAL-TR-85-3113
Author(s): Larimore, Wallace E., Mahmood, Shahjahan
Corporate Author(s): Scientific Systems, Inc.
Laboratory: Flight Dynamics Laboratory
Date of Publication: 1985-12
Pages: 204
Contract: F33615-82-C-3600
DoD Project: 2403
DoD Task: 2403N3
Identifier: ADA168016
Abstract:
The Model Algorithmic Control (MAC) method is investigated in terms of robustness and adaption to unknown or changing plants. The adaption method used is Canonical Variate Analysis (CVA) system identification. CVA is shown to provide system identification accuracy comparable to maximum likelihood and to provide an optimal selection of instrumental variables. Computationally CVA is a noniterative procedure that gives a numerically and statistically well conditioned solution to the system identification problem. A one-step-ahead MAC is explained using the classical root locus technique. Conditions are developed for robustness of the controller to perturbations in the plant due to error in plant identification. Selection of an optimal sampling rate is based upon the controlability and observability matrices. Simulations illustrating the above theory are presented using a Multi-Input Multi-Output (MIMO) missile aerodynamic model.
Provenance: University of Colorado Colorado Springs, Kraemer Family Library
Author(s): Larimore, Wallace E., Mahmood, Shahjahan
Corporate Author(s): Scientific Systems, Inc.
Laboratory: Flight Dynamics Laboratory
Date of Publication: 1985-12
Pages: 204
Contract: F33615-82-C-3600
DoD Project: 2403
DoD Task: 2403N3
Identifier: ADA168016
Abstract:
The Model Algorithmic Control (MAC) method is investigated in terms of robustness and adaption to unknown or changing plants. The adaption method used is Canonical Variate Analysis (CVA) system identification. CVA is shown to provide system identification accuracy comparable to maximum likelihood and to provide an optimal selection of instrumental variables. Computationally CVA is a noniterative procedure that gives a numerically and statistically well conditioned solution to the system identification problem. A one-step-ahead MAC is explained using the classical root locus technique. Conditions are developed for robustness of the controller to perturbations in the plant due to error in plant identification. Selection of an optimal sampling rate is based upon the controlability and observability matrices. Simulations illustrating the above theory are presented using a Multi-Input Multi-Output (MIMO) missile aerodynamic model.
Provenance: University of Colorado Colorado Springs, Kraemer Family Library