Generalized Predictive Control (GPC) describes an algorithm for the control of dynamic systems in which a control input is generated which minimizes a quadratic cost function consisting of a weighted sum of errors between desired and predicted future system output and future predicted control increments. The output predictions are obtained from an internal model of the plant dynamics. Self-tuning GPC refers to an implementation of the GPC algorithm in which the parameters of the internal model(s) are estimated on-line and the predictive control law tuned to the parameters so identified. The self-tuning GPC algorithm is applied to a problem of rotorcraft longitudinal/vertical terrain-following flight. The ability of the algorithm to tune to the initial vehicle parameters and to successfully adapt to a stability augmentation failure is demonstrated. Flight path performance is compared to a conventional, classically designed flight path control system.
[1]
Robin De Keyser,et al.
A comparative study of self-adaptive long-range predictive control methods
,
1985,
Autom..
[2]
Raman K. Mehra,et al.
Model algorithmic control (MAC); basic theoretical properties
,
1982,
Autom..
[3]
David W. Clarke,et al.
Generalized predictive control - Part I. The basic algorithm
,
1987,
Autom..
[4]
C. Thornton,et al.
Filtering and error analysis via the UDU^{T} covariance factorization
,
1978
.