Search

Data-based Predictive Control is an emerging control method that stems from Model Predictive Control (MPC). MPC computes current control action based on a prediction of the system output a number of time steps into the future and is generally derived from a known model of the system. Data-based predictive control has the advantage of deriving predictive models and controller gains from input-output data. Thus, a controller can be designed from the outputs of complex simulation code or a physical system where no explicit model exists. If the output data happens to be corrupted by periodic disturbances, the designed controller will also have the built-in ability to reject these disturbances without the need to know them. When data-based predictive control is implemented online, it becomes a version of adaptive control. The characteristics of adaptive data-based predictive control are particularly appropriate for the control of nonlinear and time-varying systems, such as Short Take-off and Landing (STOL) aircraft. STOL is a capability of interest to NASA because conceptual Cruise Efficient Short Take-off and Landing (CESTOL) transport aircraft offer the ability to reduce congestion in the terminal area by utilizing existing shorter runways at airports, as well as to lower community noise by flying steep approach and climb-out patterns that reduce the noise footprint of the aircraft. In this study, adaptive data-based predictive control is implemented as an integrated flight-propulsion controller for the outer-loop control of a CESTOL-type aircraft. Results show that the controller successfully tracks velocity while attempting to maintain a constant flight path angle, using longitudinal command, thrust and flap setting as the control inputs.

This report describes a nonlinear mapping technique where the unknown static or dynamic system is approximated by a sum of dimensionally increasing functions (one-dimensional curves, two-dimensional surfaces, etc.). These lower dimensional functions are synthesized from a set of multi-resolution basis functions, where the resolutions specify the level of details at which the nonlinear system is approximated. The basis functions also cause the parameter estimation step to become linear. This feature is taken advantage of to derive a systematic procedure to determine and eliminate basis functions that are less significant for the particular system under identification. The number of unknown parameters that must be estimated is thus reduced and compact models obtained. The lower dimensional functions (identified curves and surfaces) permit a kind of "visualization" into the complexity of the nonlinearity itself.

This paper presents a recursive algorithm for a deadbeat predictive controller design. The method combines together the concepts of system identification and deadbeat controller designs. It starts with the multi-step output prediction equation and derives the control force in terms of past input and output time histories. The formulation thus derived satisfies simultaneously system identification and deadbeat controller design requirements. As soon as the coefficient matrices are identified satisfying the output prediction equation, no further work is required to compute the deadbeat control gain matrices. The method can be implemented recursively just as any typical recursive system identification techniques.

This report describes a nonlinear mapping technique where the unknown static or dynamic system is approximated by a sum of dimensionally increasing functions (one-dimensional curves, two-dimensional surfaces, etc.). These lower dimensional functions are synthesized from a set of multi-resolution basis functions, where the resolutions specify the level of details at which the nonlinear system is approximated. The basis functions also cause the parameter estimation step to become linear. This feature is taken advantage of to derive a systematic procedure to determine and eliminate basis functions that are less significant for the particular system under identification. The number of unknown parameters that must be estimated is thus reduced and compact models obtained. The lower dimensional functions (identified curves and surfaces) permit a kind of "visualization" into the complexity of the nonlinearity itself.

Presented here is an algorithm to compute the Markov parameters of a backward-time observer for a backward-time model from experimental input and output data. The backward-time observer Markov parameters are decomposed to obtain the backward-time system Markov parameters (backward-time pulse response samples) for the backward-time system identification. The identified backward-time system Markov parameters are used in the Eigensystem Realization Algorithm to identify a backward-time state-space model, which can be easily converted to the usual forward-time representation. If one reverses time in the model to be identified, what were damped true system modes become modes with negative damping, growing as the reversed time increases. On the other hand, the noise modes in the identification still maintain the property that they are stable. The shift from positive damping to negative damping of the true system modes allows one to distinguish these modes from noise modes. Experimental results are given to illustrate when and to what extent this concept works.

A formulation is presented for the identification of a linear multivariable system from single or multiple sets of input-output data. The system input-output relationship is expressed in terms of an observer, which is made asymptotically stable by an embedded eigenvalue assignment procedure. The prescribed eigenvalues for the observer may be real, complex, mixed real and complex, or zero. In this formulation, the Markov parameters of the observer are identified from input-output data. The Markov parameters of the actual system are then recovered from those of the observer and used to obtain a state space model of the system by standard realization techniques. The basic mathematical formulation is derived, and extensive numerical examples using simulated noise-free data are presented to illustrate the proposed method.