Explore chapters and articles related to this topic
A design technique for fast sampled-data nonlinear model predictive control with convergence and stability results
Published in International Journal of Control, 2020
Andreas Steinboeck, Martin Guay, Andreas Kugi
We consider a piecewise constant input parameterisation of the form:
This means that κ can be designed using a zero-order hold. We use the linear feedback law
with the constant gain K. Here, xk = x(Tk) and k is the index of the time grid Tk. Inspired by the approach reported in Chen and Allgöwer (1998b) for continuous-time NMPC, we suggest a different systematic sampled-data control design based on the locally linearised system. Consider the Jacobian linearisation:
of the system (1) with Ac = ∂f/∂x|x = 0, u = 0, and Bc = ∂f/∂u|x = 0, u = 0. Its discrete-time counterpart found by zero-order hold reads as
with A = exp (κTAc) and B = ∫κT0exp (tAc)dtBc and is assumed to be stabilisable in . Using the steady-state LQR design method with the positive definite objective function
and the solution P of the corresponding algebraic Riccati equation, we obtain
Based on the correspondence
with
(Franklin, Powell, & Workman, 1997; Van Loan, 1978), the equivalent objective function of the sampled-data LQR design problem reads as
In Equation (17), ln ( · ) returns the logarithm of a matrix. This operation is only defined for invertible matrices.