13.07.2015 Views

Assessment and Future Directions of Nonlinear Model Predictive ...

Assessment and Future Directions of Nonlinear Model Predictive ...

Assessment and Future Directions of Nonlinear Model Predictive ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

246 L. Magni <strong>and</strong> R. Scattolini<strong>of</strong> tighter <strong>and</strong> tighter state, control <strong>and</strong> terminal constraints, so leading to veryconservative solutions or even to unfeasible problems. Algorithms with thesecharacteristics have been described in [39] for continuous-time <strong>and</strong> in [24] fordiscrete time systems. The technique presented in [24] is now briefly summarized.To this aim, the following assumption must be introduced to allow for theanalysis <strong>of</strong> the (worst-case) effects <strong>of</strong> the disturbance.Assumption 6.1. The uncertain term in (14) is bounded by γ, that is |g(·, ·, ·)| ≤γ for any x <strong>and</strong> u satisfying (4) <strong>and</strong> w ∈M W .In order to guarantee that at any future time instant in the prediction horizonthe disturbance does not cause the state constraints violation, first introduce thefollowing definition.Definition 7 (Pontryagin difference). Let A, B ⊂ R n , be two sets, then thePontryagin difference set is defined as A ∼ B = {x ∈ R n |x + y ∈ A, ∀y ∈ B} .Consider now the following sets X j = X ∼ Bγ j where Bj γ is defined as{Bγ j = z ∈ R n : |z| ≤ Lj f − 1}L f − 1 γ .Definition 8 (NRFHOCP). Consider a stabilizing auxiliary control law κ f (·)<strong>and</strong> an associated output admissible set X f . Then, given the positive integer N,the stage cost l(·, ·) <strong>and</strong> the terminal penalty V f (·), the Nominal Robust FiniteHorizon Optimal Control Problem (NRFHOCP) consists in minimizing, withrespect to u t,t+N−1 ,subject to:J(¯x, u t,t+N−1 ,N)=t+N−1∑k=tl(x(k),u(k)) + V f (x(t + N))(i) the state dynamics (3) with x(t) =¯x;(ii) the constraints u(k) ∈ U <strong>and</strong> x(k) ∈ X k−t+1 , k ∈ [t, t + N − 1], whereX k−t+1 are given in Definition 7;(iii) the terminal state constraint x(t + N) ∈ X f .From the solution <strong>of</strong> the NRFHOCP, the Receding Horizon control lawu = κ MPC (x) (18)is again obtained as in (7) <strong>and</strong> (8). Concerning the stability properties <strong>of</strong> theclosed-loop system, the following hypothesis substitutes Assumption 3.2.Assumption 6.2. Let κ f (·), V f (·), X f such that1. Φ f := {x ∈ R n : V f (x) ≤ α} ⊆X, Φ f closed, 0 ∈ Φ f ,αpositive constant2. κ f (x) ∈ U, ∀x ∈ Φ f

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!