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.

330 R. Gabasov, F.M. Kirillova, <strong>and</strong> N.M. DmitrukThe first condition gives admissibility <strong>of</strong> the control strategy while the secondone ensures its optimality.It can be shown that a control u(t|τ,y ∗ τ (·)), t ∈ [τ,t∗ ], is admissible if togetherwith u ∗ (t), t ∈ [t ∗ ,τ[, it steers the nominal systemat the instant t ∗ on the terminal setẋ 0 (t) =A(t)x 0 (t)+B(t)u(t), x 0 (t ∗ )=x 0X ∗ 0 (τ) ={x ∈ R n : g ∗i − ˆβ i (τ) ≤ h ′ ix ≤ g ∗ i − ˆα i (τ), i = 1,m}.Here the tightened terminal constraints are determined by the estimatesˆα i (τ) =maxh ′ ix, x ∈ ˆX ∗ (τ); ˆβi (τ) =minh ′ ix, x ∈ ˆX ∗ (τ); i = 1,m, (5)<strong>of</strong> the a posteriori distribution ˆX ∗ (τ) = ˆX ∗ (τ,y ∗ τ (·)) <strong>of</strong> terminal states x ∗(t ∗ )<strong>of</strong>the uncertain systemẋ ∗ (t) =A(t)x ∗ (t)+w(t), x ∗ (t ∗ )=Gz,(z,v) ∈ ˆΓ (τ).The (current) optimal open-loop control u 0 (t|τ,y ∗ τ (·)), t ∈ [τ,t∗ ], therefore, isa solution to the determined optimal control problemc ′ x(t ∗ ) → max, ẋ(t) =A(t)x(t)+B(t)u(t), x(τ) =x 0 (τ), (6)x(t ∗ ) ∈ X0 ∗(τ), u(t) ∈ U, t ∈ [τ,t∗ ].Extremal problems (5) are called the optimal observation problems accompanyingthe optimal control problem under uncertainty. Problem (6) is called theaccompanying optimal control problem.The optimal open-loop control is fed into the system until the next measurementis processed: u ∗ (t) =u 0 (t|τ,y ∗ τ(·)), t ∈ [τ + s(τ),τ + h + s(τ + h)[. Heres(τ)

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

Saved successfully!

Ooh no, something went wrong!