30.11.2014 Views

Model Predictive Control System Design and Implementation Using ...

Model Predictive Control System Design and Implementation Using ...

Model Predictive Control System Design and Implementation Using ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

8.4 CMPC with Asymptotic Stability 277<br />

for simplicity of the solution, we have assumed that the weight matrix R is a<br />

diagonal matrix.<br />

Defining the data matrices<br />

Ω =<br />

Ψ =<br />

∫ Tp<br />

0<br />

∫ Tp<br />

the quadratic cost function (8.10) becomes<br />

0<br />

φ(τ)Qφ(τ) T dτ + R L (8.11)<br />

φ(τ)Qe Aατ dτ, (8.12)<br />

J = η T Ωη +2η T Ψx(t i )+constant. (8.13)<br />

The optimal solution that minimizes the above quadratic cost function is<br />

η = −Ω −1 Ψx(t i ). (8.14)<br />

Upon obtaining η, the exponentially weighted derivative of the control signal<br />

˙u α (τ) is constructed through<br />

˙u α (τ) = [ L 1 (τ) T L 2 (τ) T ... L m (τ) T ] η. (8.15)<br />

From the receding horizon control, the optimal solution for the actual ˙u(0) is<br />

˙u(0) = ˙u α (0) = [ L 1 (0) T L 2 (0) T ... L m (0) T ] η. (8.16)<br />

Because the optimization is performed on the transformed variables x α (.) <strong>and</strong><br />

˙u α (.), when constraints are introduced, all the original constraints are required<br />

to be transformed from the variables x(.) <strong>and</strong> ˙u(.) tox α (.) <strong>and</strong> ˙u α (.). Constrained<br />

control will be discussed further in the later sections of the chapter.<br />

8.4 CMPC with Asymptotic Stability<br />

This section establishes equivalent results with LQR when exponential weighting<br />

is used. The results are investigated through two different cost functions,<br />

<strong>and</strong> we then establish that the optimal control results are identical. The results<br />

are summarized in the theorem as follows.<br />

Case A<br />

Suppose that the optimal control ˙u 1 (τ) is obtained by minimizing cost function<br />

J 1 with Q ≥ 0, <strong>and</strong> R>0<br />

J 1 =<br />

∫ ∞<br />

0<br />

[<br />

x(ti + τ | t i ) T Qx(t i + τ | t i )+ ˙u(τ) T R ˙u(τ) ] dτ, (8.17)

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

Saved successfully!

Ooh no, something went wrong!