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

Create successful ePaper yourself

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

Towards the Design <strong>of</strong> Parametric <strong>Model</strong> <strong>Predictive</strong> Controllers 203<strong>and</strong> ψ g (x(t f )) = −x(t f ) + 1. The last expression represents that at final timethe beam should be positioned at a point where x(t f ) ≥ 1. We also assume thatt 0 = 0. Although, the problem has already been solved for a fixed initial pointx 0 =[0 0] T (as for example in [3]), here the optimal solution is derived for thefirst time as a function <strong>of</strong> the initial point coordinates.The problem is first dealt in discrete-time as described in Section 2. Thecontinuous-time system is turn into a discrete-time system assuming a samplingtime ∆t such that [t 0 ,t f ] is divided in three equally spaced time intervals <strong>of</strong> ∆ti.e. [t 0 ,t f ]=3∆t. The discrete-time problemmin ∆tγ kx k+1 = x k +(2gy k ) −1/2 cosγ k ∆t , k =0, 1, 2y k+1 = y k +(2gy k ) −1/2 sinγ k ∆t , k =0, 1, 2y k − 0.5x k − 1 ≤ 0 ,k=0, 1, 2, 3x 3 ≥ 1is then solved by transforming the above problem in (5) <strong>and</strong> solving the mp-LP problem to acquire the PWA solution. The continuous-time is solved nextfollowing Algorithm 3.1. The results for both the discrete-time case <strong>and</strong> thecontinuous-time case together with a simulation for x 0 =[0 0] T are shownin Figure 1 <strong>and</strong> 2. The straight line in both diagrams represents the boundary<strong>of</strong> the linear constraint y − 0.5 − 1 ≤ 0. There are three control laws for thecontinuous-time case, depending in which region the initial state is. The controllaw in the unconstrained region (Figure 2.) is obtained by solving the followingsystem <strong>of</strong> algebraic equalities with respect to c 1 <strong>and</strong> γ0=c 2 1 − arccos(c 1√ y) − xc2 √ √1 + c 1 y sin arccos c1 yγ = −1/2(2g) 1/2 √c 1 t + arccos c 1 yIn the constrained region the control law is obtained as following. First, thefollowing system <strong>of</strong> equalities is solvedx(τ ′′ ) − (2g) 1 22c 1τ ′′ + 12c 2 1sin (2g) 1 2 c1 (t f − τ ′′ )=1− (2g) 1 22c 1t f = arccos c √ 11 y +0.5(2g) 2 c 1 τ ′′,x(τ ′′ )=0.1989gτ ′′2 +(2g 1 √0.5(2g) 1 2 )0.896 y0 τ ′′ + x 02 c 1y(τ ′′ )=(0.222(2g) 1 2 τ ′′ + √ y 0) 2, 0.46 = 0.5(2g)12 c1 (t f − τ ′′ )which is a system <strong>of</strong> five equations with five unknowns t f ,τ ′′ ,c 1 ,x(τ ′′ ),y(τ ′′ ).Then, the control to be applied is given asIf t ≤ τ ′′ then γ =0.46 = arctan(0.5) Else If t ≥ τ ′′ then γ =0.5(2g) 1 2 c 1 (t f − t)As it can be observed from Fig. 1 the approximating, discrete-time, PWAsolution is not the optimal one comparing to the optimal solution as it is givent f

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

Saved successfully!

Ooh no, something went wrong!