Revision as of 13:46, 2 July 2012 by Mboutin (Talk | contribs)

ECE Ph.D. Qualifying Exam in "Automatic Control" (AC)

Question 3, August 2011, Part 4

Part 1,2,3,4,5

 $ \color{blue}\text{4. } \left( \text{20 pts} \right) \text{ Consider the following model of a discrete-time system, } $

                    $ x\left ( k+1 \right )=2x\left ( k \right )+u\left ( k \right ), x\left ( 0 \right )=0, 0\leq k\leq 2 $

$ \color{blue}\text{Use the Lagrange multiplier approach to calculate the optimal control sequence} $

                   $ \left \{ u\left ( 0 \right ),u\left ( 1 \right ), u\left ( 2 \right ) \right \} $

$ \color{blue}\text{that transfers the initial state } x\left( 0 \right) \text{ to } x\left( 3 \right)=7 \text{ while minimizing the performance index} $
                   $ J=\frac{1}{2}\sum\limits_{k=0}^2 u\left ( k \right )^{2} $


Discussion:

This problem need a bit interpretation before getting the standard form to apply KKT condition.

Theorem about KKT, SONC, SOSC please see "Part 5".


$ \color{blue}\text{Solution 1:} $

$ \left.\begin{matrix} x\left ( 1 \right )=2x\left ( 0 \right )+u\left ( 0\right )\\ x\left ( 2 \right )=2x\left ( 1 \right )+u\left ( 1\right )\\ x\left ( 3 \right )=2x\left ( 2 \right )+u\left ( 2\right )\\ x\left ( 0 \right )=0 \end{matrix}\right\} \Rightarrow \left\{\begin{matrix} x\left ( 1 \right )=u\left ( 0 \right )\\ x\left ( 2 \right )=2u\left ( 0 \right )+u\left ( 1\right )\\ x\left ( 3 \right )=4u\left ( 0 \right )+2u\left ( 1\right )+u\left ( 2 \right )=7 \end{matrix}\right. $

$ \text{The problem is equivalent to minimize } J=\frac{1}{2}\sum\limits_{k=0}^2 u\left ( k \right )^{2} $

                                                                  $ \text{subject to } 4u \left(0 \right)+2u \left(1 \right)+u\left(2 \right)=7 $

$ \text{Let } h(u )=4u \left(0 \right)+2u \left(1 \right)+u\left(2 \right)-7 $

$ \text{FONC: } \left\{\begin{matrix} l\left(u,\lambda \right)=\nabla J\left ( u \right )+\lambda\nabla h\left( u \right)=\begin{pmatrix} u\left ( 0 \right )\\ u\left ( 1 \right )\\ u\left ( 2 \right ) \end{pmatrix}+\lambda\begin{pmatrix} 4\\ 2\\ 1 \end{pmatrix} =0\\ h(u )=4u \left(0 \right)+2u \left(1 \right)+u\left(2 \right)-7=0 \end{matrix}\right. \Rightarrow \left\{\begin{matrix} u\left(0 \right)=\frac{4}{3}\\ u\left(1 \right)=\frac{2}{3}\\ u\left(2 \right)=\frac{1}{3}\\ \lambda=-\frac{1}{3} \end{matrix}\right. $

$ \text{SOSC: } L\left( u,\lambda \right)=\nabla l\left( u,\lambda \right)=\begin{bmatrix} 1 & 0 & 0\\ 0 & 1 & 0\\ 0 & 0 & 1 \end{bmatrix}>0 $

The sequence  $ u\left( 0 \right)=\frac{4}{3},u\left( 1 \right)=\frac{2}{3},u\left( 2 \right)=\frac{1}{3} $  satisfies SOSC. It is the optimal sequence.

$ \color{red} \text{This solution did not specifically state the complete SOSC.} $


$ \color{blue}\text{Solution 2:} $

$ x\left ( 1 \right )=u\left ( 0 \right ) $

$ x\left ( 2 \right )=2u\left ( 0 \right )+u\left ( 1 \right ) $

$ x\left ( 3 \right )=4u\left ( 0 \right )+2u\left ( 1\right )+u\left ( 2 \right )=7 $

$ \text{The problem transfer to min } J\left ( u \right )=\frac{1}{2} u \left ( 0 \right )^{2}+\frac{1}{2} u \left ( 1 \right )^{2}+\frac{1}{2} u \left ( 2 \right )^{2} $

                                             $ \text{subject to } h(u )=4u \left(0 \right)+2u \left(1 \right)+u\left(2 \right)-7=0 $

$ \text{Apply KKT condition: } Dl\left( u ,\lambda \right)=DJ\left(u \right)+\lambda Dh\left(u \right)=\left[ u\left(0 \right)+4\lambda,u\left(1 \right)+2\lambda,u\left(2 \right)+\lambda \right]=0 $

                                                      $ \left\{\begin{matrix} u\left(0 \right)+4\lambda=0\\ u\left(1 \right)+2\lambda=0\\ u\left(2 \right)+\lambda=0\\ 4u\left(0 \right)+2u\left(1 \right)+u\left(2 \right)-7=0 \end{matrix}\right. \Rightarrow \left\{\begin{matrix} u\left(0 \right)=\frac{4}{3}\\ u\left(1 \right)=\frac{2}{3}\\ u\left(2 \right)=\frac{1}{3}\\ \lambda=-\frac{1}{3} \end{matrix}\right. $

$ \text{Check SOSC: } L\left( u,\lambda \right)=D^{2}l\left( u,\lambda \right)=\begin{bmatrix} 1 & 0 & 0\\ 0 & 1 & 0\\ 0 & 0 & 1 \end{bmatrix}>0 $

$ \therefore \text{For all y, } y^{T}Ly\geq 0 $

$ \therefore \text{sequence } \left\{ \frac{4}{3},\frac{2}{3},\frac{1}{3} \right\} \text{ satisfy SOSC is a strict minimizer of the problem.} $



$ \color{blue} \text{Related Problem:} $

extremize 9x1 + 3x2

$ \text{subject to } \frac{1}{2}x_{1}^{2}-x_{2}=0 $

$ \color{blue} \text{(i) Find point(s) that satisfy the FONC} $

$ \color{blue} \text{(ii) Apply the SOSC to determine the nature of the critical point(s) from the previous part} $

$ \color{blue} \text{Solution: } $

        $ l\left( x,\lambda \right)=9x_{1}+3x_{2}+\lambda\left( \frac{1}{2}x_{1}^{2}-x_{2} \right) $

Applying the FONC:

$ D_{x} l\left( x,\lambda \right) = \begin{bmatrix} 9+\lambda x_{1} & 3-\lambda \end{bmatrix} =\begin{bmatrix} 0 & 0 \end{bmatrix} $

              $ \Rightarrow \lambda^{*}=3 \text{ and } x^{*}_{1}=-3 $

           $ \Rightarrow x_{2}^{*}=0.5x_{1}^{*2}=9/2 $

$ \therefore \text{the point that satisfies the FONC is } x^{*}= \begin{bmatrix} -3\\ 9/2 \end{bmatrix} $

Check SOSC:

$ D_{x} L\left( x^{*},\lambda^{*} \right) = F\left( x^{*} \right)+ \lambda^{*} H\left( x^{*} \right) = \begin{bmatrix} 3 & 0\\ 0 & 0 \end{bmatrix} $

              $ T\left( x^{* }\right)= \left \{ y: \begin{bmatrix} -3 & -1 \end{bmatrix}y=0 \right \} = \left \{ y =\begin{bmatrix} a & -3a \end{bmatrix}^{T}, a\in\Re \right \} $

              $ \color{green} \text{For SOSC, we consider } \tilde{T}\left( x^{* },\mu^{*} \right) \text{, but here } \mu^{*} \text{ is not applicable.} $

              $ \text{Hence } y^{T}L\left ( x^{\ast },\mu ^{\ast } \right )y=\begin{bmatrix} a & -3a \end{bmatrix}\begin{bmatrix} 3 & 0\\ 0 & 0 \end{bmatrix}\begin{bmatrix} a \\ -3a \end{bmatrix}=3a^{2}>0 $

Therefore, x* is a strict local minimizer.



Automatic Control (AC)- Question 3, August 2011

Go to


Back to ECE Qualifying Exams (QE) page

Alumni Liaison

Abstract algebra continues the conceptual developments of linear algebra, on an even grander scale.

Dr. Paul Garrett