You are on page 1of 3

EE 561: Digital Control Systems

Problem Set 6 Clues

Due on Dec 6 at 5PM Problem 1 [15 Points] The state-matrices in Control Canonical form for a system are given by 6 13 10 1 0 0 , c = 0 , Cc = 1 4 3 , Dc = 0. c = 1 0 1 0 0 Convert these matrices into Modal Canonical form having the new matrices m , m , Cm and Dm . Method 1: Find out the eigenvalues of c Find out the eigenvectors m must have a block or corresponding to non-real eigenvalues j Form a transformation matrix T from the eigenvectors. Calculate m , Cm and Dm using the transformation matrix. Method 2: Write down the transfer function. Split the transfer function into partial fractions; one fraction should represent a second order system. Write down m , m , Cm and Dm where blocks in these matrices corresponding to the second order system are written in Control Canonical form. Problem 2 [20 Points]
Find out the state x = [x 1 x2 ] and the input u = [u 1 u2 ]

Fall 2013

that minimize the cost

2 2 2 J (x1 , x2 , u1 , u2 ) = 2x2 1 + 2x1 x2 + 3x2 + 5u1 + 3u1 u2 + 7u2 ,

to drive a system that is subject to the constraints x1 = 1 x2 , u1 = 1 + u2 . The only stationary point of J (x1 , x2 , u1 , u2 ) is its one global minimum. Why there are no other stationary points such as local minima/maxima, global maxima or saddle points? Dene a new cost function made by incorporating constraints
2 2 2 x1 , x2 , u1 , u2 , 1 , 2 ) = 2x2 J( 1 +2x1 x2 +3x2 +5u1 +3u1 u2 +7u2 +1 (x1 +x2 1)+2 (u1 u2 1).

Find the stationary point(s) of the new cost function through equating to zero the deriva with respect to each variable and solving the resulting system of equations. tive of J with respect to each variable. Analyze the double derivative of J

Problem 3 [30 Points] For a plant with state x(t) and state matrices F= 0 3 1 5 , G= 0 1 , H= 1 0 , J = 0,

(a) Design a controller, without using integral control, to yield a 5% overshoot and a settling time of 1 s. Find K using that same old friend Ackermann, or you can always compare the coecients but watch out for the controllability in this case. (b) Find out the steady-state error for a unit-step input, with this controller. Using u = Kx + r, nd out the new matrices F, G, H, J for the system with a reference input and the controller. Either use the fact that x = 0 at steady-state, and nd out the steady-state error. Or spend a lot of time in nding out the transfer function, and then calculate the steady-state error using nal value theorem. (c) Now design the controller using integral control. Augment the state vector with a new state xi that is output of the integral controller. Using u = Kx + ki xi + r, nd out the new matrices F, G, H, J for the system with a reference input and the controller. Find out the characteristic equation for this overall third-order system. Intelligently increase by one the order of the design specication polynomial. Find out new K and ki . (d) Find out the steady state error with this integral control, for a unit-step input. What type of system is this? You should have learned it by now. Problem 4 [35 Points] The transfer function for a system is given by G(z ) = z2 z + 0.967 . 1.67z + 0.91

(a) Find control feedback K and observer gain L that place control poles at z = 0.8 j 0.2 and observer poles at z = 0.6 j 0.3. Write down the state-matrices in a canonical form of your choice. But stick to that form to nd K and L for a consistent coupling of the observer and the controller. (b) Draw the block-diagram of the system in state-command structure.

Figure 1: State-command structure with Nx and Nu . (c) For the state-command structure, nd out Nx and Nu , and plot y (k ) for a unit step input. Using input u from the block diagram, and using the fact that x = 0 at steady-state, nd out Nx and Nu from the state-space system of equations. Using the same u and incorporating error states as wel through augmenting the state vectorl, derive new state-space representative matrices and plot y (t) on Matlab using these matrices.

= KNx + Nu , draw the new block diagram substituting N for Nx and Nu . (d) With N

. Figure 2: State-command structure with N