EEM 342: Fundamentals of Control Systems

2017-2018 Spring Semester 2nd Midterm Exam

Time: 120 minutes; Total Points: 100
One size-A4 formula sheet and a hand calculator, not capable of symbolic operations, may be used.

0. (2 points) Write the group you are registered to (not the one you are taking the lab with) on the
top-right corner (if you write it anywhere else, you will loose points) of your answer sheet.

1. (Each part worth 6 points) Consider the system described by

ÿ(t) = u(t) + 2y(t)ẏ(t) + y(t) (1)

where u(t) is the input and y(t) is the output at time t.

a) Choose a state vector for this system and write the state equations.
b) Determine all the equilibrium points of the system.
c) For each equilibrium point, determine whether or not the equilibrium point is
(i) stable in the sense of Lyapunov; (ii) asymptotically stable. Explain your answer.
d) A linearized model for this system, which is valid around y(t) = ẏ(t) = 0, can be obtained by
ignoring the higher order terms (i.e., the term 2y(t)ẏ(t) in this case) in (1). Write the state
equations of this linearized model and find its modes. Is this linearized system (i) stable in the
sense of Lyapunov; (ii) asymptotically stable? Explain your answer.
e) Find the transfer function from the input u(t) to the output y(t) of the linearized system
obtained in part (d). Find the poles of this transfer function. Is the linearized system (i)
bounded-input bounded-output stable? (ii) critically stable? Explain your answer.
f) Is it possible to stabilize (asymptotically) the linearized system obtained in part (d) by using
static output feedback u(t) = Ky(t) (where K is a constant)? If so, choose an appropriate K
to stabilize this system. If not, explain why.
g) Assume that both y(t) and ẏ(t) can be measured. Is it possible to stabilize (asymptotically) the
linearized system obtained in part (d) by using a feedback of the form u(t) = K1 y(t) + K2 ẏ(t)
(where K1 and K2 are constants). If so, choose appropriate K1 and K2 to stabilize this system.
If not, explain why.
h) Is it possible to stabilize (asymptotically) the linearized system obtained in part (d) by using
dynamic output feedback û(s) = C(s)ŷ(s), where C(s) is a rational and proper transfer function
and û and ŷ respectively denote the Laplace transforms of u and y. If so, determine such a
C(s) to stabilize this system. If not, explain why.

1
2. (40 points) Consider the following control system configuration

+ +
e - u-
- y1 = z
r - - K0 S
  - y2
6− 6−

K2 

K1 

where K0 , K1 , and K2 are constant gains to be determined and S is a system described as

" # " #
0 1 0
ẋ(t) = x(t) + u(t)
0 0 1
" #
y1 (t)
y(t) = = x(t)
y2 (t)
h i
z(t) = y1 (t) = 1 0 x(t)

Choose the gains K0 , K1 , and K2 such that (i) the overall closed-loop system is asymptotically
stable, (ii) the output z(t) can track a step reference r(t) with no steady-state error, (iii) maximum
percent overshoot in z(t) in response to a step reference r(t) is 10%, and (iv) the output z(t) settles
to 2% of its steady-state value within 10 time units.

3. (10 points) Consider the overall system of Question 2 with the gains K0 , K1 , and K2 as chosen to
satisfy the requirements (i)–(iv) given there. Determine the steady-state error in response to a

a) unit ramp reference r(t) = t, t ≥ 0;

b) unit quadratic reference r(t) = 21 t2 , t ≥ 0.

2
 SOLUTIONS

" # " #
x1 (t) y(t)
1. a) We can choose x(t) = = . Then, the state equations can be written as:
x2 (t) ẏ(t)
" #
x2 (t)
ẋ(t) = f (x(t), u(t)) =
u(t) + 2x1 (t)x2 (t) + x1 (t)

y(t) = g(x(t), u(t)) = x1 (t)

b) The only solution to"f (x#e , 0) = 0 is xe = 0 (here, of course “0” indicates the zero vector; i.e.,
0
xe = 0 means xe = ). Thus, xe = 0 (which corresponds to y(t) = ẏ(t) = 0) is the only
0
equilibrium point.
c) Let u(t) = 0, for t ≥ t1 , for some initial time t1 , and suppose that y(t1 ) and ẏ(t1 ) are both
positive. Then, no matter how small y(t1 ) and ẏ(t1 ) are, by (1), ÿ(t1 ) is also positive. This, for
t > t1 , makes ẏ(t) even more positive, which in turn makes y(t) even more positive. Thus, the
state gets away from the equilibrium point xe = 0. Therefore, the equilibrium point xe = 0 is
an unstable equilibrium point (it is neither stable in the sense of Lyapunov, nor asymptotically
stable).
d) By ignoring the higher order terms in (1), we obtain

ÿ(t) = u(t) + y(t)

" # " #
x1 (t) y(t)
Thus, by choosing x(t) = = , as above, the linearized model is obtained as:
x2 (t) ẏ(t)
" # " #
0 1 0
ẋ(t) = x(t) + u(t)
1 0 1
h i
y(t) = 1 0 x(t)

Note that this is the same as eq. (2) in Homework 3, with β = 0, 3g 3

2l = ml2 = 1. The modes
are the roots of the charecteristic equation s2 − 1 = 0, which are s1 = −1 and s2 = 1. Since
the system has one mode with positive real part (s2 = 1), it is neither stable in the sense of
Lyapunov, nor asymptotically stable.
1
e) The transfer function is G(s) = 2 . Thus, its poles are the roots of s2 − 1 = 0, which are
s −1
s1 = −1 and s2 = 1. Since the transfer function has one pole with positive real part (s2 = 1),
the system is neither bounded-input bounded-output stable nor critically stable.
f) With u(t) = Ky(t), the closed-loop dynamics equation of the linearized system is
" #
0 1
ẋ(t) = x(t)
1+K 0

whose modes are the roots of the charecteristic equation s2 − 1 − K = 0. This equation has
one positive and one negative real root, for K > −1, has a double root at zero for K = −1,
and has a couple of complex conjugate roots on the imaginary axis for K < −1. Therefore,
although it is possible to stabilize this system in the sense of Lyapunov, by choosing K < −1,
it is not possible to asymptotically stabilize this system by a static output feedback.

3
 h i
g) With u(t) = K1 y(t) + K2 ẏ(t) = K1 K2 x(t), the closed-loop dynamics equation of the
linearized system is " #
0 1
ẋ(t) = x(t)
1 + K1 K2
whose modes are the roots of the charecteristic equation s2 − K2 s − 1 − K1 = 0. By Routh-
Hurwitz criterion, both roots have negative real parts if K1 < −1 and K2 < 0. Therefore, it is
possible to asypmtotically stabilize this system by a feedback of the form u(t) = K1 y(t)+K2 ẏ(t)
(which, in fact, is static state feedback), with K1 < −1 and K2 < 0. It can be shown that,
when such a feedback is applied to the actual nonlinear system described by (1), the point
xe = 0 becomes an asymptotically stable equilibrium point.
h) Since the linearized system obtained in part (d) is a second order system with no hidden modes,
it is possible to stabilze it by a first order controller. The most general form of a first order
K1 s + K2
proper rational transfer function is C(s) = (where K1 , K2 , and p are parameters to
s+p
be chosen). When we apply this controller, the closed loop charecteristic polynomial becomes
(note that positive feedback is given in the question) d(s) = (s2 − 1)(s + p) − (K1 s + K2 ) =
s3 + ps2 − (1 + K1 )s − p − K2 . By forming the Routh table, all the roots of this polynomial
have negative real parts if and only if (i) p > 0, (ii) K2 − pK1 > 0 and (iii) −p − K2 > 0. For
example, p = 1, K2 = −2, and K1 = −3 satisfy all these conditions. Thus, the controller with
−3s − 2 1
transfer function C(s) = = −3 + stabilizes this system when implemented as
s+1 s+1
û(s) = C(s)ŷ(s). This controller can be realized as (not required for the exam)

ż(t) = −z(t) + y(t)

u(t) = z(t) − 3y(t)

" #
x(t)
where z(t) is the state of the controller. By letting x̂(t) = to represent the overall
z(t)
state of the closed-loop system, when this controller is applied to the actual nonlinear system
described by (1), it can be shown that, the point x̂e = 0 (which corresponds to y(t) = ẏ(t) =
z(t) = 0) becomes an asymptotically stable equilibrium point of the overall closed-loop system.

2. By closing the inner loop (the one with gain K2 ), the system from e to z can be described as
" # " #
0 1 0
ẋ(t) = x(t) + e(t)
0 −K0 K2 K0
h i
z(t) = 1 0 x(t)
K0
which has the transfer function GOL (s) = . Thus, the overall system is equivalent to
s(s + K0 K2 )

+ e K0
r - - GOL (s) = - z
 s(s + K0 K2 )
6−

K1 

4
 Note that the open-loop transfer function is of type-1. Therefore, given that requirement (i) is
satisfied, requirement (ii) will be satisfied if we have unity feedback. To achive unity feedback,
however, we have to choose K1 = 1. With K1 = 1, the closed-loop transfer function from r to z
K0
is GCL (s) = 2 . Therefore, requirement (i) is satisfied when K0 and K2 are both
s + K0 K2 s + K0
Kωn2
positive. By equating GCL (s) to the standard second order transfer function, 2 , we
s + 2ζωn s + ωn2
obtain K = 1 (which in fact is needed to satisfy requirement (ii)), ωn2 = K0 and 2ζωn = K0 K2 . Thus,
√ √
ωn = K0 and ζ = 12 K2 K0 . Given that requirement (i) is satisfied, to satisfy requirement (iii),
|ln(0.1)|
we must have ζ = p 2 = 0.59; and to satisfy requirement (iv), we must have (we use
π + |ln(0.1)|2
1 √
 q 
the envelope settling time) ζωn ≥ − ln 0.02 1 − ζ = 0.4127 or K0 = ωn ≥ 0.4127
2
0.59 = 0.6995.
10 √
To satisfy this, we can take K0 = 12 . Then we must have 12 K2 K0 = 2√ 1
K = 0.59 or K2 = 1.67.
2 2
1
Therefore, K0 = 2 , K1 = 1, and K2 = 1.67 satisfy all the requirements.

3. a) Since the open-loop transfer function is of type-1, the steady-state error in response to a unit
1 1
ramp reference is ess = , where Kv = lim (sGOL (s)) = . Thus, with K2 = 1.67,
Kv s→0 K2
ess = 1.67.
b) A type-1 system can not follow a quadratic reference. Thus, ess = ∞.