IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 53, NO.

9, OCTOBER 2008

[25] G. Pillonetto, Solutions of nonlinear control and estimation problems

in reproducing kernel Hilbert spaces: existence and numerical determination, Automatica, vol. 44, no. 8, pp. 21352141, Aug. 2008.
[26] R. Ravikanth and S. P. Meyn, Bounds on achievable performance in
the identification and adaptive control of time-varying systems, IEEE
Trans. Automat. Control, vol. 44, no. 4, pp. 670682, Apr. 1999.
[27] L. Schwartz, Radon Measures on Arbitrary Topological Spaces and
Cylindrical Measures. New York: Oxford University Press, 1973.
[28] H. L. Van Trees, Detection, Estimation, and Modulation Theory Part
1. New York: Wiley, 1968.
[29] Y. Z. Tsypkin and M. V. Bondarenko, An optimal algorithm for identification of rapidly time-varying systems, IEEE Trans. Automat. Control, vol. 37, no. 2, pp. 237239, Feb. 1992.
[30] A. N. Tychonov and V. Y. Arsenin, Solutions of Ill-Posed Problems.
Washington, DC: Winston/Wiley, 1977.
[31] G. Wahba, Spline Models for Observational Data. Philadelphia, PA:
SIAM, 1990.
[32] L. Y. Wang, Persistent identification of time-varying systems, IEEE
Trans. Automat. Control, vol. 42, no. 1, pp. 6682, Jan. 1997.
[33] L. Yingwei, N. Sundararajan, and P. Saratchandran, Identification of
time-varying nonlinear systems using minimal radial basis function
neural networks, Control Theory and Applications, IEE Proceedings,
vol. 144, no. 2, pp. 202208, Mar. 1997.

Unreachable Setpoints in Model Predictive Control

James B. Rawlings, Dennis Bonn, John B. Jrgensen,
Aswin N. Venkat, and Sten Bay Jrgensen

AbstractIn this work, a new model predictive controller is developed

that handles unreachable setpoints better than traditional model predictive
control methods. The new controller induces an interesting fast/slow asymmetry in the tracking response of the system. Nominal asymptotic stability
of the optimal steady state is established for terminal constraint model predictive control (MPC). The region of attraction is the steerable set. Existing
analysis methods for closed-loop properties of MPC are not applicable to
this new formulation, and a new analysis method is developed. It is shown
how to extend this analysis to terminal penalty MPC. Two examples are presented that demonstrate the advantages of the proposed setpoint-tracking
MPC over the current target-tracking MPC.
Index TermsAymptotic stability, constraints, model predictive control.

I. INTRODUCTION
Model predictive control (MPC) has been employed extensively for
constrained, multivariable control in chemical plants [1]. The main moManuscript received February 25, 2008; revised April 23, 2008. Current version published October 8, 2008. This work was supported in part by the industrial members of the TexasWisconsin Modeling and Control Consortium
and the National Science Foundation under Grant CTS-0456694, by the Computer Aided Process and Products Center, Technical University of Denmark,
and by the Danish Department of Energy under Grant EFP 273/01-0012. Recommended by Associate Editor M. Fujita.
J. B. Rawlings is with the Department of Chemical and Biological Engineering, University of Wisconsin, Madison, WI 53706-1691 USA (e-mail: [email protected]).
D. Bonn is with the CAPEC, Department of Chemical and Biochemical Engineering, Technical University of Denmark, DK 2800 Lyngby, Denmark.
J.B. Jrgensen is with the DTU Informatics, Technical University of Denmark, DK 2800 Lyngby, Denmark.
A.N. Venkat is with the Shell Global Solutions (U.S.), Inc., Houston, TX
77084 USA.
S.B. Jrgensen is with the CAPEC, Department of Chemical and Biochemical
Engineering, Technical University of Denmark, DK 2800 Lyngby, Denmark.
Digital Object Identifier 10.1109/TAC.2008.928125

2209

tivation for choosing MPC is to handle constrained systems. Essentially all current MPC theory is predicated on the assumption that the
setpoint has been transformed to a desired and reachable steady-state
target in a separate steady-state optimization [2], [3]. This assumption
allows current theory to address only transient active constraints that
occur when solving the controllers dynamic optimization to drive the
system to the desired and reachable steady state. In practice, the setpoint often becomes unreachable due to a disturbance and the choice
to transform the problem to one with a desired and reachable steady
state is a significant decision that affects controller performance. We
show in this paper that when the setpoint is unreachable, this two-level
optimization is suboptimal and does not minimize the tracking error.
To improve the performance of MPC when the setpoint is unreachable,
we propose defining system performance relative to the unreachable
setpoint rather than the reachable steady-state target. As we show in
the examples, a consequence of the unreachable setpoint formulation
is a desirable fast/slow asymmetry in the system tracking response that
depends on the system initial condition, speeding up approaches to the
unreachable setpoint, but slowing down departures from the unreachable setpoint. One technical difficulty that arises is that the proposed
MPC cost function is unbounded on the infinite horizon. A more significant problem is that the existing MPC theory for stability and convergence no longer applies because this theory is based on establishing
the controller cost function as a Lyapunov function for the closed-loop
system, but the controller cost function is not strictly decreasing in this
case.
Optimal control problems with unbounded cost are not new to control theory. The first studies of this class of problems arose in the economics literature in the 1920s [4], in which the problem of interest was
to determine optimal savings rates to maximize capital accumulation.
Since this problem has no natural final time, it was considered on the
infinite horizon. A flurry of activity in the late 1950s and 1960s led
to generalizations regarding future uncertainty, scarce resources, expanding populations, multiple products and technologies, and many
other economic considerations. Much of this work focused on establishing existence results for optimal control with infinite horizon and
unbounded cost, and the famous turnpike theorems [5] that characterize the optimal trajectory. Refs. [6] and [7] provides a comprehensive and readable overview of this research.
This class of problems was transferred to and further generalized in
the control literature. For infinite horizon optimal control of continuous
time systems, [8] established the existence of overtaking optimal trajectories. Convergence of these trajectories to an optimal steady state
is also demonstrated. Ref. [9] extended the results of [8] to infinite
horizon control of discrete time systems. Reduction of the unbounded
cost, infinite horizon optimal control problem to an equivalent optimization problem with finite costs is established. Ref. [10] provide a
comprehensive overview of these infinite horizon results.
For feedback control subject to hard input constraints, we often
employ a receding horizon implementation of a suitably chosen finite
horizon optimal control problem. The controller is implemented by
solving the open-loop control problem at each sample time as the state
becomes available and injecting only the first move of the open-loop
optimal solution. Ref. [11] provides a review of the methods for analyzing the resulting closed-loop behavior of these controllers. In this
formulation, the existence of the optimal control is not at issue because
the horizon is finite and the cost therefore bounded, but the stability
and convergence of the resulting receding horizon implementation of
the control law is the interesting question. In this paper, we address
the closed-loop properties of this receding horizon implementation for
the case of unreachable setpoints. In the formulation introduced here,

0018-9286/$25.00 2008 IEEE

2210

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 53, NO. 9, OCTOBER 2008

the stage cost of the optimal control problem measures distance from
the setpoint even if the setpoint is unreachable at steady state due to
the input constraints.
The paper is organized as follows. In Section II, basic results required to establish properties of the new controllers are given. The
new controller, termed setpoint tracking MPC, is defined in Section III.
For simplicity of exposition, a terminal state equality constraint is employed. Convergence and nominal asymptotic stability are established
for the closed-loop system. Two examples demonstrating the benefit of
the proposed setpoint tracking MPC over current approaches are presented in Section IV. Extending the asymptotic stability result to terminal penalty MPC is discussed in Section V, and conclusions of this
study are presented in Section VI.

II. PRELIMINARIES

Denote the non-negative reals by + . Denote a closed ball of radius

+ centered at a 2 n by Bb (a)

( ) = f jk 0 k  g
x

min ( ) subject to
x

( ) = 12 0
Consider a sequence ( ) 2

0
x Qx + q x + c

L x

x i

xP

Q >

= 1 ...
1
=
()
,i

=1

( )=

=1

( ( )) 0 21

=1

( + 1) =

( ) 1
P

=1

k ( ) 0 k2
x i

xP

(1)

( )=

P L xP

=1

( ( ))

the

()

x i

not

all

= 1 ...
;

;P

equal,

01

the positive function

positive
difference
Q
P
L x i
P L xP
this
difference
(or a
i
positive lower bound on this difference) proves useful in the sequel.
Lemma 2 (Basic Convex Optimization Result): Given L(x) is
a strictly convex, quadratic function of x 2 n , and optimization
problem
x i

are

2
gives
the
=1 k ( ) 0 k
(
(
))
0
(
)
.
Knowing
=1
P
i

x i

xP

min ( ) subject to 2
x

L x ;

2
2

()

(2)

Bu k

L x; u

) = 21 ( 0 sp )0 ( 0 sp ) + ( 0 sp )0 ( 0 sp )
0
0
x

Q >

R >

Q x

R u

We assume the input constraints define a nonempty polytope in

=f j

u Hu

 g

L x; u

x;u

( ( )) if and only if ( )

x i

Ax k

min (

L x i

L x i

= ( + 1)

Definition 1 (Optimal Steady State): The optimal steady state, denoted (x3 ; u3 ), is the solution to the following optimization problem

Remark 2: Because L is strictly convex, equality is achieved if and

only if x(i) = x(i + 1), i = 1; . . . P 0 1
P

, u 2 m . The cost function L(x; u) is strictly convex and

nonnegative and vanishes only at the setpoint, L(xsp ; usp ) = 0. In this
paper we employ the popular choice of a quadratic measure of distance
from the setpoint
x

L xP

( )+

x k

2  0, Jensens inequality for this

Remark 1: Since kx(i) 0 xP kQ
quadratic function follows from this result by replacing the second sum
in (1) with zero, changing the equality to an inequality and dividing by
P

The system model is assumed linear and time invariant

x i :

in which KL ; Kx  0 are the Lipschitz constants.

A more general version of this lemma is stated and proved in [13].

L x i

p;

kx3 (p) 0 x3 (0)k  Kx kpk ;

3
kL(x (p)) 0 L(x3 (0))k  KL kpk ;

with mean

;P

Ax

III. TERMINAL CONSTRAINT MPC

Then the following holds:

P L xP

L x ;

with nonempty, convex and closed (not necessarily bounded). The

(convex) feasible set for the constraints Ax = p, x 2 is assumed
nonempty for p 2 , in which contains the point p = 0.
The optimal solution x3 (p) and cost are uniquely defined and Lipschitz continuous on the set

Lemma 1 (An Equality for Quadratic Functions): Let

be a
nonempty, closed, bounded, convex subset of n (hence compact),
and let L be a strictly convex, quadratic function defined on

1
2

A proof is provided in [12, p. 7273]

.
Lemma 3 (Lipschitz Continuity): Given L(x) is a strictly convex,
quadratic function of x 2 n , and optimization problem

Bb a

If

with nonempty, convex, and closed (not necessarily bounded).

1) The solution x3 exists and is unique.
2) If x(k) is a feasible sequence (x(k) 2 , L(x(k)) bounded for
all k ) with L(x(k)) ! L(x3 ), then x(k) ! x3 .

subject to
x

Ax

Bu

We assume the model and constraints admit a steady state, i.e., the
steady-state problem is feasible. If A has no eigenvalues at unity, then
any admits a steady state. If A has eigenvalues at unity, then containing an element of the null space of B , for example, is sufficient for
a steady state to exist.
Remark 3: Positive definiteness of Q and R implies that L(x; u) is
strictly convex. Therefore (x3 ; u3 ) is unique according to Lemma 2.
A. Terminal Constraint MPC
Define the controller cost function

x;

f ( )g =001 =
u i

N
i

( + 1) =

x k

01

=0

( ( ) ( ))

L x k ;u k

( )+

Ax k

()

Bu k ;

(0) =

x:

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 53, NO. 9, OCTOBER 2008

Fig. 1. Closed-loop MPC sequence with accumulation point . Two forecast

are shown. Trajectory 1 is the truncation of the
trajectories from (2) to
optimal trajectory starting from point (1) and trajectory 2 is the complete
optimal trajectory from point (2).

For notational simplicity, we define

press the MPC control problem as

u=

fu(i)gNi=001 . Hence, we ex-

min
8(x; u)
u
subject to x(0) = x
x(k + 1) = Ax(k ) + Bu(k )
k = 0; . . . ; N 0 1
3
x (N ) = x
u (k ) 2
k = 0; . . . ; N 0 1:

The set is nonempty because it contains x3 . The set is bounded because N is finite and is bounded. All closed-loop sequences are therefore bounded because the set N is positive invariant under the MPC
control law.1
For singular A, the steerable set is not bounded. The following
lemma, however, shows that the closed-loop sequence starting in the
(unbounded) steerable set evolves in a bounded set.
Lemma 5: For singular A, there exists 0  r  n + N such that a
closed-loop sequence x(k) under terminal constraint MPC starting in
the steerable set remains in a nonempty, bounded, convex subset of n
for times k  r .
Lemma 6: Every closed-loop sequence has at least one accumulation point.
Proof: Remark 4 and Lemma 5 ensure that every closed-loop
sequence x(k) is bounded after finite k , and the Bolzano-Weierstrass
property ensures that every infinite sequence in a closed, bounded
subset of n has at least one accumulation point [14].
Let fx(k)g be a closed-loop sequence with an accumulation point
a. Say the closed-loop sequence enters B (a) at time index E +1. Define the sequence fw(k)g to be a portion of the closed-loop sequence,
w (k ) = x(k + E ), k = 1; . . . P . Because a is an accumulation point
of the closed-loop sequence, for every  > 0, we can choose E and P
large enough so that w(1); w(P ) 2 B (a) as depicted in Fig. 1. Define
the means over the closed-loop sequence to be
wP

= P1

()

w i uP
i=1

= P1

u
i=1

(w(i)):

Note also that the mean state and input satisfy

(3a)
(3b)
(3c)
(3d)
(3e)

We denote the optimal input sequence by

u0 (x) = fu0 (0; x); u0 (1; x); . . . ; u0 (N 0 1; x)g:

The MPC feedback law is the first move of this optimal sequence,
which we denote as u0 (x) = u0 (0; x), and we denote the optimal cost
by 80 (x). The closed-loop system is given by

( + 1) = Ax(k) + Bu0 (x(k))

with u(k) = u0 (x(k)) so
0
x(k + 1) = f (x(k )) f (1) = A(1) + Bu (1):
x k

Lemma 4: The optimal steady state is a fixed point of the closed-loop

system

( 3 ) = Ax3 + Bu0 (x3 ) = x3 :

f x

Proof: For x(0) = x3 in the MPC control problem, (3), both the initial
and terminal states are equal to x3 . Strict convexity of the cost function
implies the entire trajectory remains at x3 so u0 (x3 ) = u3 .
Definition 2 (Steerable Set): The steerable set N is the set of states
steerable to x3 in N steps
N

2211

= fxjx3 = AN x + AN 01 Bu(0) + 1 1 1 + Bu(N 0 1);

u(j ) 2 ; j = 0; . . . ; N 0 1g

Remark 4: For nonsingular A, the steerable set is a nonempty,

bounded, convex subset of n .

wP

= AwP + BuP + P1 (w(1) 0 w(P + 1)):

(4)

Now consider the MPC optimal open-loop trajectory from each w(k)
for k = 1; . . . ; P . The optimal open-loop input trajectories are

u0 (w) = fu0 (0; w); u0 (1; w); . . . ; u0 (N 0 1; w)g:

We wish to compare the optimal costs for two succeeding states w(k)
and w(k + 1). For w(k), the cost is 8(w(k); u0 (w(k))). For state
w (k + 1), consider first the feasible but possibly suboptimal sequence

u~ (w) = fu0 (1; w); u0 (2; w); . . . ; u0 (N 0 1; w); u3 g

which is created by dropping the first input u0 (0; w), shifting all the
other inputs forward in the sequence, and appending u3 for the final
input. The corresponding trajectory for w(2) is labeled Trajectory 1
in Fig. 1. This input evaluated at state w(k) is admissible for state
w (k + 1) because it satisfies the model, terminal constraint, and input
constraints.2 Its cost from state w(k + 1) is directly related to the optimal cost from w(k) by

8(w(k); u0 (w(k))) 0 L(w(k); u0 (w(k)))

= 8(w(k + 1); u~ (w(k))) 0 L(x3 ; u3 ):
Therefore, using the optimal sequence u0 (w(k + 1)) in place of
u~ (w(k)) in this equation produces the inequality
8(w(k + 1); u0 (w(k + 1)))  8(w(k); u0 (w(k)))
0 L(w(k); u0 (w(k))) + L(x3; u3 )
(5)
k = 1 ; . . . ; P 0 1:
This optimal trajectory for w(2) is labeled Trajectory 2 in Fig. 1.

1This positive invariance is a well-known characteristic of terminal constraint

MPC, but for those unfamiliar with MPC, it is also established later in the proof
of Theorem 1.
2This admissibility implies the positive invariance of the MPC control law
mentioned previously.

2212

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 53, NO. 9, OCTOBER 2008

It should be noted that in standard MPC with a reachable steady-state

target, the origin can be shifted to (x3 ; u3 ) and the term L(x3 ; u3 ) is
zero. That leads immediately to a cost decrease from w(k) to w(k +1)
and the cost function is a Lyapunov function for the closed-loop system.
With the unreachable setpoint, the situation is completely different. The
term 0L(w(k); u0 (w(k))) + L(x3 ; u3 ) changes sign with k on typical closed-loop trajectories. The cost decrease is lost and 80 is not a
Lyapunov function for the closed-loop system. We next prove asymptotic stability for model predictive control with unreachable setpoints
by other means.
Lemma 7: Every closed-loop sequence converges to the optimal
steady state.
Proof: Let a be an accumulation point of the closed-loop sequence. For every  > 0, we can choose E and P large enough so
that w(1); w(P ) 2 B (a). Denote the difference in the optimal costs
between points w(1) and w(P ) by dP
dP

= 80 (w(1)) 0 80 (w(P )):

(6)

Because the optimal cost is continuous in the state, dP goes to zero

with . Add the inequalities in (5) to obtain
P

( ()

L w i ;u
i=1

dP

We can choose an increasing sequence of P for which the right-hand

side is bounded above. Therefore the sequence (w(i); u0 (w(i))) converges to its mean (wP ; uP ) or we violate the inequality above. Therefore (w(i); u0 (w(i))) converges to (x3 ; u3 ).
Theorem 1 (Asymptotic Stability of Terminal Constraint MPC): The
optimal steady state is the asymptotically stable solution of the closedloop system under terminal constraint MPC. Its region of attraction is
the steerable set.
Proof: Any initial state in the steerable set defines an infinite
closed-loop sequence in N through the iteration x(k +1) = f (x(k)).
1) Lyapunov Stability: Let S (k; x0 ) denote the solution of x(k +
1) = f (x(k)) at time k with initial condition x(0) = x0 2 N .
Assume, contrary to what is to be proven, that there exists  > 0 such
that for every  > 0, kS (k; x0 ) 0 x3 k   for some finite k for some
3
3
3
x0 2 B  (x ), in which B  (x ) = B (x ) \ N , and we take the
intersection in case N does not contain a full neighborhood of x3 .
We establish a contradiction. Consider an  > 0 and x0 2 B  (x3 ).
Choose finite k that depends on x0 so that S (k; x0 ) 2 B  (x3 ) for all
k  k (x0 ). We know such a finite k exists for every x0 2 N since the
sequence S (k; x0 ) converges to x3 (Lemma 7). This setting is the same
as depicted in Fig. 1 (with a = x3 ), and we have the same inequality
derived previously in (8)

(w(i)))  P L(x3 ; u3 )

+ L(w(P ); u0 (w(P ))) 0 L(x3 ; u3 )

1 P kw(i) 0 w k2 + u0 (w(i)) 0 u 2
P Q
P R
2 i=1
 KL kw(1) 0 w(P + 1)k
+ dP + L(w(P ); u0 (w(P ))) 0 L(x3 ; u3 )

Next apply Lemma 1 to the variable (w(i); u0 (w(i))) to obtain

1 P kw(i) 0 w k2 + u0 (w(i)) 0 u 2
P Q
P R
2 i=1
 P (L(x3; u3 ) 0 L(wP ; uP ))
+ dP + L(w(P ); u0 (w(P ))) 0 L(x3 ; u3 )

(7)

Next, we show (wP ; uP ) converges to (x3 ; u3 ) with increasing P .

From (4), (wP ; uP ) converges to the set of steady states, and, if it
does not converge to (x3 ; u3 ), then, by optimality, L(xP ; uP ) achieves
values greater than L(x3 ; u3 ) for infinitely many P . Equation (7) then
provides a contradiction because the right-hand side goes to 01 for a
suitable subsequence with increasing P and the left-hand side is nonnegative. Next consider optimizing subject to the constraint given by(4)

min
L(x; u)
x;u
x

= Ax + Bu + p

1 (w(1) 0 w(P + 1))

p=
P

in which P = k , w(i) = S (i 0 1; x0 ). Since kS (i; x0 ) 0 x3 k   for

some 0 < i < k (denote one of these i as i), and kS (0; x0 ) 0 x3 k  ,
we have the following lower bound on the sum of the following two
distances from the mean:

kw(1) 0 wP k2 + w(i) 0 wP 2  12 ( 0 )2 :

2
Since Q > 0, there exists b > 0 such that kxkQ
 b kxk2 for all x,
and we have established a lower bound on the sum
P
i=1

subject to

(9)

kw(i) 0 wP kQ2 +

(w(i)) 0 uP

2
R

 2b ( 0 )2 :

Substitution into (9) gives

u

and denote this solution as (x3 (p); u3 (p)). From its construction we
know L(x3 (p); u3 (p))  L(wP ; uP ) and therefore

1 P kw(i) 0 w k2 + u0 (w(i)) 0 u 2
P Q
P R
2 i=1
 P (L(x3; u3 ) 0 L(x3(p); u3 (p)))
+ dP + L(w(P ); u0 (w(P ))) 0 L(x3 ; u3 )

Applying Lipschitz continuity, Lemma 3, gives

1 P kw(i) 0 w k2 + u0 (w(i)) 0 u 2
P Q
P R
2 i=1
 KL kw(1) 0 w(P + 1)k
+ dP + L(w(P ); u0 (w(P ))) 0 L(x3 ; u3 )

b
2
4 ( 0 )  KL kw(1) 0 w(P + 1)k
+ dP + L(w(P ); u0 (w(P ))) 0 L(x3 ; u3 )

Taking the limit as  ! 0, the left-hand side converges to b=42 > 0

and the right-hand side converges to zero, and we have established a
contradiction. Therefore, for every  > 0, there exists  > 0 such
that S (k; x0 )   for all k  0 for all x0 2 B  (x3 ), and we have
established Lyapunov stability.
2) Convergence: Convergence of x(k) to x3 for x0 in the steerable
set is established in Lemma 7. Convergence and Lyapunov stability
imply asymptotic stability and the theorem is proved.
IV. EXAMPLES

(8)

Two examples are presented to illustrate the advantages of the proposed setpoint tracking MPC framework (sp-MPC) compared to tradi-

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 53, NO. 9, OCTOBER 2008

2213

Fig. 2. Closed-loop performance of sp-MPC and targ-MPC (Example 1).

tional target tracking MPC (targ-MPC). The regulator cost function for
the new controller, setpoint-tracking MPC (sp-MPC), is

= 12

sp-MPC

01

k ( )0
x j

j =0

xsp

TABLE I
COMPARISON OF CONTROLLER PERFORMANCE (EXAMPLE 1)

2
Q

+ k ( ) 0 k + k ( + 1) 0 ( )k (10)
0,  0 and at least one of
0. This system
in which
can be put in the standard form defined for terminal constraint MPC by
( 0 1)] [3]. The regulator
augmenting the state as ~( ) = [ ( )
u j

Q >

usp

u j

R; S

u j

R; S >

x k

x k

u k

cost function in traditional target-tracking MPC (targ-MPC) is

targ-MPC

= 12

01

k ( ) 0 3k
x j

j =0

+ k ( ) 0 3 k + k ( + 1) 0 ( )k
u j

u j

u j

(11)

In the examples, the controller performance is assessed using the following three closed-loop control performance measures:

0
8 ( ) = 1 21 k ( ) 0 k + k ( + 1) 0 ( )k
0
8 ( ) = 1 21 k ( ) 0 k
8( ) = 8 ( ) + 8 ( )
1

u k

kT

u j

y k

kT

usp

u j

u j

j =0

x j

xsp

j =0

u k

y k

in which T is the process sample time. For each of the indices defined
above, we define the percentage improvement of sp-MPC compared to
targ-MPC by

1% = 8

targ-MPC

08

sp-MPC

targ-MPC

2 100

1) Example 1: The first example is the single input-single output

system

( ) = 60 0+0 592623
2 +1

G s

= 10
0 953

: s

=1

( )=

(12)

1
= 10 = 0
= 80

sampled with T
s. The input u is constrained as juj  . The de:
which corresponds to a steady-state
sired output setpoint is ysp
. The regulator parameters are Qy
,R
,
input value of 0 :
0
S
, Q
C Qy C
:
I2 . A horizon length of N
is used. Between times 50130, 200270 and 360430, a state dis0 causes the input to saturate at its lower
: ; :
turbance dx
limit. The output setpoint is unreachable under the influence of this
state disturbance dx . The closed-loop performance of sp-MPC and
targ-MPC under the described disturbance scenario are shown in Fig. 2.

= 0 25
=
+ 0 01
= [17 1 1 77]
( )

The closed-loop performance of the two control formulations are compared in Table I.
2) Comments: In the targ-MPC framework, the controller tries to
reject the state disturbance and minimize the deviation from the new
steady-state target. This requires a large, undesirable control action that
forces the input to move between the upper and lower limits of operation. The sp-MPC framework, on the other hand, attempts to minimize
the deviation from setpoint and subsequently the input just rides the
lower limit input constraint. The benefit here is that the sp-MPC controller slows down the departure from setpoint, but speeds up the approach to setpoint. The traditional targ-MPC can be tuned to be fast or
slow through relative choice of tuning parameters Q and R, but it is fast
or slow from all initial conditions, some of which lead to an approach
setpoint, but others of which lead to a departure from setpoint.
The greater cost of control action in targ-MPC is shown by the cost
index u in Table I. The cost of control action in targ-MPC exceeds
that of sp-MPC by nearly 100%. The control in targ-MPC causes the
output of the system to move away from the (unreachable) setpoint
faster than the corresponding output of sp-MPC. Since the control objective is to be close to the setpoint, this undesirable behavior is eliminated by sp-MPC.
3) Example 2: The second example is the 2 input2 output system
G s

1:5
(s+2)(s+1)
0:5
(s+0:5)(s+1)

0:75
(s+5)(s+2)
2
(s+2)(s+3)

= 0 25

(13)

:
s. The inputs u1 and
The system is sampled at the rate T
are constrained between 0 : and 0.5. The desired output setpoint
0 which corresponds to a steady-state input of
:
; :
is ysp
0 . The state and measurement evolution of the
usp
:
;0 :
system are corrupted by noise. The state and measurement noise covari04 I2 , respectively. In addition to
04 I6 and Rv
ance are Qx
random state noise, a zero-mean square-wave disturbance of magnitude
6 : ; 0 : 0 also affects the state evolution of the system. For this
u2

= [0 337 0 34]
= [0 496 0 468]
= 10
[0 03 0 03]

05

= 10

2214

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 53, NO. 9, OCTOBER 2008

Fig. 3. Closed-loop outputs of sp-MPC and targ-MPC in Example 2.

Fig. 4. Closed-loop inputs of sp-MPC and targ-MPC in Example 2.

Fig. 5. Target and disturbance estimates in Example 2.

example, the state and disturbances are estimated using a steady-state

Kalman filter. The regulator parameters are Qy = I2 , R = 0I2 ,
S = I2 , and Q = C 0 Qy C + I6 . The horizon length is N = 80.
The closed-loop performance of the sp-MPC and the targ-MPC formulations are shown in Figs. 3 and 4. Fig. 5 depicts the target and disturbance estimates for the noisy system. Table II quantifies the relative
performance of the sp-MPC and targ-MPC frameworks.
4) Comments: In this example, the input u1 is near its upper limit
at steady state. The presence of state and measurement noise causes
input u1 to saturate and consequently the setpoint to become unreachable. We note from Table II that the total cost of control action used by
targ-MPC and sp-MPC are nearly the same, but the system outputs behave quite differently. The overall controller cost of targ-MPC exceeds
that of sp-MPC by nearly 70% (Table II).

TABLE II
COMPARISON OF CONTROLLER PERFORMANCE (EXAMPLE 2)

V. TERMINAL PENALTY MPC

It is well known in the MPC literature that terminal constraint MPC
with a short horizon is not competitive with terminal penalty MPC in

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 53, NO. 9, OCTOBER 2008

terms of the size of the set of admissible initial states, and the undesirable difference between open-loop prediction and closed-loop behavior
[11]. Here we briefly outline how the previous analysis can be applied
to prove asymptotic stability of terminal penalty MPC. First define the
rotated cost Lr [9]

Lr (x; u) = L(x; u) 0 L(x3 ; u3 )

and compute the infinite horizon rotated cost-to-go under control law
uc (x) = K (x 0 x3 ) + u3 with K chosen so that A = A + BK is
asymptotically stable. A simple calculation gives

L1
r (x ) =

k=0

Lr (x(k); u(k))

x(k + 1) = Ax(k) + Bu(k); x(0) = x

in which 5 satisfies the usual Lyapunov equation and  is given by

5 = A0 5A + Q + K 0 RK
0
 = (I 0 A )01 Q(xsp 0 x3 ) + K 0 R(usp 0 u3 ) :
Note that the rotated cost-to-go satisfies Lr1 (x) = Lr1 (Ax +
Buc (x)) + Lr (x; uc (x)). Next define the terminal penalty MPC cost
8r

1
x; fu(i)gN0
i=0

N01
k=0

Notice this inequality is the same as (5) if we replace the cost function
8 in (5) by the rotated cost function 8r . The remaining steps after (5) in
the previous development can then be followed to establish asymptotic
stability of (x3 ; u3 ) for terminal penalty MPC.
If the system is unstable and feasible K does not exist to make (A +
BK ) stable given the active constraints at u3 , constraints to zero the
unstable modes at stage N are added to (14) [15]. The set of admissible
initial states is chosen to ensure positive invariance and feasibility of
control law uc . These plus compactness of ensure system evolution
in a compact set. The user has some flexibility in choosing K , such as
the simple choice K = 0. A procedure to choose feasible K that more
closely approximates the infinite horizon solution is given in [3].
VI. CONCLUSION

1
Lr1 (x) = (x 0 x3 )0 5(x 0 x3 ) 0 0 (x 0 x3 )
2

function as

2215

Lr (x(k); u(k)) + Lr1 (x(N ))

x(k + 1) = Ax(k) + Bu(k); x(0) = x

In this work, a new MPC controller was presented that handles unreachable setpoints better than standard MPC methods. The new controller was based on a cost function measuring distance from the unreachable setpoint, and, for simplicity of exposition, a finite horizon
with a terminal constraint. Nominal asymptotic stability of the optimal
steady state was established for this terminal constraint MPC. Two examples were presented that demonstrate the advantages of the proposed
setpoint-tracking MPC over the existing target-tracking MPC. Cost improvements between 50% and 70% were obtained for the two examples. It was next shown how to extend the asymptotic stability result to
the case of terminal penalty MPC by introducing rotated cost.
This new unreachable setpoint approach should also prove useful in
applications where optimization of a systems economic performance
is a more desirable goal than simple setpoint tracking.

and controller
ACKNOWLEDGMENT

min
8r (x; u)
u
subject to x(0) = x
x(k + 1) = Ax(k) + Bu(k)
k = 0; . . . ; N 0 1
k = 0 ; . . . ; N 0 1:
u (k ) 2

(14a)
(14b)

The authors would like to thank D. Q. Mayne and S. J. Wright for

helpful discussion of the ideas contained in this paper.

(14c)

REFERENCES

(14d)

We compute the change in cost along the closed-loop trajectory. For

w(k), the cost is 8r (w(k); u0 (w(k))). For state w(k + 1), consider
the candidate sequence

u~ (w) = fu0 (1; w); u0 (2; w); . . . ; u0 (N 0 1; w); uc (x0 (N; w))g

in which x0 (N; w) is the state value at stage N using the optimal input
sequence. Because we use the control law uc (x) in place of the constant
u3 as we did in terminal constraint MPC, and u3 is on the boundary of
, we must restrict the gain matrix K and initial state w such that the
control law is feasible [3] and the system is positive invariant. Computing the change in cost for this candidate sequence gives

8r (w(k + 1); u~ (w(k)))

= 8r (w(k); u0 (w(k))) 0 Lr (w(k); u0 (w(k)))
0 Lr1(x0 (N; w(k)))
+ Lr x0 (N; w(k)); uc (x0 (N; w(k)))
+ Lr1 Ax0 (N; w(k)) + Buc (x0 (N; w(k)))

The last three terms cancel and noting the optimization at stage k + 1,
we have the inequality

8r0 (w(k + 1))  8r0 (w(k)) 0 Lr (w(k); u0 (w(k)))

 8r0 (w(k)) 0 L(w(k); u0 (w(k)))
+ L (x 3 ; u 3 ):

[1] S. J. Qin and T. A. Badgwell, A survey of industrial model predictive

control technology, Control Eng. Prac, vol. 11, no. 7, pp. 733764,
2003.
[2] K. R. Muske and J. B. Rawlings, Model predictive control with linear
models, AIChE J.., vol. 39, no. 2, pp. 262287, 1993.
[3] C. V. Rao and J. B. Rawlings, Steady states and constraints in model
predictive control, AIChE J., vol. 45, no. 6, pp. 12661278, 1999.
[4] F. P. Ramsey, A mathematical theory of saving, Econ. J, vol. 38, no.
152, pp. 543559, Dec. 1928.
[5] R. P. Dorfman, P. Samuelson, and R. Solow, Linear Programming and
Economic Analysis. New York: McGraw-Hill, 1958.
[6] L. W. McKenzie, Accumulation programs of maximum utility and
the von Neumann facet, in Value, Capital, and Growth, J. N. Wolfe,
Ed. Chicago, IL: Edinburgh Univ. Press/Aldine, 1968, pp. 353383.
[7] L. W. McKenzie, Turnpike theory, Econometrica, vol. 44, no. 5, pp.
841865, Sep. 1976.
[8] W. A. Brock and A. Haurie, On existence of overtaking optimal trajectories over an infinite time horizon, Math. Oper. Res, vol. 1, pp.
337346, 1976.
[9] A. Leizarowitz, Infinite horizon autonomous systems with unbounded
cost, Appl. Math. Opt, vol. 13, pp. 1943, 1985.
[10] D. A. Carlson, A. B. Hauie, and A. Leizarowitz, Infinite Horizon Optimal Control, 2nd ed. Berlin, Germany: SpringerVerlag, 1991.
[11] D. Q. Mayne, J. B. Rawlings, C. V. Rao, and P. O. M. Scokaert,
Constrained model predictive control: Stability and optimality,
Automatica, vol. 36, no. 6, pp. 789814, 2000.
[12] O. Mangasarian, Nonlinear Programming. Philadelphia, PA: SIAM,
1994.
[13] W. W. Hager, Lipschitz continuity for constrained processes, SIAM
J. Cont. Opt, vol. 17, no. 3, pp. 321338, 1979.
[14] W. Rudin, Principles of Mathematical Analysis, 3rd ed. New York:
McGraw-Hill, 1976.
[15] K. R. Muske and J. B. Rawlings, Linear model predictive control of
unstable processes, J. Proc. Cont., vol. 3, no. 2, pp. 8596, 1993.