Ind. Eng. Chem. Res.

2006, 45, 8565-8574 8565

Robust Model Predictive Control Design for Fault-Tolerant Control of Process

Systems
Prashant Mhaskar*
Department of Chemical Engineering, McMaster UniVersity, Hamilton, ON L8S 4L7, Canada

This work considers the problem of stabilization of nonlinear systems subject to constraints, uncertainty and
faults in the control actuator. We first design a robust model predictive controller that allows for an explicit
characterization of the set of initial conditions starting from where feasibility of the optimization problem
and closed-loop stability is guaranteed. The main idea in designing the robust model predictive controller is
to employ Lyapunov-based techniques to formulate constraints that (a) explicitly account for uncertainty in
the predictive control law, without making the optimization problem computationally intractable, and (b)
allow for explicitly characterizing the set of initial conditions starting from where the constraints are guaranteed
to be initially and successively feasible. The explicit characterization of the stability region, together with the
constraint handling capabilities and optimality properties of the predictive controller, is utilized to achieve
fault-tolerant control of nonlinear systems subject to uncertainty, constraints, and faults in the control actuators.
The implementation of the proposed method is illustrated via a chemical reactor example.

1. Introduction optimization problem that requires minimizing the objective

function (and satisfying the input and state constraints) over all
A control system design for chemical processes needs to
possible realizations of the uncertainty. While the min-max
account for the inherent complexity of the process exhibited in
formulations provide a natural setting within which to address
the form of nonlinear behavior and operational issues such as
this problem, computational problems with these approaches
constraints and uncertainty, as well as to safeguard against
are well-known and stem in part from the nonlinearity of the
eventualities including (but not limited to) faults in the control
model which typically makes the optimization problem non-
actuators. The nonlinear behavior exhibited by most chemical
convex and in part from performing the min-max optimization
processes, together with the presence of constraints on the
over the nonconvex problem. Predictive control formulations
operating conditions (dictated by performance considerations
(including those that do not consider uncertainty) typically owe
or due to limited capacity of control actuators), the presence of
their stabilizing properties to some form of “stability” constraints
modeling uncertainty and disturbances, and the unavailability
that essentially require some appropriate measure of the state
of all the process states as measurements have motivated
to decrease or reach a target set by the end of the horizon. The
significant research work in the area of nonlinear control
initial feasibility of the stability constraint is typically assumed
focusing on these issues;1-10 also, see refs 11 and 12 for
to ensure stability. In refs 21 and 22, a predictive controller is
excellent reviews of results and the references therein.
designed that does not assume but provides guaranteed stabiliza-
One of the control methods suited for handling constraints
within an optimal control setting is model predictive control tion from an explicitly characterized set of initial conditions
(MPC). The main idea in the model predictive control approach under input21 and input and state constraints22 in the absence
is to set up an optimization problem to compute a manipulated of uncertainty. The stability guarantees of existing predictive
input trajectory (subject to constraints on the manipulated input control approaches for nonlinear systems with uncertainty,
variables and state variables) that minimizes a performance however, remain contingent upon the assumption of the initial
index (evaluated using the process model) over a finite horizon. feasibility of the optimization problem (or the assumption of
Only the first “piece” of the manipulated input trajectory is the knowledge of a set of initial conditions starting from where
implemented, however, and the optimization problem is solved the optimization problem is guaranteed to be feasible), and the
again at the next sampling time with the revised state values set of initial conditions, starting from where feasibility of the
(thereby imbuing the approach with the advantages of feedback optimization problem (and, therefore, stability of the closed-
control). Numerous research studies have investigated the loop system) is guaranteed, is not explicitly characterized.
stability properties of model predictive controllers for systems In ref 23 (the robust hybrid predictive control design),
without uncertainty (see, for example, the review paper ref 13). embedding the operation of predictive controllers within the
The problem of analysis and design of predictive controllers stability region of Lyapunov-based bounded robust nonlinear
for uncertain linear systems has been extensively investigated controllers (e.g., see ref 7) is utilized to achieve stability and
(see refs 13 and 14 for surveys of results in this area). explicit characterization of the stability region for the switched
For uncertain nonlinear systems, the problem of robust MPC closed-loop system. The robust hybrid predictive controllers,
design continues to be an area of ongoing research (see, for however, enable implementation of existing predictive controller
example, refs 15-20). Several robust predictive formulations formulations with an explicit characterization of the robust
utilize what we will refer to as the “min-max” approach, where stability region via switching to a fall-back controller. The
the manipulated input trajectory is computed by solving an problem of designing a robust model predictive controller that
by itself (i.e., without switching to a fall-back bounded
* To whom correspondence should be addressed. E-mail: mhaskar@ controller) guarantees stability from an explicitly characterized
mcmaster.ca. Phone: 905 5259140 x23273. Fax: 905 521-1350. set of initial conditions is not considered.
10.1021/ie060237p CCC: $33.50 © 2006 American Chemical Society
Published on Web 05/25/2006
8566 Ind. Eng. Chem. Res., Vol. 45, No. 25, 2006

The occurrence of faults in control actuators adds another that guarantees stability from an explicitly characterized set of
layer of complexity to the problem of controller design for initial conditions. The main idea in the robust model predictive
nonlinear uncertain systems. Note that the stability guarantees controller design is to employ Lyapunov-based techniques to
provided by a controller may no longer hold in the presence of formulate constraints that (a) explicitly account for uncertainty
faults in the control actuators that prevent the implementation in the predictive control law, without making the optimization
of the control action prescribed by the control law, and these problem computationally intractable, and (b) allow for explicitly
faults can have substantial negative ramifications owing to the characterizing the set of initial conditions starting from where
interconnected nature of processes.24 One approach to maintain closed-loop stability is guaranteed. The application of the robust
closed-loop stability would be to use all available (and model predictive controller is demonstrated in section 3.2. The
redundant) control actuators so that even if one of the control explicit characterization of the stability region, together with
actuators fails, the rest can maintain closed-loop stability (the the constraint handling capabilities and optimality properties
reliable control approach; e.g., see ref 25). The use of redundant of the predictive control approach, is utilized in section 4.1 to
control actuators, however, incurs (possibly) preventable opera- not only determine the appropriate backup configuration but to
tion and maintenance costs. These economic considerations also drive the system trajectory into the stability region of a
dictate the use of only as many control loops as is required at candidate backup control configuration and achieve fault-tolerant
a time. To achieve tolerance with respect to faults, control-loop control subject to failures in the primary control configuration.
Finally, in section 4.2, implementation of the proposed method
reconfiguration can be carried out in the event of failure of the
to fault-tolerant control of a chemical reactor example is
primary control configuration. Controller reconfiguration has
demonstrated.
been utilized to achieve fault-tolerance in the context of
aerospace engineering applications (see, e.g., refs 26 and 27)
and in the context of chemical process control, based on the 2. Preliminaries
assumption of a linear process description.28.29 More recently,
reconfiguration has been utilized to achieve fault-tolerant control We consider nonlinear systems with uncertain variables and
for nonlinear systems (see, for example, refs 30 and 31) where input constraints, described by
the main idea is as follows: First, backup control configurations
are identified and an explicit characterization of the stability x̆ ) f(x) + G(x)u + W(x)θ(t), u ∈ U, θ ∈ Q (1)
region associated with each control configuration is obtained
using Lyapunov-based bounded robust nonlinear controllers. where x ∈Rn denotes the vector of state variables, u ∈ Rm
Reconfiguration laws that determine, on the basis of the stability denotes the vector of constrained manipulated inputs, taking
regions, which of the available backup control configurations values in a nonempty convex subset U of Rm, where U ) {u ∈
can preserve closed-loop stability in the event of faults in the Rm: |u| e umax}, ||‚|| is the Euclidean norm of a vector, umax >
primary control configuration are subsequently devised. In ref 0 is the magnitude of input constraints, and θ(t) ) [θ1(t) ‚‚‚
31, the problem of achieving fault-tolerance in the presence of θq(t)]T ∈ Q ⊂ Rq denotes the vector of uncertain (possibly time-
uncertainty was considered, where the robust hybrid predictive varying) but bounded variables taking values in a nonempty
controller23 was used to characterize the stability region under compact convex subset of Rq and f(0) ) 0. The vector function
each control configuration. The approaches in refs 30 and 31 f(x), the matrices G(x) ) [g1(x) ‚‚‚ gm(x)] and W(x) ) [w1(x) ‚‚‚
rely on bounded controller designs to characterize the stability wq(x)], where gi(x) ∈ Rn, i ) 1 ‚‚‚ m, and wi(x) ∈ Rn, i ) 1 ‚‚‚
region under a candidate control configuration and address the q, are assumed to be sufficiently smooth on their domains of
problem of determining which backup configuration can be definition. The notation ||‚||Q refers to the weighted norm,
activated to ensure stability. The fault-tolerant capabilities of defined by ||x||Q2 ) x′Qx for all x ∈ Rn, where Q is a positive
these approaches depend on whether the state, at the time of definite symmetric matrix and x′ denotes the transpose of x.
the failure, resides in the stability region of the backup control The notation Lfh denotes the standard Lie derivative of a scalar
configuration. The problem of ensuring that it is possible to function h(‚) with respect to the vector function f(‚). Throughout
switch to such a backup configuration at the time of occurrence the manuscript, we assume that for any u ∈ U the solution of
of a fault is not addressed, which can be done via guiding the the system of eq 1 exists and is continuous for all t and we
system trajectory toward the stability region of the backup focus on the state feedback problem where measurements of
control configurations. the entire state, x(t), are assumed to be available for all t.
Guiding the system toward a precomputed target set requires
invoking the model predictive control approach that allows for 3. Model Predictive Control of Uncertain Nonlinear
incorporating both state and input constraints in the control Systems
design. A prerequisite to implementing the model predictive
control approach for this purpose is to design a robust model The objective in this section is to design a model predictive
predictive controller that guarantees closed-loop stability from controller that stabilizes nonlinear systems subject to uncertainty
an explicitly characterized set of initial conditions. In summary, and input constraints, as well as allows for an explicit
the discussion above reveals that designing a robust model characterization of the set of initial conditions starting from
predictive controller that can guarantee stability from an where closed-loop stability is guaranteed.
explicitly characterized set of initial conditions is important from 3.1. Robust Model Predictive Control. Consider the system
the perspective of guaranteeing stability for nonlinear uncertain of eq 1 and assume that the uncertain variable term is vanishing
systems in the absence of faults, as well as from the point of in the sense that W(0)θ(t) ) 0 for any θ ∈ Q (see remark 6 for
view of tackling the important problem of achieving tolerance how to handle nonvanishing disturbances), i.e., disturbances that
with respect to faults. Motivated by these considerations, we do not perturb the system from its nominal equilibrium point,
first describe the class of systems considered in section 2 and, and that a robust control Lyapunov function (RCLF32), V, exists.
then in section 3.1, design a robust model predictive controller Consider the receding horizon implementation of the control
 Ind. Eng. Chem. Res., Vol. 45, No. 25, 2006 8567

action computed by solving an optimization problem of the optimization problem continues to be feasible) and that the state
form: trajectories converge to the desired neighborhood of the origin.
Finally, in part 3, we show that the state trajectories, once they
P(x, t): min{J(x, t, u(‚))| u(‚) ∈ S} (2) reach the desired neighborhood of the origin, continue to stay
in the neighborhood.
s.t. x̆ ) f(x) + G(x)u (3) Part 1. Consider some x0 ∈ Ω under receding horizon
implementation of the predictive controller of eqs 2-6, with a
LGV(x)u(t) e LGVubc(x(t)) (4)
prediction horizon T ) N∆, where ∆ is the hold time and 1 e
N < ∞ is the number of the prediction steps. We consider the
where

( )
two constraints that may possibly lead to infeasibility, i.e., the
constraints of eq 4 and the constraint on the manipulated input
R(x) + x(R1(x))2 + (umaxâ(x))4 u. An examination of the constraint of eq 4 reveals that one
ubc ) - (LGV)T (5) possible (and always feasible) solution to this constraint is u )
(â(x)) [1 + x1 + (umaxâ(x)) ]
2 2
ubc. Furthermore, for all x(0) ∈ Ω, ||ubc|| e umax. Therefore, for
all x(0) ∈ Ω, the solution comprising of ubc as the first element
when LGV * 0 and ubc ) 0 when LGV ) 0, where R(x) ) LfV
followed by N - 1 zeros is a feasible solution to the optimization
+ (F||x|| + øθb||LWV||)(||x||/(||x|| + φ)), R1(x) ) LfV + F||x||
problem.
+ øθb||LWV||, â(x) ) ||(LGV)T||, LGV ) [Lg1V ‚‚‚ LgmV], and
Part 2. Having shown the initial feasibility of the optimization
LWV ) [Lw1V ‚‚‚ LwqV] are row vectors, θb is a positive real
problem in part 1, we now show that the implementation of the
number such that |θ(t)| e θb, for all t g 0, and F, ø, and φ are
control action computed by solving the optimization problem
adjustable parameters that satisfy F > 0, ø > 1, and φ > 0; S
of eqs 2-6 guarantees that, for a given d, if we pick a
) S(t, T) is the family of piecewise continuous functions
sufficiently small ∆ (i.e., there exists a ∆* such that if ∆ ∈ (0,
(functions continuous from the right), with period ∆, mapping
∆*]) Ω is invariant under the predictive control algorithm of
[t, t + T] into U. Equation 3 is the “nominal” nonlinear model
eqs 2-6 (this would guarantee subsequent feasibility of the
(without the uncertainty term) describing the time evolution of
optimization problem due to part 1 above) and, then, that if the
the state x. A control u(‚) in S is characterized by the sequence
optimization problem continues to be feasible, then practical
u[j] where u[j] :) u(j∆) and satisfies u(t) ) u[j] for all t ∈ [j∆,
stability (convergence to a desired neighborhood of the origin)
(j + 1)∆). The performance index is given by
for the closed-loop system is achieved.
To this end, we first note that since V(‚) is a continuous
J(x, t, u(·)) ) ∫t t+T[||xu(s; x, t)||Q2 + ||u(s)||R2] ds (6) function of the state, one can find a finite, positive real number,
δ′, such that V(x) e δ′ implies |x| e d. Now, consider a “ring”
where Q and R are positive semidefinite and strictly positive close to the boundary of Ω, described by M :) {x ∈ Rn: (cmax
definite, symmetric matrices, respectively, and xu(s; x, t) denotes - δ) e V(x) e cmax}, for a 0 e δ < cmax, with δ to be
the solution of eq 3, due to control u, with initial state x at time determined later. Let the control action be computed for some
t and T being the specified horizon. The minimizing control x(0) :) x0 ∈ M . The satisfaction of the constraint of eq 4
u0(‚) ∈ S is then applied to the plant over the interval [t, t + ∆) (guaranteed from part 1) ensures that, under implementation of
and the procedure is repeated indefinitely. u 0,
Preparatory to the formalization of the stability and feasibility
properties of the robust MPC, we recall the following result6,7 V̇(x0) ) LfV(x0) + LGV(x0)u0 + LWV(x0)θ(0) e LfV(x0) +
for the control law described by eq 5. Specifically, let Π be the LGV(x0)ubc + LWV(x0)θ(0) ) V̇bc(x0) (7)
set defined by Π(θb, umax) ) {x ∈ Rn: R1(x) e umaxâ(x)} and
assume that Ω :) {x ∈ Rn: V(x) e cmax} ⊆ Π(θb, umax) for Substituting the control law of eq 5 into the system of eq 1, it
some cmax > 0. Then, for any initial condition x0 ∈ Ω, it can be can be shown that, for all x(0) ∈ Ω and ||θ(t)|| e θb,
shown that there exists a positive real number * such that if
φ/(ø - 1) < *, the states of the closed-loop system of eqs 1 V̇bc(x) ) LfV + LGVubc + LWVθ(t) e -F*V(x) (8)
and 5 satisfy x(t) ∈ Ω ∀ t g 0 and the origin of the closed-loop
system is asymptotically stable. The stability properties of the where F* > 0. This, together with eq 7, yields V̇(x0) e -F*V(x0).
predictive controller design follow from ensuring that V̇ under Furthermore, if the control action is held constant until a time
the predictive controller is at least as negative as that under the ∆**, where ∆** is a positive real number (u(t) ) u(x0) :) u0
implementation of the controller of eq 5 (expressed in eq 4); ∀ t ∈ [0, ∆**]), then, ∀ t ∈ [0, ∆**],
feasibility of the optimization problem and stability properties
of the closed-loop system under the predictive controller are V̇(x(t)) ) LfV(x(t)) + LGV(x(t))u0 + LWV(x(t))θ(t) )
formalized in theorem 1 below. LfV(x0) + LGV(x0)u0 + LWV(x0)θ(0) + (LfV(x(t)) -
Theorem 1. Consider the constrained system of eq 1 under
the MPC law of eqs 2-6. Then, giVen any positiVe real number LfV(x0)) + (LGV(x(t))u0 - LGV(x0)u0) + LWV(x(t))θ(t) -
d, there exists a positiVe real number ∆* such that if ∆ ∈ (0, LWV(x0)θ(0) (9)
∆*] and x(0) :) x0 ∈ Ω, then the optimization problem of eqs
2-6 is guaranteed to be initially and successiVely feasible, x(t) Since x0 ∈ M ⊆ Ω, LfV(x0) + LGV(x0)u0 + LWV(x0)θ(0) e
∈ Ω ∀ t g 0 and limtf∞ sup ||x(t)|| e d. -F*V(x0). By definition, for all x0 ∈ M, V(x0) g cmax - δ;
Proof of Theorem 1. The proof of this theorem is divided therefore, LfV(x0) + LGV(x0)u0 + LWV(x0)θ(0) e -F*(cmax -
in three parts. In the first part, we show that, for all x0 ∈ Ω, the δ). Since the function f(‚) and the elements of the matrices
optimization problem of eqs 2-6 is guaranteed to be initially G(‚) and W(‚) are continuous, ||u(t)|| e umax, ||θ(t)|| e θmax,
feasible. We then show that there exists a ∆* such that if ∆ ∈ and M is bounded, and then, one can find, for all x0 ∈ M and
(0, ∆*] then Ω is invariant under receding horizon implementa- a fixed ∆**, a positive real number K1, such that ||x(t) - x0||
tion of the predictive controller of eqs 2-6 (implying that the e K1∆** for all t e ∆**.
8568 Ind. Eng. Chem. Res., Vol. 45, No. 25, 2006

Since the functions LfV(‚), LGV(‚), and LWV(‚) are lipschitz, used in the predictive controller is the nominal (without the
then, given that |x(t) - x0| e K1∆**, x0 ∈ Ω, and ||θ(t)|| e uncertainty term) nonlinear model. However, negative definite-
θmax, we have that one can find positive real numbers K2, K3, ness of V̇ during the first time step is ensured by making the V̇
and K4 such that |LfV(x(t)) - LfV(x0)| e K3K1∆**, |LGV(x(t))u0 sufficiently negative for the initial condition such that V̇
- LGV(x0)u0| e K2K1∆**, and |LWV(x(t))θ(t) - LWV(x0)θ(0)| continues to remain negative during the first time step for the
e K4K1∆**. Using these inequalities in eq 9, we get nonlinear uncertain system. The feasibility of this constraint is
ascertained via the use of the control law (and its associated
V̇(x(t)) e -F*(cmax - δ) + (K1K2 + K1K3 + K1K4)∆** region of stability) of eq 5. Several points need to be understood
(10) with regard to this particular choice of the constraint, and these
are illustrated via the nonsuitability of some other, similar-
For a choice of ∆** < (F*(cmax - δ) - )/(K1K2 + K1K3 + looking, constraints in remark 2 below.
K1K4) where  is a positive real number such that Remark 2. One possible alternative to the constraint of eq
4, for example, could have been to require the Lyapunov
 < F*(cmax - δ) (11) function to decrease at the end of the horizon (similar to the
contractive Lyapunov-based predictive controller of ref 33). In
we get that V̇(x(t)) e - < 0 for all t e ∆**. This implies that, the presence of uncertainty, it would necessitate requiring the
given δ′, if we pick δ such that cmax - δ < δ′ and find a
Lyapunov function to decay under all possible realizations of
corresponding value of ∆** then if the control action is
the uncertainty, making it a min-max type of optimization
computed for any x ∈ M and the “hold” time is less than ∆**,
problem. Note that, in the presence of contractive-type con-
we get that V̇ remains negative during this time, and therefore,
straints, the set of initial conditions from where the optimization
the state of the closed-loop system cannot escape Ω (since Ω
problem is guaranteed to be feasible is not explicitly character-
is a level set of V). This in turn implies successive feasibility
ized even in the absence of uncertainty. Therefore, even if one
of the optimization problem for all initial conditions in M and
were to implement such a constraint disregarding the compu-
that, for any initial condition, x0, such that δ < V(x0) e cmax,
tational complexity associated with solving min-max optimiza-
we have that V(x(t + ∆)) < V(x(t)). All trajectories originating
tion problems, nothing could be said about the guaranteed (initial
in Ω, therefore, converge to the set defined by Ωf :) {x ∈ Rn:
and successive) feasibility of such an optimization problem.
V(x) e cmax - δ}.
Another choice could have been, for instance, to require the
Part 3. We now show the existence of ∆′ such that, for all
Lyapunov function to decrease only during the first time step.
x0 ∈ Ωf :) {x ∈ Rn: V(x0) e cmax - δ}, we have that x(∆) ∈
While, in the absence of uncertainty, such a constraint can be
Ωu :) {x0 ∈ Rn: V(x0) e δ′}, where δ′ < cmax, for any ∆ ∈ (0,
imposed and guaranteed to be initially and successively feasible
∆′].
from an explicitly characterized set of initial conditions (see
Consider ∆′ such that
refs 21 and 22), the presence of uncertainty brings in the
following complications: if one were to persist with requiring
δ′ ) max V(x(τ)) (12)
V(x0)ecmax-δ,u∈U,θ∈Qτ∈[0,∆′] the Lyapunov function to decay during the first time step
(without requiring the constraint to be satisfied over all possible
Since V is a continuous function of x and x evolves continuously realizations of the uncertainty), such a constraint could be shown
in time, then, for any value of δ < cmax, one can choose a to be initially feasible from an explicitly characterized set of
sufficiently small ∆′ such that eq 12 holds. Let ∆* ) min- initial conditions; successive feasibility (and, therefore, stability),
{∆**, ∆′}. We now show that, for all x0 ∈ Ωu and ∆ ∈ (0, however, would not be guaranteed (due to the presence of
∆*], x(t) ∈ Ωu for all t g 0. uncertainty). The other option could be to compute by how much
For all x0 ∈ Ωu ∩ Ωf, by definition, x(t) ∈ Ωu for 0 e t e ∆ the V̇ (without considering the uncertainty) needs to be negative
(since ∆ e ∆′). For all x0 ∈ Ωu\Ωf (and therefore x0 ∈ M ), V̇ during the first time step to make sure that the V̇ for the system
< 0 for 0 e t e ∆ (since ∆ e ∆**). Since Ωu is a level set of with uncertainty also remains negative. While such a constraint
V, then x(t) ∈ Ωu for 0 e t e ∆. Either way, for all initial would guarantee stability for the uncertain system (without
conditions in Ωu, x(t) ∈ Ωu for all future times. having to solve a min-max problem), initial feasibility of the
In summary, we showed that (1) for all x(0) ∈ Ω, the optimization problem could not be guaranteed. In summary, the
optimization problem is guaranteed to be feasible, (2) the constraint of eq 4 represents one choice of stability constraint
optimization problem continues to be feasible and x(t) ∈ Ω ∀ that achieves both stability and an explicit characterization of
t g 0, all state trajectories originating in Ω converge to Ωu, the feasibility region, without requiring min-max computations.
and (3) all state trajectories originating in Ωu stay in Ωu, i.e., Remark 3. Regarding the choice of the control law of eq 5,
x(t) ∈ Ω ∀ t g 0 and limtf∞ sup ||x(t)|| e d. This completes we note that the problem of designing control laws that
the proof of theorem 1. guarantee stability in the presence of input constraints has been
Remark 1. The key step in the design of the predictive extensively studied (see, for example, refs 2, 7, 34, and 35).
controller is the appropriate choice of the stability constraint The bounded robust controller design of eq 5, proposed in ref
that is guaranteed to be initially and successively feasible from 7 (inspired by the results on bounded control in ref 2 for systems
an explicitly characterized set of initial conditions, while not without uncertainty) is an example of a controller design that
requiring the optimization to be performed over all possible (1) guarantees robust stability in the presence of constraints,
realizations of the uncertainty. The constraint of eq 4 is one and (2) allows for an explicit characterization of the closed-
such constraint. Equation 4 requires the control action to be loop stability region. The stability guarantees of the predictive
computed such that it renders the time derivative of the controller of theorem 1 are not limited to this particular choice
Lyapunov function, evaluated at the current state, to be more of controller, and any other robust controller that satisfies 1
negative than the time derivative of the Lyapunov function that and 2 above can be used to formulate the constraint of eq 4.
one would achieve if one were to implement the Lyapunov- Furthermore, referring to the control law of eq 5 (and this holds
based bounded robust controller of eq 5. Note that the model for other robust control laws as well), it is important to note
 Ind. Eng. Chem. Res., Vol. 45, No. 25, 2006 8569

that a general procedure for the construction of RCLFs for one can come up with a bound on the implement and hold time
nonlinear systems of the form of eq 1 is currently not available. such that the convergence to the desired neighborhood of the
Yet, for several classes of nonlinear systems that arise commonly origin is achieved. Similarly, under nonvanishing disturbances
in the modeling of engineering applications, it is possible to with discrete implementation, the loss in convergence to the
exploit system structure to construct RCLFs. For example, for origin is due to two factors: one due to the nonvanishing
feedback linearizable systems, quadratic Lyapunov functions disturbance and the other due to discrete implementation;
can be chosen as candidate RCLFs and made RCLFs with an however, both can be made as small as desired via the
appropriate choice of the function parameters based on the appropriate choice of parameters.
process parameters (see, for example, ref 32). Also, for nonlinear Remark 7. One of the time-consuming calculations in
systems in strict feedback form, backstepping techniques can computing solutions to nonlinear optimization problems subject
be employed for the construction of RCLFs.32 to constraints is finding a feasible solution. In this light, the
Remark 4. Note that if the problem at hand was stabilization fact that the control action computed by the control law of eq
alone, one could have used a bounded robust control design to 5 provides a feasible initial guess to the optimization problem
achieve stability and characterize the stability region. The positively impacts the computational complexity of the opti-
analytical bounded robust control designs, however, do not have mization problem. Furthermore, the fact that the stability
any mechanism for accounting for the performance criteria and guarantees provided by the optimization problem are indepen-
are not designed to be optimal with respect to a prespecified dent of the horizon used in the predictive controller can further
cost functional. Using a predictive control approach allows the alleviate the computational complexity by allowing for reduction
computation of the control action in a way that accounts for in the size of the optimization problem, without concern for
performance objectives, while, at the same time, the proposed loss of stability properties. Note, also, that the stabilizing
predictive controller does not sacrifice on the stability guarantees properties of the predictive controller are achieved via a feasible
and explicit characterization of the stability region. More solution and do not need the solution to be the optimal solution.
importantly, the flexibility associated with specifying the It is also important to keep in mind that only the nominal model
objective function in the predictive control approach allows for (without the uncertainty term) is used in specifying the objective
“guiding” the state trajectory in a desirable fashion; this function. This implies that the solution obtained is not optimal
particular strength makes the proposed predictive controller well with respect to the objective function in the presence of
suited for handling faults in the control actuators, as described uncertainty. More importantly, given that the optimization
in the next section. problem (for nonlinear systems) is typically a nonconvex
optimization problem implies that solvers would only find
Remark 5. Note that the objective behind the proposed robust
locally optimum solutions. Both of these factors, however, do
predictive control design is not that of providing robust stability
not impact negatively on the stabilizing properties of the
guarantees to various predictive controllers (achieved in ref 23
controller design since stability is dependent on the feasibility
via switching between the predictive controller and a backup
of the optimization problem (not on finding an optimal solution).
controller) but that of formulating a robust predictive controller
Furthermore, the set of initial conditions starting from where
that can guarantee closed-loop stability from an explicitly
the optimization problem is guaranteed to be feasible is explicitly
characterized set of initial conditions, without requiring any
characterized. The ability to search for a feasible solution that
switching to a fall-back controller. Note also that the fact that
is better with respect to the objective function (compared to
only practical stability (convergence to a desired neighborhood
the initial guess), however, has important implications as
of the origin) is achieved under the predictive controller is not
evidenced in the fault-tolerant capabilities of the predictive
a limitation of the approach (or the way the constraints are
controller design formulated in section 4.1.
formulated) but is due to the discrete (implement and hold) 3.2. Application to a Chemical Reactor. In this section, we
nature of the control action; even if an analytical bounded robust use a chemical reactor example (also used in ref 23) to
controller was used in an implement and hold fashion, it could demonstrate the constraint and uncertainty handling capability
also achieve only practical stability with the size of the desired of the robust predictive controller design. To this end, consider
neighborhood of the origin dictating the maximum allowable the following model of an irreversible elementary exothermic
hold time. reaction of the form
Remark 6. While, for the sake of simplicity, the results in
the present work have been presented under the assumption of k
A 98 B
vanishing disturbances, modifying the robust predictive control
design to account for nonvanishing disturbances is relatively in a well-mixed continuous stirred tank reactor:
straightforward. The line of reasoning is as follows: Let d1
denote the neighborhood of the origin that the closed-loop state dCA -E
is required to converge to, and then, pick a d2 < d1 and
parameters in the control design of eq 5 to achieve convergence
VR
dt
) F(CA0 - CA) - k0 exp ( )
C V
RT A R
dT ∆H -E Q
to the d2 neighborhood of the origin under continuous imple-
mentation. This, in turn, implies that, under continuous imple-
VR
dt
) F(TA0 - T) -
Fcp
k0 exp
RT ( )
CAVR +
Fcp
(13)
mentation of the robust predictive controller (i.e., with ∆ ) 0),
one can still achieve convergence to the same neighborhood of where CA denotes the concentration of species A, T and VR
the origin. Finally, compute ∆*, such that under sample and denote the temperature and volume of the reactor, respectively,
hold implementation, with a sampling time less than ∆*, Q denotes the rate of heat input to the reactor, k0, E, and ∆H
convergence to the neighborhood d1 is achieved. The reasoning denote the preexponential constant, the activation energy, and
can also be understood as follows: under vanishing disturbances, the enthalpy of the reaction, respectively, and cp and F denote
given an acceptable “loss” in convergence due to discrete the heat capacity and density of the fluid in the reactor,
implementation (i.e., instead of requiring convergence to the respectively. Table 1 lists the steady-state values and process
origin, to require convergence to a neighborhood of the origin), parameters. The control objective is to stabilize the reactor at
8570 Ind. Eng. Chem. Res., Vol. 45, No. 25, 2006

Table 1. Process Parameters and Steady-State Values

VR ) 0.1 m3 R ) 8.314 J/(mol K)
CA0s ) 1.0 kmol/m3 TA0s ) 310.0 K
∆Hn ) -4.78 × 104 J/mol k0 ) 72 × 109 1/min
E ) 8.314 × 104 J/mol cp ) 0.239 J/(g K)
F ) 1000.0 kg/m3 F ) 0.1 m3/min
TRs ) 395.3 K CAs ) 0.57 kmol/m3

the (open-loop) unstable equilibrium point by manipulating both

the rate of heat input/removal and the inlet reactant concentra-
tion. Defining x1 ) CA - CAs, x2 ) T - Ts, u1 ) CA0 - CA0s,
u2 ) Q, θ1(t) ) TA0 - TA0s, and θ2(t) ) ∆H - ∆Hn, where the
subscript s denotes the steady-state value and ∆Hn denotes the
nominal value of the heat of reaction, the process model of eq
13 can be cast in the form of eq 1. In all simulation runs, θ1(t)
) θ0 sin(3t), where θ0 ) 0.08TA0s, θ2(t) ) 0.5(-∆Hn), and the
manipulated input constraints were |u1| e 1.0 kmol/m3 and |u2|
e 92 kJ/s.
In the simulation example, we use a quadratic Lyapunov
function in the robust model predictive design and the param-
eters ø ) 1.01, φ ) 0.0001, and F ) 0.01 were chosen in the
control law of eq 5, with ∆ ) 0.005 min and the number of
prediction steps, N ) 6, to drive the closed-loop state trajectory
to the desired neighborhood Ωb (shown in Figure 1) of the origin
(verified via simulations); the robust feasibility (and stability)
region Ω was explicitly characterized (see Figure 1). Note that Figure 2. Evolution of the closed-loop state profiles under the predictive
controller requiring V̇ < 0 (dashed line) and V(t + ∆) < V(t) (dotted line)
the stability guarantees of the proposed predictive controller do and under the proposed predictive controller of theorem 1 (solid line).
not depend on the specific value of the horizon (or on being
“large” enough). Note also that while a simple choice of the
Lyapunov function may not yield the best estimates of the
stability region, possible conservatism of the stability region
obtained can only be gauged against the specific needs of a
particular system under consideration by analyzing whether the
given estimate of the stability region covers a “wide enough”
set of initial conditions. No meaningful comparison can be made
against the largest possible set of initial conditions starting from
where stability is achievable because such characterization is
impossible for nonlinear uncertain systems. By providing an
estimate of the stability region for nonlinear uncertain systems
under predictive control, the proposed approach provides an
opportunity to judge the stabilization capability of the predictive
controller design and modify it (if needed) via choosing a
different Lyapunov function. Note also that possibly larger
estimates of the stability region can be computed using
constructive procedures such as Zubov’s method or by using a
combination of several Lyapunov functions. The set of nonlinear
ordinary differential equations (ODEs) was integrated using the
MATLAB solver, ODE15s, and the optimization problem in
the robust MPC was solved using the MATLAB nonlinear
constrained optimization solver, fmincon.
Following up on remark 2, we first investigate the possibility
of using stability constraints other than the one of theorem 1,
Figure 3. Evolution of the closed-loop input profiles under the predictive
controller requiring V̇ < 0 (dashed line) and V(t + ∆) < V(t) (dotted line)
and under the proposed predictive controller of theorem 1 (solid line).

and specifically, we first consider using a stability constraint

that only requires V̇ to be negative. As can be seen from the
dotted lines in Figures 1 and 2, starting from an initial condition
within Ω, the optimization problem continues to be initially and
successively feasible; however, due to not accounting for the
presence of uncertainty, convergence to the desired neighbor-
hood of the origin is not achieved (the corresponding manipu-
lated input profiles are shown by dotted lines in Figure 3). As
Figure 1. Evolution of the closed-loop state trajectory under the predictive
another option, we consider a stability constraint that requires
controller requiring V̇ < 0 (dashed line) and V(t + ∆) < V(t) (dotted line) V(t + ∆) to be less than V(t). Once again, starting from an initial
and under the proposed predictive controller of theorem 1 (solid line) condition within Ω, the optimization problem continues to be
 Ind. Eng. Chem. Res., Vol. 45, No. 25, 2006 8571

feasible; however, as before, the desired convergence to the function of the “target” backup configuration; see remark 8).
origin is not achieved (dashed lines in Figures 1 and 2). Finally, Theorem 2 below formalizes the result.
if the robust model predictive control design of theorem 1 is Theorem 2. Consider the closed-loop system of eq 14 under
used, the optimization problem not only continues to be initially the primary control configuration, k(0) ) i, under the predictiVe
and successively feasible, irrespective of the choice of the control law of the following form:
horizon (i.e., the number of prediction steps N), but also achieves
convergence to the desired neighborhood of the origin (solid Pi(x, t): min{Ji(x, t, u(‚))| u(‚) ∈ S} (15)
lines in Figures 1 and 2).
s.t. x̆ ) f(x) + Gi(x)ui (16)
4. Achieving Fault-Tolerance via Robust Model
Predictive Control LGiV(x)u(t) e LGiVubc(x(t)) (17)
Having designed the robust model predictive control in the
previous section, we utilize the control design to address the with
problem of implementing fault-tolerant control in the presence
of uncertainty and faults in the control actuators. The main idea Ji(x, t, u(‚)) ) ∫t t+T∫[||xu(s; x, t)||Q2 + ||u(s)||R2] ds +
behind the achievement of fault-tolerant control is to characterize Vl(x(t + ∆)) (18)
the stability region under each candidate control configuration
and to drive the state trajectory in a way that it enters the stability Let Tfi be the earliest time that a fault occurs, and then, the
region of candidate backup configurations, allowing the pos- following switching rule,
sibility of activating a backup control configuration to achieve
stabilization in the event of a failure in the primary control
configuration. The explicit characterization of the stability region
and the ability to guide the system trajectory in a desired fashion
k(t) ) {
i, 0 e t < Tfi
j, t g Tfi, x(Tfi) ∈ Ωj } (19)

make the nonlinear robust predictive controller developed in

the previous section the ideal choice for the control law under with the control action being computed using the predictiVe
each candidate control configuration. Specifically, we consider controller of theorem 1 under configuration j guarantees that
systems represented by the following state-space description: x(t) ∈ ΩU ∀ t g 0 and limtf∞ sup ||x(t)|| e dmax.
Proof of Theorem 2. We consider the two possible cases:
x̆ ) f(x) + Gk(x)(uk + mk) + Wk(x)θk(t), uk ∈ Uk, θk ∈ Qk first, if no switching occurs and, second, if a switch occurs at
(14) a time Tfi.
Case 1. The absence of a switch implies k(t) ) i ∀ t g 0.
where x(t) ∈ Rn denotes the vector of process state variables, Note that the only difference between the control law of eqs
uk(t) ∈ [-ukmax, ukmax] ⊂ R denotes the constrained manipulated 15-17 and that of eqs 2-4 is in the objective function. The
input associated with the kth control configuration, mk(t) ∈ R stability properties of the control law depend on the stability
denotes the fault in the kth control configuration, and K ) 1, constraints, and since they are the same in both formulations,
2, ..., P denotes the set of P available control configurations. the predictive controller of eqs 15-17 can also shown to be
For each value that k assumes in K, the process is controlled robustly stabilizing with the same stability region. Since x(0) ∈
via a different manipulated input which defines a given control Ωi and control configuration i is implemented for all times in
configuration. this case, we have that x(t) ∈ Ωi ∀ t g 0 and limtf∞ sup ||x(t)||
4.1. Fault-Tolerant Model Predictive Control of Uncertain e di. Finally, since Ωi ⊆ ΩU and di e dmax, we have that x(t)
Systems. The problem that we address is that of ensuring closed- ∈ ΩU ∀ t g 0 and limtf∞ sup ||x(t)|| e dmax.
loop stability in the event of a failure in the primary control Case 2. At time Tfi, the reconfiguration takes place and a
configuration (note that this work does not focus on designing control configuration j for which x(Tfi) ∈ Ωj is activated. From
a fault-detection and isolation filter but assumes the availability this time onward, since configuration j is implemented in the
of this information; for a recent work in this direction, see ref closed-loop system for all times under the predictive controller
30). In the rest of the section, we propose an optimization-based of theorem 1 and since x(Tfi) ∈ Ωj, we have that x(t) ∈ Ωl ∀ t
fault-tolerant control structure which (1) recognizes that a g 0 and limtf∞ sup ||x(t)|| e dl. As in case 1, since Ωl ⊆ ΩU
backup control configuration needs to be activated such that and dl e dmax, we have that x(t) ∈ ΩU ∀ t g 0 and limtf∞ sup
the state of the closed-loop system at the time of the failure ||x(t)|| e dmax. This completes the proof of theorem 2.
resides in its stability region (i.e., closed-loop stability may not Remark 8. The fault-tolerant controller is implemented as
be achieved by simply activating any backup configuration) and follows.
(2) implements the robust predictive controller of section 3.1 (1) Given the nonlinear process of eq 14, identify the available
to drive the system trajectory to reside in the stability region of control configurations k ) 1, ..., P, and for each control
some backup configuration should a failure occur. configuration, design the robust predictive controllers of eqs
To this end, consider the nonlinear uncertain system of eq 2-4 and calculate an estimate of the stability regions Ωk, k )
14, where under each control configuration, k, robust model 1, ..., P.
predictive controllers of the form of eqs 2-6 have been designed (2) Given any x0 ∈ Ωi, compute Vj(x0)/cmax j , for all j * i, i.e,
to drive the state trajectory to a desired neighborhood of the compute an estimate of the “distance” of the initial state from
origin, dk, and the stability regions under each control config- the boundary of the stability region of the candidate control
uration, Ωj, j ) 1, ..., P, have been explicitly characterized. Let configurations. Note that Vj(x0)/cmax
j e 1 implies that the initial
dmax ) maxj)1,...,P dj, where dj was defined in theorem 1, ∆ e condition is in the stability region of the jth control configu-
∆/min ≡ minj)1,...,P ∆/j , and ΩU ) ∪j)1P
Ωj. Let k(0) ) i for some ration. If the initial condition is outside the stability region,
i ∈ K, x(0) :) x0 ∈ Ωi, and l * i be such that Dl ) minj∈K Vj(x0)/cmax
j > 1 and the bigger the ratio is, the farther the initial
Vj(x0)/cmax
j (Vl, therefore, denotes the value of the Lyapunov condition is from the stability region of a candidate backup
8572 Ind. Eng. Chem. Res., Vol. 45, No. 25, 2006

configuration. Denote the candidate backup configuration which

is closest to the initial conditions as the target configuration l.
(3) Initialize the closed-loop system under the robust predic-
tive controller of eqs 15-17.
(4) At any time Tfi that a fault occurs, out of the available
backup configurations check whether the state of the closed-
loop system resides in the stability region estimate under the
candidate control configuration (i.e., to check if x(Tfi) ∈ Ωj).
Pick the control configuration for which the state of the closed-
loop system resides in its stability region and apply the robust
predictive controller using this control configuration to achieve Figure 4. Evolution of the closed-loop state trajectory under the primary
configuration in the absence of faults (dashed line) and in the presence of
closed-loop stability. a fault but without accounting for it in the design of the controller for the
Remark 9. The choice of lth configuration, for which primary configuration (dotted line) and under the proposed fault-tolerant
Vl(x0)/cmax
l is the minimum, and its use within the objective controller of theorem 2 (solid line).
function and in the switching logic can be explained as
follows: Note that Vl(x0)/cmax is only one measure of how far/ configuration, similar performance considerations may be
l
close the initial condition is from the stability region of a employed in picking the backup configuration that is imple-
candidate control configuration; this measure being the smallest mented in the closed-loop to achieve fault-tolerance.
for a given candidate control configuration does not provide Remark 12. Note that the proposed fault-tolerant control
any guarantees whether this candidate control configuration can structure differs from other reconfiguration-based fault-tolerant
be reached fastest under the implementation of the currently control methods (e.g., ref 30) in the control designs it uses under
active control configuration (and subject to uncertainty). Hence, each control configuration and, as a result, in the fault-tolerant
no claim is made that this is the “best” choice of the measure capability. Specifically, the robust predictive control design
and that this is the candidate control configuration that should provides an explicit characterization of the robust stability region
be used within the objective function, but it is simply a choice. under each backup configuration (this characterization deter-
Another option could be to change the optimization problem in mines which backup control configuration should be imple-
a way that the control action is computed to minimize the time mented in the closed-loop) and the fault-tolerant controller uses
that it takes for the closed-loop state to reach the stability region the guiding capabilities of the predictive control approach in
of the backup configuration. Solving such an optimization taking precautionary action (via driving the state trajectory into
problem online, however, would be an incredibly difficult task the stability region of the backup configurations) to enable fault-
and impossible to implement online if the uncertainty were also tolerant control, as opposed to only Verifying whether fault-
to be explicitly taken into consideration. Another seemingly tolerant control can be implemented.
attractive possibility could be to incorporate, as a constraint, 4.2. Application to a Chemical Reactor. Consider once
that the closed-loop state should enter the stability region of a again the chemical reactor of section 3.2, and let the robust
candidate backup configuration by a certain time. Computing predictive controller be designed as before. The problem we
such a time, however, would once again be a difficult task for consider in this section is the following: if, at some time during
nonlinear systems with uncertainty. As opposed to all of these operation, the control actuator manipulating the heat input to
alternatives, the option used, that of incorporating the require- the reactor Q were to fail, how could we use a backup control
ment in the objective function, provides a simple tool, that may configuration to maintain closed-loop stability. For the purpose
reduce the time that the closed-loop state trajectory takes to of illustration, let the backup control configuration be one that
enter the stability region of the backup configuration and may uses the inlet concentration (as before) and temperature as the
result in enabling fault-tolerant control (see the simulation manipulated variables, with constraints |u2| e 100K.
example for a demonstration). As in section 3.2, quadratic Lyapunov functions were used
Remark 10. Note that the proposed controller owes its fault- in the robust model predictive designs under configurations 1
tolerant properties to both nonlinear control tools and the and 2, and for configuration 2, the parameters ø ) 1.5, φ )
flexibility provided by the optimization-based predictive control 0.0001, and F ) 0.05 were chosen in the control law of eq 5,
approach. In setting up the objective function, the formulation with ∆ ) 0.005 and the number of prediction steps, N ) 6, to
of the penalty on the Lyapunov function value of a candidate drive the closed-loop state trajectory to the neighborhood Ωb
backup configuration (not of the currently active configuration) (shown in Figure 4); the robust feasibility (and stability) regions
is facilitated by the characterization of the stability regions in Ωp (the stability region of the primary configuration) and Ωs
the form of level sets of the Lyapunov functions, which in turn (the stability region of the secondary configuration) were
exploits a Lyapunov-based analysis. The ability to incorporate explicitly characterized (see Figure 4).
these considerations in the objective function is possible due to The first simulation run, shown by the dashed lines in Figures
the use of the optimization-based approach and allows for the 4-6, demonstrates the scenario where the primary configuration
controller to take preventive measures (via guiding the state does not fail and, starting from an initial condition within the
trajectory inside the stability region of the backup control stability region of the primary control configuration, closed-
configuration) to achieve fault-tolerant control. loop stability is achieved. The second run (shown by dotted
Remark 11. Note that, other than proximity of the initial lines in Figures 4-6) considers, once again, the implementation
condition to the stability region, additional considerations, such of the predictive control design of theorem 1 (not of theorem
as ease of operation or cost of operation, could be incorporated 2), wherein the possibility of a fault is not taken into
in picking the “preferred” backup configuration (i.e., the one consideration; however, a fault occurs in the primary configu-
which is incorporated in the cost function in theorem 2). Note ration at t ) 0.075 min. At the time of the fault (at t ) 0.075
also that in the event that, at the time of the failure, the closed- min), the state of the closed-loop system is outside the stability
loop state resides in the stability region of more than one control region of the backup control configuration, and switching to
 Ind. Eng. Chem. Res., Vol. 45, No. 25, 2006 8573

the fault once again takes place at the same time (at t ) 0.075
min) and the system is initialized from the same initial condition,
switching to a backup control configuration is able to achieve
closed-loop stability. Note that, at the time of the fault, the state
of the closed-loop system is still outside the stability region of
the backup control configuration, but it is closer than it is in
the first simulation run. The simulation result also underscores
an important point that the stability region computed is only an
estimate of the set of points starting from where the robust
predictive controller can guarantee closed-loop stability. Note
that it provides guarantees for all initial conditions within the
set; however, it is also possible to achieve closed-loop stability
from points outside this set.
In summary, implementing the robust predictive control
design that accounts for the occurrence of faults guides the
system trajectory in a way that it goes faster toward the stability
region of the candidate backup configuration and, together with
verifying the presence of the closed-loop state in the stability
region of the backup configuration (necessary to achieve fault-
tolerance), incorporates appropriate penalties in the objective
function that increase the chances of achieving fault-tolerant
control.
Figure 5. Evolution of the closed-loop state profiles under the primary
configuration in the absence of faults (dashed line) and in the presence of 5. Conclusions
a fault but without accounting for it in the design of the controller for the
primary configuration (dotted line) and under the proposed fault-tolerant
controller of theorem 2 (solid line). In this work, we considered the problem of stabilization of
nonlinear systems subject to uncertainty and constraints and
the backup configuration does not achieve closed-loop stability. designed a robust model predictive controller that guarantees
Finally, when the fault-tolerant controller of theorem 2 is stabilization from an explicitly characterized set of initial
implemented, the control action computed by the controller in conditions. We then utilized the optimality and constraint
the primary configuration guides the system trajectory faster handling properties of the proposed predictive controller to
toward the stability region of the backup control configuration design a fault-tolerant controller that achieves fault-tolerance
(solid lines in Figures 4-6). Compared to the control action through controller reconfiguration. The fault-tolerant controller
implemented when not accounting for the possibility of the uses the knowledge of the stability regions of the backup control
occurrence of a fault, the control action computed by the fault- configurations to guide the state trajectory within the stability
tolerant controller of theorem 2 uses up as much control action region of the backup control configurations to enhance the fault-
as is available; see the inset of Figure 6. As a result, even though tolerance capabilities. The implementation of the proposed

Figure 6. Evolution of the closed-loop input profiles under the primary configuration in the absence of faults (dashed line) and in the presence of a fault
but without accounting for it in the design of the controller for the primary configuration (dotted line) and under the proposed fault-tolerant controller of
theorem 2 (solid line).
8574 Ind. Eng. Chem. Res., Vol. 45, No. 25, 2006

robust predictive controller as well as the application to fault- (19) Wang, Y. J.; Rawlings, J. B. A new robust model predictive control
tolerant control was demonstrated via a chemical reactor method I: theory and computation. J. Process Control 2004, 14, 231.
(20) Sakizlis, V.; Kakalis, N. M. P.; Dua, V.; Perkins, J. D.; Pistiko-
example.
poulos, E. N. Design of robust model-based controllers via parametric
programming. Automatica 2004, 40, 189.
Literature Cited (21) Mhaskar, P.; El-Farra, N. H.; Christofides, P. D. Predictive control
of switched nonlinear systems with scheduled mode transitions. IEEE Trans.
(1) Kravaris, C.; Palanki, S. Robust nonlinear state feedback under Autom. Control 2005, 50, 1670.
structured uncertainty. AIChE J. 1988, 34, 1119. (22) Mhaskar, P.; El-Farra, N. H.; Christofides, P. D. Stabilization of
(2) Lin, Y.; Sontag, E. D. A universal formula for stabilization with nonlinear systems with state and control constraints using Lyapunov-based
bounded controls. Syst. Control Lett. 1991, 16, 393. predictive control. Syst. Control Lett. 2006, in press.
(3) Valluri, S.; Soroush, M. Analytical control of SISO nonlinear
(23) Mhaskar, P.; El-Farra, N. H.; Christofides, P. D. Robust hybrid
processes with input constraints. AIChE J. 1998, 44, 116.
predictive control of nonlinear systems. Automatica 2005, 41, 209.
(4) Agamennoni, O.; Figueroa, J. L.; Palazoglu, A. Robust controller
design under highly structured uncertainty. Int. J. Control 1998, 70, 721. (24) Ydstie, E. B. New vistas for process control: Integrating physics
(5) Kapoor, N.; Daoutidis, P. Stabilization of nonlinear processes with and communication networks. AIChE J. 2002, 48, 422.
input constraints. Comput. Chem. Eng. 2000, 24, 9. (25) Yang, G. H.; Wang, J. L.; Soh, Y. C., Reliable H∞ control design
(6) El-Farra, N. H.; Christofides P. D. Integrating robustness, optimality, for linear systems. Automatica 2001, 37, 717.
and constraints in control of nonlinear processes. Chem. Eng. Sci. 2001, (26) Patton, R. J. Fault-tolerant control systems: The 1997 situation.
56, 1841. In Proceedings of the IFAC Symposium SAFEPROCESS 1997; Hull: United
(7) El-Farra, N. H.; Christofides, P. D. Bounded robust control of Kingdom, 1997; p 1033.
constrained multivariable nonlinear processes. Chem. Eng. Sci. 2003, 58, (27) Zhou, D. H.; Frank, P. M. Fault diagnostics and fault tolerant
3025. control. IEEE Trans. Aerospace Electron. Syst. 1998, 34, 420.
(8) Mhaskar, P.; El-Farra, N. H.; Christofides, P. D. Hybrid predictive (28) Bao, J.; Zhang, W. Z.; Lee, P. L. Decentralized fault-tolerant control
control of process systems. AIChE J. 2004, 50, 1242. system design for unstable processes. Chem. Eng. Sci. 2003, 58, 5045.
(9) Elisante, E.; Rangaiah, G. P.; Palanki, S. Robust controller synthesis (29) Wu, N. E. Coverage in fault-tolerant control. Automatica 2004, 40,
for multivariable nonlinear systems with unmeasured disturbances. Chem. 537.
Eng. Sci. 2004, 59, 977. (30) Mhaskar, P.; Gani, A.; El-Farra, N. H.; McFall, C.; Christofides,
(10) El-Farra, N. H.; Mhaskar, P.; Christofides, P. D. Output feedback P. D.; Davis, J. F. Integrated Fault Detection and Fault-Tolerant Control of
control of switched nonlinear systems using multiple Lyapunov functions. Process Systems. AIChE J. 2006 52, 2129.
Syst. Control Lett. 2005, 54, 1163. (31) Mhaskar, P.; Gani, A.; Christofides, P. D. Fault-tolerant control of
(11) Bequette, W. B. Nonlinear control of chemical processes: A review. nonlinear processes: Performance-based reconfiguration and robustness.
Ind. Eng. Chem. Res. 1991, 30, 1391. Int. J. Rob. Nonlinear Control 2006, 16, 91.
(12) Christofides, P. D.; El-Farra, N. H. Control of Nonlinear and Hybrid
(32) Freeman, R. A.; Kokotovic, P. V. Robust Nonlinear Control
Process Systems: Designs for Uncertainty, Constraints and Time-Delays;
Design: State-Space and LyapunoV Techniques; Birkhauser: Boston, 1996.
Springer: New York, 2005.
(13) Mayne, D. Q.; Rawlings, J. B.; Rao, C. V.; Scokaert, P. O. M. (33) Kothare, S. L. D.; Morari, M. Contractive model predictive control
Constrained model predictive control: Stability and optimality. Automatica for constrained nonlinear systems. IEEE Trans. Autom. Control 2000, 45,
2000, 36, 789. 1053.
(14) Bemporad, A.; Morari, M. Robust model predictive control: A (34) Teel, A. Global stabilization and restricted tracking for multiple
survey. In Robustness in Identification and Control, Lecture Notes in Control integrators with bounded controls. Syst. Control Lett. 1992, 18, 165.
and Information Sciences; Garulli, A., Tesi, A., Vicino, A., Eds.; Springer: (35) Liberzon, D.; Sontag, E. D.; Wang, Y. Universal construction of
Berlin, 1999; Vol. 245, p 207. feedback laws achieving ISS and integral-ISS disturbance attenuation. Syst.
(15) Michalska, H.; Mayne, D. Q. Robust receding horizon control of Control Lett. 2002, 46, 111.
constrained nonlinear systems. IEEE Trans. Autom. Control 1993, 38, 1623. (36) Dubljevic, S.; Kazantzis, N. A new Lyapunov design approach for
(16) Sarimveis, H.; Genceli, H.; Nikolaou, M. Design of robust nonlinear systems based on Zubov’s method. Automatica 2002, 38, 1999.
nonsquare constrained model-predictive control. AIChE J. 1996, 42, 2582.
(17) Magni, L.; Nicolao, G.; Scattolini, R.; Allgower, F. Robust model ReceiVed for reView February 24, 2006
predictive control for nonlinear discrete-time systems. Int. J. Rob. Nonlinear ReVised manuscript receiVed April 20, 2006
Control 2003, 13, 229. Accepted April 21, 2006
(18) Langson, W.; Chryssochoos, I.; Rakovic, S. V.; Mayne, D. Q.
Robust model predictive control using tubes. Automatica 2004, 40, 125. IE060237P