Beruflich Dokumente
Kultur Dokumente
TABLE I UGES protocols (cf. [6]), which exclude the commonly used
REFERENCES THAT STUDY MULTIPLE NETWORK- Round-Robin (RR) protocol.
INDUCED IMPERFECTIONS SIMULTANEOUSLY
One of the main contributions of this paper is that we ex-
plicitly construct a continuum of Lyapunov functions based on
the standard delay-free conditions as adopted in [5], [11], [38],
[39], [51], [52]. This continuum of Lyapunov functions leads
to tradeoff curves between the maximally allowable transmis-
sion interval (MATI) and the maximally allowable delay (MAD)
of a NCS often requires tradeoffs between the different types. guaranteeing stability of the NCS. These tradeoff curves will de-
For instance, reducing quantization errors (and thus transmit- pend on the specific communication protocol used, so that they
ting larger or more packets) typically results in larger transmis- even allow for the comparison of different protocols. In addi-
sion delays. To support the designers in making these tradeoffs, tion to stability, which is only a basic property that has to be
tools are needed that provide quantitative information on the satisfied by the control loop, there are often additional require-
consequences of each of the possible choices. However, less ments with respect to the performance of the NCS. This paper
results are available that study combinations of these imper- also studies the performance in terms of gains between spe-
fections. References that simultaneously consider two types of cific exogenous inputs (e.g., disturbances) and controlled out-
network-induced limitations are [5], [9], [13], [18], [25], [29], puts of the system. We will show how performance of the
.[31], [39], [48], [51], [52] which are categorized in Table I. NCS depends on the MATI, the MAD and the protocol used,
Moreover, [37] consider imperfections of type (i), (iii), (v), [8], leading to tradeoff curves as well. This design methodology and
[33], [34] study simultaneously type (ii), (iii), (iv), [38] focusses the method to compute the tradeoff curves will be demonstrated
on type (ii), (iii), (v), and [15] incorporates type (i), (ii), and (iv). on the case study of the batch reactor that has developed over the
In addition some of the approaches mentioned in Table I that years as a benchmark system for NCS, see e.g., [5], [38], [39].
study varying transmission intervals and/or varying communi- Next to stability and performance, also semiglobal practical
cation delays can be extended to include type (ii) phenomena stability results will be presented that can be obtained under
as well by modeling dropouts as prolongations of the maximal weaker conditions.
transmission interval or delay (cf. also Remark II.4 below). The paper is organized as follows. The current section will
Another paper that studies three different types of network end with introducing some notational conventions and con-
imperfections is written by Chaillet and Bicchi [6]. This paper cepts. Next, in Section II we present a general NCS modeling
studies NCS involving both variable delays, variable trans- framework that extends the NCS models in [5], [11], [38], [39],
mission intervals and communication constraints, and uses a [51], [52] to include both communication constraints as well as
method for delay compensation. The delay compensation is varying transmission delays and transmission intervals. In Sec-
based on sending a larger control packet to the plant containing tion III we will transform this new NCS model into the hybrid
not just one control value at one particular time instant, but system framework as introduced in [17] as this will facilitate
containing a control signal valid for a given future time horizon. further analysis. Also the stability and performance concepts as
For this particular control scheme, [6] provides bounds on the used in this paper are defined in this section. In Section IV we
tolerable delays and transmission intervals such that stability will derive the Lyapunov-based conditions that determine both
of the NCS is guaranteed. Also in the present paper we will the maximally allowable transmission interval (MATI) and the
study NCS corrupted by varying delays, varying transmission maximally allowable delay (MAD) guaranteeing global asymp-
intervals and communication constraints, while packet dropouts totic stability and performance. We also present the results
can be included as well (in the way explained in Remark II.4 on semiglobal practical stability in this section. In Section V
below). In other words, this paper considers network-induced we show how the Lyapunov functions can be constructed on
imperfections of type (iii), (iv) and (v). After developing a the basis of the widely adopted non-delay conditions in [5],
novel NCS model incorporating all these types of network [11], [38], [39], [51], [52] and show that the non-delay case is a
phenomena, we will present allowable bounds on delays and particular case of general framework. To demonstrate how the
transmission intervals guaranteeing both stability and perfor- developed methods can be used for explicitly computing MATI
mance of the NCS. However, in contrast with [6], we consider and MAD guaranteeing stability or certain performance, we
the more basic emulation approach in the spirit of [5], [11], apply the framework to the benchmark problem of the batch
[38], [39], [51], [52], which encompasses no specific delay reactor [5], [38], [39]. Finally, we state the conclusions and our
compensation schemes. The work in [6] is of interest, as it aims ideas for future work.
at allowing larger delays by including specific delay compen- The following notational conventions will be used in this
sation schemes, at the cost of sending larger control-packets paper. will denote all nonnegative integers, denotes all
and requiring time-stamping of messages. The features of positive integers, denotes the field of all real numbers and
compensation and time-stamping of messages are not needed denotes all nonnegative reals. By and we denote
in our framework. Another distinction with [6] is related to the the Euclidean norm and the usual inner product of real vectors,
admissible protocols that schedule which node is allowed to respectively. For a collection of real vectors with
transmit its packet at a transmission time. Our work applies , we denote the column vector ob-
for all protocols satisfying the UGES property (see below tained by stacking the vectors , on top of each
for an exact definition) and not only for so-called invariably other by . For a symmetric matrix ,
HEEMELS et al.: NETWORKED CONTROL SYSTEMS WITH COMMUNICATION CONSTRAINTS 1783
denotes the largest eigenvalue of . By and we denote the identical to the usual norm in case a function is defined on a
logical or and and, respectively. A function subset of .
is said to be of class if it is continuous, zero at zero and strictly
increasing. It is said to be of class if it is of class and it is II. NCS MODEL AND PROBLEM STATEMENT
unbounded. A function is said to be of In this section, we introduce the model that will be used to de-
class if it is continuous, is of class for each scribe NCS including both communication constraints as well
and is nonincreasing and satisfies as varying transmission intervals and transmission delays. This
for each . A function is model will form an extension of the NCS models used before
said to be of class if it is continuous, and for each , in [38], [39] that were motivated by the work in [52]. All these
and belong to class . We write for previous models did not include transmission delays. We con-
the standard exponential function. sider the continuous-time plant
We recall now some definitions given in [17] that will be used
for developing a hybrid model of a NCS later. For the motivation (3)
and more details on these definitions, one can consult [17].
that is sampled. Here, denotes the state of the plant,
Definition I.1: A compact hybrid time domain is a set
denotes the most recent control values available at the
with and
plant, is a disturbance input and is the output
. A hybrid time domain is a set such
of the plant. The controller1 is given by
that is a compact hybrid time domain
for each . (4)
Definition I.2: A hybrid trajectory is a pair (dom , ) con-
sisting of a hybrid time domain dom and a function defined where the variable is the state of the controller,
on dom that is absolutely continuous in on is the most recent output measurement of the plant that
for each . is available at the controller and denotes the control
Definition I.3: For the hybrid system given by the state input. The functions , are assumed to be continuous and
space , the input space and the data , where and are assumed to be continuously differentiable. At times
is continuous, is locally , , (parts of) the input at the controller and/or the
bounded, and and are subsets of , a hybrid trajectory output at the plant are sampled and sent over the network. The
(dom , ) with : dom is a solution to for a locally tranmission/sampling times satisfy
integrable input function if and there exists a such that the transmission intervals
1) For all and for almost all dom satisfy for all ,
, we have and . where denotes the maximally allowable transmission in-
2) For all dom such that dom , we have terval (MATI). At each transmission time , , the pro-
and . tocol determines which of the nodes is granted
Hence, the hybrid systems that we consider are of the form access to the network. Each node corresponds to a collection of
sensors or actuators. The sensors/actuators corresponding to the
node that is granted access collect their values of the entries in
or that will be sent over the communication net-
work. They will arrive after a transmission delay of time units
We sometimes omit the time arguments and write at the controller or actuator. This results in updates of the cor-
responding entries in or at times , . The situ-
ation described above is illustrated for and in Fig. 1 for the
situation of two nodes that obtain access to the network in an
(1)
alternating fashion.
It is assumed that there are bounds on the maximal delay in
where we denoted as . We also note that typi-
the sense that , , where
cally and, in this case, if we have that
is the maximally allowable delay (MAD). To be more precise,
either a jump or flow is possible, the latter only if flowing keeps
we adopt the following standing assumption.
the state in . Hence, the hybrid model (1) may have non-unique
Standing Assumption II.1: The transmission times satisfy
solutions.
, and the delays satisfy
In addition, for , , we introduce the norm
, , where
of a function defined on a hybrid time domain dom
is arbitrary.
with possibly and/or , by
The latter condition implies that each transmitted packet ar-
rives before the next sample is taken. This assumption indicates
(2) that we are considering the so-called small delay case as op-
posed to the large delay case, where delays can be larger than
the transmission interval. The inequalities and
provided the right-hand side exists and is finite. In case is 1Extensions of the theory presented below to the case of time-dependent sys-
finite, we say that . Note that this definition is essentially tems (3) and time-dependent controllers (4) are straightforward.
1784 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 55, NO. 8, AUGUST 2010
(7a)
(7b)
Fig. 1. Illustration of a typical evolution of y and y^ in case of two nodes that
get access to the network in an alternating manner. where with , , are
appropriately defined functions depending on , , and
can be taken non-strict with the understanding and . See [38] for the explicit expressions of and
that in case the update instant coincides with the next , which also reveal how we use the differentiability conditions
transmission instant , the update is performed before the on and imposed earlier.
next sample is taken. The updates at satisfy Standing Assumption II.2: and are continuous and is
locally bounded.
Observe that the system
(5)
(8)
where denotes the vector with and
. Hence, with . If the is the closed-loop system (3) and (4) without the network (i.e.,
NCS has nodes, then the error vector can be partitioned and in (3) and (4)).
as . The functions and are now up- The problem that we consider in this paper is formulated as
date functions that are related to the protocol that determines follows.
on the basis of and the networked error which node is Problem II.3: Suppose that the controller (4) was designed
granted access to the network. Typically when the -th node gets for the plant (3) rendering the closed-loop (3) and (4) without the
access to the network at some transmission time we have network, i.e., and , (or equivalently, (8))
that the corresponding part in the error vector has a jump at stable in some sense. Determine the value of and so
. In most situations, the jump will actually be to zero, that the NCS given by (7) is stable as well when the transmission
since we assume that the quantization effects are negligible. For intervals and delays satisfy Standing Assumption II.1
instance, when is sampled at time , then we have that Remark II.4: Of course, there are certain extensions that can
. However, we allow for more freedom in be made to the above setup. The inclusion of packet dropouts
the protocols by allowing general functions . See [38], [39] for is relatively easy, if one models them as prolongations of the
more details. We will refer to as the protocol. transmission interval. Indeed, if we assume that there is a bound
In between the updates of the values of and , the network is on the maximum number of successive dropouts, the sta-
assumed to operate in a zero order hold (ZOH) fashion, meaning bility bounds derived below are still valid for the MATI given
that the values of and remain constant in between the up- by , where is the obtained value
dating times and for all : for the dropout-free case. Another extension of the framework
in this paper could be the inclusion of quantization effects. This
(6) step is more involved. It might be based on recent work in [37]
that unifies the areas of networked and quantized control sys-
To compute the resets of at the update times , tems without communication delays. It can be envisioned that
we proceed as follows: the results presented here can be combined with the framework
in [37] leading to an overall methodology capable of handling
all types of networked phenomena that were mentioned in the
introduction. The specific conditions and types of quantizers for
which this methodology is effective are subject of future re-
search and some preliminary results are reported in [21]. An-
other possible extension of interest is the consideration of the
large-delay case (in which the delays can be larger than the
transmission interval). This would require a more involved NCS
model that does not have the periodicity between transmission
and update events as implied by Standing Assumption II.1. This
In the third equality we used that due to is a hard problem, which will be considered in future research.
the zero order hold character of the network. We also implicitly
HEEMELS et al.: NETWORKED CONTROL SYSTEMS WITH COMMUNICATION CONSTRAINTS 1785
III. REFORMULATION IN A HYBRID SYSTEM FRAMEWORK , because it will simplify the analysis later. By taking
To facilitate the stability analysis, we transform the above the hybrid system above is in the
NCS model into the hybrid system framework as developed in form (1).
[17]. This hybrid systems framework was also employed in [5], Definition III.1: For the hybrid system with ,
where a similar model was obtained without the incorporation of the set given by is
delays. To do so, we introduce the auxiliary variables , said to be uniformly globally asymptotically stable (UGAS) if
, and to reformulate the model in there exists a function such that, for each
terms of flow equations and reset equations. The variable is , and any initial condition , ,
an auxiliary variable containing the memory in (7b) storing the , , ,
value for the update of at the update instant with3
, is a counter keeping track of the transmission, is a , all corresponding solutions satisfy
timer to constrain both the transmission interval as well as the
transmission delay2 and is a Boolean keeping track whether the
next event is a transmission event or an update event. To be pre- (13)
cise, when the next event will be related to transmission
and when the next event will be an update. The Boolean for all in the solutions domain. The set is uniformly
will be used to guarantee in the model below that the transmis- globally exponentially stable (UGES) if can be taken of the
sion and update events are alternating in the sense that before a form for some and
next sample is taking the previous update is implemented in the .
NCS. Remark III.2: The factor multiplying in the right-hand
The hybrid system is now given by the flow equations side of (13) is motivated by the fact that for small values of
solutions of exist that have many resets without
progressing too fast. Actually, the limit case has livelock
solutions that are only resetting (with remaining 0 and
(9) ) as discussed above. The scales the right-hand side of (13)
for this effect.
We also introduce the concept of uniform semiglobal prac-
tical asymptotical stability.
and the reset equations are obtained by combining the trans- Definition III.3: For the hybrid system with ,
mission reset relations, active at the transmission instants the set is said to be uniformly semiglobally practically asymp-
, and the update reset relations, active at the update totically stable (USPAS) with respect to and , if there
instants , given by exists such that for any pair of positive numbers
there exist and such that
when for each , each initial condition ,
(10) , , , ,
with
with the mapping given by the transmission resets (when , ,
) , and each corresponding solution we have
(11)
(14)
and the update resets (when )
for all in the solutions domain.
(12)
In the presence of disturbance inputs in we might
Two comments on this model are in order. First of all, the role be interested in reducing its influence on a particular controlled
of is to exclude (instantaneous Zeno) solutions to output variable
satisfying , and
for or with , when (15)
. Also when similar solutions exist that
in terms of the induced gain, as formally defined below. The
are only resetting (sometimes called livelock in hybrid systems
hybrid model expanded with the output (15) is denoted
theory [50]). However, can be taken arbitrarily small to
by .
still allow for small transmission intervals.
Definition III.4: Consider with and let be
Secondly, the choice for when is irrelevant from a
given. The hybrid system is said to be stable with gain
modeling point of view. However, it was selected here as
, if there is a -function such that for any ,
2We could also have introduced two timers, one corresponding to the trans- any input and any initial condition ,
mission interval and one to the transmission delay. However, it turns out that
the NCS can be described using only one timer, which has the advantage of re- 3Note that the next condition is just saying that (0; 0) 2 C [ D in the
sulting in a more compact hybrid model. terminology of (1).
1786 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 55, NO. 8, AUGUST 2010
(16) (22a)
(22b)
IV. STABILITY AND PERFORMANCE ANALYSIS
where and , are the real constants
In this section we focus on the analysis of UGAS and UGES,
as given in Condition IV.1. Observe that the solutions to these
USPAS, and stability.
differential equations are strictly decreasing as long as
A. Stability Analysis , .
Theorem IV.2: Consider the system that satisfies Con-
In order to guarantee UGAS or UGES, we assume the exis- dition IV.1. Suppose satisfy
tence of a Lyapunov function for the reset equa-
tions (11) and (12) satisfying (23a)
(17a) (23b)
(17b) for solutions and of (22) corresponding to certain chosen
initial conditions , , with
for all and all and the bounds , and as in Condition IV.1. Then
for the system with the set is UGAS in the
(18)
sense of Definiton III.1. If in addition, there exist strictly positive
for all , and for some functions real numbers , , , and such that ,
and and . , , ,
In Section V we will show how a function satisfying (17) and , , then this set is UGES.
and (18) can be derived from the generally accepted conditions Proof: The solutions to the differential (22) are scaled
on the protocol as used for the delay-free case in [5], [38]. as and for new functions , . The
To solve Problem II.3, we extend (17) and (18) to the following differential (22) and the conditions (23) transform into
condition.
Condition IV.1: There exist a function (24)
with locally Lipschitz for all
and , a locally Lipschitz function , and
-functions , , and , continuous functions (25a)
, positive definite functions and and constants (25b)
, , for , and such that:
for all and all (17) holds and (18) holds We consider now the function
for all ;
for all , , , and almost (26)
all it holds that
and show that this constitutes a suitable Lyapunov function for
the system , which can be used to conclude UGAS and
UGES under the stated conditions.
(19) Below, with some abuse of notation, we consider the quantity
for all , , and almost all with
Similarly, when we have that and obtain, Condition IV.4: There are a function
using (12) with locally Lipschitz for all
and , -functions , , , and such
that
for all and all (17) holds and (18) holds
for all ;
for all , , all and almost all
We also have, for all and almost all , that
it holds that
(28)
(27)
where . Hence, for the delay-free case
we recover the results in [5] as a special case.
(35)
(31)
for all , , , and almost
all
(32)
constructing Lyapunov and storage functions as in Condition for almost all and all and all ,
IV.1 and Condition IV.6, respectively. For the delay-free case, we can derive functions and satisfying Condition IV.1
one considers in [5], [38] protocols satisfying the following and Condition IV.6, respectively. The constants in (44) satisfy
condition: , where is sufficiently small.
Condition V.1: The protocol given by is UGES (uniformly Remark V.2: Condition V.1 and inequality (44) are essen-
globally exponentially stable), meaning that there exists a func- tially the same as in [5] with . The constant
tion that is locally Lipschitz in its second is selected small to sacrifice only a little of the
argument such that gain from to . In [38, Thm. 4] the inequality (44) was
actually formulated in terms of an gain, while we use a
(38a) Lyapunov-based formulation here. As gains are established
(38b) often using Lyapunov functions, this seems to be a natural refor-
mulation (see also [5, Rem. 2]). The only additional condition
for constants and . we add here is (39), which holds for all protocols considered in
Additionally we assume here that [5], [11], [38], [39], [51], [52] as is demonstrated in Lemma V.4
below for the RR and TOD protocols.
(39) Theorem V.3: Consider the systems and , re-
spectively, such that
for some constant4 and that for almost all and Condition V.1, (39) with and (40) with constant
all hold;
Equation (41) is satisfied for some function
(40) (and (42) for some function
, respectively) and ;
for some constant . For all protocols discussed in [5], there exists a locally Lipschitz continuous function
[38], [39], [51], [52] such constants exist. In Lemma V.4 below, satisfying the bounds (43) for some
we specify appropriate values for these constants in case of the -functions , , and (44) with and
often used Round Robin (RR) and the Try-Once-Discard (TOD) (and (45) with ,
protocols (see [38], [52] for their definitions). We also assume and , respectively).
the growth condition on the NCS model (7) Then, the functions given by
(41)
(45)
4In principle this constant can be taken non-negative. However, as all proto-
cols available in the literature satisfy 1, we take 1 to reduce some
notational burden later.
1790 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 55, NO. 8, AUGUST 2010
Using (38b) and (39), this follows trivially. The condition (17b)
is identical to
(52)
(53)
(49)
Case 2: and
(54)
Case 3: and
Proof: The constants , , are derived for both pro- globally asymptotically stable. Then, the function given by
tocols in [38] and corresponding Lyapunov functions . For (46) satisfies Condition IV.4 with , ,
the RR protocol can be taken with for some positive constants and and
, , and is the as in Condition V.1.
number of nodes in the network, see Example 3 in [38]. This The conditions in Theorem V.5 are precisely those used in
implies that the delay-free case for obtaining USPAS as adopted in [5, Thm.
2]. To obtain the constants , , and for the RR or
TOD protocol again Lemma V.4 above can be employed.
(56)
1792 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 55, NO. 8, AUGUST 2010
(57a)
We take and
in (41). To verify (45) we
compute simultaneously an upper bound on the gain from
where we used that as both matrices to and on the gain
are diagonal. Using this sharper result in (49)(52) instead of from to by finding Pareto minimal values for and subject
(40) and (41), we obtain the improved values , to the matrix inequalities in the matrix and multiplier
, and , similar to [5] as given in (58), shown at the bottom of the page. These matrix
and [38], which leads for the delay-free case to inequalities demonstrate that there will be a tradeoff between
(recovering the result in [5], which outperforms the values performance in terms of the gain from to reflected in
found in [38], [39], [51]). The tradeoff curve between MATI on the one hand and the size of and on the other
and MAD is also given in Fig. 4. In this figure also the delay-free as reflected in . Recall that directly influences MATI and
case with and is visualized. These MAD through and in (48). The tradeoff between
tradeoff curves can be used to impose conditions or select a
and can still be made by varying and whilst
suitable network with certain communication delay and band-
guaranteeing a certain gain for the NCS.
width requirements (note that MATI is inversely proportional
To make these observations quantitative, we fix at various
to the bandwidth).
values and search for the smallest value of such that there
Also different protocols can be compared with respect to each
exist and satisfying (58). Note that (58) is an LMI
other. In Fig. 4, it is seen that for the task of stabilization of
the unstable batch reactor the TOD protocol outperforms the when is fixed and hence, can be solved efficiently. As a lower
RR protocol in the sense that it can allow for larger delays bound on we take the gain from to of the system
and larger transmission intervals. The difference between the without the network . This value is found by solving
tradeoff curves for different protocols is caused by different the standard -gain/ LMI for linear systems (see e.g., [1])
values for the parameters , , and (see Lemma using, again, the SEDUMI solver and the YALMIP interface.
V.4), which in turn induce different values for the parameters This yields . Using this as a lower bound on the con-
, , and (see Theorem V.3) and thus different solu- sidered values of the desirable performance level in terms of the
tions to the differential equations (22). This results in different gain , we search minimal values for (corresponding to the
combinations of and that guarantee stability due to selected value of ) under the feasibility of the LMI (57). These
Theorem IV.2. minimal values yield the Pareto optimal curves for as
shown in Fig. 5. This figure demonstrates that the minimal value
B. Gain Analysis of approaches 15.9165 for large values of , which was the
To apply the developed framework for gain analysis, the gain from to as computed
effect of the additional disturbance input on the controlled in Section VI-A for the stability analysis. This is expected as the
variable is studied. value 15.9165 corresponds to the smallest value of found in
(58)
1794 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 55, NO. 8, AUGUST 2010
VII. CONCLUSION
In this paper, we presented a framework for studying the
stability of a NCS, which involves communication constraints
(only one node accessing the network per transmission), varying
transmission intervals and varying transmission delays. Based
on a newly developed model, a Lyapunov-based characteriza-
tion of stability was provided and explicit bounds on the max-
imally allowable transmission interval (MATI) and the max-
imally allowable delay (MAD) were obtained that guarantee
stability of the NCS. We explicitly showed how a continuum
of Lyapunov functions can be constructed from the commonly
adopted conditions for the delay-free case. The application of
Fig. 6. Tradeoff curves between MATI and MAD for various levels of the L the results on a benchmark example showed how tradeoff curves
gain of the NCS with the TOD protocol. between MATI and MAD can be computed providing designers
of NCS with proper tools to support their design choices. In-
terestingly, recently developed improvements in [5] leading to
sharper bounds for the MATI (the non-delay case) are a spe-
Section VI-A if only stability is required (without any additional cial case of this more general framework. Additionally, we ana-
performance conditions, so the gain approaches in- lyzed the performance of NCS and provided the theoretical
finity). The other extreme, when approaches infinity, recovers framework that shows how MATI, MAD and can be traded
the situation where the values approach the asymptotic value quantitatively against each other. Under weaker conditions, we
of being the optimal network-free gain from provided also semiglobal practical stability results.
to . Future work will involve the consideration of the large delay
As we can see, a smaller gain requires a larger value case (delays larger than the transmission interval) and the devel-
of , which will in turn result in smaller values for and opment of an analysis framework that also includes quantization
. We will demonstrate this for the TOD protocol. The re- effects. Extending the current framework so as to include quan-
sults for the RR protocol can be obtained using the same ap- tization effects would provide the means to study NCS incor-
proach. The computed values of are now used in (45) and porating all the types of network phenomena mentioned in the
the values for the constants , , and can be obtained introduction (as packet dropouts can be modeled using a prolon-
from Lemma V.4, as in the stability analysis above. Then The- gation of the MATI as discussed in Remark II.4). We foresee that
orem V.3 can be applied to construct a continuum of suitable such an extension would be extremely valuable to network and
storage functions for the closed-loop NCS system choosing a control system designers, provided that the extended framework
certain combination of in a similar manner as for the sta- leads to quantitative tradeoff curves between the various net-
bility analysis. This leads to combinations of work parameters (MATI, MAD, quantization error, bandwidth,
such that the NCS has an gain from to smaller than for etc.) and the performance of the overall control loop. Some pre-
transmission intervals smaller than and communication liminary results in this direction can be found in [21].
delays smaller than . The tradeoff curves for various levels REFERENCES
of the performance are provided in Fig. 6. For the value [1] S. Boyd, L. E. Ghaoui, E. Feron, and V. Balakrishnan, Linear Matrix
of we (almost) recover the tradeoff plot as in Fig. 4 Inequalities in Control Theory, volume 15 of Studies in Applied Math-
for the TOD protocol, because, loosely speaking, the gain ematics. Philadelphia, PA: SIAM, 1994.
[2] R. Brockett, Stabilization of motor networks, in Proc. 34th IEEE
requirement is very mild as is a relatively large value Conf. Decision Control, 1995, vol. 2, pp. 14841488.
that practically approaches the condition that the NCS should be [3] R. W. Brockett and D. Liberzon, Quantized feedback stabilization
UGES only. In the other extreme, if very high requirements are of linear systems, IEEE Trans. Autom. Control, vol. 45, no. 7, pp.
12791289, Jul. 2000.
given with respect to robustness to disturbances (in the sense [4] B. Brogliato, R. Lozano, B. Maschke, and O. Egeland, Dissipative
of a very low gain from to approaching the network-free Systems Analysis and Control: Theory and Applications. New York:
gain ), the values for MATI and MAD that guar- Springer, 2000.
[5] D. Carnevale, A. R. Teel, and D. Nesic, A Lyapunov proof of improved
antee this gain are approaching 0. This is clearly shown in
maximum allowable transfer interval for networked control systems,
Fig. 6 as the tradeoff curve for the value corre- IEEE Trans. Autom. Control, vol. 52, no. 5, pp. 892897, May 2007.
sponds to values of and close to zero. The limit case [6] A. Chaillet and A. Bicchi, Delay compensation in packet-switching
would actually correspond to and . networked controlled sytems, in Proc. IEEE Conf. Decision Control,
2008, pp. 36203625.
Hence, the control and network engineers have to make clear [7] F. H. Clarke, Y. S. Ledyaev, and R. J. Stern, Asymptotic stability and
design tradeoffs between MATI, MAD, robustness in terms of smooth Lyapunov functions, J. Differential Equations, vol. 149, pp.
performance and the choice of the protocol. The provided 69114, 1998.
[8] M. B. G. Cloosterman, L. Hetel, N. van de Wouw, W. P. M. H. Heemels,
framework supports the engineers to make these design choices J. Daafouz, and H. Nijmeijer, Controller Synthesis for Networked
in a quantified manner. Control Systems, Automatica, vol. 46, no. 10, Oct. 2010.
HEEMELS et al.: NETWORKED CONTROL SYSTEMS WITH COMMUNICATION CONSTRAINTS 1795
[9] M. B. G. Cloosterman, N. van de Wouw, W. P. M. H. Heemels, and [32] L. A. Montestruque and P. Antsaklis, Stability of model-based
H. Nijmeijer, Stabilization of networked control systems with large networked control systems with time-varying transmission times,
delays and packet dropouts, in Proc. Amer. Control Conf., 2008, pp. IEEE Trans. Autom. Control, vol. 49, no. 9, pp. 15621572, Sep.
49914996. 2004.
[10] M. B. G. Cloosterman, N. van de Wouw, W. P. M. H. Heemels, and [33] P. Naghshtabrizi and J. P. Hespanha, Designing an observer-based
H. Nijmeijer, Stability of networked control systems with uncertain controller for a network control system, in Proc. Conf. Decision Con-
time-varying delays, IEEE Trans. Autom. Control, vol. 54, no. 7, pp. trol Eur. Control Conf., Seville, Spain, 2005, pp. 848853.
15751580, Jul. 2009. [34] P. Naghshtabrizi and J. P. Hespanha, Stability of network control sys-
[11] D. B. Dacic and D. Nesic, Quadratic stabilization of linear networked tems with variable sampling and delays, in Proc. Annu. Allerton Conf.
control systems via simultaneous protocol and controller design, Au- Commun., Control, Comp., 2006, [CD ROM].
tomatica, vol. 43, pp. 11451155, 2007. [35] P. Naghshtabrizi, J. P. Hespanha, and A. R. Teel, Stability of delay
[12] D. F. Delchamps, Stabilizing a linear system with quantized state impulsive systems with application to networked control systems, in
feedback, IEEE Trans. Autom. Control, vol. 35, no. 8, pp. 916924, Proc. Amer. Control Conf., New York, 2007, pp. 48994904.
Aug. 1990. [36] G. N. Nair and R. J. Evans, Stabilizability of stochastic linear systems
[13] M. C. F. Donkers, L. Hetel, W. P. M. H. Heemels, N. van de Wouw, with finite feedback data rates, SIAM J. Control Optim., vol. 43, pp.
and M. Steinbuch, Stability analysis of networked control systems 413436, 2004.
using a switched linear systems approach, in Hybrid Systems: Compu- [37] D. Nesic and D. Liberzon, A unified framework for design and anal-
tation and Control, Lecture Notes in Computer Science. New York: ysis of networked and quantized control systems, IEEE Trans. Autom.
Springer Verlag, 2009, pp. 150164. Control, vol. 54, no. 4, pp. 732747, Apr. 2009.
[14] H. Fujioka, Stability analysis for a class of networked/embedded con- [38] D. Nesic and A. R. Teel, Input-output stability properties of networked
trol systems: A discrete-time approach, in Proc. Amer. Control Conf., control systems, IEEE Trans. Autom. Control, vol. 49, no. 10, pp.
2008, pp. 49975002. 16501667, Oct. 2004.
[15] H. Gao, T. Chen, and J. Lam, A new delay system approach to net- [39] D. Nesic and A. R. Teel, Input-to-state stability of networked control
work-based control, Automatica, vol. 44, no. 1, pp. 3952, 2008. systems, Automatica, vol. 40, pp. 21212128, 2004.
[16] R. H. Gielen, S. Olaru, M. Lazar, W. P. M. H. Heemels, N. van de [40] H. Rehbinder and M. Sanfridson, Scheduling of a limited commu-
Wouw, and S.-I. Niculescu, On polytopic inclusions as a modeling nication channel for optimal control, Automatica, vol. 40, no. 3, pp.
framework for systems with time-varying delays, Automatica, vol. 46, 491500, 2004.
no. 3, pp. 615619, Mar. 2010. [41] B. Sinopoli, L. Schenato, M. Franceschetti, K. Poolla, M. I. Jordan, and
[17] R. Goebel and A. R. Teel, Solution to hybrid inclusions via set and S. S. Sastry, Kalman filtering with intermittent observations, IEEE
graphical convergence with stability theory applications, Automatica, Trans. Autom. Control, vol. 49, no. 9, pp. 14531464, Sep. 2004.
vol. 42, pp. 573587, 2006. [42] S. C. Smith and P. Seiler, Estimation with lossy measurements: Jump
[18] V. Gupta, T. H. Chung, B. Hassibi, and R. M. Murray, On a stochastic estimators for jump systems, IEEE Trans. Autom. Control, vol. 48, no.
sensor selection algorithm with applications in sensor scheduling and 12, pp. 21632171, Dec. 2003.
sensor coverage, Automatica, vol. 42, no. 2, pp. 251260, 2006. [43] J. Sturm, Using SeDuMi 1.02, a MATLAB toolbox for optimization
[19] W. M. Haddad, V. Chellaboina, and S. G. Nersesov, Impulsive over symmetric cones, Optim. Methods Software, 1112, Special issue
and Hybrid Dynamical Systems: Stability, Dissipativity, and Con- Interior Point Methods, pp. 625653, 1999.
trol. Princeton, NJ: Princeton Univ. Press, 2006. [44] M. Tabbara, D. Nesic, and A. R. Teel, Stability of wireless and wire-
[20] W. P. M. H. Heemels, R. J. A. Gorter, A. van Zijl, P. P. J. v. d. Bosch, line networked control systems, IEEE Trans. Autom. Control, vol. 52,
S. Weiland, W. H. A. Hendrix, and M. R. Vonder, Asynchronous mea- no. 9, pp. 16151630, Sep. 2007.
surement and control: A case study on motor synchronisation, Control [45] S. Tatikonda and S. K. Mitter, Control under communication con-
Eng. Prac., vol. 7, no. 12, pp. 14671482, 1999. straints, IEEE Trans. Autom. Control, vol. 49, no. 7, pp. 10561068,
[21] W. P. M. H. Heemels, D. Nesic, A. R. Teel, and N. van de Wouw, Net- Jul. 2004.
worked and quantized control systems with communication delays, in [46] A. Teel and L. Praly, Tools for semiglobal stabilization by partial
Proc. Joint IEEE Conf. Decision Control (CDC) 28th Chinese Control state and output feedback, SIAM J. Control Optim., vol. 33, no. 5, pp.
Conf., 2009, pp. 79297935. 14431488, 1995.
[22] W. P. M. H. Heemels, H. Siahaan, A. Juloski, and S. Weiland, Control [47] Y. Tipsuwan and M.-Y. Chow, Control methodologies in networked
of quantized linear systems: An l -optimal control approach, in Proc. control systems, Control Eng. Practice, vol. 11, pp. 10991111,
Amer. Control Conf., Denver, CO, 2003, pp. 35023507. 2003.
[23] J. P. Hespanha, P. Naghshtabrizi, and Y. Xu, A survey of recent results [48] N. van de Wouw, P. Naghshtabrizi, M. Cloosterman, and J. P. Hes-
in networked control systems, Proc. IEEE, vol. 95, no. 1, pp. 138162, panha, Tracking control for sampled-data systems with uncertain
Jan. 2007. time-varying sampling intervals and delays, Int. J. Robot. Nonlin.
[24] L. Hetel, J. Daafouz, and C. Iung, Stabilization of arbitrary switched Control, vol. 20, no. 4, pp. 387411, 2010.
linear systems with unknown time-varying delays, IEEE Trans. [49] A. J. van der Schaft, L2-Gain and Passivity in Nonlinear Control, 2nd
Autom. Control, vol. 51, no. 10, pp. 16681674, Oct. 2006. ed. New York: Springer, 1999.
[25] L. Hetel, J. Daafouz, and C. Iung, Analysis and control of LTI and [50] A. J. van der Schaft and J. M. Schumacher, An Introduction to Hy-
switched systems in digital loops via an event-based modeling, Int. J. brid Dynamical Systems, volume 251 of LNCIS. New York: Springer-
Control, vol. 81, no. 7, pp. 11251138, 2008. Verlag, 2000.
[26] D. Hristu and K. Morgansen, Limited communication control, Syst. [51] G. C. Walsh, O. Belidman, and L. G. Bushnell, Asymptotic behavior
Control Lett., vol. 37, no. 4, pp. 193205, Jul. 1999. of nonlinear networked control systems, IEEE Trans. Autom. Control,
[27] C.-Y. Kao and B. Lincoln, Simple stability criteria for systems with vol. 46, no. 7, pp. 10931097, Jul. 2001.
time-varying delays, Automatica, vol. 40, pp. 14291434, 2004. [52] G. C. Walsh, O. Belidman, and L. G. Bushnell, Stability analysis of
[28] D. Liberzon, On stabilization of linear systems with limited informa- networked control systems, IEEE Trans. Control Syst. Technol, vol.
tion, IEEE Trans. Autom. Control, vol. 48, no. 2, pp. 304307, Feb. 10, no. 3, pp. 438446, May 2002.
2003. [53] J. C. Willems, Dissipative dynamical systems, Archive Rational
[29] D. Liberzon, Quantization, time delays, and nonlinear stabilization, Mech. Anal., vol. 45, pp. 321393, 1972.
IEEE Trans. Autom. Control, vol. 51, no. 7, pp. 11901195, Jul. [54] T. C. Yang, Networked control system: A brief survey, Proc. Inst.
2006. Elect. Eng., vol. 153, no. 4, pp. 403412, Jul. 2006.
[30] J. Lfberg, YALMIP: A toolbox for modeling and optimization in [55] L. Zhang, Y. Shi, T. Chen, and B. Huang, A new method for stabiliza-
MATLAB, in Proc. CACSD Conf., 2004, pp. 284289. tion of networked control systems with random delays, IEEE Trans.
[31] A. S. Matveev and A. V. Savkin, The problem of state estimation Autom. Control, vol. 50, no. 8, pp. 11771181, Aug. 2005.
via asynchronous communication channels with irregular transmission [56] W. Zhang, M. S. Branicky, and S. M. Phillips, Stability of networked
times, IEEE Trans. Autom. Control, vol. 48, no. 4, pp. 670676, Apr. control systems, IEEE Control Syst. Mag., vol. 21, no. 1, pp. 8499,
2003. 2001.
1796 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 55, NO. 8, AUGUST 2010
W. P. Maurice H. Heemels (M06) received the Nathan van de Wouw (M08) received the M.Sc.
M.Sc. degree in mathematics (with honors) and the degree (with honors) and Ph.D. degree in mechan-
Ph.D. degree (with honors) from the Eindhoven ical engineering from the Eindhoven University of
University of Technology (TU/e), Eindhoven, The Technology, Eindhoven, the Netherlands, in 1994 and
Netherlands, in 1995 and 1999, respectively. 1999, respectively.
From 2000 to 2004 he was with the Electrical Engi- From 1999 until 2010, he was with the Dynamics
neering Department, TU/e, as an Assistant Professor and Control Group, Department of Mechanical
and from 2004 to 2006 with the Embedded Systems Engineering, Eindhoven University of Technology
Institute (ESI) as a Research Fellow. Since 2006, he as an Assistant/Associate Professor. In 2000, has
has been with the Department of Mechanical Engi- been working at Philips Applied Technologies,
neering, TU/e, where he is currently a Full Professor Eindhoven, The Netherlands, and, in 2001, he has
in the Hybrid and Networked Systems Group. He held visiting research posi- been working at the Netherlands Organisation for Applied Scientific Research
tions at the Swiss Federal Institute of Technology (ETH), Zurich, Switzerland (TNO), Delft, The Netherlands. He has held positions as a Visiting Professor
(2001) and at the University of California at Santa Barbara (2008). In 2004, he at the University of California Santa Barbara, in 2006 and 2007 and at the
was also at the Research & Development laboratory, Oc, Venlo, The Nether- University of Melbourne, Australia, in 2009 and 2010. He has published a large
lands. Currently, he is an Associate Editor for AUTOMATICA and Nonlinear Anal- number of journal and conference papers and co-authored the books Uniform
ysis: Hybrid Systems. His current research interests include hybrid and non- Output Regulation of Nonlinear Systems: A convergent Dynamics Approach
smooth dynamical systems, networked control systems and constrained systems (Boston, MA: Birkhauser, 2005) and Stability and Convergence of Mechanical
including model predictive control. Systems with Unilateral Constraints (New York: Springer-Verlag, 2008). His
Dr. Heemels serves in the international program committees of various current research interests are the analysis and control of nonlinear/non-smooth
conferences. systems and networked control systems.
Andrew R. Teel (S91M92SM99F02) re- Dragan Nesic (SM01F08) received the B.E. de-
ceived the A.B. degree in Engineering Sciences from gree from The University of Belgrade, Yugoslavia,
Dartmouth College, Hanover, NH, in 1987, and the in 1990, and the Ph.D. degree from Systems Engi-
M.S. and Ph.D. degrees in electrical engineering neering, RSISE, Australian National University, Can-
from the University of California, Berkeley, in 1989 berra, Australia in 1997.
and 1992, respectively. He is currently a Professor in the Department of
He was a Postdoctoral Fellow at the Ecole des Electrical and Electronic Engineering (DEEE), Uni-
Mines de Paris, Fontainebleau, France. In September versity of Melbourne, Australia. From 1997 to 1999,
of 1992 he joined the faculty of the Electrical he held postdoctoral positions at the University of
Engineering Department, University of Minnesota, Melbourne, Australia, the Universit Catholique de
Minneapolis, where he was an Assistant Professor Louvain, Louvain la Neuve, Belgium and the Uni-
until September of 1997. In 1997, he joined the faculty of the Electrical and versity of California, Santa Barbara, CA. Since February 1999 he has been with
Computer Engineering Department, University of California, Santa Barbara, The University of Melbourne. He is an Associate Editor for Automatica, Systems
where he is currently a Professor. and Control Letters, and the European Journal of Control. His research inter-
Dr. Teel received NSF Research Initiation and CAREER Awards, the 1998 ests include networked control systems, discrete-time, sampled-data and contin-
IEEE Leon K. Kirchmayer Prize Paper Award, the 1998 George S. Axelby Out- uous-time nonlinear control systems, input-to-state stability, extremum seeking
standing Paper Award, the first SIAM Control and Systems Theory Prize in control, applications of symbolic computation in control theory, and so on.
1998, and the 1999 Donald P. Eckman Award and the 2001 O. Hugo Schuck Dr. Nesic received the Humboldt Research Fellowship funded by the
Best Paper Award from the American Automatic Control Council. Alexander von Humboldt Foundation, Germany, in 2003. He was an Australian
Professorial Fellow from 2004 to 2009 (which is a research position funded by
the Australian Research Council). He is a Fellow of IEAust and a Distinguished
Lecturer of CSS of the IEEE. He served as an Associate Editor for the IEEE
TRANSACTIONS ON AUTOMATIC CONTROL.