676 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 14, NO.

5, OCTOBER 2006

A Survey on Analysis and Design of Model-Based

Fuzzy Control Systems
Gang Feng, Senior Member, IEEE

Abstract—Fuzzy logic control was originally introduced and de- people have devoted a great deal of time and effort to both
veloped as a model free control design approach. However, it un- theoretical research and application techniques of fuzzy logic
fortunately suffers from criticism of lacking of systematic stability controllers. This can be witnessed by a number of excellent
analysis and controller design though it has a great success in in-
dustry applications. In the past ten years or so, prevailing research books and tutorial articles on the topic; see, for example, [7],
efforts on fuzzy logic control have been devoted to model-based [164], [165], [210], [236], [240], [241], [251], [259], [269],
fuzzy control systems that guarantee not only stability but also per- [299], [300], and [318]. Much success has also been achieved
formance of closed-loop fuzzy control systems. This paper presents in applying FLC to various areas including power systems [1],
a survey on recent developments (or state of the art) of analysis [88], [99], [149]; telecommunications [5], [45], [49], [131],
and design of model based fuzzy control systems. Attention will
be focused on stability analysis and controller design based on the [169], [343]; mechanical/robotic systems [8], [10], [18], [40],
so-called Takagi–Sugeno fuzzy models or fuzzy dynamic models. [102], [109], [118], [138], [139], [180], [182], [204], [247],
Perspectives of model based fuzzy control in future are also dis- [262], [284], [289], [292], [294], [314], [319]; automobile [16],
cussed. [102], [116], [185], [205], [218], [222], [260]; industrial/chem-
Index Terms—Adaptive control, control theory, fuzzy control, ical processes [22], [41], [90], [111], [129], [137], [146], [153],
fuzzy models, fuzzy systems, Lyapunov functions, robustness, sta- [162], [199], [200], [229], [248], [258], [271], [279], [288]; air-
bility, stabilization, Takagi–Sugeno (T–S) fuzzy models. crafts [58], [73], [130], [161]; motors [9], [100], [143]; medical
services [158], [248], [345]; consumer electronics [106], [156],
I. INTRODUCTION [170], [172], [219], [255], [263], [311]; and other areas such as
chaos control [52], [183] and nuclear reactors [17], [217].

F UZZY sets and systems have gone through substantial de-

velopment since the introduction of fuzzy set theory by
Zadeh [331]–[335] about four decades ago. They have found a
The basic structure of a fuzzy control system consists of
four conceptual components: knowledge base, fuzzification
interface, inference engine, and defuzzification interface [164],
great variety of applications ranging from control engineering, [165]. Fig. 1 shows the block diagram of a fuzzy control system.
qualitative modelling, pattern recognition, signal processing, in- The knowledge base contains all the controller knowledge
formation processing, machine intelligence, decision making, and it comprises a fuzzy control rule base and a data base. The
management, finance, medicine, motor industry, robotics, and data base is the declarative part of the knowledge base which
so on [7], [11], [14], [16], [65], [154], [164], [165], [210], [241], describes definition of objects (facts, terms, concepts) and defi-
[248], [258], [261], [277], [289], [328], [334], [348]. In partic- nition of membership functions used in the fuzzy control rules.
ular, fuzzy logic control (FLC), as one of the earliest applica- The fuzzy control rule base is the procedural part of the knowl-
tions of fuzzy sets and systems, has become one of the most edge base which contains information on how these objects can
successful applications. In fact, FLC has proven to be a suc- be used to infer new control actions. The inference engine is
cessful control approach to many complex nonlinear systems or a reasoning mechanism which performs inference procedure
even nonanalytic systems. It has been suggested as an alterna- upon the fuzzy control rules and given conditions to derive rea-
tive approach to conventional control techniques in many cases. sonable control actions. It is the central part of a fuzzy con-
The first fuzzy logic control system is developed by Mam- trol system. The fuzzification interface (or fuzzifier) defines a
dani and Assilian [199], [200], where control of a small steam mapping from a real-valued space to a fuzzy space, and the de-
engine is considered. The fuzzy control algorithm consists of a fuzzification interface (or defuzzifier) defines a mapping from
set of heuristic control rules, and fuzzy sets and fuzzy logic are a fuzzy space defined over an output universe of discourse to a
used, respectively, to represent linguistic terms and to evaluate real-valued space. The fuzzifier converts a crisp value to a fuzzy
the rules. Since then, fuzzy logic control has attracted great at- number while the defuzzifier converts the inferred fuzzy conclu-
tention from both academic and industrial communities. Many sion to a crisp value.
Based on the differences of fuzzy control rules and their
Manuscript received January 19, 2006; revised April 22, 2006 and July 25, generation methods, approaches to fuzzy logic control can be
2006. The work described in this paper was supported in part by a grant from
City University of Hong Kong under Project SRG 7001954, and in part grants
roughly classified into the following categories: i) Conven-
from Australian Research Council, the University of New South Wales, and the tional fuzzy control; ii) fuzzy proportional-integral-derivative
Research Grants Council of Hong Kong. (PID) control; iii) neuro-fuzzy control; iv) fuzzy-sliding mode
The author is with the Department of Manufacturing Engineering and En- control; v) adaptive fuzzy control; and vi) Takagi–Sugeno (T–S)
gineering Management, City University of Hong Kong, Kowloon, Hong Kong
(e-mail: [email protected]). model-based fuzzy control. However, it should be noted that the
Digital Object Identifier 10.1109/TFUZZ.2006.883415 overlapping among these categories is inevitable. For example,

1063-6706/$20.00 © 2006 IEEE

FENG: A SURVEY ON ANALYSIS AND DESIGN OF MODEL-BASED FUZZY CONTROL SYSTEMS 677

Fig. 1. Basic structure of fuzzy control systems.

conventional fuzzy control can be adaptive, fuzzy PID control These methods of conventional fuzzy control are essentially
can be tuned by neuro-fuzzy systems, or neuro-fuzzy control heuristic and model free. The fuzzy control “IF-THEN” rules
is adaptive in nature in many cases. Publications on the topic are obtained based on an operator’s control action or knowl-
of fuzzy logic control are so huge that an exhaustive list is edge. It is obvious that the design method works well only in
impossible. Instead, only a very selective list, in fact a small the case where an operator plays an important role in control-
portion of them, is given in the end of this paper. Many excellent ling the system. Even though the performance of such control
works are unfortunately missed. Moreover, this survey paper is scheme is generally satisfactory, stability issue of the closed
not able to cover all these categories of fuzzy logic control in loop fuzzy control system is often criticized in the earlier de-
detail. Instead it will briefly review all these categories in the velopment of these methods though the authors in [20] provide
next section and then focus on T–S model-based fuzzy control a stability analysis of fuzzy control systems via a heuristic ap-
in more detail in the rest of this paper. Therefore, the main proach. Moreover, design of such control systems suffers from
purpose of this paper is to survey state of the art of approaches lack of systematic and consistent approaches. Thus great efforts
to systematic analysis and design of model based, in particular, have been devoted to stability analysis and controller design
T–S model-based fuzzy control systems which have been issues of conventional fuzzy control systems, and various ap-
developed during the last few years. proaches have been developed. The key idea of these approaches
The rest of the paper is organized as follows. Section II is to regard a fuzzy controller as a nonlinear controller and
briefly reviews the general approaches to fuzzy logic control. embed the stability and/or control design problem of fuzzy con-
Section III formulates T–S fuzzy models and discusses their trol systems into conventional nonlinear system stability theory.
universal function approximation capability. Section IV sum- The typical approaches include describing function approach
marizes main results on stability analysis of T–S fuzzy systems. [136], cell-state transition [132], Lure’s system approach [59],
Sections V–VII present control design approaches based on [208], Popov’s theorem [91], circle criterion [226], [244], [252],
common (or global) quadratic Lyapunov functions, piecewise conicity criterion [69], sliding mode control [120], and hyper-
quadratic Lyapunov functions, and fuzzy (or non-quadratic) stability [21], [226], among others. However, a general system-
Lyapunov functions, respectively. Concluding remarks, per- atic theory for stability analysis and control design of conven-
spectives and challenges of model based fuzzy control in future tional fuzzy control systems is still out of reach. Additional ref-
are discussed in Section VIII. erences on the topic of conventional fuzzy control can be found
in [5], [62], [66], [88], [99], [108], [158], [160]–[162], [164],
[165], [173], [176], [233], [247], [278], [325], and [343].
II. BRIEF REVIEW OF FUZZY LOGIC CONTROL
B. Fuzzy PID Control
A. Conventional Fuzzy Control (Mamdani Type Fuzzy Control)
Conventional PID controllers are still the most widely
Mamdani and Assilian’s fuzzy control [199], [200], which adopted method in industry for various control applications,
is classified as Type-I fuzzy control systems by Sugeno [259], due to their simple structure, ease of design, and low cost in
has been replicated for many different control processes. For implementation. However, PID controllers might not perform
example, the authors in [137] develop a fuzzy control algorithm satisfactorily if the system to be controlled is of highly non-
for a warm water plant. Ostergaard [229] presents results of linear and/or uncertain nature. On the other hand, conventional
experiments with a fuzzy control algorithm for a small scale fuzzy control has long been known for its ability to handle
heat exchanger. There are many other applications of conven- nonlinearities and uncertainties through use of fuzzy set theory.
tional fuzzy control, including robot [10], [289], [314], [319], It is thus believed that by combining these two techniques
stirred tank reactor [146], traffic junction [237], steel furnace together a better control system can be achieved.
[153], cement kilns [288], automobile [16], [218], [260], waste- The name of fuzzy PID control has been widely used in lit-
water treatment [279], aircraft [58], [161], missile autopilot erature with all sorts of different meanings. For example, the
[73], motor [100], network traffic management and congestion authors in [114] suggest that if a fuzzy controller is designed
control [131], [169], bioprocesses [111], fusion welding [15], (or implied equivalently) to generate control actions within PID
and so on. In addition, fuzzy control has been widely used in concepts like a conventional PID controller, then it is called the
various consumer electronic devices such as video cameras, fuzzy PID controller. In this aspect, the conventional fuzzy con-
washing machines, TV, and sound systems in the late 1980s troller developed by Mamdani and Assilian [199], [200] is in
and early 1990s [106], [156], [170], [172], [219], [255], [263], fact a two-input fuzzy PI controller. Moreover, this conventional
[311]. fuzzy controller can be further classified as the “direct-action”
678 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 14, NO. 5, OCTOBER 2006

type [201] of fuzzy PID controllers, since its fuzzy inference de- [63], [72], [90], [123], [124], [138], [163], [177], [178], [186],
duces a control action output directly to control a system. In con- [187], [193], [205], [209], [217], [271], [294], [305], [306],
trast with “direct-action,” another type of fuzzy PID controllers [342]. The process of fuzzy reasoning is realized by neural net-
is classified as “gain-scheduling” [107], [344], for the reason works, whose connection weights correspond to the parameters
that controller gains change as operating condition or dynamics of fuzzy reasoning [38], [123], [124], [135], [187], [220], [231],
of a system varies. [232], [264]. Using back-propagation type, or reinforcement
Generally the “direct-action” type of fuzzy PID controllers type, or any other type neuro network learning algorithms, a
is able to do as well as conventional PID controllers. However, neuro-fuzzy control system can identify fuzzy control rules
high cost of setting up a fuzzy control system would usually dis- and learn (tune) membership functions of the fuzzy reasoning,
courage replacing a conventional PID controller with a “direc- and thus realize the neuro-fuzzy control. An excellent survey is
tion-action” fuzzy PID controller. As suggested by Chiu [57], given in [212] for neuro-fuzzy rule generation in a more gen-
it is the “gain-scheduling” type of fuzzy PID controllers that is eral setting of soft computation. Other topics of interest related
more likely to gain acceptance from industry. In addition, it is to this class of control scheme include tuning parameters in
shown that many fuzzy PID controllers are nonlinear PID con- neuro-fuzzy controller via genetic algorithm [72], [209], [249],
trollers and perform better than conventional PID controllers in [306], tuning PID controllers via fuzzy neural networks [250],
most cases [53], [57], [107], [113], [114], [179], [202], [203], self-organizing or adaptive neuro-fuzzy control [63], [177],
[215], [270], [316], [344]. [178], [193], [217], [294], and input–output stability analysis
Other topics of interest related to fuzzy PID control include based on small gain theorem [89]. Additional references on
tuning of fuzzy PID parameters [202], [215], [310], [316], op- the topic of neuro-fuzzy control can be found in [64], [89],
timal fuzzy PID controller based on genetic algorithm [113], [93], [149], [174], [185], [189], [222], [223], [227], [292], and
[270], realization of conventional PID controllers by fuzzy con- [329]. It should be noted that the T–S fuzzy model is one of the
trol method [213], improved robust fuzzy-PID controller with general fuzzy systems used to realize the neuro-fuzzy control
optimal fuzzy reasoning [179], and stability of fuzzy PID con- in this category, for example, see [129], [287], [292], and [329].
trollers [54], [253]. The author in [53] gives excellent overview One of the main advantages of neuro-fuzzy control is that
on fuzzy PID controllers in general, including adaptive fuzzy it does not basically require information on the mathematical
PID control and applications of fuzzy PID control. One major model of a system to be controlled. Thus this class of fuzzy
limitation of fuzzy PID control is the difficulty of its systematic control offers a new avenue in solving many difficult control
design with consistent and guaranteed performance. Additional problems in real life where the mathematical model of a system
references on the topic of fuzzy PID can be found in [8], [108], might be hard, if not impossible, to obtain. However, one of its
[182], [211], [216], [230], [250], [254], [262], and [325]. major limitations is the systematic analysis of stability of the
closed loop control systems and convergence of the learning
algorithms in the context of the closed loop control systems.
C. Neuro-Fuzzy Control

Neuro control, more precisely neural network control, and D. Fuzzy Sliding-Mode Control
fuzzy control are two of the most popular intelligent control It is well known that sliding-mode control provides a
techniques. They are similar in many ways. For example, both robust approach to controlling nonlinear systems with un-
of them are basically model-free control techniques, both are certainties [290], [349]. Its salient features include good
able to store knowledge and use it to make control decisions, control performance for nonlinear systems, applicability to
and both are able to provide robustness of control to certain ex- multiple-input–multiple-output (MIMO) systems, and most
tent with respect to system variations and external disturbances. importantly, robustness to parameter changes and/or external
However, the two techniques are different in their ways to obtain disturbances. It however often results in chattering phenomena
knowledge. Neuro control acquires knowledge mainly through due to its discontinuous switching which arises from its digital
data training (or learning). This could be an advantage as it lets implementation. Although a fuzzy controller is shown to be
the data “speak” for itself, but sometimes a disadvantage if the similar to a modified sliding mode controller [234], the key
training data set does not fully represent the domain of interest. idea of fuzzy sliding model control is to combine or integrate
Fuzzy control, in particular conventional fuzzy control, on the fuzzy control and sliding mode control in such a way that the
other hand mainly obtains qualitative and imprecise knowledge advantages of both techniques can be realized. One approach
via an operator or expert’s perspective. is that a sliding mode controller is equipped with capability
As the two control techniques complement to each other, of handling fuzzy linguistic qualitative information [50], [94],
that is, neuro control providing learning capabilities and high [235]. A direct benefit of such control is that fuzzy logic can
computation efficiency in parallel implementation, and fuzzy effectively eliminate chattering through construction of fuzzy
control providing a powerful framework for expert knowledge boundary layers which replace crisp switching surfaces [94],
representation, the combination or integration of the two tech- [101], [121]. Another approach is to design fuzzy control sys-
niques have attracted lots of attention from control community. tems in a way of conventional fuzzy control, fuzzy PID control,
A typical combination of these two techniques is the so-called or model based fuzzy control, and then to add a supervisory
neuro-fuzzy control, which is basically a fuzzy control aug- sliding model controller to not only guarantee stability but also
mented by neural networks to enhance its characteristics like improve robust performance of the closed-loop control systems
flexibility, data processing capability, and adaptability [17], [80], [206], [298].
FENG: A SURVEY ON ANALYSIS AND DESIGN OF MODEL-BASED FUZZY CONTROL SYSTEMS 679

Another important advantage of fuzzy sliding mode control is likely only semiglobal stabilization can be achieved if no
that stability analysis and controller design issue of fuzzy con- supplementary control strategy is employed.
trol systems can be addressed within the framework of sliding Other topics of interest include improved adaptive fuzzy
mode control [50], [55], [120], [235], [257], and the well devel- control schemes with smaller number of tuning parameters or
oped techniques of sliding mode control can be applied [290], better performance [87], [320], [321], robust adaptive fuzzy
[349]. Other topics of interest in fuzzy sliding mode control in- controller with various kinds of performances with respect to
clude using genetic algorithms to tune fuzzy membership func- external disturbances [39], [44], [97], fuzzy model reference
tions of such controllers [50], [188], decoupling of the high-di- adaptive control [95], [150], [324], using genetic algorithms
mensional systems into subsystems with lower dimensionality to adaptively tuning membership functions [190], and self-or-
[196], use of adaptive fuzzy systems in parameter tuning of ganizing schemes to tune fuzzy membership functions [4],
sliding-mode controllers [68], and adaptive fuzzy sliding mode [189]. Fusion of adaptive techniques and sliding mode control
control [13], [55], [63], [67], [116]–[118], [274], [280]. In ad- techniques are presented in [13], [55], [63], [67], [116]–[118],
dition, the authors in [134] present an excellent survey on the [274], and [280]. Comparison of adaptive fuzzy control to
fusion of computationally intelligent methodologies, including conventional adaptive control is reported in [228]. Additional
fuzzy logic, and sliding model control. Additional references references on the topic of adaptive fuzzy control can be found
on the topic of fuzzy sliding mode control can be found in [9], in [1], [5], [49], [64], [68], [93], [96], [139], [174], [193], [205],
[119], [180], [184]–[186], [305], [309], and [313]. [217], [222], [223], [227], [238], [242], [257], [281], [306],
[315], [330], and [345].
E. Adaptive Fuzzy Control
F. T–S Model-Based Fuzzy Control
Adaptive control refers to the control of partially known T–S fuzzy model [265], also called the Type-III fuzzy model
systems with some kind of adaptation mechanism. Most works by Sugeno [259], is in fact a fuzzy dynamic model [25], [28],
in adaptive control are based on the assumption of linear or [29]. This model is based on using a set of fuzzy rules to describe
simplified non-linear mathematical models of systems to be a global nonlinear system in terms of a set of local linear models
controlled. In fact, adaptive control of linear systems and certain which are smoothly connected by fuzzy membership functions.
special classes of nonlinear systems has been well developed This fuzzy modelling method offers an alternative approach to
from the late 1970s to the 1990s [96], [122], [155], while describing complex nonlinear systems [28], [71], [127], [269],
adaptive control of general nonlinear systems still presents a [326], [336], and drastically reduces the number of rules in mod-
challenge to control community. Nevertheless, mathematical elling higher order nonlinear systems [259]. Consequently, T–S
models might not be available for many complex systems in fuzzy models are less prone to the curse of dimensionality than
practice, and the adaptive control problem of these systems is other fuzzy models. More importantly, T–S fuzzy models pro-
far from being satisfactorily resolved. vide a basis for development of systematic approaches to sta-
Following the similar idea in neural networks [246] for their bility analysis and controller design of fuzzy control systems in
universal function approximation capability, it is shown [301] view of powerful conventional control theory and techniques.
that a fuzzy system is capable of approximating any smooth A great number of theoretical results on function approxima-
nonlinear functions over a convex compact region. Other ex- tion, stability analysis, and controller synthesis have been devel-
cellent works on the topic of function approximation of fuzzy oped for T–S fuzzy models during the last ten years or so. T–S
systems can be found in [326] and [336]–[339]. Based on this fuzzy models are shown to be universal function approximators
function approximation capability of fuzzy systems, the author in the sense that they are able to approximate any smooth non-
in [298] presents an adaptive fuzzy controller for affine non- linear functions to any degree of accuracy in any convex com-
linear systems with unknown functions. Fuzzy basis function pact region [28], [71], [127], [269], [326], [336]. This result
provides a theoretical foundation for using T–S fuzzy models
based fuzzy systems are used to represent those unknown non-
to represent complex nonlinear systems approximately. Based
linear functions. The parameters of the fuzzy systems including
on the differences of design approaches, the methods for sta-
membership functions characterizing linguistic terms in fuzzy
bility analysis and control design of T–S fuzzy systems can be
rules are updated according to some adaptive laws which are
roughly classified into the following six categories: i) simple
derived based on Lyapunov stability theory. Since then, a great local controller design and stability checking; ii) stabilization
number of works on adaptive fuzzy control have been reported, with/without various performance indexes such as and
see for example, [4], [18], [23], [41], [44], [70], [87], [97], control based on a nominal linear model and a single quadratic
[103], [150], [166], [171], [243], [256], [280], [291], [320], Lyapunov function; iii) stabilization with/without various per-
[321], [340], [341], and [346]. The key idea of these works is to formance indexes based on a common quadratic Lyapunov func-
use fuzzy systems to approximate unknown nonlinear functions tion; iv) stabilization with/without various performance indexes
in nonlinear systems and to represent the fuzzy systems in the based on a piecewise quadratic Lyapunov function; v) stabiliza-
form of linear regression with respect to unknown parameters tion with/without various performance indexes based on a fuzzy
and then to apply the well developed adaptive control tech- Lyapunov function; and vi) adaptive control when parameters
niques. However, it should be noted that some kinds of robust of T–S fuzzy models are unknown.
approaches have to be adopted for adaptive fuzzy control due to The first category of methods is proposed in the earlier stage
the inherent approximation errors between the approximating of developments [24], [25], [30], [80], [133]. Its basic idea is
fuzzy systems and the original nonlinear functions, and most to design a feedback controller for each local model, to obtain
680 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 14, NO. 5, OCTOBER 2006

a global controller by combining the local controllers in cer- developments of continuous time counterparts have also been
tain way, and then to use some stability criteria to check sta- widely reported in literature.
bility of the resulting closed-loop fuzzy control system. Unfor-
tunately, this kind of methods suffers from a problem that the III. T–S MODEL AND UNIVERSAL FUNCTION APPROXIMATION
design process is not constructive in general and many steps of T–S fuzzy models or so-called fuzzy dynamic models can be
trial and error might be needed before an acceptable controller used to represent complex MIMO systems with both fuzzy infer-
design can be obtained. The main idea of the second category of ence rules and local analytic linear dynamic models as follows:
methods is to represent a T–S fuzzy model as a nominal linear
model with uncertainties around the equilibrium of the system,
which include all the nonlinearities of the T–S fuzzy model, IF is and is
and then to recast the control problem as a robust linear con-
THEN
trol problem with uncertainties [74], [83], [145]. In this way,
many available robust control synthesis approaches can be di-
rectly applied to or further developed for the T–S fuzzy systems. (3.1)
However, this kind of methods tends to be conservative since
one nominal model has to be assumed which might not be the where denotes the th fuzzy inference rule, the number of
case for many complex highly nonlinear systems, thus has not inference rules, are the fuzzy sets,
become a mainstream of research efforts in model based fuzzy the state vector, the input vector, the
control. output vector, and the matrices of the th local
The basic idea of categories iii)–v) of methods is to design model, and some measurable variables
a feedback controller for each local model and to construct a of the system, for example, the state variables. It is also assumed
global controller from the local controllers in such a way that without loss of generality that the origin is the equilibrium of the
global stability with/without various performance indexes of T–S fuzzy system (3.1).
the closed-loop fuzzy control system is guaranteed. The major It is noted that the local model in terms of
techniques that have been used include quadratic stabilization, in (3.1) only represents the properties of the system in a local
linear matrix inequalities (LMIs), Lyapunov stability theory, region and thus is referred to as the fuzzy local model.
bilinear matrix inequalities, and so on. The third category of By using a standard fuzzy inference method, that is, using a
methods is most popular to date [2], [3], [6], [12], [33]–[37], singleton fuzzifier, product fuzzy inference, and center-average
[40], [42], [46], [47], [56], [61], [109], [110], [112], [115], defuzzifier, the T–S fuzzy model in (3.1) can be rewritten as
[119], [125], [130], [140]–[144], [147], [148], [151], [159], [269]
[167], [168], [175], [181], [183], [191], [192], [195], [197],
[198], [206], [214], [221], [239], [267]–[269], [273], [275],
[276], [282]–[286], [293], [302]–[304], [312], [317], [322],
[323], [327]. It, however, requires that a common quadratic
(3.2)
Lyapunov function can be found for all the local subsystems in
a T–S fuzzy model, and this proves to be conservative in many where
cases. As a less conservative alternative, the fourth category of
methods, at the same time, has also been well developed [26],
[27], [29], [51], [52], [74], [76]–[78], [82]–[86], [104], [105],
[128], [157], [224], [272], [295]–[297]. The fifth category of
methods has attracted some attention recently but it presents
more challenges or difficulties [60], [98], [266], [307], [347]. (3.3)
The sixth category of methods is to deal with control of T–S
fuzzy systems when parameters of T–S fuzzy models are un-
known. The most works to date however are quite preliminary is the normalized membership function satisfying
in the sense that they only consider unknown parameters in
local linear models by assuming that the number of fuzzy rules
and membership functions are all known a priori [75], [81],
[126], [144], [238].
All these results on various approaches to fuzzy logic
control, in particular on approaches to T–S model-based fuzzy
control demonstrate that these methods provide systematic
tools for analysis and design of fuzzy control systems, and (3.4)
that conventional linear system control theories can be suitably
utilized and developed for analysis and design of model based
fuzzy control systems. In the next few sections, the more and is the grade of membership of in the fuzzy set .
detailed survey on the T–S fuzzy model based approaches It should be noted that the previous model is a nonlinear
will be presented. For the sake of presentation simplicity only model in nature since the membership functions are nonlinear
developments of discrete time T–S fuzzy systems will be functions of the premise variables which contain some or all of
focused in this paper. However, it should be noted that the the state variables in general. The previous T–S fuzzy model is
FENG: A SURVEY ON ANALYSIS AND DESIGN OF MODEL-BASED FUZZY CONTROL SYSTEMS 681

in fact the state space fuzzy model. Similarly, the input–output tion of the nonlinear system. The latter is particularly impor-
fuzzy model can also be defined [28]. tant in control design. However, this becomes a difficult mul-
Remark 3.1: T–S fuzzy models include two kinds of tiobjective identification problem. It has been shown that con-
knowledge: one is qualitative knowledge represented by fuzzy strained and regularized identification methods may improve in-
IF–THEN rules, and the other is quantitative knowledge rep- terpretability of constituent local models as local linearizations,
resented by local linear models. T–S fuzzy models have a and locally weighted least squares method may explicitly ad-
compatible structure with a two level control system with the dress the tradeoff between the local and global accuracy of T–S
lower level providing basic feedback control and the higher fuzzy models.
level providing supervisory control or scheduling. By using Before we present a survey on approaches to stability anal-
T–S fuzzy models, one can formulate these two kinds of knowl- ysis and controller synthesis of T–S fuzzy systems in the next
edge into a unified mathematical framework. This framework few sections, we first give a result on universal function approx-
provides a possibility for developing a systematic analysis and imation of T–S fuzzy models.
design method for complex nonlinear control systems.
Remark 3.2: T–S fuzzy models are, to certain extent, sim- A. Universal Function Approximation
ilar to the concept of typical piecewise linear approximation Consider a general nonlinear discrete-time system described
methods in nonlinear control. Control of a nonlinear system by by a state–space model of the form
piecewise linearization is approached by linearizing the system
around a number of nominal operating points, and then applying
linear feedback control methods to each local linear model (3.5)
[152]. However, analysis of the resulting closed-loop control where is the state vector and the input
system is in general difficult and stability or performance of vector of the system. The function satisfies the
the system can hardly be guaranteed due to the approximation. following assumption.
On the other hand, T–S fuzzy models consist of a set of local Assumption 3.1: There exists an equilibrium
linear models smoothly connected by membership functions such that and , that is, has the second-
yielding global models of the systems. Thus, T–S fuzzy models order continuous derivative with respect to and .
provide a way of designing controllers based on local linear Let be the set of all systems of the form (3.5) satisfying
models and analysing stability or performance based on the the Assumption 3.1. Let be the set of all T–S fuzzy models
global nonlinear model, and this also provides a framework to of the form (3.1) or (3.2). It is also assumed that
consolidate the general industrial practice of nonlinear control here, that is, the membership functions depend on the system
system designs such as gain scheduling control. state variables. It has been shown [28], [31] that T–S fuzzy
Remark 3.3: When only the constant term in the consequent models are universal function approximators in the sense that
part of (3.1) is present, the fuzzy model is called Type II fuzzy given any there exists a fuzzy model
model by Sugeno in [259]. It is a simplified special case of more that will approximate
general T–S fuzzy models. Sugeno [259] presents a detailed sta- to any degree of accuracy in any convex compact region. More
bility study for this type of fuzzy models by using Lyapunov sta- accurately, let and be compact sets in , respec-
bility theory, and also gives an excellent survey on the stability tively, and
issues of more general fuzzy control systems.
Identification of T–S fuzzy models has attracted great atten-
tion from control community and a number of results have been
obtained [7], [28], [127], [214], [276], [287], [329]. There are (3.6)
basically two kinds of approaches, one is to linearize the orig- be the -metric, then is a metric space. The
inal nonlinear system in a number of operating points when the following theorem shows that is dense in
model of the system is known, which is straightforward, and .
the other is based on the data generated from the original non- Theorem 3.1 [28], [31]: For any given on the
linear system when its model is unknown (or in the form of black compact set and arbitrary , there exists
box). The authors in [28] present an approach to identification an such that
of T–S fuzzy models, including identification of the number of
fuzzy rules (or the number of local linear models) and parame-
ters of fuzzy membership functions by using a fuzzy clustering (3.7)
method, and identification of parameters of local linear models
by using a least squares method. The objective is to minimize Remark 3.4: It is noted that there are many other results
the global nonlinear prediction error between T–S fuzzy models on universal function approximation property of other kinds of
and the corresponding original nonlinear systems. The authors fuzzy systems, for example, the fuzzy systems with fuzzy basis
in [127] present a study of interpretation capability of T–S fuzzy functions [301], [337]–[339]. It is also recently proved that the
models and propose a method for their identification. The key more commonly used T–S fuzzy model in (3.1) with is
idea is to achieve not only accurate global nonlinear prediction also a universal function approximator [269].
but also at the same time accurate local models in the sense that Remark 3.5: It should be noted that the result in Theorem 3.1
the local models are close approximations to the local lineariza- only concerns the approximation between two static nonlinear
682 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 14, NO. 5, OCTOBER 2006

functions, that is, and . However, the error be- ii) if there exists a positive–definite matrix such that the
tween the states of two dynamic systems described in (3.5) and following LMIs are satisfied:
(3.2), which are corresponding to and , might
grow as time goes. Therefore, much care has to be taken in
dealing with the approximation between two dynamic systems (4.5)
instead of two static functions.
For the ease of presentation, we will mainly use the more
Remark 4.1: The equivalence of (4.4) and (4.5) can be easily
commonly used T–S fuzzy models, that is, the model (3.1) with
established by using the Schur complement together with
in the next few sections, though it is shown that T–S
[19]. However, the form of (4.5) is more suitable to con-
fuzzy models with affine terms has much improved function ap-
troller synthesis which can be observed in the next section. It is
proximation capabilities [71].
also noted that (4.5) implies that its feasible solution of is pos-
itive definite. However, the term of “positive definite” instead of
IV. STABILITY ANALYSIS “symmetric” will still be used in subsequent presentation of the-
orems to avoid any possible confusion.
Consider the T–S fuzzy model in (3.1) with and
Remark 4.2: Conditions (4.4) or (4.5) are linear matrix in-
as follows:
equalities in the variable or , respectively. The feasibilities
of these LMIs, as well as other LMIs in the rest of this section
IF is and is and the subsequent sections, are easy to be tested by the avail-
THEN able software package LMI Toolbox [92].
(4.1) Remark 4.3: It has been noted that common quadratic Lya-
punov functions tend to be conservative, and even worse, might
which can also be described by not exist for many complex highly nonlinear systems as demon-
strated in [77], and [128]. This is one of the main limitations of
this kind of approaches.
(4.2)
B. Analysis Based on Piecewise Quadratic
Lyapunov Functions
Stability analysis of T–S fuzzy systems has been pursued Due to the drawback of common quadratic Lyapunov func-
mainly based on Lyapunov stability theory but with different tions, it is thus desirable to develop less conservative stability
Lyapunov functions. One of them is the so-called common results for T–S fuzzy systems. Piecewise quadratic Lyapunov
(or global) quadratic Lyapunov functions, another one is the functions are one of the options available. In order to facili-
so-called piecewise quadratic Lyapunov functions, and the tate development of approaches based on piecewise quadratic
third one is the so-called fuzzy (or non-quadratic) Lyapunov Lyapunov functions, one needs partition of the premise variable
functions. In the rest of this section we will present stability space, or partition of the state space in the case of .
analysis results of T–S fuzzy systems based on these Lyapunov The following partition will be referred to be the first kind in
functions, respectively. sequel [26], [27], [29].
Define regions in the premise variable space as follows:
A. Analysis Based on Common Quadratic Lyapunov Functions
One of the first results on stability analysis based on common
quadratic Lyapunov functions is suggested in [268], and since (4.6)
then numerous modifications and improved methods have been
proposed. By defining a Lyapunov function candidate as Then the global model of the T–S fuzzy system (4.2) can be
expressed in each local region as

(4.3) (4.7)

where the matrix is positive definite, the following result can where
be readily obtained [268].
Theorem 4.1: The T–S fuzzy system (4.1), or equivalently,
(4.2) is globally exponentially stable
i) if there exists a positive definite matrix such that the
following linear matrix inequalities (LMIs) are satisfied:
Remark 4.4: It is noted that the number of regions in this
kind of partition is the same as the number of fuzzy rules or the
(4.4) number of local linear models, and that the fuzzy model (4.7)
is different from the local model in the T–S fuzzy model (4.1)
or, equivalently; because it considers all interactions among the local models of
FENG: A SURVEY ON ANALYSIS AND DESIGN OF MODEL-BASED FUZZY CONTROL SYSTEMS 683

(4.1) in terms of uncertainty and is in fact the global fuzzy of the resulting stability analysis since the worst case of uncer-
model (4.2) expressed in the local region . tainties is considered as shown in (4.8). Authors in [26], [27],
For purpose of stability analysis and stabilization, we intro- and [29] present some approaches to searching for approximate
duce the following upper bounds for the uncertainty term of the upper bounds instead of the worst case bounds for these uncer-
fuzzy model (4.7): tainties.
Another approach to space partition, which is referred to be
(4.8) the second kind in sequel, is suggested by [128]. The parti-
tion is based on the natural induction of the fuzzy system (4.1)
Remark 4.5: It is noted that there are many ways to obtain
into a number of polyhedral regions of the
these upper bounds, the interested readers can refer to [26], [27],
premise variable space. The regions consist of crisp (operating)
and [29] for details.
and fuzzy (interpolation) regions. The crisp region is defined as
In addition, we define a set that represents all possible
the region where for some , all other mem-
system transitions among regions, that is
bership functions evaluate to zero. The system dynamic of a
crisp region is given by one of local models of the fuzzy system
(4.1). On the other hand, the fuzzy region is defined as the re-
(4.9) gion where and the system dynamic is given
Then, we are ready to present a stability result based on the by a convex combination of several local linear models. In the
following piecewise quadratic Lyapunov function candidate, extreme case where all the regions of a T–S fuzzy system are
crisp, that is, for some and all other membership
functions are equal to zero, then the global fuzzy model (4.2)
(4.10) becomes a piecewise linear system

Theorem 4.2 [77]: The T–S fuzzy system (4.1), or equiva- (4.15)
lently (4.7) is globally exponentially stable Remark 4.8: It is noted that the number of regions in the set
i) if there exists a set of positive–definite matrices for the second kind of space partition is different from the
such that the following LMIs are satisfied: number of the local linear models in general which is the number
of regions in the first kind of space partition approach.
With such a partition, we can rewrite the global fuzzy model
(4.2) in each region as a convex combination of linear models
(4.11)

(4.12) (4.16)
or, equivalently; with . For each region ,
ii) if there exists a set of positive–definite matrices the set contains the indexes for the system matrices used
such that the following LMIs are satisfied: in the interpolation within that region. For a crisp region,
contains a single element.
It is noted that in comparison with the global fuzzy model
(4.2), the fuzzy model (4.16) is described in each local region.
This is similar to (4.7) for the first kind of space partition ap-
(4.13) proach.
Similar to (4.9), we define a set that represents all possible
transitions among regions of the system (4.16), that is,

(4.14)
(4.17)
Remark 4.6: Conditions (4.11) and (4.12) or, equivalently,
(4.13) and (4.14) are LMIs in the variables or, equiv- Then, we are ready to present a stability result based on a sim-
alently, . A solution to these inequalities ilar piecewise quadratic Lyapunov function candidate defined in
ensures , or equivalently (4.10).
to be a Lyapunov function for the system. Theorem 4.3 [295]: The T–S fuzzy system (4.1), or equiva-
The LMIs in (4.11) or (4.13) guarantee that the function de- lently (4.16) is globally exponentially stable
creases along all system trajectories within each region. The i) if there exists a set of positive–definite matrices
LMIs in (4.12) or (4.14) guarantee that the function decreases such that the following LMIs are satisfied:
when the system transits from one region to another.
(4.18)
Remark 4.7: It is noted that the uncertainty terms are intro-
duced in (4.7) for the partition of the premise variable space (4.19)
defined in (4.6), and this would normally lead to conservatism or, equivalently;
684 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 14, NO. 5, OCTOBER 2006

ii) if there exists a set of positive–definite matrices more involved or difficult than that for the discrete time coun-
such that the following LMIs are satisfied: terparts due to the fact that the membership functions are in gen-
eral functions of system states and the derivative of fuzzy Lya-
punov functions involves derivatives of the membership func-
(4.20)
tions and, thus, the derivatives of the system states. As a result,
the derivative of the fuzzy Lyapunov function becomes much
more complex, in fact it becomes a nonlinear function in terms
(4.21) of the system matrices which thus leads to difficulty in stability
analysis.
Remark 4.9: It is noted that when the positive–definite ma- Stability analysis of T–S fuzzy models with affine terms, that
trices in (4.18) and (4.19) [or, equivalently, (4.20) and (4.21)] is is, for some in (4.1), is much more involved. The anal-
chosen as a common one, that is, ysis based on common quadratic Lyapunov functions is sug-
(or, equivalently, ), then the gested in [140] and [141], and the analysis based on piecewise
result of Theorem 4.3 reduces to that of Theorem 4.1. It thus can quadratic Lyapunov functions is suggested in [77], [128], and
be easily seen that common quadratic Lyapunov functions are a [295]. Here, for illustration, we present a result based on the
special case of the more general piecewise quadratic Lyapunov second kind of space partition and piecewise quadratic Lya-
functions, and the latter is less conservative. However, compu- punov functions.
tation cost of the latter would be higher in general. In this case, we define as the set of indexes for the
regions that contain the origin and the set of indexes
C. Analysis Based on Fuzzy Lyapunov Functions for the regions that do not contain the origin. For convenient
notation, we also define
In addition to stability results based on common or piecewise
quadratic Lyapunov functions, the authors in [60], [98], [266],
[307], and [347] present stability results based on the so-called
nonquadratic or fuzzy Lyapunov functions defined as (4.25)

where it is noted that for all with . Then,

(4.22) using this notation, similar to (4.16), the fuzzy system (4.1) with
can be expressed as

Theorem 4.4 [347]: The T–S fuzzy system (4.1), or equiva-

lently (4.2) is globally exponentially stable
i) if there exists a set of positive–definite matrices (4.26)
such that the following LMIs are satisfied: The following Lyapunov function candidate is used,

(4.23) (4.27)

or, equivalently;
ii) if there exists a set of positive–definite matrices This function combines the power of quadratic Lyapunov func-
such that the following LMIs are satisfied: tions near the equilibrium point with the flexibility of piecewise
affine functions in the large.
Since the matrix or is only used to describe the Lya-
punov function in the local region , the -procedure [19] can
(4.24) be used to reduce conservatism of stability results. To this end,
the characteristics or information of the local regions can be
Remark 4.10: It is noted that (4.24) implies that its feasible utilized. As partitions of premise variable space induced from
solution of is positive definite. It is also noted that when the fuzzy membership functions are polyhedra, the matrices
positive definite matrices in (4.22) and (4.23) [or, equivalently, can be constructed for each region such that
(4.24)] is chosen as a common one, that is,
(or, equivalently, ), then the
result of Theorem 4.4 reduces to that of Theorem 4.1. It thus
can be easily seen that common quadratic Lyapunov functions
are a special case of the more general fuzzy Lyapunov func- where with being an matrix,
tions, and the latter is less conservative. However, similar to being an vector, and moreover for .
piecewise quadratic Lyapunov functions, computation cost of It should be noted that the above vector inequality is defined as
the latter would be much higher. element-wise, that is, each entry of the vector is nonnegative.
Remark 4.11: In the case of continuous time T–S fuzzy sys- Remark 4.12: A systematic procedure for constructing these
tems, stability analysis via fuzzy Lyapunov functions is much matrices for a given T–S fuzzy system can be found
FENG: A SURVEY ON ANALYSIS AND DESIGN OF MODEL-BASED FUZZY CONTROL SYSTEMS 685

in [128]. The procedure is directly based on the information in In the following sections, three categories of controller
the fuzzy rule base. synthesis approaches will be presented for T–S fuzzy systems
Then, we are ready to present the following stability result with local linear (instead of affine) models based on common
[77], [296]. quadratic Lyapunov functions, piecewise quadratic Lyapunov
Theorem 4.5: The T–S fuzzy system (4.1) with functions, and fuzzy Lyapunov functions respectively. For sake
or, equivalently, (4.26) is globally exponentially stable, if of clarity of presentation, only stabilization controller synthesis
there exist symmetric matrices, , and will be considered.
symmetric matrices such that have
nonnegative entries, and the following LMIs are satisfied: V. STABILIZATION BASED ON COMMON QUADRATIC
LYAPUNOV FUNCTIONS
Controller synthesis of T–S fuzzy systems based on common
(4.28)
quadratic Lyapunov functions has been well developed during
the last few years. The problem can be efficiently solved by LMI
(4.29) techniques. In this section, some basic stabilization results will
(4.30) be presented.
Consider the T–S fuzzy system without affine terms in (3.1)
rewritten as
(4.31)
IF is AND is
(4.32) THEN (5.1)

(4.33)

or, equivalently
(4.34)

(4.35) (5.2)
where , and are defined in (3.3).
where for (4.34), Two kinds of control schemes are mainly used, one is the
for (4.35). smooth fuzzy control scheme defined as
Remark 4.13: The previous conditions are linear matrix
inequalities in the variables , and . A solu- IF is AND is
tion to these inequalities ensures defined in (4.27) to be THEN (5.3)
a Lyapunov function for the T–S fuzzy system. The LMIs in
(4.28) and (4.30) for each region guarantee that the function
is positive definite, the LMIs in (4.29) and (4.31) guarantee which can be rewritten as
that the function decreases within each region, and the LMIs
in (4.32)–(4.35) guarantee that the function decreases when
the system transits from one region to another. In addition,
(5.4)
, and
in these LMIs are the terms of the -procedure used
to reduce conservatism of the stability result. It should also be and the other is the switching control scheme defined as
noted that the matrices are not required
to be positive definite.
Remark 4.14: It is noted that most results on controller syn- (5.5)
thesis of T–S fuzzy systems with affine local models as in (4.1)
can only be cast as a solution to bilinear matrix inequalities due where the space partition of the first kind is adopted here. The
to the S-procedure [78], [82], [140], [141], which is not a convex first control scheme is often called the parallel distributed com-
programming problem and might lead to difficulty in searching pensation while the second one is called the local compensation.
for a feasible solution. Some attempts have been made to cast The closed-loop fuzzy control system consisting of the T–S
the controller synthesis as a solution to linear matrix inequali- fuzzy system (5.2) and the smooth controller (5.4) can be de-
ties but the results appear to be very restrictive in the sense that scribed as
the matrices to characterize the regions for the S-procedure has
to be a square and invertible matrix [76]. In other words, if the
local models of T–S fuzzy systems are affine instead of linear, (5.6)
the controller synthesis warrants further study.
686 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 14, NO. 5, OCTOBER 2006

By defining a Lyapunov function candidate as Theorem 5.3: The closed-loop fuzzy control system (5.14)
is globally exponentially stable, if there exist a positive–definite
matrix and a set of matrices such that the following
(5.7) LMIs are satisfied:
where the matrix is positive definite, one can easily obtain
the following stabilization result [267], [269].
Theorem 5.1: The closed-loop fuzzy control system (5.6) is
globally exponentially stable, if there exist a positive–definite
matrix and a set of matrices such that the following (5.16)
LMIs are satisfied: Moreover, the controller gain for each local region is given by

(5.8)
(5.17)
Moreover, the controller gains are given by
Remark 5.2: By comparing the results in Theorems 5.1–5.3,
one notices that the numbers of LMIs in (5.8), (5.10)–(5.12), and
(5.9) (5.16) are , and , respectively. It
can be easily seen that the computation cost of (5.16) is smaller
Remark 5.1: It is noted that a number of improved results than that of (5.10)-(5.12), and much smaller than that of (5.8).
have been obtained which aim either to reduce the number of However, it should also be noted that the price of such computa-
LMIs in (5.8), or to reduce its conservatism, or both [142], [192], tion efficiency in (5.16) is the introduction of uncertainty terms
[198], [267], [275], [286], [304]. One typical result is given as as described in (5.14) which might lead to some extent of con-
follows [142]. servatism. In addition, the control law used in Theorem 5.3 is a
Theorem 5.2: The closed-loop fuzzy control system (5.6) is switching control law while that for Theorems 5.1 and 5.2 is a
globally exponentially stable, if there exist a positive–definite continuous control law.
matrix , a set of matrices , a set of positive–definite Remark 5.3: The stabilization results in Theorems 5.1–5.3
matrices , and a set of symmetric matrices have been extended to: i) robust controller designs of T–S fuzzy
such that the following LMIs are satisfied: systems with various performance indexes such as and
with respect to external disturbances [34], [36], [46], [47], [110],
[168], [192], [197], [283], [312], [317]; ii) controller designs
of uncertain T–S fuzzy systems with or without various perfor-
(5.10)
mance indexes [6], [34], [167], [168], [317]; and iii) controller
designs of time-delay T–S fuzzy systems with or without var-
ious performance indexes [33], [42], [56], [112], [168], [317],
[323].
Remark 5.4: Observer designs, filter designs, and output
(5.11) feedback control designs with or without various performance
indexes have also been well developed for T–S fuzzy systems
based on common quadratic Lyapunov functions and LMIs
.. .. .. .. (5.12) [3], [6], [12], [45], [47], [56], [151], [168], [192], [197], [198],
. . . .
[221], [267], [275], [282], [285], [312], [317].
A typical result on observer design of T–S fuzzy systems
Moreover, the controller gains are given by based on common quadratic Lyapunov functions is given below
[198].
Consider the following observer rule:
(5.13)
IF is AND is
On the other hand, with the switching control law (5.5) the
closed-loop fuzzy control system can be described as THEN
(5.18)

(5.14)
By defining the following upper bounds for the uncertainties which can be rewritten as
as those in (3.8) for

(5.15)
one has the following result. (5.19)
FENG: A SURVEY ON ANALYSIS AND DESIGN OF MODEL-BASED FUZZY CONTROL SYSTEMS 687

where is the estimated state vector, the estimated the closed-loop fuzzy control system can be described by
output vector, the observer gains to be determined,
and the rest of the variables and matrices are as defined in (3.1).
Then, the fuzzy observer error dynamic equation consisting (6.2)
of (5.1) and (5.18) can be given as
where the upper bounds for and are given in (4.8) and
(5.15), respectively. By using a piecewise quadratic Lyapunov
(5.20) function candidate of the form

where , and the following result is readily (6.3)

obtained [198].
Theorem 5.4: The fuzzy observer error system (5.20) is glob- one has the following stabilization result for the T–S fuzzy
ally exponentially stable, if there exist a positive–definite matrix system described in (5.1).
and a set of matrices such that the following LMIs Theorem 6.1 [52]: The closed-loop fuzzy control system
are satisfied: (6.2) is globally exponentially stable, if there exist a set of
positive–definite matrices and a set of matrices
such that the following LMIs are satisfied:

(5.21)

(6.4)

(5.22)
Moreover, the observer gains are given by
(6.5)
(5.23) Moreover, the controller gain for each local region is given by

Remark 5.5: It has been shown in [198] that the separa-

tion principle does hold for T–S fuzzy systems in the case of (6.6)
common quadratic Lyapunov functions. That is, controller de-
sign and observer design can be independently carried out based With space partition of the second kind and the switching
on common quadratic Lyapunov functions, and the resulting controller defined as
closed-loop fuzzy control system, with estimated state variables
to be used for state feedback control, will be asymptotically (6.7)
stable.
Most works to date on controller synthesis of various T–S the closed-loop fuzzy control system can be described by
fuzzy systems have been mainly based on common quadratic
Lyapunov functions as in the case of stability analysis. How-
ever, it has been noted that common quadratic Lyapunov func-
tions tend to be conservative and might not exist for many highly (6.8)
nonlinear systems. Thus, piecewise quadratic Lyapunov func-
where the set is defined in (4.16). By using the same piece-
tions have been suggested to be an alternative approach which
wise quadratic Lyapunov function candidate as in (6.3) one has
has demonstrated to be less conservative [77], [128]. In the next
the following stabilization result for the T–S fuzzy system de-
section, stabilization approaches to T–S fuzzy systems based on
scribed in (5.1).
piecewise quadratic Lyapunov functions will be outlined.
Theorem 6.2 [295]: The closed-loop fuzzy control system
(6.8) is globally exponentially stable, if there exist a set of posi-
VI. STABILIZATION BASED ON PIECEWISE QUADRATIC tive–definite matrices and a set of matrices,
LYAPUNOV FUNCTIONS such that the following LMIs are satisfied:
In this section, we will present stabilization methods based on
piecewise quadratic Lyapunov functions which are suggested in
[29], [52], [76], [77], [78], and [85]. With space partition of the
(6.9)
first kind and the switching controller defined as

(6.1) (6.10)
688 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 14, NO. 5, OCTOBER 2006

Moreover, the controller gain for each local region is given by Then the fuzzy observer error dynamic equation can be de-
scribed as

(6.11)

The stabilization results in Theorems 6.1 and 6.2 can be fur-

ther improved in the sense of less conservatism by introducing (6.16)
extra slack variables in LMIs as in [225]. It is noted where , and the following result is readied
that the matrices are not even required to be sym- obtained.
metric. One improved result for Theorem 6.2 is summarized in Theorem 6.4: The fuzzy observer error system (6.16) is glob-
the following theorem. ally exponentially stable, if there exist a set of positive definite
Theorem 6.3: The closed-loop fuzzy control system (6.8) is matrices and sets of matrices such that
globally exponentially stable, if there exist a set of positive–def- the following LMIs are satisfied:
inite matrices , sets of matrices , and
such that the following LMIs are satisfied:

(6.17)

(6.12)
(6.18)
Moreover, the observer gain for each region is given by
(6.13)
Moreover, the controller gain for each local region is given by
(6.19)

(6.14) Remark 6.3: Similar to the case of common quadratic Lya-

punov functions, it has been shown in [297] that the separa-
Remark 6.1: The conservatism and the number of LMIs in tion principle does hold for T–S fuzzy systems in the case of
Theorems 6.1–6.3 can be further reduced by employing the sim- piecewise quadratic Lyapunov functions. That is, controller de-
ilar technique as in [142, Th. 5.2]. sign and observer design can be independently carried out based
Remark 6.2: Similar to the common quadratic Lyapunov on piecewise quadratic Lyapunov functions, and the resulting
function case, stabilization controller designs based on piece- closed loop fuzzy control system, with estimated state variables
wise quadratic Lyapunov functions have been extended to: i) to be used for state feedback control, will be asymptotically
robust controller designs of T–S fuzzy systems with various stable.
performance indexes such as and with respect to
external disturbances [27], [52], [78], [82], [84], [105], [295], VII. STABILIZATION BASED ON FUZZY LYAPUNOV FUNCTIONS
[296]; ii) controller designs of uncertain T–S fuzzy systems In this section, we will present methods based on fuzzy (or
[74], [83]; and iii) controller designs of time-delay T–S fuzzy nonquadratic) Lyapunov functions which is suggested in [60],
systems [51]. Observer designs, filter designs, and output [98], [266], and [307]. With a controller defined in (5.3) and
feedback control designs with or without various performance rewritten as
indexes have also been developed for T–S fuzzy systems based
on piecewise quadratic Lyapunov functions and linear matrix
inequalities [51], [52], [79], [104]. (7.1)
A typical result on observer design of T–S fuzzy systems
based on piecewise quadratic Lyapunov functions is given
below. the closed loop fuzzy control system consisting of the T–S fuzzy
Consider the following piecewise fuzzy observer of the form: system (5.2) and the smooth controller (7.1) can be described as

(7.2)

By defining a Lyapunov function candidate as

(6.15)
where is the estimated state vector, the estimated (7.3)
output vector, the observer gain for each region ,
and the rest of the variables and matrices are as defined in (3.1) where the matrix is positive definite, one has the following
and (4.16). stabilization result for the T–S fuzzy system described in (5.1).
FENG: A SURVEY ON ANALYSIS AND DESIGN OF MODEL-BASED FUZZY CONTROL SYSTEMS 689

Theorem 7.1: The closed loop fuzzy control system (7.2) is

globally exponentially stable, if there exist a set of positive–def-
.. .. .. .. (7.11)
inite matrices , sets of matrices , and . . . .
such that the following LMIs are satisfied:
Remark 7.1: Not like the common or piecewise quadratic
Lyapunov function case, the controller synthesis based on fuzzy
Lyapunov functions has not been well studied yet though ap-
(7.4)
proaches to controller synthesis have been presented in [60]
Moreover, the controller gains are given by and [347]. By comparing (5.8) and (7.4), one notices that the
number of LMIs in (7.4) would be much larger than that of (5.8)
when the fuzzy rule number is large. This may lead to diffi-
(7.5) culty in many practical situations.
Remark 7.2: Similar to the case of common/piecewise
It is noted that the number of LMIs and the conservatism in quadratic Lyapunov functions, it is expected that the separation
Theorem 7.1 can be reduced by following the similar idea as in principle also holds for T–S fuzzy systems in the case of fuzzy
[142, Th. 6.2]. The result is given here. Lyapunov functions though its proof is not available yet in open
Theorem 7.2 [98]: The closed-loop fuzzy control system literature.
(7.2) is globally exponentially stable, if there exist a set Remark 7.3: As indicated in Section IV, the main difficulty
of positive–definite matrices , sets of ma- in using fuzzy Lyapunov functions arises when the continuous
trices , a set of positive definite time systems are considered. It appears that the difficulty cannot
matrices , and a set of symmetric matrices be easily overcome and will present a considerable challenge for
such that the following LMIs are satisfied: community of fuzzy logic control.

VIII. CONCLUSION
This paper presents a brief survey on analysis and design
(7.6) methods of model based fuzzy control systems. The particular
attention is given to the so-called T–S models or fuzzy dy-
namic models which consist of a family of local linear models
smoothly combined with fuzzy membership functions. The is-
sues of universal function approximation, stability analysis, and
(7.7) controller synthesis of T–S fuzzy models have been addressed.
It is shown that T–S fuzzy models are able to approximate
any smooth nonlinear functions to any degree of accuracy in
.. .. .. .. (7.8) any convex compact region. Approaches to stability analysis
. . . . and controller synthesis of T–S fuzzy systems based on either
common quadratic Lyapunov functions, piecewise Lyapunov
functions, or fuzzy (or nonquadratic) Lyapunov functions are
For the observer design, consider the smooth observer defined
then presented. It is believed that these methods provide a
in (5.18) and the resulting closed-loop observer error system
systematic approach to analysis and design of model based
defined in (5.20). Then the following result on observer design
fuzzy control systems, and they probably suggest an efficient
of T–S fuzzy systems based on fuzzy Lyapunov functions can
alternative way to solve more difficult general nonlinear control
be easily established.
problems.
Theorem 7.3: The fuzzy observer error system (5.20) is glob-
The approaches based on common quadratic Lyapunov func-
ally exponentially stable, if there exist a set of positive–definite
tions have been well developed for both discrete time and con-
matrices , sets of matrices , a set of
tinuous time T–S fuzzy systems, while the approaches based on
positive–definite matrices , and a set of symmetric
piecewise quadratic Lyapunov functions have had more chal-
matrices such that the following LMIs are
lenges. The early works [27], [29], [74], [83], [84], [104], and
satisfied:
[105] on controller synthesis based on piecewise quadratic Lya-
punov functions suffer from a drawback of some extra restrictive
boundary conditions which are in general hard to test or sat-
isfy. Those restrictive conditions have been removed from the
(7.9)
most recent works on stability analysis and controller synthesis
based on novel piecewise quadratic Lyapunov functions [51],
[52], [76]–[78], [82], [128], [295], [296]. However, many issues
still need to be addressed for these approaches based on piece-
wise quadratic Lyapunov functions or fuzzy Lyapunov func-
(7.10) tions. For example, if local models are affine for T–S fuzzy sys-
690 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 14, NO. 5, OCTOBER 2006

tems, the best available stabilization results are given by solving membership functions, fuzzy inferences, etc., in the con-
a set of bilinear matrix inequalities which is more difficult to text of stochastic variables and systems. Some results on
solve and computationally much more expensive. Is it possible probabilistic fuzzy logic and probabilistic fuzzy systems
to cast the problem as one by solving a set of linear matrix in- provide a possible way to address the issues on this topic
equalities instead? How less conservative are the approaches [194], [207].
based on piecewise quadratic Lyapunov functions or fuzzy Lya- In addition to the aforementioned challenges, the authors in
punov functions compared to the approaches based on common [245] have also listed a number of open problems and perspec-
quadratic Lyapunov functions? Which kind of approaches, the tives in a more general setting of fuzzy systems and control,
approaches based on piecewise quadratic Lyapunov functions such as fuzzy system identification, adaptive fuzzy control, ex-
or the approaches based on fuzzy Lyapunov functions, is less pert control, supervision and diagnosis, as well as the integration
conservative? How to further reduce the conservatism of these of fuzzy logic techniques and other artificial intelligence tech-
approaches? niques for the more sophisticated intelligent control systems. All
Moreover, there are many other challenges which have not these challenges also provide great opportunities for our fuzzy
been well studied in model based fuzzy control. logic control community.
• What kind of nonlinear systems can be well represented by
T–S fuzzy models? How to identify T–S fuzzy models to ACKNOWLEDGMENT
facilitate both accurate approximation and effective con-
The author would like to thank the Associate Editor and
troller synthesis? The authors in [127] and [339] have shed
anonymous reviewers for their constructive comments which
some light on the issues.
have improved the presentation of this paper. He would also
• Are there any other better techniques which can be used
like to thank his colleagues and students: Prof. S. G. Cao and
for stability analysis and controller synthesis of T–S fuzzy
Prof. N. W. Rees; Dr. C. K. Chak, Dr. Z. X. Han, Dr. L. Wang,
systems in addition to common/piecewise quadratic Lya-
Dr. C. L. Chen, Dr. M. Cheng, Dr. X. P. Guan, and Dr. S.
punov functions or fuzzy Lyapunov functions?
S. Zhou; and M. Chen, T. J. Zhang, and B. Yang for their
• How can one use as much information of T–S fuzzy models
collaboration and contribution in the area of fuzzy logic control
as possible to achieve more effective controller synthesis
over the last few years.
and/or better performance? It is known that much informa-
tion on fuzzy membership functions have not been used
in the approaches based on common/piecewise quadratic REFERENCES
Lyapunov functions or fuzzy Lyapunov functions. In fact, [1] T. Abdelazim and O. P. Malik, “An adaptive power system stabilizer
using on-line self-learning fuzzy systems,” in Proc. IEEE Power En-
only little information of membership functions has been gineering Society General Meeting, Toronto, ON, Canada, 2003, pp.
used in space partitions for the approaches based on piece- 1715–1720.
wise quadratic Lyapunov functions. [2] M. Akar and U. Ozguner, “Decentralized techniques for the analysis
and control of Takagi-Sugeno fuzzy systems,” IEEE Trans. Fuzzy Syst.,
• How can one design adaptive controllers for T–S fuzzy vol. 8, no. 6, pp. 691–704, Dec. 2000.
systems if information/parameters of the models are not [3] A. Akhenak, M. Chadli, J. Ragot, and D. Maquin, “Design of robust ob-
known a priori, including information on the number of server for uncertain Takagi-Sugeno models,” in Proc. IEEE Int. Conf.
Fuzzy Systems, Budapest, Hungary, 2004, pp. 1327–1330.
fuzzy rules, the shape and parameters of membership func- [4] H. C. Andersen, A. Lotfi, and A. C. Tsoi, “A new approach to adaptive
tions, and the parameters of local models? The issue of fuzzy control: The controller output error method,” IEEE Trans. Syst.,
adaptive control becomes prohibitively difficult if all the Man, Cybern., vol. 27, no. 4, pp. 686–691, Aug. 1997.
[5] Y. H. Aoul, A. Nafaa, D. Negru, and A. Mehaoua, “FAFC: Fast adap-
information has to be identified online. Some preliminary tive fuzzy AQM controller for TCP/IP networks,” in Proc. IEEE Global
results have been obtained where only the parameters of Telecom. Conf., Dallas, TX, 2004, pp. 1319–1323.
local models are assumed unknown [75], [81], [126], [144], [6] W. Assawinchaichote, S. K. Nguang, and P. Shi, “H-infinity output
feedback control design for uncertain fuzzy singularly perturbed sys-
[238]. tems: An LMI approach,” Automatica, vol. 40, no. 12, pp. 2147–2152,
• Whether can fuzzy controllers, which are designed to sta- Dec. 2004.
bilize T–S fuzzy models, stabilize the original nonlinear [7] R. Babuska, Fuzzy Modeling for Control. Boston, MA: Kluwer,
1998.
systems, even if the T–S fuzzy models universally approx- [8] Y. Bai, H. Q. Zhuang, and Z. S. Roth, “Fuzzy logic control to suppress
imate the original nonlinear systems? If it is possible, how noises and coupling effects in a laser tracking system,” IEEE Trans.
can one design the fuzzy controller to achieve such stabi- Control Syst. Technol., vol. 13, no. 1, pp. 113–121, Jan. 2005.
[9] F. Barrero, A. Gonzalez, A. Torralba, E. Galvan, and L. G. Franquelo,
lization of the original nonlinear systems? “Speed control of induction motors using a novel fuzzy sliding-mode
• Whether does there exist a fuzzy controller to stabilize a structure,” IEEE Trans. Fuzzy Syst., vol. 10, no. 3, pp. 375–383, Jun.
given nonlinear system if it can be stabilized by a smooth 2002.
[10] I. Baturone, F. J. Moreno-Velo, S. Sanchez-Solano, and A. Ollero, “Au-
controller? This is usually called the universal fuzzy control tomatic design of fuzzy controllers for car-like autonomous robots,”
problem. How can it be designed if it exists? Some prelim- IEEE Trans. Fuzzy Syst., vol. 12, no. 4, pp. 447–465, Aug. 2004.
inary results have been obtained in [31] and [32]. [11] R. E. Bellman and L. A. Zadeh, “Decision making in a fuzzy environ-
ment,” Manage. Sci., vol. 17, pp. 141–164, 1970.
• Whether can stability analysis and controller synthesis re- [12] P. Bergsten, R. Palm, and D. Driankov, “Observers for Takagi–Sugeno
sults on deterministic T–S fuzzy systems be extended to fuzzy systems,” IEEE Trans. Syst., Man, Cybern., B, Cybern., vol. 32,
stochastic T–S fuzzy systems? Several authors have made no. 1, pp. 114–121, Feb. 2002.
[13] R. G. Berstecher, R. Palm, and H. D. Unbehauen, “An adaptive fuzzy
attempts to address these issues [43], [48], [308]. However, sliding-mode controller,” IEEE Trans. Ind. Electron., vol. 48, no. 1, pp.
to make such results useful one needs to define fuzzy sets, 18–31, Feb. 2001.
FENG: A SURVEY ON ANALYSIS AND DESIGN OF MODEL-BASED FUZZY CONTROL SYSTEMS 691

[14] J. C. Bezdek, J. M. Keller, R. Krishnapuram, and N. R. Pal, Fuzzy [41] B. Chen and X. Liu, “Fuzzy approximate disturbance decoupling of
Models and Algorithms for Pattern Recognition and Image Pro- MIMO nonlinear systems by backstepping and application to chemical
cessing. Boston, MA: Kluwer, 1999. processes,” IEEE Trans. Fuzzy Syst., vol. 13, no. 6, pp. 832–847, Dec.
[15] Z. Bingul, G. E. Cook, and A. M. Strauss, “Application of fuzzy logic 2005.
to spatial thermal control in fusion welding,” IEEE Trans. Ind. Appl., [42] B. Chen and X. P. Liu, “Delay-dependent robust H-infinity control for
vol. 36, no. 6, pp. 1523–1530, Dec. 2000. T–S fuzzy systems with time delay,” IEEE Trans. Fuzzy Syst., vol. 13,
[16] P. P. Bonissone, V. Badami, K. H. Chiang, P. S. Khedkar, K. W. Mar- no. 4, pp. 544–556, Aug. 2005.
celle, and M. J. Schutten, “Industrial applications of fuzzy logic at Gen- [43] B. S. Chen, B. K. Lee, and L. B. Guo, “Optimal tracking design for
eral Electric,” Proc. IEEE, vol. 38, no. 3, pp. 450–465, Mar. 1995. stochastic fuzzy systems,” IEEE Trans. Fuzzy Syst., vol. 11, no. 6, pp.
[17] M. Boroushaki, M. B. Ghofrani, C. Lucas, and M. J. Yazdanpanah, 796–813, Dec. 2003.
“Identification and control of a nuclear reactor core (VVER) using re- [44] B. S. Chen, C. H. Lee, and Y. C. Chang, “H tracking design of uncer-
current neural networks and fuzzy systems,” IEEE Trans. Nucl. Sci., tain nonlinear SISO systems: Adaptive fuzzy approach,” IEEE Trans.
vol. 50, no. 1, pp. 159–174, Feb. 2003. Fuzzy Syst., vol. 4, no. 1, pp. 32–43, Feb. 1996.
[18] R. Boukezzoula, S. Galichet, and L. Foulloy, “Observer-based fuzzy [45] B. S. Chen, C. L. Tsai, and D. S. Chen, “Robust H and mixed
adaptive control for a class of nonlinear systems: Real-time implemen- H =H filters for equalization designs of nonlinear communication
tation for a robot wrist,” IEEE Trans. Control Syst. Technol., vol. 12, systems: Fuzzy interpolation approach,” IEEE Trans. Fuzzy Syst., vol.
no. 3, pp. 340–351, May 2004. 11, no. 3, pp. 384–398, Jun. 2003.
[19] S. Boyd, L. El Ghaoui, E. Feron, and V. Balakrishnan, Linear Matrix [46] B. S. Chen, C. S. Tseng, and H. J. Uang, “Robustness design of non-
Inequalities in Systems and Control Theory. Philadelphia, PA: SIAM, linear dynamic systems via fuzzy linear control,” IEEE Trans. Fuzzy
1994. Syst., vol. 7, no. 5, pp. 571–585, Oct. 1999.
[20] M. Braae and D. A. Rutherford, “Theoretical and linguistic aspects of [47] ——, “Mixed H =H fuzzy output feedback control design for non-
the fuzzy logic controller,” Automatica, vol. 15, pp. 553–577, 1979. linear dynamic systems: An LMI approach,” IEEE Trans. Fuzzy Syst.,
[21] G. Calcev, R. Gorez, and M. De Neyer, “Passivity approach to fuzzy vol. 8, no. 3, pp. 249–265, Jun. 2000.
control systems,” Automatica, vol. 34, no. 3, pp. 339–344, Mar. 1998. [48] ——, “Fuzzy differential games for nonlinear stochastic systems:
[22] R. J. G. B. Campello, L. A. C. Meleiro, and W. C. Amaral, “Control of Suboptimal approach,” IEEE Trans. Fuzzy Syst., vol. 10, no. 2, pp.
a bioprocess using orthonormal basis function fuzzy models,” in Proc. 222–233, Apr. 2002.
IEEE Int. Conf. Fuzzy Systems, Budapest, Hungary, 2004, pp. 801–806. [49] B. S. Chen, Y. S. Yang, B. K. Lee, and T. H. Lee, “Fuzzy adaptive
[23] J. Campos and F. L. Lewis, “Deadzone compensation in discrete time predictive flow control of ATM network traffic,” IEEE Trans. Fuzzy
using adaptive fuzzy logic,” IEEE Trans. Fuzzy Syst., vol. 7, no. 6, pp. Syst., vol. 11, no. 4, pp. 568–581, Aug. 2003.
697–707, Dec. 1999. [50] C. L. Chen and M. H. Chang, “Optimal design of fuzzy sliding mode
[24] S. G. Cao, N. W. Rees, and G. Feng, “Stability analysis of fuzzy control control: A comparative study,” Fuzzy Sets Syst., vol. 93, pp. 37–48,
systems,” IEEE Trans. Syst., Man, Cybern., B, Cybern., vol. 26, no. 1, 1998.
pp. 201–204, Feb. 1996. [51] C. L. Chen, G. Feng, and X. P. Guan, “Delay-dependent stability anal-
[25] ——, “Analysis and design of fuzzy control systems using dynamic ysis and controller synthesis for discrete time T–S fuzzy systems with
fuzzy global models,” Fuzzy Sets Syst., vol. 75, pp. 47–62, 1995. time delays,” IEEE Trans. Fuzzy Syst., vol. 13, no. 5, pp. 630–643, Oct.
[26] ——, “Stability analysis and design for a class of continuous-time 2005.
fuzzy control systems,” Int. J. Control, vol. 64, pp. 1069–1087, 1996. [52] C. L. Chen, G. Feng, D. Sun, and Y. Zhu, “H-infinity output feedback
[27] ——, “H control of nonlinear continuous-time systems based on dy- control of discrete-time fuzzy systems with application to chaos con-
namic fuzzy models,” Int. J. Syst. Sci., vol. 27, pp. 821–830, 1996. trol,” IEEE Trans. Fuzzy Syst., vol. 13, no. 4, pp. 531–543, Aug. 2005.
[28] ——, “Analysis and design for a class of complex control sys- [53] G. Chen, “Conventional and fuzzy PID controllers: An overview,” Int.
tems—Part I: Fuzzy modeling and identification,” Automatica, vol. J. Intell. Control Syst., vol. 1, pp. 235–246, 1996.
33, pp. 1017–1028, 1997. [54] G. Chen and H. Ying, “Stability analysis of nonlinear fuzzy PI control
[29] ——, “Analysis and design for a class of complex control sys- systems,” in Proc. 3rd Int. Conf. Fuzzy Logic Applications, Houston,
tems—Part II: Fuzzy controller design,” Automatica, vol. 33, pp. TX, 1993, pp. 128–133.
1029–1039, 1997. [55] J. Y. Chen, “Rule regulation of fuzzy sliding mode controller design:
[30] ——, “Analysis and design of fuzzy control systems using dynamic Direct adaptive approach,” Fuzzy Sets Syst., vol. 120, pp. 159–168,
fuzzy state space models,” IEEE Trans. Fuzzy Syst., vol. 7, no. 2, pp. 2001.
192–200, Apr. 1999. [56] S. S. Chen, Y. C. Chang, S. F. Su, S. L. Chung, and T. T. Lee, “Robust
[31] ——, “Universal fuzzy controllers for a class of nonlinear systems,” static output-feedback stabilization for nonlinear discrete-time systems
Fuzzy Sets Syst., vol. 122, pp. 117–123, 2001. with time delay via fuzzy control approach,” IEEE Trans. Fuzzy Syst.,
[32] ——, “Mamdani-type fuzzy controllers are universal fuzzy con- vol. 13, no. 2, pp. 263–272, Apr. 2005.
trollers,” Fuzzy Sets Syst., vol. 123, no. 3, pp. 359–367, Nov. 2001. [57] S. Chiu, “Using fuzzy logic in control applications: Beyond fuzzy PID
[33] Y. Y. Cao and P. M. Frank, “Analysis and synthesis of nonlinear time- control,” IEEE Control Syst. Mag., vol. 18, no. 5, pp. 100–104, Oct.
delay systems via fuzzy control approach,” IEEE Trans. Fuzzy Syst., 1998.
vol. 8, no. 2, pp. 200–211, Apr. 2000. [58] S. Chiu, S. Chand, D. Moore, and A. Chaudhary, “Fuzzy logic for con-
[34] Y. Y. Cao and P. M. Frank, “Robust H-infinity disturbance attenua- trol of roll and moment for a flexible wing aircraft,” IEEE Control Syst.
tion for a class of uncertain discrete-time fuzzy systems,” IEEE Trans. Mag., vol. 11, no. 1, pp. 42–48, Jan. 1991.
Fuzzy Syst., vol. 8, no. 4, pp. 406–415, Aug. 2000. [59] K. H. Cho, C. W. Kim, and J. T. Lim, “On stability analysis of nonlinear
[35] Y. Y. Cao and Z. L. Lin, “Robust stability analysis and fuzzy-sched- plants with fuzzy logic controllers,” in Proc. 5th IFSA World Congr.,
uling control for nonlinear systems subject to actuator saturation,” Seoul, Korea, 1993, pp. 1094–1097.
IEEE Trans. Fuzzy Syst., vol. 11, no. 1, pp. 57–67, Feb. 2003. [60] D. J. Choi and P. G. Park, “H-infinity state-feedback controller design
[36] B. Castillo-Toledo and J. A. Meda-Campana, “The fuzzy discrete-time for discrete-time fuzzy systems using fuzzy weighting-dependent Lya-
robust regulation problem: An LMI approach,” IEEE Trans. Fuzzy punov functions,” IEEE Trans. Fuzzy Syst., vol. 11, no. 2, pp. 271–278,
Syst., vol. 12, no. 3, pp. 360–367, Jun. 2004. Apr. 2003.
[37] M. Chadli, D. Maquin, and J. Ragot, “Stabilisation of Takagi-Sugeno [61] F. Cuesta, F. Gordillo, J. Aracil, and A. Ollero, “Stability analysis of
models with maximum convergence rate,” in Proc. IEEE Int. Conf. nonlinear multivariable Takagi-Sugeno fuzzy control systems,” IEEE
Fuzzy Systems, Budapest, Hungary, 2004, pp. 1323–1326. Trans. Fuzzy Syst., vol. 7, no. 5, pp. 508–520, Oct. 1999.
[38] S. Chakraborty, K. Pal, and N. R. Pal, “A neuro-fuzzy framework for [62] E. Czogala and W. Pedrycz, “On identification in fuzzy systems and its
inferencing,” Neural Netw., vol. 15, pp. 247–261, 2002. applications in control problems,” Fuzzy Sets Syst., vol. 6, pp. 73–83,
[39] Y. C. Chang, “Adaptive fuzzy-based tracking control for nonlinear 1981.
SISO systems via VSS and H approaches,” IEEE Trans. Fuzzy Syst., [63] F. P. Da and W. Z. Song, “Fuzzy neural networks for direct adaptive
vol. 9, no. 2, pp. 278–292, Apr. 2001. control,” IEEE Trans. Ind. Electron., vol. 50, no. 3, pp. 507–513, Jun.
[40] Y. C. Chang and B. S. Chen, “Intelligent robust tracking controls for 2003.
holonomic and nonholonomic mechanical systems using only position [64] Y. X. Diao and K. M. Passino, “Adaptive neural/fuzzy control for in-
measurements,” IEEE Trans. Fuzzy Syst., vol. 13, no. 4, pp. 491–507, terpolated nonlinear systems,” IEEE Trans. Fuzzy Syst., vol. 10, no. 5,
Aug. 2005. pp. 583–595, Oct. 2002.
692 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 14, NO. 5, OCTOBER 2006

[65] S. Dutta, “Fuzzy logic applications: Technological and strategic is- [91] E. Furutani, M. Saeki, and M. Araki, “Shifted Popov criterion and sta-
sues,” IEEE Trans. Eng. Manage., vol. 40, no. 3, pp. 237–254, Aug. bility analysis of fuzzy control systems,” in Proc. 23rd IEEE Conf. De-
1993. cision Control, Tuscon, AZ, 1992, pp. 2790–2795.
[66] C. R. Edgar and B. E. Postlethwaite, “MIMO fuzzy internal model con- [92] P. Gahinet, A. Nemirovski, A. Laub, and M. Chilali, The LMI Control
trol,” Automatica, vol. 36, no. 6, pp. 867–877, June 2000. Toolbox. Natick, MA: The Mathworks, Inc., 1995.
[67] A. L. Elshafei, “Adaptive fuzzy control of nonlinear systems via a vari- [93] Y. Gao and M. J. Er, “Online adaptive fuzzy neural identification and
able-structure algorithm,” in Proc. IEEE Int. Symp. Intelligent Control, control of a class of MIMO nonlinear systems,” IEEE Trans. Fuzzy
Vancouver, BC, Canada, 2002, pp. 620–625. Syst., vol. 11, no. 4, pp. 462–477, Aug. 2003.
[68] K. Erbatur and O. Kaynak, “Use of adaptive fuzzy systems in parameter [94] S. Glower and J. Munighan, “Designing fuzzy controllers from a vari-
tuning of sliding-mode controllers,” IEEE/ASME Trans. Mechatron., able structures standpoint,” IEEE Trans. Fuzzy Syst., vol. 5, no. 1, pp.
vol. 6, no. 4, pp. 474–482, Dec. 2001. 138–144, Feb. 1997.
[69] A. Espada and A. Barreiro, “Robust stability of fuzzy control systems [95] N. Golea, A. Golea, and K. Benmahammed, “Fuzzy model reference
based on conicity conditions,” Automatica, vol. 35, no. 4, pp. 643–654, adaptive control,” IEEE Trans. Fuzzy Syst., vol. 10, no. 4, pp. 436–444,
Apr. 1999. Aug. 2002.
[70] N. Essounbouli, A. Hamzaoui, and K. Benmahammed, “Adaptation [96] G. C. Goodwin and K. S. Sin, Adaptive Filtering, Prediction, and Con-
algorithm for robust fuzzy controller of nonlinear uncertain systems,” trol. Englewood Cliffs, NJ: Prentice-Hall, 1984.
in Proc. IEEE Conf. Control Applications, Istanbul, Turkey, 2003, pp. [97] X. P. Guan and C. L. Chen, “Adaptive fuzzy control for chaotic systems
386–391. with H-infinity tracking performance,” Fuzzy Sets Syst., vol. 139, no.
[71] C. Fantuzzi and R. Rovatti, “On the approximation capabilities of the 1, pp. 81–93, Oct. 2003.
homogeneous Takagi-Sugeno model,” in Proc. 5th IEEE Int. Conf. [98] T. M. Guerra and L. Vermeiren, “LMI-based relaxed nonquadratic
Fuzzy Systems, New Orleans, LA, 1996, pp. 1067–1072. stabilization conditions for nonlinear systems in the Takagi–Sugeno’s
[72] W. A. Farag, V. H. Quintana, and G. Lambert-Torres, “A genetic-based form,” Automatica, vol. 40, no. 5, pp. 823–829, May 2004.
neuro-fuzzy approach for modeling and control of dynamical systems,” [99] T. Guesmi, H. H. Adballah, and A. Toumi, “Transient stability fuzzy
IEEE Trans. Neural Netw., vol. 9, no. 5, pp. 756–767, Sep. 1998. control approach for power systems,” in Proc. IEEE Int. Conf. Indus-
[73] S. S. Farinwata, D. Pirovolou, and G. J. Vachtsevanos, “On input-output trial Technology, Hammamet, Tunisia, 2004, pp. 1676–1681.
stability analysis of a fuzzy controller for a missile autopilot’s yaw [100] P. Guillemin, “Fuzzy logic applied to motor control,” IEEE Trans. Ind.
axis,” in Proc. 3rd IEEE Int. Conf. Fuzzy Systems, Orlando, FL, 1994, Appl., vol. 32, no. 1, pp. 51–56, Jan. 1996.
pp. 930–935. [101] Q. P. Ha, Q. H. Nguyen, D. C. Rye, and H. F. Durrant-Whyte, “Fuzzy
[74] G. Feng, “Approaches to quadratic stabilization of uncertain fuzzy sliding mode controllers with applications,” IEEE Trans. Ind. Electron.,
dynamic systems,” IEEE Trans. Circuits Syst. I, vol. 48, no. 6, pp. vol. 48, no. 1, pp. 38–46, Jan. 2001.
760–769, Jun. 2001. [102] H. A. Hagras, “A hierarchical type-2 fuzzy logic control architecture
[75] ——, “An approach to adaptive control of fuzzy dynamic systems,” for autonomous mobile robots,” IEEE Trans. Fuzzy Syst., vol. 12, no.
IEEE Trans. Fuzzy Syst., vol. 10, no. 2, pp. 268–275, Apr. 2002. 4, pp. 524–539, Aug. 2004.
[76] ——, “Controller synthesis of fuzzy dynamic systems based on piece- [103] H. Han, C. Y. Su, and Y. Stepanenko, “Adaptive control of a class of
wise Lyapunov functions,” IEEE Trans. Fuzzy Syst., vol. 11, no. 10, pp. nonlinear systems with nonlinearly parameterized fuzzy approxima-
605–612, Oct. 2003. tors,” IEEE Trans. Fuzzy Syst., vol. 9, no. 2, pp. 315–323, Apr. 2001.
[77] ——, “Stability analysis of discrete time fuzzy dynamic systems based [104] Z. X. Han, G. Feng, B. L. Walcott, and J. Ma, “Dynamic output feed-
on piecewise Lyapunov functions,” IEEE Trans. Fuzzy Syst., vol. 12, back controller design for fuzzy systems,” IEEE Trans. Syst., Man, Cy-
no. 1, pp. 22–28, Feb. 2004. bern., B, Cybern., vol. 30, no. 1, pp. 204–210, Feb. 2000.

1
[78] ——, “H-infinity controller design of fuzzy dynamic systems based on [105] Z. X. Han, B. L. Walcott, Y. M. Zhang, and G. Feng, “State feed-
piecewise Lyapunov functions,” IEEE Trans. Syst., Man, Cybern., B., back H controller design for fuzzy dynamic systems,” J. Intell. Fuzzy
Cybern., vol. 34, no. 1, pp. 283–292, Feb. 2004. Syst., vol. 8, pp. 261–273, 2000.
[79] ——, “Robust H-infinity filtering of fuzzy dynamic systems,” IEEE [106] T. Haruki and K. Kikuchi, “Video camera system using fuzzy logic,”
Trans. Aerospace Electron. Syst., vol. 41, no. 2, pp. 658–670, Apr. IEEE Trans. Consumer Electron., vol. 38, no. 3, pp. 624–634, Aug.
2005. 1992.
[80] G. Feng, S. G. Cao, N. W. Rees, and C. K. Chak, “Design of fuzzy [107] S. Z. He, S. Tan, F. L. Xu, and P. Z. Wang, “Fuzzy self-tuning of PID
control systems with guaranteed stability,” Fuzzy Sets Syst., vol. 85, controllers,” Fuzzy Sets Syst., vol. 56, pp. 37–46, 1993.
pp. 1–10, 1997. [108] S. Z. He, S. Tan, C. C. Han, and P. Z. Wang, “Control of dynamic
[81] G. Feng, S. G. Cao, and N. W. Rees, “Stable adaptive control of fuzzy processes using an online rule-adaptive fuzzy control system,” Fuzzy
dynamic systems,” Fuzzy Sets Syst., vol. 131, pp. 217–224, 2002. Sets Syst., vol. 54, pp. 11–22, Feb. 1993.
[82] G. Feng, C. L. Chen, D. Sun, and X. P. Guan, “H-infinity controller syn- [109] S. K. Hong and R. Langari, “Robust fuzzy control of a magnetic bearing
thesis of fuzzy dynamic systems based on piecewise Lyapunov func- system subject to harmonic disturbances,” IEEE Trans. Control Syst.
tions and bilinear matrix inequalities,” IEEE Trans. Fuzzy Syst., vol. Technol., vol. 8, no. 2, pp. 366–371, Mar. 2000.
13, no. 1, pp. 94–103, Feb. 2005. [110] ——, “An LMI-based H-infinity fuzzy control system design with TS
[83] G. Feng and J. Ma, “Quadratic stabilization of uncertain discrete-time framework,” Inform. Sci., vol. 123, no. 3–4, pp. 163–179, 2000.
fuzzy dynamic systems,” IEEE Trans. Circuits Syst. I, vol. 48, no. 11, [111] J. I. Horiuchi and M. Kishimoto, “Application of fuzzy control to in-
pp. 1337–1343, Nov. 2001. dustrial bioprocesses in Japan,” Fuzzy Sets Syst., vol. 128, no. 1, pp.
[84] G. Feng and D. Sun, “Generalized H controller synthesis of fuzzy dy- 117–124, May 2002.
namic systems based on piecewise Lyapunov functions,” IEEE Trans. [112] F. H. Hsiao, C. W. Chen, Y. W. Liang, S. D. Xu, and W. L. Chiang,
Circuits Syst. I, vol. 49, no. 12, pp. 1843–1850, Dec. 2002. “T–S fuzzy controllers for nonlinear interconnected systems with mul-
[85] M. Feng and C. J. Harris, “Piecewise Lyapunov stability conditions of tiple time delays,” IEEE Trans. Circuits Syst. I, vol. 52, no. 9, pp.
fuzzy systems,” IEEE Trans. Syst., Man, Cybern., B, Cybern., vol. 31, 1883–1893, Sep. 2005.
pp. 259–262, Apr. 2001. [113] B. G. Hu, G. K. I. Mann, and R. G. Gosine, “New methodology for
[86] M. Feng and C. J. Harris, “Feedback stabilization of fuzzy systems via analytical and optimal design of fuzzy PID controllers,” IEEE Trans.
linear matrix inequalities,” Int. J. Syst. Sci., vol. 32, pp. 221–231, 2001. Fuzzy Syst., vol. 7, no. 5, pp. 521–539, Oct. 1999.
[87] K. Fischle and D. Schroder, “An improved stable adaptive fuzzy con- [114] ——, “A systematic study of fuzzy PID controllers—Function based
trol method,” IEEE Trans. Fuzzy Syst., vol. 7, no. 1, pp. 27–40, Feb. evaluation approach,” IEEE Trans. Fuzzy Syst., vol. 9, no. 5, pp.
1999. 699–712, Oct. 2001.
[88] A. Flores, D. Saez, J. Araya, M. Berenguel, and A. Cipriano, “Fuzzy [115] L. Hu and B. Huang, “Multirate robust digital control for fuzzy systems
predictive control of a solar power plant,” IEEE Trans. Fuzzy Syst., vol. with periodic Lyapunov function,” IEEE Trans. Fuzzy Syst., vol. 13, no.
13, no. 1, pp. 58–68, Feb. 2005. 4, pp. 436–443, Aug. 2005.
[89] M. French and E. Rogers, “Input/output stability theory for direct [116] S. J. Huang and W. C. Lin, “Adaptive fuzzy controller with sliding
neuro-fuzzy controllers,” IEEE Trans. Fuzzy Syst., vol. 6, no. 3, pp. surface for vehicle suspension control,” IEEE Trans. Fuzzy Syst., vol.
331–345, Aug. 1998. 11, no. 4, pp. 550–559, Aug. 2003.
[90] C. W. Frey and H. B. Kuntze, “A neuro-fuzzy supervisory control [117] C. L. Hwang, “A novel Takagi–Sugeno-based robust adaptive fuzzy
system for industrial batch processes,” IEEE Trans. Fuzzy Syst., vol. sliding-mode controller,” IEEE Trans. Fuzzy Syst., vol. 12, no. 5, pp.
9, no. 4, pp. 570–577, Aug. 2001. 676–687, Oct. 2004.
FENG: A SURVEY ON ANALYSIS AND DESIGN OF MODEL-BASED FUZZY CONTROL SYSTEMS 693

[118] C. L. Hwang and C. Y. Kuo, “A stable adaptive fuzzy sliding-mode [144] S. W. Kim, Y. W. Cho, and M. Park, “A multirule-base controller using
control for affine nonlinear systems with application to four-bar linkage the robust property of a fuzzy controller and its design method,” IEEE
systems,” IEEE Trans. Fuzzy Syst., vol. 9, no. 2, pp. 238–252, Apr. Trans. Fuzzy Syst., vol. 4, pp. 315–327, Aug. 1996.
2001. [145] W. C. Kim, S. C. Ahn, and W. H. Kwon, “Stability analysis and stabi-
[119] C. L. Hwang and H. Y. Lin, “A fuzzy decentralized variable structure lization of fuzzy state space models,” Fuzzy Sets Syst., vol. 71, no. 1,
tracking control with optimal and improved robustness designs: Theory pp. 131–142, 1995.
and applications,” IEEE Trans. Fuzzy Syst., vol. 12, no. 5, pp. 615–630, [146] P. J. King and E. H. Mamdani, “The application of fuzzy control sys-
Oct. 2004. tems to industrial process,” Automatica, vol. 13, pp. 235–242, 1977.
[120] G. C. Hwang and S. C. Lin, “A stability approach to fuzzy control [147] K. Kiriakidis, “Robust stabilization of the Takagi–Sugeno fuzzy model
design for nonlinear systems,” Fuzzy Sets Syst., vol. 48, pp. 179–287, via bilinear matrix inequalities,” IEEE Trans. Fuzzy Syst., vol. 9, no. 2,
1992. pp. 269–277, Apr. 2001.
[121] Y. R. Hwang and M. Tomizuka, “Fuzzy smoothing algorithms for vari- [148] K. Kiriakos, “Fuzzy model-based control of complex systems,” IEEE
able structure systems,” IEEE Trans. Fuzzy Syst., vol. 2, no. 4, pp.
Trans. Fuzzy Syst., vol. 6, no. 4, pp. 517–529, Nov. 1998.
277–284, Nov. 1994.
[149] H. S. Ko and T. Niimura, “Power system stabilization using fuzzy-
[122] P. A. Ioannou and J. Sun, Stable and Robust Adaptive Control. En-
glewood Cliffs, NJ: Prentice-Hall, 1995. neural hybrid intelligent control,” in Proc. IEEE Int. Symp. Intelligent
[123] J. S. R. Jang, “ANFIS: Adaptive network-based fuzzy inference Control, Vancouver, BC, Canada, 2002, pp. 879–884.
system,” IEEE Trans. Syst., Man, Cybern., vol. 23, pp. 665–685, [150] T. J. Koo, “Stable model reference adaptive fuzzy control of a class of
May/Jun. 1993. nonlinear systems,” IEEE Trans. Fuzzy Syst., vol. 9, no. 4, pp. 624–636,
[124] J. S. R. Jang and C. T. Sun, “Neuro-fuzzy modeling and control,” Proc. Aug. 2001.
IEEE, vol. 83, no. 3, pp. 378–406, Mar. 1995. [151] P. Korba, R. Babuska, H. B. Verbruggen, and P. M. Frank, “Fuzzy gain
[125] J. Joh, Y. H. Chen, and R. Langari, “On the stability issues of linear scheduling: Controller and observer design based on Lyapunov method
Takagi-Sugeno fuzzy models,” IEEE Trans. Fuzzy Syst., vol. 6, no. 3, and convex optimization,” IEEE Trans. Fuzzy Syst., vol. 11, no. 3, pp.
pp. 402–410, Aug. 1998. 285–298, Jun. 2003.
[126] T. A. Johansen, “Fuzzy model based control: Stability robustness [152] A. Kordon, P. S. Dhurjati, Y. O. Fuentes, and B. A. Ogunnaike, “An
and performance issues,” IEEE Trans. Fuzzy Syst., vol. 2, no. 1, pp. intelligent parallel control system structure for plants with multiple op-
221–233, Feb. 1994. erating regimes,” J. Process Control, vol. 9, pp. 453–460, 1999.
[127] T. A. Johansen, R. Shorten, and R. Murray-Smith, “On the interpre- [153] R. J. Kornblum and M. Tribus, “The use of Bayesian inference in
tation and identification of dynamic Takagi–Sugeno models,” IEEE the design of an endpoint control system for the basic oxygen steel
Trans. Fuzzy Syst., vol. 8, no. 3, pp. 297–313, Jun. 2000. furnace,” IEEE Trans. Syst., Man, Cybern., vol. SMC-6, no. 2, pp.
[128] M. Johansson, A. Rantzer, and K. E. Arzen, “Piecewise quadratic sta- 339–348, Mar./Apr. 1970.
bility of fuzzy systems,” IEEE Trans. Fuzzy Syst., vol. 7, no. 6, pp. [154] B. Koska, Neural Networks and Fuzzy Systems. Englewood Cliffs,
713–722, Dec. 1999. NJ: Prentice-Hall, 1992.
[129] C. F. Juang and C. H. Hsu, “Temperature control by chip-implemented [155] M. Krstic, I. Kanellakopoulos, and P. Kokotovic, Nonlinear and Adap-
adaptive recurrent fuzzy controller designed by evolutionary algo- tive Control Design. New York: Wiley, 1995.
rithm,” IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 52, no. 11, pp. [156] S. Kumar, “A review of smart volume controllers for consumer elec-
2376–2384, Nov. 2005. tronics,” IEEE Trans. Consumer Electron., vol. 51, no. 2, pp. 600–605,
[130] B. Kadmiry and D. Driankov, “A fuzzy gain-scheduler for the attitude
May 2005.
control of an unmanned helicopter,” IEEE Trans. Fuzzy Syst., vol. 12,
[157] C. C. Kung, T. H. Chen, and C. H. Chen, “H-infinity state feedback
no. 4, pp. 502–515, Aug. 2004.
controller design for T–S fuzzy systems based on piecewise Lyapunov
[131] A. Kandel, O. Manor, Y. Klein, and S. Fluss, “ATM traffic management
and congestion control using fuzzy logic,” IEEE Trans. Syst., Man, Cy- function,” in Proc. 14th IEEE Int. Conf. Fuzzy Systems, Reno, NV, May
bern. C, Appl. Rev., vol. 29, no. 3, pp. 474–480, Aug. 1999. 2005, pp. 708–713.
[132] H. Kang, “Stability and control of fuzzy dynamic systems via cell-state [158] H. F. Kwok, D. A. Linkens, M. Mahfouf, and G. H. Mills, “SIVA: A
transitions in fuzzy hypercubes,” IEEE Trans. Fuzzy Syst., vol. 1, no. hybrid knowledge-and-model-based advisory system for intensive care
4, pp. 267–279, Nov. 1993. ventilators,” IEEE Trans. Inform. Technol. Biomed., vol. 8, no. 2, pp.
[133] G. Kang and W. Lee, “Design of fuzzy state controllers and observers,” 161–172, Jun. 2004.
in Proc. Int. Joint Conf. 4th FUZZ-IEEE/2nd IFES, Yokohama, Japan, [159] H. K. Lam, F. H. F. Leung, and P. K. S. Tam, “Nonlinear state feedback
1995, pp. 1355–1360. controller for nonlinear systems: Stability analysis and design based on
[134] O. Kaynak, K. Erbatur, and M. Ertugnrl, “The fusion of computation- fuzzy plant model,” IEEE Trans. Fuzzy Syst., vol. 9, no. 4, pp. 657–661,
ally intelligent methodologies and sliding-mode control—A survey,” Aug. 2001.
IEEE Trans. Ind. Electron., vol. 48, no. 1, pp. 4–17, Feb. 2001. [160] R. Langari and M. Tomizuka, “Stability of fuzzy linguistic control sys-
[135] J. M. Keller, R. R. Yager, and H. Tahani, “Neural network implemen- tems,” in Proc. IEEE Conf. Decision and Control, Honolulu, HI, 1990,
tation of fuzzy logic,” Fuzzy Sets Syst., vol. 45, pp. 1–12, 1992. pp. 2185–2190.
[136] W. J. M. Kickert and E. H. Mamdani, “Analysis of a fuzzy logic con- [161] L. I. Larkin, “A fuzzy logic controller for aircraft flight control,” in In-
troller,” Fuzzy Sets Syst., vol. 1, pp. 29–44, 1978. dustrial Applications of Fuzzy Control, M. Sugeno, Ed. Amsterdam,
[137] W. J. M. Kickert and H. R. Van Nauta Lemke, “Application of a The Netherlands: North-Holland, 1985, pp. 87–104.
fuzzy logic controller in a warm water plant,” Automatica, vol. 12, pp. [162] P. M. Larsen, “Industrial applications of fuzzy logic control,” Int. J.
301–308, 1976. Man Mach. Stud., vol. 12, pp. 3–10, 1980.
[138] K. Kiguchi, T. Tanaka, and T. Fukuda, “Neuro-fuzzy control of a [163] B. Lazzerini, L. M. Reyneri, and M. Chiaberge, “A neuro-fuzzy ap-
robotic exoskeleton with EMG signals,” IEEE Trans. Fuzzy Syst., vol. proach to hybrid intelligent control,” IEEE Trans. Ind. Appl., vol. 35,
12, no. 4, pp. 481–490, Aug. 2004. no. 2, pp. 413–425, Mar./Apr. 1999.
[139] E. Kim, “Output feedback tracking control of robot manipulators with [164] C. C. Lee, “Fuzzy logic in control systems: Fuzzy logic con-
model uncertainty via adaptive fuzzy logic,” IEEE Trans. Fuzzy Syst.,
troller—Part I,” IEEE Trans. Syst., Man, Cybern., vol. 20, no. 2, pp.
vol. 12, no. 3, pp. 368–378, Jun. 2004.
404–418, Mar./Apr. 1990.
[140] E. Kim and D. Kim, “Stability analysis and synthesis for an affine fuzzy
[165] ——, “Fuzzy logic in control systems: Fuzzy logic controller—Part II,”
system via LMI and ILMI: Discrete case,” IEEE Trans. Syst., Man,
Cybern., B, Cybern., vol. 31, no. 1, pp. 132–140, Feb. 2001. IEEE Trans. Syst., Man, Cybern., vol. 20, no. 2, pp. 419–435, Mar./Apr.
[141] E. Kim and S. Kim, “Stability analysis and synthesis for an affine 1990.
fuzzy control system via LMI and ILMI: Continuous case,” IEEE [166] H. Lee and M. Tomizuka, “Robust adaptive control using a universal
Trans. Fuzzy Syst., vol. 10, no. 3, pp. 391–400, Jun. 2002. approximator for SISO nonlinear systems,” IEEE Trans. Fuzzy Syst.,
[142] E. Kim and H. Lee, “New approaches to relaxed quadratic stability vol. 8, no. 1, pp. 95–106, Feb. 2000.
condition of fuzzy control systems,” IEEE Trans. Fuzzy Syst., vol. 8, [167] H. J. Lee, J. B. Park, and G. Chen, “Robust fuzzy control of nonlinear
no. 5, pp. 523–534, Oct. 2000. systems with parametric uncertainties,” IEEE Trans. Fuzzy Syst., vol.
[143] E. Kim and S. Lee, “Output feedback tracking control of MIMO sys- 9, no. 2, pp. 369–379, Apr. 2001.
tems using a fuzzy disturbance observer and its application to the speed [168] K. R. Lee, J. H. Kim, and E. T. Jeung, “Output feedback robust H
control of a PM synchronous motor,” IEEE Trans. Fuzzy Syst., vol. 13, control of uncertain fuzzy dynamic systems with time-varying delay,”
no. 6, pp. 725–741, Dec. 2005. IEEE Trans. Fuzzy Syst., vol. 8, no. 6, pp. 657–664, Dec. 2000.
694 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 14, NO. 5, OCTOBER 2006

[169] S. H. Lee and J. T. Lim, “Multicast ABR service in ATM networks [194] Z. Liu and H. X. Li, “A probabilistic fuzzy logic system for modeling
using a fuzzy-logic-based consolidation algorithm,” Proc. Inst. Elect. and control,” IEEE Trans. Fuzzy Syst., vol. 13, no. 6, pp. 848–859, Dec.
Eng.—Commun., vol. 148, pp. 8–13, 2001. 2005.
[170] S. H. Lee and Z. Bien, “Design of expandable fuzzy inference pro- [195] J. C. Lo and Y. M. Chen, “Stability issues on Takagi-Sugeno fuzzy
cessor,” IEEE Trans. Consumer Electron., vol. 40, no. 2, pp. 171–175, model—Parametric approach,” IEEE Trans. Fuzzy Syst., vol. 7, no. 5,
May 1994. pp. 597–607, Oct. 1999.
[171] Y. G. Lee and S. H. Zak, “Uniformly ultimately bounded fuzzy adap- [196] J. C. Lo and Y. H. Kuo, “Decoupled fuzzy sliding mode control,” IEEE
tive tracking controllers for uncertain systems,” IEEE Trans. Fuzzy Trans. Fuzzy Syst., vol. 6, no. 3, pp. 426–435, Aug. 1998.
Syst., vol. 12, no. 6, pp. 797–811, Dec. 2004. [197] J. C. Lo and M. L. Lin, “Observer-based robust H-infinity control for
[172] Y. M. Lee, S. I. Jang, K. W. Chung, D. Y. Lee, W. C. Kim, and C. W. fuzzy systems using two-step procedure,” IEEE Trans. Fuzzy Syst., vol.
Lee, “A fuzzy-control processor for automatic focusing,” IEEE Trans. 12, no. 3, pp. 350–359, Jun. 2004.
Consumer Electron., vol. 40, no. 2, pp. 138–144, May 1994. [198] X. Ma, Z. Sun, and Y. He, “Analysis and design of fuzzy controller and
[173] T. Leephakpreeda, “H-infinity stability robustness of fuzzy control sys- fuzzy observer,” IEEE Trans. Fuzzy Syst., vol. 6, no. 1, pp. 41–51, Feb.
tems,” Automatica, vol. 35, no. 8, pp. 1467–1470, Aug. 1999. 1998.
[174] Y. G. Leu, W. Y. Wang, and T. T. Lee, “Observer-based direct adap- [199] E. H. Mamdani, “Application of fuzzy algorithms for simple dynamic
tive fuzzy-neural control for nonaffine nonlinear systems,” IEEE Trans. plant,” Proc. Inst. Elect. Eng., vol. 121, pp. 1585–1588, 1974.
Neural Netw., vol. 16, no. 4, pp. 853–861, Jul. 2005. [200] E. H. Mamdani and S. Assilian, “An experiment in linguistic synthesis
[175] F. H. F. Leung, H. K. Lam, P. K. S. Tam, and Y. S. Lee, “Stable fuzzy with a fuzzy logic controller,” Int. J. Man Mach. Stud., vol. 7, pp. 1–13,
controller design for uncertain nonlinear systems: Genetic algorithm 1975.
approach,” in Proc. 12th IEEE Int. Conf. Fuzzy Systems, St Louis, MO, [201] G. K. I. Mann, B. G. Hu, and R. G. Gosine, “Analysis of direct action
2003, pp. 500–505. fuzzy PID controller structures,” IEEE Trans. Syst., Man, Cybern., B,
[176] F. L. Lewis and K. Liu, “Towards a paradigm for fuzzy logic control,” Cybern., vol. 29, no. 3, pp. 371–388, Jun. 1999.
Automatica, vol. 32, no. 2, pp. 167–181, 1996. [202] ——, “Two-level tuning of fuzzy PID controllers,” IEEE Trans. Syst.,
[177] C. S. Li and C. Y. Lee, “Self-organizing neuro-fuzzy system for con- Man, Cybern., B, Cybern., vol. 31, no. 2, pp. 263–269, Apr. 2001.
trol of unknown plants,” IEEE Trans. Fuzzy Syst., vol. 11, no. 1, pp. [203] G. K. I. Mann and R. G. Gosine, “Three-dimensional min-max-gravity
135–150, Feb. 2003. based fuzzy PID inference analysis and tuning,” Fuzzy Sets Syst., vol.
[178] C. S. Li, C. Y. Lee, and K. H. Cheng, “Pseudoerror-based self-orga- 156, pp. 300–323, 2005.
nizing neuro-fuzzy system,” IEEE Trans. Fuzzy Syst., vol. 12, no. 6, [204] A. Mannani and H. A. Talebi, “A fuzzy Lyapunov-based control
strategy for a macro-micro manipulator,” in Proc. IEEE Conf. Control
pp. 812–819, Dec. 2004.
Applications, Istanbul, Turkey, 2003, pp. 368–373.
[179] H. X. Li, L. Zhang, K. Y. Cai, and G. R. Chen, “An improved robust
[205] J. Mar and F. J. Lin, “An ANFIS controller for the car-following colli-
fuzzy-PID controller with optimal fuzzy reasoning,” IEEE Trans. Syst.,
sion prevention system,” IEEE Trans. Veh. Technol., vol. 50, no. 4, pp.
Man, Cybern., B, Cybern., vol. 35, no. 6, pp. 1283–1294, Dec. 2005.
1106–1113, Jul. 2001.
[180] T. H. S. Li, S. J. Chang, and W. Tong, “Fuzzy target tracking control
[206] J. A. Meda-Campana and B. Castillo-Toledo, “On the output regula-
of autonomous mobile robots by using infrared sensors,” IEEE Trans.
tion for TS fuzzy models using sliding modes,” in Proc. Amer. Control
Fuzzy Syst., vol. 12, no. 4, pp. 491–501, Aug. 2004. Conf., Portland, OR, 2005, pp. 4062–4067.
[181] T. H. S. Li and K. J. Lin, “Stabilization of singularly perturbed fuzzy [207] A. H. Meghdadi and M. R. Akbarzadeh, “Probablistic fuzzy logic and
systems,” IEEE Trans. Fuzzy Syst., vol. 12, no. 5, pp. 579–595, Oct. probabilistic fuzzy systems,” in Proc. IEEE Int. Conf. Fuzzy Systems,
2004. Melbourne, Australia, 2001, pp. 1127–1130.
[182] W. Li, X. G. Chang, J. Farrell, and F. M. Wahl, “Design of an enhanced [208] C. Melin and B. Vidolov, “Passive two-rule-based fuzzy logic con-
hybrid fuzzy P+ID controller for a mechanical manipulator,” IEEE trollers: Analysis and application to stabilization,” in Proc. 3rd IEEE
Trans. Syst., Man, Cybern., B, Cybern., vol. 31, no. 6, pp. 938–945, Int. Conf. Fuzzy Systems, Orlando, FL, 1994, pp. 947–950.
Dec. 2001. [209] P. Melin and O. Castillo, “Intelligent control of complex electrochem-
[183] K. Y. Lian, C. S. Chiu, T. S. Chiang, and P. Liu, “LMI-based fuzzy ical systems with a neuro-fuzzy-genetic approach,” IEEE Trans. Ind.
chaotic synchronization and communications,” IEEE Trans. Fuzzy Electron., vol. 48, no. 5, pp. 951–955, Oct. 2001.
Syst., vol. 9, no. 4, pp. 539–553, Aug. 2001. [210] J. M. Mendel, “Fuzzy logic systems for engineering: A tutorial,” Proc.
[184] C. Lin, Q. G. Wang, and T. H. Lee, “Stabilization of uncertain fuzzy IEEE, vol. 83, no. 2, pp. 345–377, Mar. 1995.
time-delay systems via variable structure control approach,” IEEE [211] D. Misir, H. A. Malki, and G. Chen, “Design and analysis of a fuzzy
Trans. Fuzzy Syst., vol. 13, no. 6, pp. 787–798, Dec. 2005. proportional-integral-derivative controller,” Fuzzy Sets Syst., vol. 79,
[185] C. M. Lin and C. F. Hsu, “Self-learning fuzzy sliding-mode control for pp. 297–314, 1996.
antilock braking systems,” IEEE Trans. Control Syst. Technol., vol. 11, [212] S. Mitra and Y. Hayashi, “Neuro-fuzzy rule generation: Survey in soft
no. 2, pp. 273–278, Mar. 2003. computing framework,” IEEE Trans. Neural Netw., vol. 11, no. 3, pp.
[186] ——, “Supervisory recurrent fuzzy neural network control of wing 748–768, May 2000.
rock for slender delta wings,” IEEE Trans. Fuzzy Syst., vol. 12, no. 5, [213] M. Mizumoto, “Realization of PID controls by fuzzy control methods,”
pp. 733–742, Oct. 2004. Fuzzy Sets Syst., vol. 70, pp. 171–182, 1995.
[187] C. T. Lin and C. S. G. Lee, “Reinforcement structure/parameter [214] S. Mollov, R. Babuska, J. Abonyi, and H. B. Verbruggen, “Effective
learning for neural-network-based fuzzy logic control systems,” IEEE optimization for fuzzy model predictive control,” IEEE Trans. Fuzzy
Trans. Fuzzy Syst., vol. 2, no. 1, pp. 46–63, Feb. 1994. Syst., vol. 12, no. 5, pp. 661–675, Oct. 2004.
[188] S. C. Lin and Y. Y. Chen, “Design of self learning fuzzy sliding mode [215] R. K. Mudi and N. R. Pal, “A robust self-tuning scheme for PI- and
controllers based on genetic algorithms,” Fuzzy Sets Syst., vol. 86, pp. PD- type fuzzy controllers,” IEEE Trans. Fuzzy Syst., vol. 7, no. 1, pp.
139–153, 1997. 2–16, Feb. 1999.
[189] W. S. Lin and C. H. Tsai, “Self-organizing fuzzy control of multi-vari- [216] ——, “A note on fuzzy PI-type controllers with resetting action,” Fuzzy
able systems using learning vector quantization network,” Fuzzy Sets Sets Syst., vol. 121, no. 1, pp. 149–159, 2001.
Syst., vol. 124, pp. 197–212, 2001. [217] S. R. Munasinghe, M. S. Kim, and J. J. Lee, “Adaptive neurofuzzy
[190] B. D. Liu, C. Y. Chen, and J. Y. Tsao, “Design of adaptive fuzzy logic controller to regulate UTSG water level in nuclear power plants,” IEEE
controller based on linguistic-hedge concepts and genetic algorithms,” Trans. Nucl. Sci., vol. 52, no. 1, pp. 421–429, Feb. 2005.
IEEE Trans. Syst., Man, Cybern., B, Cybern., vol. 31, no. 1, pp. 32–53, [218] S. Murakami and M. Maeda, “Application of fuzzy controller to auto-
Feb. 2001. mobile speed control system,” in Industrial Applications of Fuzzy Con-
[191] H. P. Liu, F. C. Sun, and Z. Q. Sun, “Stability analysis and synthesis of trol, M. Sugeno, Ed. Amsterdam, The Netherlands: North-Holland,
fuzzy singularly perturbed systems,” IEEE Trans. Fuzzy Syst., vol. 13, 1985, pp. 105–124.
no. 2, pp. 273–284, Apr. 2005. [219] N. Nakagaki, Y. Bando, T. Mori, S. Torikoshi, and S. Suzuki, “Wide
[192] X. Liu and Q. Zhang, “New approaches to controller designs based on aspect TV receiver with aspect detection and non-linear control for
fuzzy observers for T–S fuzzy systems via LMI,” Automatica, vol. 39, picture quality,” IEEE Trans. Consumer Electron., vol. 40, no. 3, pp.
no. 9, pp. 1571–1582, 2003. 743–752, Aug. 1994.
[193] X. J. Liu, F. Lara-Rosano, and C. W. Chan, “Model-reference adaptive [220] H. R. Nerenji and P. Khedkar, “Learning and tuning fuzzy logic con-
control based on neurofuzzy networks,” IEEE Trans. Syst., Man, Cy- trollers through reinforcements,” IEEE Trans. Neural Netw., vol. 3, no.
bern., C, Appl. Rev., vol. 34, no. 3, pp. 302–309, Aug. 2004. 5, pp. 724–740, Sep. 1992.
FENG: A SURVEY ON ANALYSIS AND DESIGN OF MODEL-BASED FUZZY CONTROL SYSTEMS 695

[221] S. K. Nguang and W. Assawinchaichote, “ H filtering for fuzzy dy- [248] H. Seker, M. O. Odetayo, D. Petrovic, and R. N. G. Naguib, “A fuzzy
namical systems with pole placement constraints,” IEEE Trans. Cir- logic based-method for prognostic decision making in breast and
cuits Syst. I, Reg. Papers, vol. 50, no. 11, pp. 1503–1508, Nov. 2003. prostate cancers,” IEEE Trans. Inform. Technol. Biomed., vol. 7, no.
[222] A. H. Niasar, H. Moghbeli, and R. Kazemi, “Yaw moment control via 2, pp. 114–122, Jun. 2003.
emotional adaptive neuro-fuzzy controller for independent rear wheel [249] T. L. Seng, M. Bin Khalid, and R. Yusof, “Tuning of a neuro-fuzzy
drives of an electric vehicle,” in Proc. IEEE Conf. Control Applications, controller by genetic algorithm,” IEEE Trans. Syst., Man, Cybern., B,
Istanbul, Turkey, 2003, pp. 380–385. Cybern., vol. 29, no. 2, pp. 226–236, Apr. 1999.
[223] H. N. Nounou and K. M. Passino, “Stable auto-tuning of adaptive [250] J. C. Shen, “Fuzzy neural networks for tuning PID controller for plants
fuzzy/neural controllers for nonlinear discrete-time systems,” IEEE with underdamped responses,” IEEE Trans. Fuzzy Syst., vol. 9, no. 2,
Trans. Fuzzy Syst., vol. 12, no. 1, pp. 70–83, Feb. 2004. pp. 334–342, Apr. 2001.
[224] H. Ohtake, K. Tanaka, and H. O. Wang, “Piecewise nonlinear control,” [251] C. W. de Silva, Intelligent Control: Fuzzy Logic Applications. New
in Proc. 42nd IEEE Conf. Decision and Control, Maui, HI, Dec. 2003, York: CRC, 1995.
pp. 4735–4740. [252] S. Singh, “Stability analysis of discrete fuzzy control systems,” in Proc.
[225] M. C. de Oliveira, J. Bernussou, and J. C. Geromel, “A new discrete 2nd IEEE Int. Conf. Fuzzy Systems, San Diego, CA, 1992, pp. 527–534.
time robust stability condition,” Syst. Control Lett., vol. 36, no. 2, pp. [253] K. C. Sio and C. K. Lee, “Stability of fuzzy PID controllers,” IEEE
135–141, 1999. Trans. Syst., Man, Cybern., A, Syst. Humans, vol. 28, no. 4, pp.
[226] H. P. Opitz, “Fuzzy control and stability criteria,” in Proc. EUFIT’93, 490–495, Jul. 1998.
Aachen, Germany, 1993, pp. 130–136. [254] S. Skoczowski, S. Domek, K. Pietrusewicz, and B. Broel-Plater, “A
[227] R. Ordonez and P. M. Passino, “Stable multi-input multi-output adap- Method for improving the robustness of PID control,” IEEE Trans. Ind.
tive fuzzy neural control,” IEEE Trans. Fuzzy Syst., vol. 7, no. 3, pp. Electron., vol. 52, no. 6, pp. 1669–1676, Dec. 2005.
345–353, Jun. 1999. [255] M. L. Smith, “Sensors, appliance control, and fuzzy logic,” IEEE Trans.
[228] R. Ordonez, J. Zumberge, J. T. Spooner, and K. M. Passino, “Adaptive Ind. Appl., vol. 30, no. 2, pp. 305–310, Mar./Apr. 1994.
fuzzy control: Experiments and comparative analyses,” IEEE Trans. [256] C. Y. Su, M. Oya, and H. Hong, “Stable adaptive fuzzy control of non-
Fuzzy Syst., vol. 5, no. 2, pp. 167–188, May 1997. linear systems preceded by unknown backlash-like hysteresis,” IEEE
[229] J. J. Ostergaard, “Fuzzy logic control of a heat exchanger process,” in Trans. Fuzzy Syst., vol. 11, no. 1, pp. 1–8, Feb. 2003.
Fuzzy Automata and Decision Processes, M. M. Gupta, G. N. Saridis, [257] J. P. Su, T. M. Chen, and C. C. Wang, “Adaptive fuzzy sliding mode
and B. R. Gaines, Eds. Amsterdam, The Netherlands: North-Holland, control with GA-based reaching laws,” Fuzzy Sets Syst., vol. 120, pp.
1977, pp. 285–320. 145–158, 2001.
[230] K. Pal, R. K. Mudi, and N. R. Pal, “A new scheme for fuzzy rule- [258] M. Sugeno, Industrial Applications of Fuzzy Control. New York: El-
based system identification and its application to self-tuning fuzzy con-
sevier, 1985.
trollers,” IEEE Trans. on Syst., Man, Cybern., B, Cybern., vol. 32, no.
[259] M. Sugeno, “On stability of fuzzy systems expressed by fuzzy rules
4, pp. 470–482, Aug. 2002.
with singleton consequents,” IEEE Trans. Fuzzy Syst., vol. 7, no. 2, pp.
[231] K. Pal and N. R. Pal, “A neuro-fuzzy system for inferencing,” Int. J.
201–224, Apr. 1999.
Intell. Syst., vol. 14, pp. 1155–1182, 1999.
[260] M. Sugeno and M. Nishida, “Fuzzy control of model car,” Fuzzy Sets
[232] K. Pal, N. R. Pal, and J. M. Keller, “Some neural net realizations of
Syst., vol. 16, pp. 103–113, 1985.
fuzzy reasoning,” Int. J. Intell. Syst., vol. 13, pp. 859–886, 1998.
[261] M. Sugeno and T. Yasukawa, “A fuzzy-logic-based approach to quali-
[233] R. Palm, “Tuning of scaling factors in fuzzy controllers using correla-
tion functions,” in Proc. IEEE Int. Conf. Fuzzy Systems, San Francisco, tative modeling,” IEEE Trans. Fuzzy Syst., vol. 1, no. 1, pp. 7–31, Feb.
CA, 1993, pp. 691–696. 1993.
[234] R. Palm, “Sliding mode fuzzy control,” in Proc. 1st IEEE Int. Conf. [262] Y. L. Sun and M. J. Er, “Hybrid fuzzy control of robotics systems,”
Fuzzy Systems, San Diego, CA, 1992, pp. 519–526. IEEE Trans. Fuzzy Syst., vol. 12, no. 6, pp. 755–765, Dec. 2004.
[235] R. Palm, “Robust control by fuzzy sliding mode,” Automatica, vol. 30, [263] H. Takagi, “Application of neural networks and fuzzy logic to con-
pp. 1429–1437, 1994. sumer products,” in Proc. Int. Conf. on Industrial Electronics, Con-
[236] R. Palm, D. Driankov, and H. Hellendoorn, Model Based Fuzzy Con- trol, Instrumentation, and Automation, San Diego, CA, Nov. 1992, pp.
trol. New York: Springer-Verlag, 1996. 1629–1633.
[237] C. P. Pappis and E. H. Mamdani, “A fuzzy logic controller for a traffic [264] H. Takagi, N. Suzuki, T. Koda, and Y. Kojima, “Neural networks de-
junction,” IEEE Trans. Syst., Man, Cybern., vol. 7, no. 10, pp. 707–717, signed on approximate reasoning architecture and their applications,”
Oct. 1977. IEEE Trans. Neural Netw., vol. 3, no. 5, pp. 752–760, Sep. 1992.
[238] C. W. Park and Y. W. Cho, “T–S model based indirect adaptive fuzzy [265] T. Takagi and M. Sugeno, “Fuzzy identification of systems and its ap-
control using online parameter estimation,” IEEE Trans. Syst., Man, plications to modeling and control,” IEEE Trans. Syst., Man, Cybern.,
Cybern., B, Cybern., vol. 34, no. 6, pp. 2293–2302, Dec. 2004. vol. SMC-15, no. 1, pp. 116–132, Jan./Feb. 1985.
[239] Y. M. Park, M. J. Tahk, and H. C. Bang, “Design and analysis of op- [266] K. Tanaka, T. Hori, and H. O. Wang, “A multiple Lyapunov function
timal controller for fuzzy systems with input constraint,” IEEE Trans. approach to stabilization of fuzzy control systems,” IEEE Trans. Fuzzy
Fuzzy Syst., vol. 12, no. 6, pp. 766–779, Dec. 2004. Syst., vol. 11, no. 4, pp. 582–589, Aug. 2003.
[240] K. Passino and S. Yurkovich, Fuzzy Control. Reading, MA: Addison- [267] K. Tanaka, T. Ikeda, and H. O. Wang, “Fuzzy regulators and fuzzy
Wesley, 1998. observers: Relaxed stability conditions and LMI-based designs,” IEEE
[241] W. Pedrycs, Fuzzy Control and Fuzzy Systems. Somerset, U.K.: Re- Trans. Fuzzy Syst., vol. 6, no. 2, pp. 250–265, May 1998.
search Studies Press, Ltd., 1993. [268] K. Tanaka and M. Sugeno, “Stability analysis and design of fuzzy con-
[242] H. Pomares, I. Rojas, J. Gonzalez, M. Damas, B. Pino, and A. Prieto, trol systems,” Fuzzy Sets Syst., vol. 12, pp. 135–156, 1992.
“Online global learning in direct fuzzy controllers,” IEEE Trans. Fuzzy [269] K. Tanaka and H. O. Wang, Fuzzy Control Systems Design and Anal-
Syst., vol. 12, no. 2, pp. 218–229, Apr. 2004. ysis: A LMI Approach. New York: Wiley, 2001.
[243] R. Y. Qi and M. A. Brdys, “Adaptive fuzzy modelling and control [270] K. S. Tang, K. F. Man, G. Chen, and S. Kwong, “An optimal fuzzy
for discrete-time nonlinear uncertain systems,” in Proc. Amer. Control PID controller,” IEEE Trans. Ind. Electron., vol. 48, no. 4, pp. 757–765,
Conf., Portland, OR, 2005, pp. 1108–1113. Aug. 2001.
[244] K. S. Ray and D. D. Majumder, “Application of circle criteria for sta- [271] T. Tani, S. Murakoshi, and M. Umano, “Neuro-fuzzy hybrid control
bility analysis of linear SISO and MIMO system associated with fuzzy system of tank level in petroleum plant,” IEEE Trans. Fuzzy Syst., vol.
logic controller,” IEEE Trans. Syst., Man, Cybern., vol. SMC-14, no. 4, no. 3, pp. 360–368, Aug. 1996.
4, pp. 345–349, Mar./Apr. 1984. [272] T. Taniguchi and M. Sugeno, “Stabilization of nonlinear systems based
[245] A. Sala, T. M. Guerra, and R. Babuska, “Perspectives of fuzzy systems on piecewise Lyapunov functions,” in Proc. 13th IEEE Int. Conf. on
and control,” Fuzzy Sets Syst., vol. 156, pp. 432–444, 2005. Fuzzy Systems, Budapest, Hungary, Jul. 2004, pp. 1607–1612.
[246] R. M. Sanner and J. E. Slotine, “Gaussian networks for direct adaptive [273] C. W. Tao and J. S. Taur, “Robust fuzzy control for a plant with fuzzy
control,” IEEE Trans. Neural Netw., vol. 3, no. 6, pp. 837–863, Jun. linear model,” IEEE Trans. Fuzzy Syst., vol. 13, no. 1, pp. 30–41, Feb.
1992. 2005.
[247] V. Santibanez, R. Kelly, and M. A. Llama, “A novel global asymptotic [274] C. W. Tao, J. S. Taur, and M. L. Chan, “Adaptive fuzzy terminal sliding
stable set-point fuzzy controller with bounded torques for robot ma- mode controller for linear systems with mismatched time-varying un-
nipulators,” IEEE Trans. Fuzzy Syst., vol. 13, no. 3, pp. 362–372, Jun. certainties,” IEEE Trans. Syst., Man, Cybern., B, Cybern., vol. 34, no.
2005. 1, pp. 255–262, Feb. 2004.
696 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 14, NO. 5, OCTOBER 2006

[275] M. C. M. Teixeira, E. Assuncao, and R. G. Avellar, “On relaxed LMI- [302] W. J. Wang and W. W. Lin, “Decentralized PDC for large-scale T–S
based designs for fuzzy regulators and fuzzy observers,” IEEE Trans. fuzzy systems,” IEEE Trans. Fuzzy Syst., vol. 13, no. 6, pp. 779–786,
Fuzzy Syst., vol. 11, no. 5, pp. 613–623, Oct. 2003. Dec. 2005.
[276] M. C. M. Teixeira and S. H. Zak, “Stabilizing controller design for [303] W. J. Wang and L. Luoh, “Stability and stabilization of fuzzy large-
uncertain nonlinear systems using fuzzy models,” IEEE Trans. Fuzzy scale systems,” IEEE Trans. Fuzzy Syst., vol. 12, no. 3, pp. 309–315,
Syst., vol. 7, no. 2, pp. 133–142, Apr. 1999. Jun. 2004.
[277] H. N. Teodorescu, L. C. Jain, and A. Kandel, Fuzzy and Neuro-Fuzzy [304] W. J. Wang and C. H. Sun, “Relaxed stability and stabilization condi-
Systems in Medicine. Boca Raton, FL: CRC, 1998. tions for a T–S fuzzy discrete system,” Fuzzy Sets Syst., vol. 156, no.
[278] R. M. Tong, “A control engineering review of fuzzy systems,” Auto- 2, pp. 208–225, Dec. 2005.
matica, vol. 13, pp. 559–568, 1977. [305] W. Y. Wang, M. L. Chan, C. C. J. Hsu, and T. T. Lee, “H tracking-
[279] R. M. Tong, M. B. Beck, and A. Latten, “Fuzzy control of the activated based sliding mode control for uncertain nonlinear systems via an adap-
sludge wastewater treatment process,” Automatica, vol. 6, pp. 695–701, tive fuzzy-neural approach,” IEEE Trans. Syst., Man, Cybern., B, Cy-
1980. bern., vol. 32, no. 4, Aug. 2002.
[280] S. C. Tong and H. X. Li, “Fuzzy adaptive sliding-mode control for [306] W. Y. Wang, C. Y. Cheng, and Y. G. Leu, “An online GA-based
MIMO nonlinear systems,” IEEE Trans. Fuzzy Syst., vol. 11, no. 3, pp. output-feedback direct adaptive fuzzy-neural controller for uncertain
354–360, Jun. 2003. nonlinear systems,” IEEE Trans. Syst., Man, Cybern., B, Cybern., vol.
[281] D. L. Tsay, H. Y. Chung, and C. J. Lee, “The adaptive control of 34, no. 1, pp. 334–345, Feb. 2004.
nonlinear systems using the Sugeno-type of fuzzy logic,” IEEE Trans. [307] Y. Wang, Z. Q. Sun, and F. C. Sun, “Stability analysis and control of
discrete-time fuzzy systems: A fuzzy Lyapunov function approach,”
1
Fuzzy Syst., vol. 7, no. 2, pp. 225–229, Apr. 1999.
[282] C. S. Tseng and B. S. Chen, “H fuzzy estimation for a class of non- in Proc. 5th Asian Control Conf., Melbourne, Australia, 2004, pp.
linear discrete-time dynamic systems,” IEEE Trans. Signal Process., 1855–1860.
vol. 49, no. 11, pp. 2605–2619, Nov. 2001. [308] Z. D. Wang, D. W. C. Ho, and X. H. Liu, “A note on the robust stability
[283] C. S. Tseng, B. S. Chen, and H. J. Uang, “Fuzzy tracking control de- of uncertain stochastic fuzzy systems with time-delays,” IEEE Trans.
sign for nonlinear dynamic systems via T–S fuzzy model,” IEEE Trans. Syst., Man, Cybern., A, Syst. Humans., vol. 34, no. 4, pp. 570–576, Jul.
Fuzzy Syst., vol. 9, no. 3, pp. 381–392, Jun. 2001. 2004.
[284] A. Tsourdos, J. T. Economou, A. B. White, and P. C. K. Luk, “Control [309] L. K. Wong, F. H. F. Leung, and P. K. S. Tam, “A fuzzy sliding con-
design for a mobile robot: A fuzzy LPV approach,” in Proc. IEEE Conf. troller for nonlinear systems,” IEEE Trans. Ind. Electron., vol. 48, no.
Control Applications, Istanbul, Turkey, 2003, pp. 552–557. 1, pp. 32–37, Feb. 2001.
[285] H. D. Tuan, P. Apkarian, T. Narikiyo, and M. Kanota, “New fuzzy con- [310] Z. W. Woo, H. Y. Chung, and J. J. Lin, “A PID type fuzzy controller
trol model and dynamic output feed back parallel distributed compen- with self-tuning scaling factors,” Fuzzy Sets Syst., vol. 115, pp.
sation,” IEEE Trans. Fuzzy Syst., vol. 12, no. 1, pp. 13–21, Feb. 2004. 321–326, 2000.
[311] C. J. Wu and A. H. Sung, “The application of fuzzy logic to JPEG,”
[286] H. D. Tuan, P. Apkarian, T. Narikiyo, and Y. Yamamoto, “Parame-
IEEE Trans. Consumer Electron., vol. 40, no. 4, pp. 976–984, Nov.
terized linear matrix inequality techniques in fuzzy control system de-
1994.
sign,” IEEE Trans. Fuzzy Syst., vol. 9, no. 2, pp. 324–332, Apr. 2001.
[312] H. N. Wu and H. Y. Zhang, “Reliable mixed L-2/H-infinity fuzzy static
[287] S. G. Tzafestas and K. C. Zikidis, “NeuroFAST: On-line neuro-fuzzy
output feedback control for nonlinear systems with sensor faults,” Au-
ART-based structure and parameter learning TSK model,” IEEE Trans.
tomatica, vol. 41, no. 11, pp. 1925–1932, Nov. 2005.
Syst., Man, Cybern., B, Cybern., vol. 31, no. 5, pp. 797–802, Oct. 2001.
[313] J. C. Wu and T. S. Liu, “A sliding mode approach to fuzzy control
[288] I. G. Umbers and P. J. King, “An analysis of human-decision making
design,” IEEE Trans. Control Syst. Technol., vol. 4, no. 2, pp. 141–150,
in cement kiln control and the implications for automation,” Int. J. Man
Mar. 1996.
Mach. Stud., vol. 12, pp. 11–23, 1980.
[314] J. Xiao, J. Z. Xiao, N. Xi, R. L. Tummala, and R. Mukherjee, “Fuzzy
[289] M. Uragami, M. Mizumoto, and K. Tanaka, “Fuzzy robot controls,” controller for wall-climbing microrobots,” IEEE Trans. Fuzzy Syst.,
Cybern., vol. 6, pp. 39–64, 1976. vol. 12, no. 4, pp. 466–480, Aug. 2004.
[290] V. I. Utkin, Sliding Modes in Control Optimization. Berlin, Germany: [315] C. Xu and Y. C. Shin, “Design of a multilevel fuzzy controller for non-
Springer-Verlag, 1992. linear systems and stability analysis,” IEEE Trans. Fuzzy Syst., vol. 13,
[291] D. Velez-Diaz and Y. Tang, “Adaptive robust fuzzy control of nonlinear no. 6, pp. 761–778, Dec. 2005.
systems,” IEEE Trans. Syst., Man, Cybern., B, Cybern., vol. 34, no. 3, [316] J. X. Xu, C. C. Hang, and C. Liu, “Parallel structure and tuning of a
pp. 1596–1601, Jun. 2004. fuzzy PID controller,” Automatica, vol. 36, pp. 673–684, 2000.
[292] R. J. Wai and P. C. Chen, “Intelligent tracking control for robot manipu- [317] S. Y. Xu and J. Lam, “Robust H-infinity control for uncertain discrete-
lator including actuator dynamics via TSK-type fuzzy neural network,” time-delay fuzzy systems via output feedback controllers,” IEEE Trans.
IEEE Trans. Fuzzy Syst., vol. 12, no. 4, pp. 552–560, Aug. 2004. Fuzzy Syst., vol. 13, no. 1, pp. 82–93, Feb. 2005.
[293] H. O. Wang, K. Tanaka, and M. F. Griffin, “An approach to fuzzy con- [318] R. R. Yager and D. P. Filev, Essentials of Fuzzy Modeling and Con-
trol of nonlinear systems: Stability and design issues,” IEEE Trans. trol. New York: Wiley, 1994.
Fuzzy Syst., vol. 4, no. 1, pp. 14–23, Feb. 1996. [319] S. X. Yang, H. Li, M.Q. H. Meng, and P. X. Liu, “An embedded fuzzy
[294] J. S. Wang and C. S. G. Lee, “Self-adaptive recurrent neuro-fuzzy con- controller for a behavior-based mobile robot with guaranteed perfor-
trol of an autonomous underwater vehicle,” IEEE Trans. Robot. Au- mance,” IEEE Trans. Fuzzy Syst., vol. 12, no. 4, pp. 436–446, Aug.
tomat., vol. 19, no. 2, pp. 283–295, Apr. 2003. 2004.
[295] L. Wang and G. Feng, “Piecewise H-infinity controller design of dis- [320] Y. S. Yang and J. S. Ren, “Adaptive fuzzy robust tracking controller
crete time fuzzy systems,” IEEE Trans. Syst., Man, Cybern., B, Cy- design via small gain approach and its application,” IEEE Trans. Fuzzy
bern., vol. 34, no. 1, pp. 682–686, Feb. 2004. Syst., vol. 11, no. 6, pp. 783–795, Dec. 2003.
[296] L. Wang, G. Feng, and T. Hesketh, “Piecewise generalized H con- [321] Y. S. Yang and C. J. Zhou, “Adaptive fuzzy H-infinity stabilization
troller synthesis of discrete time fuzzy systems,” in Proc. Inst. Elect. for strict-feedback canonical nonlinear systems via backstepping and
Eng. Control Theory Applications, Sep. 2004, vol. 151, no. 5, pp. small-gain approach,” IEEE Trans. Fuzzy Syst., vol. 13, no. 1, pp.
554–560, Part D. 104–114, Feb. 2005.
[297] L. Wang, G. Feng, and T. Hesketh, “Piecewise output feedback con- [322] Z. M. Yeh, “A systematic method for design of multivariable fuzzy
troller synthesis of discrete time fuzzy systems,” in Proc. IEEE Conf. logic control systems,” IEEE Trans. Fuzzy Syst., vol. 7, no. 6, pp.
Decision and Control, Maui, HI, 2003, pp. 4741–4746. 741–752, Dec. 1999.
[298] L. X. Wang, “Stable adaptive fuzzy control of nonlinear systems,” IEEE [323] Z. Yi and P. A. Heng, “Stability of fuzzy control systems with bounded
Trans. Fuzzy Syst., vol. 1, no. 2, pp. 146–155, May 1993. uncertain delays,” IEEE Trans. Fuzzy Syst., vol. 10, no. 1, pp. 92–97,
[299] L. X. Wang, Adaptive Fuzzy Systems and Control: Design and Stability Feb. 2002.
Analysis. Englewood Cliffs, NJ: Prentice-Hall, 1994. [324] T. K. Yin and C. S. G. Lee, “Fuzzy model-reference adaptive control,”
[300] L. X. Wang, A Course in Fuzzy Systems and Control. London, U.K.: IEEE Trans. Syst., Man, Cybern., vol. 25, no. 12, pp. 1606–1615, Dec.
Prentice-Hall, 1997. 1995.
[301] L. X. Wang and J. M. Mendel, “Fuzzy basis functions, universal [325] H. Ying, “The simplest fuzzy controllers using different inference
approximation, and orthogonal least squares learning,” IEEE Trans. methods are different nonlinear proportional-integral controllers with
Neural Netw., vol. 3, no. 5, pp. 807–814, Sep. 1992. variable gains,” Automatica, vol. 29, pp. 1579–1589, 1993.
FENG: A SURVEY ON ANALYSIS AND DESIGN OF MODEL-BASED FUZZY CONTROL SYSTEMS 697

[326] ——, “General SISO Takagi–Sugeno fuzzy systems with linear rule [343] R. T. Zhang and Y. A. Phillis, “Fuzzy control of queueing systems
consequent are universal approximators,” IEEE Trans. Fuzzy Syst., vol. with heterogeneous servers,” IEEE Trans. Fuzzy Syst., vol. 7, no. 1, pp.
6, no. 4, pp. 582–587, Nov. 1998. 17–26, Feb. 1999.
[327] J. Yoneyama, M. Nishikawa, H. Katayama, and A. Ichikawa, “Output [344] Z. Y. Zhao, M. Tomizuka, and S. Isaka, “Fuzzy gain-scheduling of
stabilization of Takagi-Sugeno fuzzy systems,” Fuzzy Sets Syst., vol. PID controllers,” IEEE Trans. Syst., Man, Cybern., vol. 23, no. 5, pp.
111, pp. 253–266, 2000. 1392–1398, Sep./Oct. 1993.
[328] L. X. Yu and Ya. Qi. Zhang, “Evolutionary fuzzy neural networks for [345] H. Zheng and K. Y. Zhu, “A fuzzy controller-based multiple-model
hybrid financial prediction,” IEEE Trans. Syst., Man, Cybern., C, App. adaptive control system for blood pressure control,” in Proc. 8th Conf.
Rev., vol. 35, no. 2, pp. 244–249, May 2005. Control, Automation, Robotics and Vision, Kunming, China, 2004, pp.
[329] W. Yu and X. O. Li, “Fuzzy identification using fuzzy neural networks 1353–1358.
with stable learning algorithms,” IEEE Trans. Fuzzy Syst., vol. 12, no. [346] S. S. Zhou, G. Feng, and C. B. Feng, “Robust control for a class of
3, pp. 411–420, Jun. 2004. uncertain nonlinear systems: Adaptive fuzzy approach based on back-
[330] W. S. Yu and C. J. Sun, “Fuzzy model based adaptive control for a stepping,” Fuzzy Sets Syst., vol. 151, no. 1, pp. 1–20, Apr. 2005.
class of nonlinear systems,” IEEE Trans. Fuzzy Syst., vol. 9, no. 3, pp. [347] S. S. Zhou, G. Feng, J. Lam, and S. Y. Xu, “Robust H-infinity control
413–425, Jun. 2001. for discrete fuzzy systems via basis-dependent Lyapunov functions,”
[331] L. A. Zadeh, “Fuzzy sets,” Inform. Control, vol. 8, pp. 338–353, 1965. Inform. Sci., vol. 174, no. 3–4, pp. 197–217, Aug. 2005.
[332] ——, “Fuzzy algorithm,” Inform. Control, vol. 12, pp. 94–102, 1968. [348] H. J. Zimmermann, Fuzzy Set Theory and Its Application, 2nd ed.
[333] ——, “Similarity relations and fuzzy orderings,” Inform. Sci., vol. 3, Boston, MA: Kluwer, 1991.
pp. 177–200, 1971. [349] A. S. Zinober, Variable Structure and Lyapunov Control. Berlin, Ger-
[334] ——, “Outline of a new approach to the analysis of complex systems many: Springer-Verlag, 1994.
and decision processes,” IEEE Trans. Syst., Man, Cybern., vol. SMC-3,
no. 1, pp. 28–44, Jan. 1973.
[335] L. A. Zadeh, “The concept of a linguistic variable and its application
to approximate reasoning: I, II, III,” Inform. Sci., vol. 8, pp. 199–251,
1975.
[336] K. Zeng, N. Y. Zhang, and W. L. Xu, “A comparative study on suf- Gang Feng (S’90–M’92–SM’95) received the
ficient conditions for Takagi–Sugeno fuzzy systems as universal ap- B.Eng. and M.Eng. degrees in automatic control (of
proximators,” IEEE Trans. Fuzzy Syst., vol. 8, no. 6, pp. 773–780, Dec. electrical engineering) from Nanjing Aeronautical
2000. Institute, Nanjing, China, in 1982 and in 1984,
[337] X. J. Zeng and M. G. Singh, “Approximation theory of fuzzy systems- respectively, and the Ph.D. degree in electrical
SISO case,” IEEE Trans. Fuzzy Syst., vol. 2, no. 2, pp. 162–176, May engineering from the University of Melbourne,
1994. Melbourne, Australia, in 1992.
[338] ——, “Approximation theory of fuzzy systems-MIMO case,” IEEE He has been an Associate Professor and then Pro-
Trans. Fuzzy Syst., vol. 3, no. 2, pp. 219–235, May 1995. fessor at City University of Hong Kong since 2000,
[339] ——, “Approximation accuracy analysis of fuzzy systems as function and was a Lecturer/Senior Lecturer at the School
approximators,” IEEE Trans. Fuzzy Syst., vol. 4, no. 1, pp. 44–63, Feb. of Electrical Engineering, University of New South
1996. Wales, Australia, from 1992 to 1999. He was a visiting Fellow at the National
[340] H. G. Zhang and L. L. Cai, “Nonlinear adaptive control using the University of Singapore (1997) and Aachen Technology University, Germany
Fourier integral and its application to CSTR systems,” IEEE Trans. (1997–1998). He has authored and/or coauthored numerous referred technical
Syst., Man, Cybern., B, Cybern., vol. 32, no. 3, pp. 367–372, Jun. 2002. papers. His current research interests include robust adaptive control, signal
[341] H. G. Zhang, L. L. Cai, and Z. Bien, “A fuzzy basis function vector- processing, piecewise linear systems, and intelligent systems and control.
based multivariable adaptive controller for nonlinear systems,” IEEE Dr. Feng was awarded an Alexander von Humboldt Fellowship in
Trans. Syst., Man, Cybern., B, Cybern., vol. 30, no. 1, pp. 210–217, 1997–1998. He is a an Associate Editor of the IEEE TRANSACTIONS ON
Feb. 2000. AUTOMATIC CONTROL, the IEEE TRANSACTIONS ON FUZZY SYSTEMS, the
[342] J. Zhang, “Modeling and optimal control of batch processes using re- IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS, PART C, and the
current neuro-fuzzy networks,” IEEE Trans. Fuzzy Syst., vol. 13, no. 4, Journal of Control Theory and Applications. He was an Associate Editor of the
pp. 417–426, Aug. 2005. Conference Editorial Board of the IEEE Control System Society.