PDF Event Triggered Sliding Mode Control A New Approach To Control System Design 1St Edition Bijnan Bandyopadhyay Ebook Full Chapter

Event Triggered Sliding Mode Control A
New Approach to Control System

Design 1st Edition Bijnan
Bandyopadhyay
Visit to download the full and correct content document:
https://textbookfull.com/product/event-triggered-sliding-mode-control-a-new-approach
-to-control-system-design-1st-edition-bijnan-bandyopadhyay/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Sliding Mode Control Methodology in the Applications of

Industrial Power Systems Jianxing Liu
https://textbookfull.com/product/sliding-mode-control-
methodology-in-the-applications-of-industrial-power-systems-
jianxing-liu/
Process control system fault diagnosis: a Bayesian

approach First Edition Gonzalez
https://textbookfull.com/product/process-control-system-fault-
diagnosis-a-bayesian-approach-first-edition-gonzalez/
Photovoltaic Power System Modeling Design and Control

1st Edition Weidong Xiao
https://textbookfull.com/product/photovoltaic-power-system-
modeling-design-and-control-1st-edition-weidong-xiao/
Process Control: A Practical Approach Myke King
https://textbookfull.com/product/process-control-a-practical-
approach-myke-king/
Automation in Textile Machinery: Instrumentation and
Control System Design Principles 1st Edition L. Ashok
Kumar
https://textbookfull.com/product/automation-in-textile-machinery-
instrumentation-and-control-system-design-principles-1st-edition-
l-ashok-kumar/
Statistical Process Control: A Pragmatic Approach 1st

Edition Stephen Mundwiller
https://textbookfull.com/product/statistical-process-control-a-
pragmatic-approach-1st-edition-stephen-mundwiller/
Virtual Equivalent System Approach for Stability

Analysis of Model based Control Systems Weicun Zhang
https://textbookfull.com/product/virtual-equivalent-system-
approach-for-stability-analysis-of-model-based-control-systems-
weicun-zhang/
Statistical Process Control: A Pragmatic Approach 1st

Edition Stephen Mundwiller (Author)
https://textbookfull.com/product/statistical-process-control-a-
pragmatic-approach-1st-edition-stephen-mundwiller-author/
Gaosu Startup Control Theory and Application Technology

Gaosu Qidong Kongzhi Lilun He Yingyong Jishu Control
System and Energy System 1st Edition Yaobao Yin
https://textbookfull.com/product/gaosu-startup-control-theory-
and-application-technology-gaosu-qidong-kongzhi-lilun-he-
yingyong-jishu-control-system-and-energy-system-1st-edition-
Studies in Systems, Decision and Control 139
Bijnan Bandyopadhyay
Abhisek K. Behera
Event-Triggered
Sliding Mode
Control
A New Approach to Control System
Design
Studies in Systems, Decision and Control
Volume 139
Series editor
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
e-mail: kacprzyk@ibspan.waw.pl
The series “Studies in Systems, Decision and Control” (SSDC) covers both new
developments and advances, as well as the state of the art, in the various areas of
broadly perceived systems, decision making and control- quickly, up to date and
with a high quality. The intent is to cover the theory, applications, and perspectives
on the state of the art and future developments relevant to systems, decision
making, control, complex processes and related areas, as embedded in the fields of
engineering, computer science, physics, economics, social and life sciences, as well
as the paradigms and methodologies behind them. The series contains monographs,
textbooks, lecture notes and edited volumes in systems, decision making and
control spanning the areas of Cyber-Physical Systems, Autonomous Systems,
Sensor Networks, Control Systems, Energy Systems, Automotive Systems,
Biological Systems, Vehicular Networking and Connected Vehicles, Aerospace
Systems, Automation, Manufacturing, Smart Grids, Nonlinear Systems, Power
Systems, Robotics, Social Systems, Economic Systems and other. Of particular
value to both the contributors and the readership are the short publication timeframe
and the world-wide distribution and exposure which enable both a wide and rapid
dissemination of research output.
More information about this series at http://www.springer.com/series/13304

Bijnan Bandyopadhyay Abhisek K. Behera
•
Event-Triggered Sliding
Mode Control
A New Approach to Control System Design
123
Bijnan Bandyopadhyay Abhisek K. Behera
Systems and Control Engineering Systems and Control Engineering
Indian Institute of Technology Bombay Indian Institute of Technology Bombay
Mumbai, Maharashtra Mumbai, Maharashtra
India India
ISSN 2198-4182 ISSN 2198-4190 (electronic)

Studies in Systems, Decision and Control
ISBN 978-3-319-74218-2 ISBN 978-3-319-74219-9 (eBook)
https://doi.org/10.1007/978-3-319-74219-9
Library of Congress Control Number: 2017963863
© Springer International Publishing AG, part of Springer Nature 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
Printed on acid-free paper
This Springer imprint is published by the registered company Springer International Publishing AG part
of Springer Nature
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Dedicated to my Ph.D. guide, late Dr. S. S.
Lamba, former Professor, IIT Delhi
Bijnan Bandyopadhyay
Dedicated to my parents and gurus

Abhisek K. Behera
Preface
Variable structure systems using sliding mode control (SMC) was originated in
USSR in the late fifties to stabilize the uncertain dynamical systems with relay as a
feedback control law. It gained popularity outside USSR only after the late sev-
enties due to a survey paper in English by Prof. Vadim I. Utkin. Since then, it has
become now a well-established robust control technique to deal with the uncer-
tainties in the plant, and achieving the system stability. A vast number of scientific
publications and the practical applications of this control technique have made
SMC as an important area in the control literatures.
The attention on design of SMC in discrete-time domain was paid by many
researchers soon after the importance of microprocessor and computer/processors
are realized in control applications in the early eighties. The first and important
observation in the discrete-time design is that no exact sliding mode is achieved as
in the continuous-time counterpart. A new notion of sliding mode is introduced
which is known a quasi-sliding mode (QSM). This has led to the development of
discrete-time SMC as an important area in SMC due to its practical importance.
Many design approaches have been proposed to improve the performance of SMC
for the sampled-data system.
In this monograph, a new approach to design SMC is presented using a novel
implementation strategy, namely event-triggering. In this strategy, the control is
updated whenever a certain stabilizing condition is violated, and hence, the system
stability is always maintained. Due to the need-based control strategy, it finds a
major application in spatially distributed control systems to reduce the communi-
cation among different sensor and actuator ends. So, the resources of the systems
are optimally used. The event-triggering-based design of SMC not only gives the
robust performance but also ensures minimal use of resources in the control system.
This monograph presents the recent results on event-triggered SMC for robust
stabilization of dynamical systems. In the first part of the monograph, the prelim-
inaries on sampled-data systems with an introduction to event-triggered control are
presented to familiarize the readers the event-triggering-based design of control
law. In addition to this, a brief introduction to SMC and its design are also dis-
cussed. Then, the design of event-triggered SMC for both linear and nonlinear
vii
viii Preface
systems is given in Chaps. 2 and 3. The event-triggered SMC is presented for linear
systems guaranteeing the semi-global and global stability in Chap. 2. However,
only local stabilization result is discussed for the nonlinear systems, which is
presented in detail in Chap. 3. In the event-triggered control, the state trajectory is
continuously measured to evaluate the triggering condition which may not be
practical in all applications. So, Chaps. 4 and 5 present few variants of
event-triggered control, namely self-triggered and discrete event-triggered control,
respectively. In the self-triggering strategy, the triggering strategy is developed
without using the continuous state measurements. On the other hand, the periodic
state measurements are used in discrete event-triggered control and the control is
updated when it is violated at some periodic instants. In recent time, there has been
a considerable amount of interest in stability of quantized control system. The final
chapter presents the design of event-triggered SMC with quantized state mea-
surements. It is our belief that the material of the monograph would serve its
purpose and explore the new challenges on the topic.
This work would have not been completed without the support and encour-
agement from many of our friends and colleagues. The authors would like to thank
Indian Institute of Technology Bombay for providing the conducive environment to
carry out the research reported in the monograph. Finally, we extend our gratitude
to our family for their love, support and understanding throughout the process of
this endeavour.
Mumbai, India Bijnan Bandyopadhyay

December 2017 Abhisek K. Behera
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Computer-Controlled Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1 Basic Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.2 Design Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Event-Triggered Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 Preliminary Idea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.2 Stability of Event-Triggered Systems . . . . . . . . . . . . . . . . 8
1.2.3 Need of Event-Triggered Control . . . . . . . . . . . . . . . . . . . 11
1.3 Sliding Mode: An Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.3.1 Dynamics During Sliding Mode . . . . . . . . . . . . . . . . . . . . 14
1.3.2 Design of Sliding Mode Control . . . . . . . . . . . . . . . . . . . . 17
1.4 Discrete-Time Sliding Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.4.1 Switching-Based Reaching Law . . . . . . . . . . . . . . . . . . . . 22
1.4.2 Switching-Free Reaching Law . . . . . . . . . . . . . . . . . . . . . 23
1.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.6 Notes and References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2 Event-Triggered Sliding Mode Control for Linear Systems . . . . . . . 27
2.1 System Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2 Event-Triggered Sliding Mode Control . . . . . . . . . . . . . . . . . . . . 28
2.2.1 Stability of Sliding Motion . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2.2 Stability of Event-Triggered System . . . . . . . . . . . . . . . . . 33
2.2.3 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.3 Global Event-Triggered Sliding Mode Control . . . . . . . . . . . . . . . 37
2.3.1 Global Event-Triggering Rule . . . . . . . . . . . . . . . . . . . . . . 37
2.3.3 Global Stability of Event-Triggered System . . . . . . . . . . . . 41
ix
x Contents
2.4 Event-Triggered Sliding Mode Control for Multivariable

Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.4.1 Event-Triggered Design of SMC . . . . . . . . . . . . . . . . . . . 45
2.4.2 Event-Triggering Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3 Event-Triggered Sliding Mode Control for Nonlinear Systems . . . . . 55
3.2 Event-Triggered Sliding Mode Control . . . . . . . . . . . . . . . . . . . . 57
3.2.1 Stability of Sliding Motion . . . . . . . . . . . . . . . . . . . . . . . . 60
3.3 Event-Triggering Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.3.1 Design of Event-Triggering Scheme with Constraints . . . . 63
3.4 Event-Triggering with Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.5 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.5.1 Without Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.5.2 With Delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4 Self-Triggered Sliding Mode Control for Linear Systems . . . . . . . . . 77
4.2 Self-Triggering Scheme Without Delay . . . . . . . . . . . . . . . . . . . . 78
4.3 Self-Triggering Scheme with Delay . . . . . . . . . . . . . . . . . . . . . . . 81
4.3.1 Design of Self-Triggered Sliding Mode Control . . . . . . . . . 85
4.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5 Discrete Event-Triggered Sliding Mode Control for Linear
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.2 Discrete-Time Sliding Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.2.1 Bartoszewicz’s Reaching Law . . . . . . . . . . . . . . . . . . . . . 96
5.2.2 Design of Discrete-Time Sliding Mode Control . . . . . . . . . 97
5.3 Discrete Event-Triggered Sliding Mode Control: State
Feedback Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 98
5.3.1 Event-Triggered Bartoszewicz’s Reaching Law . . ....... 99
Contents xi
5.3.2 Event-Triggering Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

5.4 Discrete Event-Triggered Sliding Mode Control: Output
Feedback Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.4.1 Multirate Output Feedback Technique . . . . . . . . . . . . . . . . 103
5.4.2 Multirate-Based Event-Triggered Discrete-Time
Sliding Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6 Event-Triggered Sliding Mode Control with Quantized State
Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.1.1 Quantizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.2 Design of Sliding Mode Control . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.3 Design of Event-Triggered Sliding Mode Control . . . . . . . . . . . . . 117
6.3.1 Design of Event-Triggering Rule . . . . . . . . . . . . . . . . . . . 119
6.4 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Acronyms
A-D Analog-to-digital
D-A Digital-to-analog
DTSM Discrete-time sliding mode
ETCS Event-triggered control systems
FOS Fast output sampling
ISS Input-to-state stable
LTI Linear time-invariant
MIMO Multiple-input multiple-output
MROF Multirate output feedback
QSM Quasi-sliding mode
SISO Single-input single-output
SMC Sliding mode control
VSS Variable structure system
xiii
Symbols
R Set of real numbers

R0 Set of nonnegative real numbers
Z Set of integers
Z0 Set of nonnegative integers
Rn n-dimensional vector space over R
ða bÞðxÞ Composition of the two functions a and b
xy Dot product of any two vectors x and y in Rn
P [ ð Þ0 Positive (semi)-definite matrix P
Q⊤ Transpose of any matrix Q
jj Absolute value of a scalar variable ‘’
k: k1 1-norm of a finite-dimensional vector ‘’
k: k Euclidean (2-) norm of a finite-dimensional vector ‘’ or a matrix ‘’
of appropriate dimension
kmaxðminÞ fg Largest (smallest) eigenvalue of a square matrix ‘’
rf ðxÞ Gradient of a real-valued function f ðxÞ
sup (inf) Supremum or least upper bound (infimum or greatest lower bound)
b xc Floor function that returns largest integer less than or equal to x
ln Natural logarithm with base e (¼ 2:71828)
K Set of strictly increasing and continuous real-valued functions
defined on the nonnegative interval with zero at zero
K1 Set of unbounded class-K functions
sign Signum function
FðxÞ Set-valued map of the vector field f ðxÞ at the point of discontinuity x
in Filippov’s inclusion
co Convex closure
Bd ðxÞ Open ball of radius d centred at x
lðÞ Lebesgue measure of a set ‘’
v Sensitivity of the quantizer
M Saturation level of the quantizer
xv
xvi Symbols
Ti Inter-event time/time interval between two consecutive triggering

instants
s Constant sampling period for the discrete-time systems
N An integer greater than or equal to the observability index of the
system
Chapter 1
Introduction
This chapter briefly introduces the readers to the preliminary ideas on design and
analysis of computer-controlled systems and then sliding mode control (SMC). In
general, computer-controlled systems consists of both continuous and discrete-time
systems that interact among themselves through the feedback channel to achieve cer-
tain objectives. Different available classical techniques, namely emulation, discrete-
time and hybrid approaches, are summarized here with their own advantages and
disadvantages. In almost all these techniques, the periodic sampling interval is often
used to design and analyse the sampled-data systems for its simplicity and easier in
design. On the other hand, aperiodic control implementation is desired in sampled-
data systems to reduce the periodic computational burden and cost associated with
the implementation. However, this introduces few difficulties in analysing closed
loop system stability. A novel sampling strategy known as event-triggered control
is introduced here where the control is updated whenever it is demanded. In this
technique, the time instants for updating the control signal is determined using some
rule that ensures the stability of the system. So, this strategy maintains the system
stability while reducing extra burden on the system.
The design of SMC is also presented in this chapter to familiarize the readers. This
is a robust controller that stabilizes the plant in the presence of external disturbances.
The sliding motion and SMC are briefly elaborated to understand sliding mode
with discontinuous control action. This is followed by the design of SMC for linear
systems. The discrete realization of SMC, unfortunately, does not yield sliding motion
exactly due to discrete nature of control is also discussed. Some control design
techniques are reviewed for discrete-time sliding mode.
© Springer International Publishing AG, part of Springer Nature 2018 1

B. Bandyopadhyay and A. K. Behera, Event-Triggered Sliding Mode Control, Studies
in Systems, Decision and Control 139, https://doi.org/10.1007/978-3-319-74219-9_1
2 1 Introduction
1.1 Computer-Controlled Systems
Computer-controlled systems, in general, are broadly defined as a control system that

allows interaction of both analog system and digital controller through computer
or any other digital platform. Such systems are ubiquitous in almost all fields of
engineering due to rapid advancement of digital technology. The plant represents
the continuous-time system, while the control signal is a discrete in nature that is
applied to the plant. The control signal is updated at the discrete instants only and,
however, is held constant in between two consecutive sampling instants. This is why,
it is also called as sampled-data system. The plant dynamics evolves in open loop
manner between two discrete instants as there is no control variable. So, the interest
is mainly focused on the plant behaviour at the discrete instants only leading to
the so-called discrete-time systems. In other words, the system dynamics evolves
at discrete instants with the control input applied at these instants. This is not only
easy to analyse but also one of the simplest ways of implementing the control law to
achieve a certain objective compared to its analog counterparts.
In actual practice, computer-controlled systems have facilitated the control design
problems to a great extent by introducing the digital control signal. However, it should
not be thought of as an ideal control design and implementation scenario for a physi-
cal plant. Rather, it is a limiting approach of the analog control system (both plant and
control evolve in continuous-time) and all the analyses are carried out for the analog
control system in sampled-data fashion. The control law is designed for any plant
using system dynamics which may be defined on continuous- and/or discrete-time
domain. However, in the present case only, we focus on continuous-time plant. So,
the control is designed from the continuous time plant which achieves desired perfor-
mance of the system when it is implemented in continuous manner. But, in discrete
implementation, the continuous-time performance may also be achieved if it is imple-
mented at a faster sampling rate. So, the computer-controlled system yet provides an
alternative to analog control implementation subject to some desired performance.
In spite of this, there are numerous challenges associated with computer-controlled
systems that make the researchers to rethink possible new techniques for control
implementation and design philosophy.
1.1.1 Basic Architecture
The basic architecture of a computer-controlled system is shown in Fig. 1.1. The

plant, which evolves in continuous-time, interacts with the controller only at discrete
instants. The plant states are sampled at discrete instants and converted into a digital
signals. The whole process of converting analog-to-digital conversion is represented
by the analog-to-digital (A-D) converter. These digital signals are then processed
through a control algorithm to generate a new set of digital signals, known as discrete
control signals. We see that there is same number of discrete control signals as the
1.1 Computer-Controlled Systems 3
Fig. 1.1 Basic architecture

D-A Analog Plant A-D
of a computer-controlled
systems
Clock
Control Law/
Algorithm
samples of plant states. Now, the discrete signal is processed by digital-to-analog (D-
A) converter to generate a continuous-time signal between two consecutive sampling
instants. But, this continuous signal is an approximation of analog control signal. The
overall process from A-D to D-A is coordinated by a clock that synchronizes all the
tasks carried out at A-D, control synthesis and D-A converter. Here, the control signal
may be designed in the continuous-time frame work or in discrete domain depending
on the design requirements.
Both the converters, often, operate in periodic time interval in the sampled-data
system. The information is sampled at time instants known as sampling instant, and
the interval between successive sampling instants is called as sampling period. For
every sampling instant, the control signal is computed and the same is applied to the
plant by its approximate analog signal through D-A converter. The control signal is
held constant in every sampling interval making the system as open loop, and hence,
the system evolves in open loop manner. This is one of the differences that makes
the computer-controlled systems different from other feedback systems. There has
been many techniques available to explore the analysis and design of such system.
But, yet the full potential of this system needs to be investigated for effective use of
computer-controlled systems. For instance, the effect of sampling interval on system
performance, the absence of clock that synchronizes both A-D and D-A converters,
etc. are need to be investigated. Of all these, sampling interval variation is one of
the most important and challenging problems in computer-controlled system. Too
fast sampling is sometime unnecessary, while the slow sampling may deteriorate the
system performance. So, it is always desirable to have an optimal sampling interval
that stabilizes the computer-controlled system.
1.1.2 Design Techniques
Many techniques have been used to analyse and design of computer-controlled

systems. The closed loop system is hybrid in nature involving both continuous
and discrete dynamics; however, the standard available technique deals with either
continuous-time design or discrete-time design. That means design the control either
4 1 Introduction
using continuous-time or discrete-time model and apply it to the continuous-time

plant. In the same manner, the stability analysis of the system is carried out using
either of these models depending on control signal. Broadly speaking, there are three
methods available for analysing the computer-controller system, namely emulation-
based approach, discrete-time approach and hybrid system approach. Also, in most
applications the constant sampling period is used for implementing the discrete con-
troller.
1.1.2.1 Emulation-Based Approach
It may also be seen as continuous-time approach design to sampled-data system. Here,

the performance specifications are in continuous-time domain since the plant and
controller both are in continuous-time domain. The basic idea is first to ignore the A-D
and D-A converters in Fig. 1.1 and follow the design steps of any stabilizing controller.
The continuous-time controller is then approximated by replacing derivatives with
finite-differences and continuous-time signals with sampled values at that instants.
Though it is an approximation, the satisfactory performance of the sampled-data
system is still achieved by this approximation. Here, the controller designed from
continuous-time plant emulates the behaviour of continuous-time plant in spite of
discrete implementations, and hence, this is known as emulation-based approach.
However, the main issue in this approach is that the sampling period is not taken
into account in the design of controller. It is very natural that the emulated controller
gives the stability of sampled-data system for a range of sampling period only while
it destabilizes for other values of sampling period.
1.1.2.2 Discrete-Time Approach
This approach is simple and is based on the discrete-time model of the plant. The
continuous plant interacts with digital controller through A-D and D-A converters.
Thus, the controller sees the plant as discrete-time model through these converters.
The discrete-time representation of the plant is obtained by combining the plant with
A-D and D-A converters. There are numerous approaches available for discretizing
the plant for a given sampling period, and the popular among them are Euler dis-
cretization, zero-order hold (ZOH) discretization, etc. The closed loop response is
analysed only at discrete instants since it ignores the inter-sampling behaviour of
the plant. On the other hand, the stability of the system is analysed by ignoring the
dynamics between two sampling instants. The wide application of this technique is
found in many slow processes where it is enough to study the system behavior only
at some periodic intervals. However, the well-known shortcomings of this technique
are that no inter-sampling dynamics of the plant can be analysed and selection of a
suitable sampling interval that captures the undesirable phenomenon in the system.
1.1 Computer-Controlled Systems 5
1.1.2.3 Hybrid Approach
As the name suggests, in this technique, both the continuous and discrete behaviours
are analysed for the sampled-data system without representing the system by some
approximated system dynamics. This is why, it is also known as direct design
approach to sampled-data system. Due to this hybrid nature of the system, the design
and stability methods are complicated than earlier two methods.
It is to be noted that variable sampling period may also be used for designing
the controller for sampled-data systems. However, it involves many design issues
for analysing the stability due to restricted mathematical tools and/or time-varying
nature of system dynamics. Nevertheless, many attempts have been made for sta-
bilizing the sampled-data systems with aperiodic control sampling process. Event-
triggering strategy is one of such techniques that generates nonuniform sampling
instants while ensuring system stability. In this book, only event-triggered technique
will be emphasized for computer-controlled systems.
1.2 Event-Triggered Control
Event-triggering strategy is a control implementation technique that is motivated

from the Lebesgue sampling. In this case, sampling period is not constant but is
determined by the evolution of system trajectory satisfying some stability condi-
tion. To understand the concept of event-triggered control, first Lebesgue sampling
technique is discussed.
In classical sampled-data system, generally constant sampling period is chosen
and the control is implemented once this constant time period is elapsed. This is
generally referred as Reimann sampling as shown in Fig. 1.2a. For any constant hR >
0, the sampling of the continuous-time signal ψ(t) takes place at every hR intervals
of time. It does not monitor the evolution of signal ψ(t). As a result of this, the
important concern in Reimann sampling is the proper selection of sampling period
to capture the transient behaviour of the plant. On the other hand, the signal may
also be sampled at the time instants when state evolution from its immediate past
sampled value crosses a certain threshold value (say, hL ) as the time does in Reimann
sampling. This is known as Lebesgue sampling which is shown in Fig. 1.2b. Though
both the sampling techniques are different, these have similarity in the sampling
mechanism. In the former, the time is measured while state is monitored in the latter
case. However, in doing so many advantages are obtained in the case of Lebesgue
sampling. For example, it is not necessary most of the time to update the control
signal frequently at periodic interval. So, Lebesgue sampling gives sampling instant
whenever it is needed subject to some satisfactory system performance.
In case the function ψ(t) is a finite-dimensional vector for any fixed t, the sampling
instant may be decided by observing the individual state evolution of the vector-
valued function. Then, for any given constants hLi > 0 with i = 1, 2, . . . , n, the
individual state evolution, denoted by hLi (t), is observed for generating sampling
6 1 Introduction
Fig. 1.2 Comparison of ψ (t)

Reimann and Lebesgue
samplings
hR t
(a) Reimann sampling
ψ (t)
hL
t
(b) Lebesgue sampling
instant of the corresponding state. However, it is very complicated and difficult to

analyse the stability of closed loop system. Another school of thought is to sample all
the states simultaneously whenever certain condition is violated. Thus, this strategy
does not necessitate individual sampling of the state at different time instants. Due to
this, the latter is more convenient for implementing practically than the earlier one.
1.2 Event-Triggered Control 7
1.2.1 Preliminary Idea
Event-triggered control is one of such techniques that generates the sampling instant
(also called as triggering instant) for sampling and updating the control signal. To pro-
vide a preliminary idea on event-triggered control, we consider a nonlinear dynamical
system
ẋ = f (x, u), x(0) = x0 ∈ Rn (1.1)
where the function f (·, ·) is Lipschitz with respect to both the arguments u ∈ Rm . Let
there exists a continuous feedback control law u(x) = π(x) such that the dynamics
ẋ = f (x, π(x))
is asymptotically stable. It is assumed that the control is implemented digitally to

the plant. So, the control signal π(x) is computed for every sampling instant and is
applied to the plant at these discrete instants. Then, the system becomes open loop
between two consecutive sampling instants. However, due to this, the discrete error
is introduced in the plant, defined by e(t) = x(ti ) − x(t) with e(ti ) = x(ti ) − x(ti ) = 0
where t ∈ [ti , ti+1 ). This error appears in the plant due to discrete implementation of
continuous-time control, but its value is zero if the control is continuously updated
as in analog implementation.
Further, we assume that the system (1.1) is input-to-state stable (ISS) with respect
to the error e(t). That means there exists a continuously differentiable Lyapunov
function V : Rn → R≥0 such that
a(x) ≤ V (x) ≤ a(x) (1.2)

∇V (x) · f (x, π(x + e)) < −a(x) + γ (e) (1.3)
for some class-K∞ 1 functions a, a, a, and class-K function γ . Here, the notation
‘·’ denotes inner (scalar) product. Event-triggering strategy is developed for deter-
mining the sampling instants such that desired stability is achieved. In this case, the
asymptotic stability of the system is desired with the discrete implementation of the
control law. So, the obvious condition for which this holds is γ (e) < σ a(x) for
some σ ∈ (0, 1). This can be simplified further, by assuming a−1 and γ are Lipschitz
on some compacts, as σ x > Le e, where Le is an appropriate constant. Thus, the
triggering instant may be generated by executing the following,
ti+1 = inf {t > ti : σ x(t) ≤ Le e(t)} . (1.4)
1 Any function a is said to be class-K if it is continuous, strictly increasing, zero at zero. Again, it
is said to be class-K∞ if it belongs to class-K and is unbounded. Clearly, class-K∞ functions are
subsets of class-K functions.
8 1 Introduction
This is known as triggering rule for event-triggered control π(x). It is seen that
this ensures Le e < σ x which implies that γ (e) < σ a(x) also holds. This
implies from (1.2) and (1.3) that
V̇ < −(1 − σ )a(x)

≤ −(1 − σ ) a ◦ a−1 (V )
= −(1 − σ )a (V )
<0
where K∞ a := a ◦ a−1 which is the composition of two functions a and a−1 .

This shows that closed loop system is asymptotically stable with the control applied
at discrete instants. Moreover, in the event-triggering technique the inter-sampling
behaviour is considered for stability of the closed loop system.
It is worth mentioning here that the event-triggered control scheme not only guar-
antees system stability but also ensures the convergence of inter-sampling behaviour.
From the above, it is observed that the Lyapunov function decreases continuously
and goes to zero as the time tends to infinity. This is one of the important properties
of the event-triggered control system (ETCS).
1.2.2 Stability of Event-Triggered Systems
In this technique, the triggering instants are generated by the triggering rule at which
the control signal is updated and this results the closed loop system stability. However,
it might happen that the control signal is not updated at the triggering instant when
triggering instants are too close to each other. This demands fast execution of control
tasks, or even in worst-case continuous-time like execution which is not possible
by digital processor. It may be noted here that in periodic execution such situation
does not arise due to each sampling/triggering instant that occurs after every constant
sampling period.
Let {ti }i∈Z≥0 be sequence of triggering instants generated by some stabilizing
triggering rule. We define Ti = ti+1 − ti as the inter-event/execution time for any
given triggering sequence {ti }i∈Z≥0 . For stability of the event-triggered system, the
inter-event time must be strictly greater than zero, i.e. Ti > aT for all i ∈ Z≥0 and
some positive constant aT . This guarantees the Zeno-free execution of triggering
sequence. The positive inter-event time ensures control is updated after every finite-
time interval only. This is essential for the processor to execute the control task and
update the control signal. In other words, it can be said that {ti }i∈Z≥0 is an increasing
sequence, i.e. t0 < t1 < t2 < · · · such that ti+1 > ti + aT . Such a triggering sequence
is feasible for implementing the control practically to ensure the stability of closed
loop system. The triggering instants generated by some triggering rule that is not
necessarily satisfying the above property would make the event-triggered system
unstable.
Example 1.1 Consider a scalar nonlinear control system as
ẋ = x2 + u
where x ∈ {[−c, c] : c ∈ R>0 } which is a compact set. Any stabilizing controller can
be designed for the above system to ensure the asymptotic stability. Let u = −x2 −kx
be a feedback control which ensures the asymptotic stability of the system with k > 0.
This control is applied to the above system at discrete instants only such that closed
loop system is stable. So, the discrete-time control is given as
u(t) = −x2 (ti ) − kx(ti ), t ∈ [ti , ti+1 ), i ∈ Z≥0 .
It can be shown that the closed loop system with the above discrete control is
ISS with respect to the error. Choose V (x) = 21 x2 . Then, with some calculation, we
arrive at

V̇ (x(t)) = x(t) x2 (t) − x2 (ti ) − kx(ti )
= −x(t) (x(t) + x(ti )) e(t) − kx(t)x(ti )
≤ 2c|e(t)||x(t)| − kx(t)(e(t) + x(t))
= 2c|e(t)||x(t)| − kx(t)e(t) − kx2 (t)
≤ 2c|e(t)||x(t)| + k|x(t)||e(t)| − kx2 (t)
≤ (2c + k)|e(t)||x(t)| − kx2 (t).
Here, we use the fact |x| ≤ c and x(ti ) = e(t) + x(t). Now applying Young’s
inequality2 to the first term (ε = 2c+k
k
), we obtain
k (2c + k)2
V̇ (t) ≤ − |x(t)|2 + |e(t)|2
2 2k
= −a(|x(t)|) + γ (|e(t)|)
where a(r) and γ (r) are given as
k 2 (2c + k)2 2
a(r) = r and γ (r) = r .
2 2k
Thus, the triggering rule is designed according to (1.4) which stabilizes the system
and is given by
2 Young’s inequality for exponent two states that for any nonnegative real numbers p, q and every
ε > 0, the following holds
p2 εq2
pq ≤ + .
2ε 2
10 1 Introduction

k
ti+1 = inf t > ti : σ |x(t)| ≤ |e(t)|
2c + k
for some σ ∈ (0, 1). This triggering rule ensures |e(t)| < σ 2c+k k
|x(t)| for all time
and thus implies that V̇ < 0 for all time. Hence, the closed loop system becomes
asymptotically stable even if the control is applied at the discrete instants generated
by the triggering sequence {ti }i∈Z≥0 . It can also be established that the triggering rule
does not have a Zeno triggering sequence. For x ∈ {[−c, c] : c ∈ R>0 }\{0}, one can
write

d |e(t)| 1 d d
= 2 |e(t)| |x(t)| − |x(t)| |e(t)|
dt |x(t)| x (t) dt dt

1
≤ 2 d e(t) |x(t)| − d |x(t)| |e(t)|
x (t) dt dt

1
≤ 2 d x(t) |x(t)| + d x(t) |e(t)|
x (t) dt dt

1 d
= 2 (|x(t)| + |e(t)|) x(t) .
x (t) dt
Now, using the fact |x| ≤ c and the control law u(t) = −x2 (ti ) − kx(ti ) for all
t ∈ [ti , ti+1 ) and i ∈ Z≥0 , one can easily obtain that

d
x(t) = x2 (t) − x2 (ti ) − kx(ti )
dt
= |−(x(t) + x(ti )) e(t) − k (e(t) + x(t))|
= |x(t) + x(ti )| |e(t)| + k |e(t) + x(t)|
≤ 2c|e(t)| + k|e(t)| + k|x(t)|
< (2c + k)|e(t)| + (2c + k)|x(t)|
= (2c + k)(|e(t)| + |x(t)|).
Using this in the following, it can be written as
d |e(t)| 1
< 2 (|x(t)| + |e(t)|)2 (2c + k)
dt |x(t)| x (t)
2
|e(t)|
= (2c + k) +1 .
|x(t)|
The solution to the above differential inequality can be obtained using comparison
Lemma. Then, the solution to the above differential can be obtained as
|e(t)|
≤ μ(t), t ∈ [ti , ti+1 )
|x(t)|
where μ(t) satisfies the differential equation μ̇ = (2c + k)(1 + μ)2 with the initial
|e(ti )|
condition |x(ti )|
= μ(ti ) = 0. Then, corresponding to triggering mechanism for this
system, the lower bound of the inter-event time is obtained as
σk
Ti > > 0.
(2c + (1 + σ )k)(2c + k)
This shows that the inter-event time is lower bounded by a positive quantity which
is strictly greater than zero. Indeed, it is necessary to eliminate the Zeno execution
of triggering sequence and ultimately to guarantee the system stability.
In the numerical simulation, the values of c, k and σ are selected as 5, 1 and
0.85, respectively. The initial condition is taken as x(0) = 4. The response of the
system is shown in Fig. 1.3. It is seen that state trajectory goes to zero as time tends
to infinity. The varying sampling interval or inter-event time generated by executing
the triggering rule is shown in Fig. 1.4. It is seen that the inter-event time is lower
bounded from zero which is given by 0.0065 and it increases to a value as high as
0.072. The plot of control signal is also shown in Fig. 1.5. As the sampling interval
increases, the control signal also remains constant until the next sampling instant is
generated. The plot of Lyapunov function is given in Fig. 1.6. It is seen that event-
triggered control implementation achieves asymptotic stability of the closed loop
system with guaranteed convergence of inter-sampling behaviour of the system.
1.2.3 Need of Event-Triggered Control
Event-triggered control is one of the aperiodic control implementation strategies in

digital platform that ensures closed loop system stability. Unlike periodic sampling,
Fig. 1.3 Response of the

system
12 1 Introduction
Fig. 1.4 Variation of

sampling interval or
inter-event time
Fig. 1.5 Event-triggered

control signal
Fig. 1.6 Time evolution of

Lyapunov function
here sampling instants are determined whenever it is demanded subject to system

stability. Otherwise, no control signal is updated. So, if there is no triggering for long
duration, then the control is not updated while maintaining the stability of the system.
Such a need-based strategy is more useful in resource-constrained systems such as
networked control system, embedded control systems. In these systems, frequent
state transmission in periodic interval is sometimes not desirable due to bandwidth
constraints of the communication channel.
In the event-triggered control strategy, there is less computational burden than that
of periodic control implementation subject to some satisfactory system performance.
This is mainly because the control signal is computed only when an event is triggered.
This is more appealing in many practical applications to avoid control computation
whenever the states do not change rapidly. Moreover, it enables the digital processors
to work in parallel scheduling tasks by minimizing the periodic execution.
Another important aspect of event-based technique is the scheduling of sensors in
practical systems. Sensors are the integral part of the control systems which measure
the state/output trajectory and transmit it to the control end. In case the data is to be
transmitted over lossy channel, it is desirable to have a control policy that requires
minimal state transmission. So, in such cases, scheduling of sensors is so designed
that it can decide the transmission instant. Event-triggering strategy is very much
helpful in scheduling these sensors.
Apart from these, this strategy is successfully deployed in many other applications
such as in signal processing, state estimation. However, the discussion of these topics
is the beyond the scope of this book, so here only the control aspect of event-triggered
technique is exploited.
1.3 Sliding Mode: An Introduction
Sliding mode has its root in the variable structure system (VSS) where the system
structure changes during the evolution of the system dynamics. In VSS, the structure
of the system is changed or switched such that the closed loop system is asymp-
totically stable. However, in SMC the system trajectory is forced to remain on a
predesigned manifold by the action of discontinuous control signal. In this case, the
vector fields on both the sides of sliding (we also refer it as ‘switching’) manifold
act towards this manifold and thus maintain the trajectory along it. This eventually
ensures sliding takes place and is called sliding mode. Since the system dynamics
switches between two structures while sliding mode is enforced, SMC is referred as
a special class of VSS where switching takes place along a predesigned switching
manifold. Due to this, overall system becomes discontinuous on this manifold. It is
not necessarily required that sliding mode is enforced in the system from the very
beginning, but it must start in some finite-time. Otherwise, system cannot be said to
be in sliding mode. In controlled dynamical system, the control law is often designed
that brings sliding mode in the system and is called as sliding mode control. This
control is responsible for sliding mode.
14 1 Introduction
To illustrate the concept of sliding mode, consider a controlled nonlinear dynam-

ical system as
ẋ = f (x, u), x ∈ Rn and u ∈ R. (1.5)
The vector field f : Rn × R → Rn given in (1.5) is Lipschitz with respect to

its first argument. Let the sliding manifold be given for any continuous switching
function, denoted by s(x) which maps Rn to R, as
S = x ∈ Rn : s(x) = 0 .
Similarly, we also define S + = {x ∈ Rn : s(x) > 0} and S − = {x ∈ Rn :

s(x) < 0}. The control law that brings sliding mode in the system (1.5) is given
by
u+ (x) if s(x) > 0,

u(x) = (1.6)
u− (x) if s(x) < 0.
The control functions u+ and u− are continuous, and also u+ = u− . Clearly, this
implies that the control u is discontinuous on s = 0. The corresponding resulting
vector fields are f + (x, u+ ) and f − (x, u− ) due to the control signals u+ and u− ,
respectively. So, the vector field f (·, ·) is also discontinuous on S . The existence of
solutions of the closed loop system (1.5) with control law (1.6) cannot be explained
using the classical existence theorem. Indeed, in this case, the closed loop system
becomes discontinuous and is referred as system with discontinuous right-hand side.
Though there are many techniques available for defining solutions of such systems,
in this book we understood the solutions of the system in Filippov’s sense [1].
1.3.1 Dynamics During Sliding Mode
The existence of solutions of the dynamical system (1.5) is briefly discussed in

Filippov’s sense. The dynamical system is first replaced by a set-valued function,
called differential inclusion, on the zero (Lebesgue) measure set where solution is not
defined in classical sense. We denote it by F (x, u). Thus, the differential inclusion
for the system (1.5) is written as
ẋ ∈ F (x, u). (1.7)
The set-valued map F coincides with f ; i.e., F contains only one element f ,
whenever the latter is continuous on its respective domain. In other words, the set-
valued map represents the vector fields of the dynamical system at the points of
discontinuity. One of the main reasons to replace the dynamical equation (1.5) by
1.3 Sliding Mode: An Introduction 15
an inclusion (1.7) is that to capture all the vector fields of the system at the point of
discontinuity. Another way of interpretation is that the right-hand side is enlarged
such that all vector fields in the vicinity of manifold are contained in the inclusion.
Once that is accomplished, the most obvious question is under what conditions the
solution of the system (1.7) exists. Before that, we briefly discuss how the set-valued
map can be constructed from an dynamical system (1.5) that has discontinuous right-
hand side.
Since the inclusion, given in (1.7), captures all the vector fields in some sufficiently
small δ-neighbourhood of s(x) = 0, the set-valued map is constructed by collecting
the convex combination of the vector fields in that neighbourhood. Thus, the limiting
vector fields in the small neighbourhood of the domains S + and S − for any x ∈ S
are obtained as
lim f (y, u+ (y)) = f + (x, u+ (x)), lim f (y, u− (y)) = f − (x, u− (x)).
y∈S + y∈S −
y→x y→x
Then, the set-valued map F (x, u) is obtained by collecting all the vector fields
pointing the line segment joining the end points of the vector fields in S + and S − ,
i.e. f + and f − . Note that every trajectory in S + crosses S before reaching S − and
vice versa. This implies that the line segment joining f + and f − also intersects S .
If Tx S represents the tangent vector of S at the point x, then the line segment also
intersects Tx S at some point. Thus, this intersection point gives the end point of the
vector f 0 (x, u) such that
ẋ = f 0 (x, u) (1.8)
holds for the point x ∈ S . Similarly, it can be defined for other points on S . This
gives the motion of the system trajectory on the sliding manifold. In other words,
the function x(t) satisfies (1.8) is a solution of the system (1.5). This is because this
solution also satisfies the inclusion (1.7) as f 0 is contained in the inclusion. Also,
f − = f 0 and f + = f 0 ; the motion of the system remains tangent to S and is called
sliding motion. Thus, during sliding mode, the system trajectory is a solution to (1.8)
which is also a solution of (1.5) due to (1.7).
Now, we discuss how to find the velocity vector f 0 during sliding mode. Define
fN and fN− as the projections of the vectors f + and f − , respectively, onto the normal
+
to the surface S at the point x ∈ S . Then, the vector field f 0 is calculated using
(1.8) as
fN−
f 0 = θ f + + (1 − θ )f − , θ= , θ ∈ [0, 1]. (1.9)
fN− − fN+
This shows the vector field during sliding mode is a convex combination of the
vector fields f + and f − such that the trajectory is tangent to surface S . The value
of θ given in (1.9) is obtained from the relation
Another random document with
no related content on Scribd:
A common fish (A. cuchia) in Bengal, remarkable for its singular
respiratory apparatus. It has only three branchial arches, with
rudimentary branchial laminæ, and with very narrow slits between
the arches. To supplement this insufficient respiratory apparatus, a
lung-like sac is developed on each side of the body behind the head,
opening between the hyoid and first branchial arch. The interior of
the sac is abundantly provided with blood-vessels, the arterial
coming from the branchial arteries, whilst those issuing from it unite
to form the aorta. A. cuchia approaches the Eels in having the
humeral arch not attached to the skull.
Monopterus.—Vent in the posterior half of the body, which is
naked. Three branchial arches with rudimentary gills, but without
breathing sac.
One species (M. javanicus), which is extremely common in the
East Indian Archipelago and in the eastern parts of the Continent.
Upwards of three feet long.
Symbranchus.—Vent in the posterior half of the body, which is
naked. Four branchial arches with well developed gills.
Three species, of which one (S. marmoratus) is extremely
common in tropical America, and the other (S. bengalensis) not less
so in the East Indies.
Chilobranchus.—Vent in the anterior half of the length of the
body, which is naked. Vertical fins reduced to a simple cutaneous fold,
without rays.
A small fish (Ch. dorsalis) from North Western Australia and
Tasmania.
Thirty-First Family—Murænidæ.
Body elongate, cylindrical or “band-shaped, naked or with
rudimentary scales. Vent situated at a great distance from the head.
Ventral fins none. Vertical fins, if present, confluent, or separated by
the projecting tip of the tail. Sides of the upper jaw formed by the
tooth-bearing maxillaries, the fore part by the intermaxillary, which is
more or less coalescent with the vomer and ethmoid. Humeral arch
not attached to the skull. Stomach with a blind sac; no pyloric
appendages. Organs of reproduction without efferent ducts.
The “Eels” are spread over almost all fresh waters and seas of
the temperate and tropical zones; some descend to the greatest
depths of the oceans. The young of some have a limited pelagic
existence. (Leptocephali, see p. 179.) At Monte Bolca fossil remains
are very numerous, belonging to recent genera, Anguilla,
Sphagebranchus, and Ophichthys; even larval Leptocephales have
been preserved. Anguilla has been found also in the chalk of Aix and
Oeningen.
In the majority of the species the branchial openings in the
pharynx are wide slits (Murænidæ platyschistæ); in others, the true
Murænæ, (Murænidæ engyschistæ) they are narrow.
Nemichthys.—Exceedingly elongate, band-shaped; tail tapering
into a point. Vent approximate to the pectorals, but the abdominal
cavity extending far behind the vent. Jaws produced into a long
slender bill, the upper part being formed by the vomer and
intermaxillaries. The inner surface of the bill covered with small tooth-
like asperities. Eye large. The nostrils of each side are close together,
in a hollow before the eye. Gill-openings wide, nearly confluent.
Pectoral and vertical fins well developed.
This very singular type is a deep-sea form, occurring at depths of
from 500 to 2500 fathoms. The two species known have hitherto
been found in the Atlantic only.
Cyema.—This genus combines the form of the snout of
Nemichthys, with the soft and shorter body of a Leptocephalus; but
the gill-openings are very narrow and close together on the abdominal
surface. Vent in about the middle of the length of the body; vertical fins
well developed, confined to, and surrounding, the tail. Pectoral fins
well developed. Eye very small.
Known from two specimens only, 4½ inches long, dredged in
depths of 1500 and 1800 fathoms in the Pacific and Antarctic
Oceans.
Saccopharynx.—Deep-sea Congers, with the muscular system
very feebly developed, with the bones very thin, soft, and wanting in
inorganic matter. Head and gape enormous. Snout very short,
pointed, flexible, like an appendage overlapping the gape. Maxillary
and mandibulary bones very thin, slender, arched, armed with one or
two series of long, slender, curved, widely set teeth, their points being
directed inwards; palate toothless. Gill-openings wide, at some
distance from the head, at the lower part of the sides; gills very
narrow, free, and exposed. Trunk of moderate length. Stomach
distensible in an extraordinary degree. Vent at the end of the trunk.
Tail band-like, exceedingly long, tapering in a very fine filament.
Pectoral small, present. Dorsal and anal fins rudimentary.
This is another extraordinary form of Deep-sea Eels; the
muscular system, except on the head, is very feebly developed; the
bones are as thin, soft, and wanting in inorganic matter, as in the
Trachypteridæ. This fish is known from three specimens only, which
have been found floating on the surface of the North Atlantic, with
their stomachs much distended, having swallowed some other fish,
the weight of which many times exceeded that of their destroyer. It
attains to the length of several feet.
Synaphobranchus.—Gill-openings ventral, united into a
longitudinal slit between the pectoral fins, separate internally. Pectoral
and vertical fins well developed. Nostrils lateral, the anterior
subtubular, the posterior round, before the lower half of the eye. Cleft
of the mouth very wide; teeth small; body scaly. Stomach very
distensible.
Deep-sea Congers, with well-developed muscular system,
spread over all oceans, and occurring in depths of from 345 to 2000
fathoms. Four species are known. Probably attaining to the same
length as the Conger.
Anguilla.—Small scales imbedded in the skin. Upper jaw not
projecting beyond the lower. Teeth small, forming bands. Gill-openings
narrow, at the base of the pectoral fins. The dorsal fin commences at
a considerable distance from the occiput.
Some twenty-five species of “Eels” are known from the
freshwaters and coasts of the temperate and tropical zones; none
have been found in South America or the west coast of North
America and West Africa. The following are the most noteworthy:—
The common European species (A. anguilla) is spread over Europe
to 64° 30´ lat. N., and all round the Mediterranean area, but is not
found either in the Danube or in the Black and Caspian Seas; it
extends across the Atlantic to North America. The form of the snout
varies much, and some naturalists have believed that specimens
with a broad and obtuse snout were specifically distinct from those
with pointed snout. However, every degree of breadth of the snout
may be observed; and a much safer way of recognizing this species,
and distinguishing it from other European Eels, is the forward
position of the dorsal fin; the distance between the commencement
of the dorsal and anal fins being as long as, or somewhat longer
than, the head. Eels grow generally to a length of about three feet,
but the capture of much larger examples is on record. Their mode of
propagation is still unknown. So much only is certain that they do not
spawn in fresh water, that many full-grown individuals, but not all,
descend rivers during the winter months, and that some of them at
least must spawn in brackish water or in deep water in the sea; for in
the course of the summer young individuals from three to five inches
long ascend rivers in incredible numbers, overcoming all obstacles,
ascending vertical walls or floodgates, entering every larger and
smaller tributary, and making their way even over terra firma to
waters shut off from all communication with rivers. Such
immigrations have been long known by the name of “Eel-fairs.” The
majority of the Eels which migrate to the sea appear to return to
fresh water, but not in a body, but irregularly, and throughout the
warmer part of the year. No naturalist has ever observed these
fishes in the act of spawning, or found mature ova; and the organs of
reproduction of individuals caught in fresh water are so little
developed and so much alike, that the female organ can be
distinguished from the male only with the aid of a microscope.
The second species found in Great Britain, on the coasts of
Europe generally, in China, New Zealand, and the West Indies, is (A.
latirostris) the “Grig” or “Glut,” which prefers the neighbourhood of
the sea to distant inland-waters, and in which the dorsal fin begins
farther backwards, the distance between the commencement of the
dorsal and anal fins being shorter than the head; its snout seems to
be always broad. On the American side of the Atlantic other species,
beside A. anguilla are found in abundance: A. bostoniensis, A.
texana. The largest Eels occur in lakes of the islands of the Indo-
Pacific, and they play a conspicuous part in the mythology of the
South-Sea Islanders and Maories; individuals of from eight to ten
feet in length have been seen, and referred to several species, as A.
mauritiana, fidjiensis, obscura, aneitensis, etc.
Conger.—Scaleless. Cleft of the mouth wide, extending at least to
below the middle of the eye. Maxillary and mandibulary teeth arranged
in series, one of which contains teeth of equal size, and so closely set
as to form a cutting edge. No canine teeth. Vomerine band of teeth
short. Pectoral and vertical fins well developed, the dorsal
commencing behind the root of the pectoral. Gill-openings large,
approximate to the abdomen. The posterior nostril opposite to the
upper or middle part of the orbit, the anterior in a tube. Eyes well
developed.
The “Congers” are marine Eels; the best known species (C.
conger) seems to be almost cosmopolitan, and is plentiful all round
Europe, at St. Helena, in Japan, and Tasmania. It attains to a length
of eight feet, and thrives and grows rapidly even in confinement,
which is not the case with the freshwater Eel. Three other species
are known, of which C. marginatus from the Indian Ocean, is the
most common. Leptocephalus morrisii is an abnormal larval
condition of the Conger.
Genera allied to Conger are Poeciloconger, Congromurcæna,
Uroconger, and Heteroconger.
Murænesox.—Scaleless. Snout produced. Jaws with several
series of small closely set teeth, anteriorly with canines; vomer with
several long series of teeth, the middle of which is formed by large
conical or compressed teeth. Gill-openings wide, approximate to the
abdomen. Pectoral and vertical fins well developed, the dorsal
beginning above the gill-opening. Two pairs of nostrils, the posterior
opposite to the upper part or middle of the eye.
Four species from tropical seas, M. cinereus being very common
in the Indian Ocean, and attaining to a length of six feet.
Nettastoma.—Scaleless. Snout much produced, depressed.
Jaws and vomer with bands of card-like teeth, those along the median
line of the vomer being somewhat the larger. Vertical fins well
developed; pectorals none. Gill-openings of moderate width, open.
Nostrils on the upper surface of the head, valvular; the anterior near to
the end of the snout, the posterior above the anterior angle of the eye.
This genus lives at some depth, the Japanese species (N.
parviceps) having been obtained at 345 fathoms. N. melanurum from
the Mediterranean, seems to inhabit a similar depth. Hyoprorus is its
Leptocephalid form.
Genera allied to Murcænesox are Saurenchelys, Oxyconger,
Hoplunnis, and Neoconger; in all these the nostrils have a superior
or lateral position. In other genera the nostrils perforate the upper lip,
as in Myrus, Myrophis, Paramyrus, Chilorhinus, Murænichthys, and
Ophichthys, the last genus deserving of particular mention on
account of its great range and common occurrence.
Ophichthys.—Nostrils labial; extremity of the tail free, not
surrounded by a fin.
More than eighty species are known, many of which are
abundant on the coasts of the tropical and sub-tropical zones. They
do not attain to a large size, but many must be extremely voracious
and destructive to other fishes, if we draw an inference from the
formidable dentition with which their jaws and palate is armed. Other
species have much more feeble, and some even obtuse teeth, better
adapted for seizing Crustaceans than vigorous and slippery fishes.
Some have rudimentary pectoral fins or lack them altogether. Many
are highly ornamented with bands or spots, the coloration being
apparently very constant in the several species.
Fig. 303.—Ophichthys crocodilinus,

from the Indo-Pacific.
Moringua.—Body scaleless, cylindrical, with the trunk much
longer than the tail. Pectorals none or small; vertical fins but little
developed, limited to the tail. Posterior nostrils in front of the small
eye. Cleft of the mouth narrow; teeth uniserial. Heart placed far behind
the branchiæ. Gill-openings rather narrow, inferior.
Six species from freshwaters, brackish water, and the coasts of
India to the Fiji Islands.
Muræna.—Scaleless. Teeth well developed. Gill-openings and
clefts between the branchial arches narrow. Pectoral fins none; dorsal
and anal fins well developed. Two nostrils on each side of the upper
surface of the snout; the posterior a narrow round foramen, with or
without tube; the anterior in a tube.
Fig. 304.—Head of a Muræna.
Fig. 305.—Muræna pavonina, from Southern Seas.

The Murænas are as abundantly represented in the tropical and
sub-tropical zones, and have nearly the same range, as Ophichthys.
The number of species known exceeds eighty. The majority are
armed with formidable pointed teeth, well suited for seizing other fish
on which they prey. Large specimens thus armed readily attack
persons in and out of the water; and as some species attain a length
of some six or eight feet, they are justly feared by fishermen. The
minority of species have obtuse and molar-like teeth, their food
consisting chiefly of Crustaceans and other hard-shelled animals.
Most of the Murænas are beautifully coloured and spotted, some in a
regular and constant manner, whilst in others the pattern varies in a
most irregular fashion: they have quite the appearance of snakes.
The Muræna of the Ancient Romans is Muræna helena, which is not
confined to the Mediterranean, but also found in the Indian Ocean
and on the coast of Australia. Its skin is of a rich brown, beautifully
marked with large yellowish spots, each of which contains smaller
brown spots.
Fig. 306.—Muræna picta, from the Indo-Pacific.

Gymnomuræna differs from Muræna in having the fins reduced to
a short rudiment near the end of the tail. Six species are known
growing to a length of eight feet.
Fig. 307.—Gymnomuræna vittata, from Cuba.
Myroconger and Enchelycore belong to the same sub-family as
Muræna, but the former is provided with pectoral fins, and in the
latter the posterior nostril is a long slit, and not round as in the other
genera.
FIFTH ORDER—LOPHOBRANCHII.
The gills are not laminated, but composed of small rounded lobes
attached to the branchial arches. Gill-cover reduced to a large simple
plate. Air-bladder simple, without pneumatic duct. A dermal skeleton
composed of numerous pieces arranged in segments, replaces more
or less soft integuments. Muscular system not much developed.
Snout prolonged. Mouth terminal, small, toothless, formed as in
Acanthopterygians.
Fig. 308.—Gills of Hippocampus abdominalis.
First Family—Solenostomidæ.
Gill-openings wide. Two dorsal fins, the rays of the anterior not
articulated. All the other fins well developed.
One living genus only is known, which was preceded in the
tertiary epoch by Solenorhynchus (Monte Postale).
Solenostoma.—Snout produced into a long tube. Body
compressed, with very short tail. All parts covered with thin skin,
below which there is a dermal skeleton formed by large star-like
ossifications. The soft dorsal and anal fins on elevated bases; caudal
fin long. Ventral fins inserted opposite to the anterior dorsal, close
together, seven-rayed; they are free in the male, but in the female
their inner side coalesces with the integuments of the body, a large
pouch for the reception of the eggs being formed thereby. Air-bladder
and pseudobranchiæ absent. Branchiostegals four, very thin.
Intestinal tract very simple, with a stomachic dilatation, without pyloric
appendages. Ova very small.
The dermal skeleton of this singular type is formed by star-like
ossifications, four in each horizontal and vertical series on the side of
the fore part of the trunk; each consists of four or three radiating
branches by which it joins the neighbouring bones; on the hind part
of the trunk and tail the series are diminished to two. The dorsal and
abdominal profiles in front of the fins are protected by similar bones.
The vertebral column is composed of eighteen abdominal and fifteen
caudal vertebræ, the vertebræ gradually decreasing in length
backwards, so that the shortness of the tail is caused not only by the
smaller number of vertebræ, but also by their much lesser length.
Neural and hæmal spines are developed. The pelvis consists of two
pairs of cartilaginous laminæ, the convex margin of the anterior
fitting into an angle of a dermal bone which separates the pelvis from
the well-ossified humeral arch.
The singular provision for the retention and protection of the eggs
has been described above (p. 162, figs. 73 and 74), and we have
only to repeat here that it is the female which takes care of the
progeny, and not the male as in the following family. Two or three
small species are known from the Indian Ocean; they are beautifully
marked, especially the male, which also appears to be of smaller
size in this genus than the female.
Second Family—Syngnathidæ.
Gill-openings reduced to a very small opening near the upper
posterior angle of the gill-cover. One soft dorsal fin; no ventrals, and,
sometimes, one or more of the other fins are also absent.
Small marine fishes, which are abundant on such parts of the
coasts of the tropical and temperate zones as offer by their
vegetation shelter to these defenceless creatures. They are bad
swimmers (the dorsal fin being the principal organ of locomotion),
and frequently and resistlessly carried by currents into the open
ocean or to distant coasts. All enter brackish water, some fresh
water. The strata of Monte Bolca and Licata (Sicily) have, yielded
evidence of their existence in the tertiary epochs; beside species of
Siphonostoma and Syngnathus (Pseudosyngnathus), remains of an
extinct genus, Calamostoma, allied to Hippocampus, but with a
distinct caudal fin, have been found. On their propagation see p.
163, Fig. 76.
A. Syngnathina.—The tail is not prehensile, and generally
provided with a caudal fin.—Pipe-Fishes.
Siphonostoma.—Body with distinct ridges, the upper caudal ridge
continuous with the lateral line, but not with the dorsal ridge of the
trunk. Pectoral and caudal fins well developed; dorsal fin of moderate
length, opposite to the vent. Humeral bones movable, not united into a
“breast-ring.” Males with an egg-pouch on the tail, the eggs being
covered by cutaneous folds.
Two species, of which S. typhle is common on the British, and
generally distributed on the European coasts.
Syngnathus.—Body with the ridges more or less distinct, the
dorsal ridge of the trunk not being continuous with that of the tail.
Pectoral fins well developed; caudal present. Dorsal fin opposite or
near to the vent. Humeral bones firmly united into the breast-ring.
Egg-pouch as in Siphonostoma.
The distribution of this genus nearly coincides with that of the
family, some fifty species being known. S. acus, the great Pipe-fish
(see Fig. 75, p. 163), is one of the most common European fishes,
extending across the Atlantic and southwards to the Cape of Good
Hope; it attains a length of 18 inches. Another very common species,
frequently met at sea, and spread over nearly all the tropical and
sub-tropical seas, is S. pelagicus, agreeably marked with alternate
brown and silvery cross-bars.
Doryichthys.—Body with the ridges well developed. Pectoral and
caudal fins present. Dorsal fin long or of moderate length, opposite to
the vent. Humeral bones firmly united. Males with the lower ridges of
the abdomen dilated, the dilated parts forming a broad groove for the
reception of the ova.
In these Pipe-fishes the ova are not received in a completely
closed pouch, but glued on to the surface of the abdomen. Twenty
species from tropical seas.
Nerophis.—Body smooth, rounded, with scarcely any of the
ridges distinct. Pectoral fin none, caudal absent or rudimentary, the tail
tapering into a point. Dorsal fin of moderate length, opposite to the
vent. The ova are attached to the soft integument of the abdomen of
the male, and are not covered by lateral folds of the skin.
Seven species from the European seas and the Atlantic. N.
æquoreus (Ocean Pipe-fish), N. ophidion (Straight-nosed Pipe-fish),
and N. lumbriciformis (Little Pipe-fish), are common on the British
coasts.
Protocampus.—The whole dermal skeleton is covered with skin.
A broad cutaneous fold runs along the back in front and behind the
dorsal; a similar fold along the abdomen. Pectoral fin none; caudal
very small.
The single species of this remarkable genus, P. hymenolomus,
occurs in the Falkland Islands. It may be regarded as an embryonal
form of Nerophis, the median skin-folds being evidently remains of
the fringe which surrounds the body of the embryo.
The other genera belonging to this group are, Icthyocampus,
Nannocampus, Urocampus, Leptoichthys, Coelonotus, and
Stigmatophora.
Hippocampina.—The tail is prehensile, and invariably without
caudal fin.—Sea-horses.
Gastrotokeus.—Body depressed, the lateral line running along
the margin of the abdomen. Shields smooth. Tail shorter than the
body. Pectoral fins. No pouch is developed for the ova, which are
imbedded in the soft integument of the abdomen of the male.
Gastrotokeus biaculeatus, very common in the Indian Ocean to
the coasts of Australia.
Solenognathus.—Body compressed, deeper than broad. Shields
hard, rugose, with round or oval interannular plates; and without
elongate processes. Tail shorter than the body. Pectoral fins.
Three species, from the Chinese and Australian Seas; they are
the largest of Lophobranchs, S. hardwickii, attaining to a length of
nearly two feet.
Fig. 309.—Phyllopteryx eques.
Phyllopteryx.—Body compressed, or as broad as deep. Shields
smooth, but some or all of them are provided with prominent spines or
processes on the edges of the body; some of the processes with
cutaneous filaments. A pair of spines on the upper side of the snout
and above the orbit. Tail about as long as the body. Pectoral fins. The
ova are imbedded in soft membrane on the lower side of the tail,
without a pouch being developed.
Three species from the coasts of Australia. The protective

resemblances with which many Lophobranchs are furnished, attain
to the highest degree of development in the fishes of this genus. Not
only their colour closely assimilates that of the particular kind of
seaweed which they frequent, but the appendages of their spines
seem to be merely part of the fucus to which they are attached. They
attain a length of 12 inches.
Hippocampus.—Trunk compressed, more or less elevated.
Shields with more or less prominent tubercles or spines. Occiput
compressed into a crest, terminating at its supero-posterior corner in a
prominent knob (coronet). Pectoral fins. The males carry the eggs in a
sac at the base of the tail, opening near the vent.
A singular resemblance of the head and fore part of the body to
that of a horse, has given to these fishes the name of “Sea-horses.”
They are abundant between and near the tropics, becoming scarcer
in higher latitudes. Some twenty species are known, some of which
have a wide geographical range, as they are often carried to great
distances with floating objects to which they happen to be attached.
—Acentronura is a genus closely allied to Hippocampus.
SIXTH ORDER—PLECTOGNATHI.
Teleosteous fishes with rough scales, or with ossifications of the
cutis in the form of scutes or spines; skin sometimes entirely naked.
Skeleton incompletely ossified, with the vertebræ in small number.
Gills pectinate; a narrow gill-opening in front of the pectoral fins.
Mouth narrow; the bones of the upper jaw generally firmly united. A
soft dorsal fin, belonging to the caudal portion of the vertebral
column, opposite to the anal; sometimes elements of a spinous
dorsal besides. Ventral fin none, or reduced to spines. Air-bladder
without pneumatic duct.
First Family—Sclerodermi.
Snout somewhat produced; jaws armed with distinct teeth in
small number. Skin with scutes or rough. The elements of a spinous
dorsal and ventral fins generally present.
Marine fishes of moderate or small size, very common in the
tropical zone, but scarcer in higher latitudes. They have been found
in three localities of tertiary strata, viz., at Monte Bolca, where a
species of Ostracion occurs, and in the Schists of Glaris, from which
two genera have been described, Acanthoderma and
Acanthopleurus, closely allied to Balistes and Triacanthus.
Glyptocephalus from the Isle of Sheppey has the skull of a Balistes,
but its body is covered with tubercles arranged in regular series. The
Scleroderms may be divided into three very natural groups:—
A. Triacanthina.—The skin is covered with small, rough, scale-
like scutes. A spinous dorsal fin with from four to six spines. A pair of
strong, movable ventral spines, joined to the pelvic bone.
To this group belong the genera Triacanthodes, Hollardia, and
Triacanthus, represented by five species, of which Triacanthus
brevirostris from the Indian Ocean is the most common.
B. Balistina.—Body compressed, covered with movable scutes or
rough. Spinous dorsal reduced to one, two, or three spines. Ventral
fins reduced to a single pelvic prominence, or entirely absent.
To this group belong the genera Balistes, Monacanthus, and
Anacanthus, the last genus being distinguished by a barbel at the
lower jaw.
Fig. 310.—Balistes vidua.

Balistes, or the “File-fishes” proper, inhabit the tropical and sub-
tropical seas; shoals of young are not rarely met with in mid-ocean.
Some thirty species are known, many attaining a length exceeding
two feet; but the majority are much smaller, and frequently beautifully
and symmetrically marked. Both jaws are armed with eight strong
incisor-like and obliquely truncated teeth, by which these fishes are
enabled to break off pieces of corals on which they feed, or to chisel
a hole into the hard shell of Mollusca, in order to extract the soft
parts. They destroy an immense number of Mollusks, thus becoming
most injurious to the pearl-fisheries. The first of their three dorsal
spines is very strong, roughened in front like a file, and hollowed out
behind to receive the second much smaller spine, which, besides,
has a projection in front, at its base, fitting into a notch of the first.
Thus these two spines can only be raised or depressed
simultaneously, and the first cannot be forced down, unless the
second has been previously depressed. The latter has been
compared to a trigger, hence a second name, “Trigger-fish,” has
been given to these fishes. Some species are armed with a series of
short spines or tubercles on each side of the tail. Two species (B.
maculatus and B. capriscus), common in the Atlantic, sometimes
wander to the British coasts.
The Monacanthus are similarly distributed as the Balistes, and
still more abundant, some fifty species being known. Their dentition
is very similar, but they possess one dorsal spine only, and their
rough scales are so small as to give a velvety appearance to the skin
(Figs. 17 and 18, p. 48). Adult males of some of the species possess
a peculiar armature on each side of the tail, which in females is
much less developed or entirely absent. This armature may consist
either in simple spines arranged in rows, or in the development of
the minute spines of the scales into long stiff bristles, so that the
patch on each side of the tail looks like a brush.
C. Ostraciontina.—The integuments of the body form a hard
continuous carapace, consisting of hexagonal scutes juxtaposed in
mosaic-fashion. A spinous dorsal and ventral fins are absent; but
sometimes indicated by protuberances.
The “Coffer-fishes” (Ostracion) are too well known to require a
lengthened description. Only the snout, the bases of the fins, and the
hind part of the tail are covered with soft skin, so as to admit of free
action of the muscles moving these parts. The mouth is small, the
maxillary and intermaxillary bones coalescent, each jaw being armed
with a single series of small slender teeth. The short dorsal fin is
opposite to the equally short anal. The vertebral column consists of
fourteen vertebræ only, of which the five last are extremely short, the
anterior elongate. Ribs none. The carapaces of some species are
three-ridged, of others four- and five-ridged, of some provided with
long spines. Twenty-two species from tropical and sub-tropical seas
are known.
Second Family—Gymnodontes.
Body more or less shortened. The bones of the upper and lower
jaw are confluent, forming a beak with a trenchant edge, without
teeth, with or without median suture. A soft dorsal, caudal and anal
are developed, approximate. No spinous dorsal. Pectoral fins; no
ventrals.
Marine fishes of moderate or small size from tropical and sub-
tropical seas. A few species live in fresh water. Fossil remains of
Diodon are not scarce at Monte Bolca and Licata; a distinct genus,
Enneodon, has been described from Monte Postale. The
Gymnodonts may be divided into three groups:
A. Triodontina.—Tail rather long, with a separate caudal fin.
Abdomen dilatable into a very large, compressed, pendent sac, the
lower part of which is merely a flap of skin, into which the air does not
penetrate, the sac being capable of being expanded by the very long
pelvic bone. The upper jaw divided by a median suture, the lower
simple.
A single genus and species (Triodon bursarius) from the Indian
Ocean.
B. Tetrodontina.—Tail and caudal fin distinct. Part of the
œsophagus much distensible, and capable of being filled with air. No
pelvic bone.
“Globe-fishes” have a short, thick, cylindrical body, with well
developed fins. It is covered with thick scaleless skin, in which,
however, spines are imbedded of various sizes. The spines are very
small, and but partially distributed over the body in some species,
whilst in others they are very large, and occupy equally every part of
the body. These fishes have the power of inflating their body by filling
their distensible œsophagus with air, and thus assume a more or
less globular form. The skin is, then, stretched to its utmost extent,
and the spines protrude and form a more or less formidable
defensive armour, as in a hedgehog; therefore they are frequently
called “Sea-hedgehogs.” A fish thus blown out turns over and floats
belly upwards, driving before the wind and waves. However, it is
probable that the spines are a protection not only when the fish is on
the surface and able to take in air, but also when it is under water.
Some Diodonts, at any rate, are able to erect the spines about the
head by means of cutaneous muscles; and, perhaps, all fill their
stomach with water instead of air, for the same purpose and with the
same effect. In some Diodonts the spines are fixed, erect, not
movable. The Gymnodonts generally, when taken, produce a sound,
doubtless by the expulsion of air from the œsophagus. Their
vertebral column consists of a small number of vertebræ, from 20 to
29, and their spinal chord is extremely short. All these fishes have a
bad reputation, and they are never eaten; indeed, some of them are
highly poisonous, and have caused long continued illness and death.
Singularly, the poisonous properties of these fishes vary much as
regards intensity, only certain individuals of a species, or individuals
from a certain locality, or caught at a certain time of the year, being
dangerous. Therefore it is probable that they acquire their poisonous
quality from their food, which consists in corals and hard-shelled
Mollusks and Crustaceans. Their sharp beaks, with broad
masticating posterior surface, are admirably adapted for breaking off
branchlets of coral-stocks, and for crushing hard substances.
Fig. 311.—Jaws of Tetrodon.

Fig. 312.—Tetrodon margaritatus.
Tetrodon (including Xenopterus).—Both the upper and lower
jaws are divided into two by a mesial suture.
Extremely numerous in tropical and sub-tropical zones, more
than sixty species being known. In some of the species the dermal
spines are extremely small, and may be absent altogether. Many are
highly ornamented with spots or bands. A few species live in large
rivers—thus T. psittacus from Brazil; T. fahaka, a fish well known to
travellers on the Nile, and likewise abundant in West African rivers;
T. fluviatilis from brackish water and rivers of the East Indies. The
species figured is one of the smallest, about six inches long, and
common in the Indo-Pacific.
Diodon.—Jaws without mesial suture, so that there is only one
undivided dental plate above and one below.
In these fishes, as well as in some closely allied genera, the
dermal spines are much more developed than in the Tetrodonts; in
some the spines are erectile, as in Diodon, Atopomycterus,
Trichodiodon, and Trichocyclus; in others they are stiff and
immovable, as in Chilomycterus and Dicotylichthys. Seventeen
species are known, of which Diodon hystrix is the most common as it
is the largest, growing to a length of two feet. It is spread over the
Tropical Atlantic as well as Indo-Pacific, as is also a smaller, but
almost equally common species, Diodon maculatus.

PDF Event Triggered Sliding Mode Control A New Approach To Control System Design 1St Edition Bijnan Bandyopadhyay Ebook Full Chapter

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PDF Event Triggered Sliding Mode Control A New Approach To Control System Design 1St Edition Bijnan Bandyopadhyay Ebook Full Chapter

Uploaded by

Copyright:

Available Formats

Event Triggered Sliding Mode Control A

New Approach to Control System

Sliding Mode Control Methodology in the Applications of

Process control system fault diagnosis: a Bayesian

Photovoltaic Power System Modeling Design and Control

Process Control: A Practical Approach Myke King

Statistical Process Control: A Pragmatic Approach 1st

Virtual Equivalent System Approach for Stability

Statistical Process Control: A Pragmatic Approach 1st

Gaosu Startup Control Theory and Application Technology

More information about this series at http://www.springer.com/series/13304

ISSN 2198-4182 ISSN 2198-4190 (electronic)

© Springer International Publishing AG, part of Springer Nature 2018

Printed on acid-free paper

Dedicated to my parents and gurus

Mumbai, India Bijnan Bandyopadhyay

2.4 Event-Triggered Sliding Mode Control for Multivariable

5.3.2 Event-Triggering Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

R Set of real numbers

Ti Inter-event time/time interval between two consecutive triggering

© Springer International Publishing AG, part of Springer Nature 2018 1

1.1 Computer-Controlled Systems

Computer-controlled systems, in general, are broadly defined as a control system that

1.1.1 Basic Architecture

The basic architecture of a computer-controlled system is shown in Fig. 1.1. The

Fig. 1.1 Basic architecture

1.1.2 Design Techniques

Many techniques have been used to analyse and design of computer-controlled

using continuous-time or discrete-time model and apply it to the continuous-time

1.1.2.1 Emulation-Based Approach

It may also be seen as continuous-time approach design to sampled-data system. Here,

1.1.2.2 Discrete-Time Approach

1.1.2.3 Hybrid Approach

1.2 Event-Triggered Control

Event-triggering strategy is a control implementation technique that is motivated

Fig. 1.2 Comparison of ψ (t)

(a) Reimann sampling

instant of the corresponding state. However, it is very complicated and difficult to

1.2.1 Preliminary Idea

ẋ = f (x, u), x(0) = x0 ∈ Rn (1.1)

is asymptotically stable. It is assumed that the control is implemented digitally to

a(x) ≤ V (x) ≤ a(x) (1.2)

ti+1 = inf {t > ti : σ x(t) ≤ Le e(t)} . (1.4)

V̇ < −(1 − σ )a(x)

where K∞ a := a ◦ a−1 which is the composition of two functions a and a−1 .

1.2.2 Stability of Event-Triggered Systems

Example 1.1 Consider a scalar nonlinear control system as

u(t) = −x2 (ti ) − kx(ti ), t ∈ [ti , ti+1 ), i ∈ Z≥0 .

where a(r) and γ (r) are given as

ε > 0, the following holds

Using this in the following, it can be written as

1.2.3 Need of Event-Triggered Control

Event-triggered control is one of the aperiodic control implementation strategies in

Fig. 1.3 Response of the

Fig. 1.4 Variation of

Fig. 1.5 Event-triggered

Fig. 1.6 Time evolution of

here sampling instants are determined whenever it is demanded subject to system

1.3 Sliding Mode: An Introduction

To illustrate the concept of sliding mode, consider a controlled nonlinear dynam-

ẋ = f (x, u), x ∈ Rn and u ∈ R. (1.5)

The vector field f : Rn × R → Rn given in (1.5) is Lipschitz with respect to

Similarly, we also define S + = {x ∈ Rn : s(x) > 0} and S − = {x ∈ Rn :

a(x) ≤ V (x) ≤ a(x) (1.2)

ti+1 = inf {t > ti : σ x(t) ≤ Le e(t)} . (1.4)

V̇ < −(1 − σ )a(x)

where K∞ a := a ◦ a−1 which is the composition of two functions a and a−1 .