A study on the dynamic analysis, motion optimization ang control of a 3 joint fish robot

공학박사학위논문

3축 물고기 로봇에 대한 동적 해석,
운동 최적화 및 제어에 관한 연구
A Study on the Dynamic Analysis, Motion
Optimization, and Control of a 3-Joint Fish Robot

울산대학교대학원
기계자동차공학과
VO TUONG QUAN

1


3축 물고기 로봇에 대한 동적 해석,
운동 최적화 및 제어에 관한 연구
A Study on the Dynamic Analysis, Motion
Optimization, and Control of a 3-Joint Fish Robot

지도교수

이병룡

이 논문을 공학박사 학위 논문으로 제출함

2010 년 05 월

울산대학교대학원
기계자동차공학과
VO TUONG QUAN

2


VO TUONG QUAN 의 공학박사학위 논문을 인준함

울산대학교대학원
2010 년 05 월

3


A Study on the Dynamic Analysis, Motion
Optimization, and Control of a 3-Joint Fish Robot

This certifies that the dissertation of
VO TUONG QUAN is approved by

School of Mechanical and Automotive Engineering
University of Ulsan, Ulsan, Korea

4


A Study on the Dynamic Analysis, Motion
Optimization, and Control of a 3-Joint Fish Robot
VO TUONG QUAN

A thesis submitted to the School of Mechanical and Automotive Engineering in
fulfillment of the thesis requirements for the Degree of Doctor of Philosophy in
the Graduate School, University of Ulsan, Ulsan, Korea.

May 2010
5


Acknowledgements
This dissertation is a product of research conducted in the Intelligent Control and
Mechatronics Lab (IML), Mechanical and Automotive Engineering Department, University of
Ulsan, Korea.
First of all, I would like to express my deepest gratitude and my heartfelt thanks to my
advisor, Professor. Byung Ryong Lee, for his mentor-ship, supports and instructions of this
dissertation and also during the time I live and study in the IML.
Second, I would like to thank to all members in the Interlligent Control and
Mechatronics Lab especially Mr. Hyoung Seok Kim, Mr. Cho Hyo Seung, Mr. Heo Nam
Geon for their kindly help during the time I study and carry on this project in the IML.
Third, I would like to send my deeply thanks to all Professors who have read and given
me advises for my PhD dissertation.
Finally, I would like to send my deepest thanks to my big family and especially for my
small family (my wife Hanh Dung and my little son Chinh Minh “Cu Bin”). I am indebted to
my big family, to my wife and my little son for their sweet love and support. All of their moral
supports are unvaluable and lead to my successful today.
One more time, I would like to send the deepest thank to my advisor Professor, to IML,
to Mechanical and Automotive Engineering Department and also to University of Ulsan for
their supports and giving me a good chance to study in Korea.

Best Regards,
Vo Tuong Quan

6i

Abstract
In the past, most of underwater robot researches are focus on ROVs (Remotely Operated
Vehicles), AUVs (Autonomous Underwater Vehicles), UUVs (Unmanned Underwater
Vehicles) or underwater manipulator, and etc. Nowadays, underwater robot research field are
mostly concerned on the biomimetic underwater robot. One of the most favorite robots is fish
robot. Fish robot, which is a new type of underwater biomimetic robot, has attracted great
attention because of its silence in moving, flexible in moving and energy efficiency compared
to the conventional propeller-oriented propulsive mechanism.
Firstly, the purpose of this paper is the study on the mechanical structure of a 3-joint
Carangiform fish robot. The dynamic system of this robot is analyzed in two cases of concept.
The first concept is that the anterior part (head and body) of fish robot is supposed to be a rigid
body and does not move in the dynamic analyzing process. The second concept is that the
anterior part of fish robot is supposed to move or undulate in the dynamic analyzing process.
Besides, the influences of fluid are also discussed in the fish robot’s dynamic. Secondly,
based on the dynamic system of fish robot, its maximum straight velocity is optimized by the
combination of Genetic Algorithm (GA) and Hill Climbing Algorithm (HCA). Then, the
optimization by GA-HCA is also used to find the control parameters that can make fish robot
swim with a smooth motion or smooth gait like a real fish. Lastly, some controllers such as:
Hebbian Neural Network PID (HNNPID), Fuzzy PID (FPID), conventional PID, Sliding
Mode Controller (SMC) and Fuzzy Sliding Mode Controller (FSMC) are designed in direction
control of fish robot. There are two kinds of direction control problem are concerned. They are
the tracking control along straight path and the tracking control for turning motion. In theses
controllers, the strong point or the effectiveness of the HNNPID controller and FSMC are
proved to be stronger than other controllers in direction control of fish robot. The simulation
results are carried out by Matlab program. And, some primitive experiments are implemented
to check the operation of our fish robot.

Keywords: Fish robot, Carangiform, Dynamic, optimal, GA, HCA, HNNPID, FPID, PID,
SMC, FSMC, heading, turning,

ii
7


Table of contents
Acknowledgements

i

Abstract

ii

Table of contents

iii

List of figures

vi

List of tables

xii

Chapter 1 Introduction
1.1 General researches about fish robot in recent years & some
typical types of fish robot.

1

1.1.1 General introduction

1

1.1.2 Analysis of some current researches about Carangiform fish robot.

7

1.2 Fish robot locomotion.

10

1.3 Scope and contribution of this dissertation.

12

1.4 Outline of this dissertation

13

Chapter 2 Dynamic Analysis of a 3-Joint Fish Robot
2.1 The analytical model of 3-joint fish robot.

15

2.2 The dynamic analysis based on first concept (θ 0 = 0) .

18

2.3 The dynamic analysis based on second concept (θ 0 ≠ 0) .

21

2.4 Motion equation of fish robot.

23

Chapter 3 Motion Optimization of a 3-Joint Fish Robot
3.1 The Genetic Algorithm (GA).

25

3.2 The Hill Climbing Algorithm (HCA).

28

3.3 The combination of GA-HCA.

29

3.4 The application of GA-HCA in maximum straight velocity of fish robot.

30

3.4.1 The implementation of GA-HCA.

33

3.4.2 The influence of input torque functions’ parameters on fish robot velocity.

34

iii
8


3.4.2.1 Influence of amplitudes on fish robot velocity.

34

3.4.2.2 Influence of frequencies on fish robot velocity.

36

3.4.2.3 Influence of phase difference on fish robot velocity.

38

3.4.3 Optimal results generated by GA-HCA.

40

3.4.3.1 Different frequencies case.

40

3.4.3.2 Same frequencies case.

43

3.4.4 Experimental Results

45

3.5 The application of GA-HCA in smooth gait of fish robot.
3.5.1 Optimal results generated by GA-HCA and Non-optimal results

50
54

3.5.1.1 Different frequencies

54

3.5.1.1.1 Optimal case

54

3.5.1.1.2 Non-optimal case

58

3.5.1.2 Same frequencies case

61

3.5.1.2.1 Optimal case

61

3.5.1.2.2 Non-optimal case

64

3.5.2 Experimental results

68

3.5.2.1 Different frequencies

69

3.5.2.2 Same frequencies

72

Chapter 4 Heading and Turning Control of the Robot using Intelligent Controls
4.1 The sensory system.

75

4.2 The heading operation and turning operation of fish robot.

76

4.3 The dynamic system used in the direction control problem.

77

4.4 The Disturbances Effect to Fish Robot

77

4.5 The Fuzzy-PID (FPID) Controller

78

4.5.1 Tracking Control along a Straight Path

82

4.5.2 Tracking Control for Turning Motion

83

4.6 The Hebbian Neural Network-PID (HNNPID) Controller

85

4.6.1 Hebbian Neural Network

85

4.6.2 HNNPID Controller Design

86

4.6.3 Tracking Control along a Straight Path

88

iv
9


4.6.4 Tracking Control for Turning Motion

89

4.7 The Conventional PID Controller

90

4.7.1 Tracking Control along a Straight Path

91

4.7.2 Tracking Control for Turning Motion

92

4.8 The Comparison of PID Based Controllers

93

4.8.1 The Comparison in Following a Straight Path

94

4.8.2 The Comparison in Turning Motion

95

Chapter 5 Heading and Turning Control of the Robot Using Sliding Mode
and Fuzzy Sliding Mode Controllers
5.1 The General about Sliding Mode Controller (SMC) and
Fuzzy Sliding Mode Controller (FSMC)

99

5.2 The Sliding Mode Controller (SMC)

100

5.2.1 Tracking Control along a Straight Path

100

5.2.2 Tracking Control for Turning Motion

104

5.3 The Fuzzy Sliding Mode Controller (FSMC)

106

5.3.1 Tracking Control along a Straight Path

109

5.3.2 Tracking Control for Turning Motion

110

5.4 The Comparison between SMC and FSMC

111

5.4.1 The Comparison in Following a Straight Path

112

5.4.2 The Comparison in Turning Motion

113

Chapter 6 Conclusion and Future Works
6.1 Conclusion

115

6.2 Future Works

116

Bibliography

118

Appendix

128

10v


List of figures
Fig. 1.1 Fish robot Charlie I developed at MIT.

2

Fig. 1.2 Fish robot RoboTuna II developed at MIT.

2

Fig. 1.3 Fish robot RoboPike developed at MIT.

3

Fig. 1.4 Fish robot developed at Caltech.

3

Fig. 1.5 Fish robot’s turning mode proposed by Koichi HIRATA et al. [7].

4

Fig. 1.6 Fish robots developed by Koichi HIRATA et al..

4

Fig. 1.7 Fish robot with new actuator developed by [14], [15].

5

Fig. 1.8 Fish robot developed in Essex University, UK.

5

Fig. 1.9 Fish robot proposed by Guo Jenhwa, National Taiwan University.

6

Fig. 1.10 Fishes robot developed by Junzhil et al., and Tianmiao Wang et al..

6

Fig. 1.11 Fish robots (cuttlefish and knifefish) researched by K. H. Low et al. [28].

7

Fig. 1.12 Serial chain of links used in representing the body/fin BCF swimmer [28].

10

Fig. 1.13 Swimming modes of fishes using BCF propulsion. Hatched areas
show the propulsive segment that contributes to thrust generation [27].

11

Fig. 1.14 Swimming modes of fishes using MPF propulsion [27].

12

Fig. 1.15 Fish morphology used to identify fins in different swimming modes [27].

12

Fig. 2.1 Carangiform locomotion style.

15

Fig. 2.2 General analytical model of 3-Joint Crangifrom fish robot.

16

Fig. 2.3 The analytical model of first concept of dynamic analysis.

16

Fig. 2.4 Distribution of forces on Carngiform fish robot.

17

Fig. 2.5 Model of inertial fluid force and lift force in first concept.

18

Fig. 2.6 (a) Relationship between U and U m . (b) The calculation of attack angle α .

19

Fig. 2.7 Model of inertial fluid force and lift force in second concept.

21

Fig. 3.1 The process of GA [66].

27

Fig. 3.2 Hill Climbing graph demonstration [12].

29

Fig. 3.3 General algorithm of the optimizing process.

31

Fig. 3.4 Fitness function of GA and HCA.

32

Fig. 3.5 HCA diagram in maximum straight velocity problem.

33

vi
11

Fig. 3.6 (a) The influence of amplitudes on velocity with f1 = f2 = 0.2Hz, β = 300.
(b) Divergence case.

35

Fig. 3.7 The influence of amplitudes on velocity, f1 = f2 = 0.6Hz, β = 300.

36

Fig. 3.8 (a) The influence of frequencies on velocity with A1 = A2 = 3Nm, β = 300.
(b) Divergence case.

37

Fig. 3.9 The influence of frequencies on velocity with A1 = A2 = 1Nm, β = 300.

38

Fig. 3.10 The influence of phase difference on velocity with A1 = A2 = 1.5Nm,
f1 = f2 = 0.3Hz.

39

Fig. 3.11 The influence of phase difference on velocity with A1 = A2 = 3Nm,
f1 = f2 = 0.7Hz.

39

Fig. 3.12 Simulation of fish robot links oscillation and
their angular velocities using Eq. (3.5)

41

Fig. 3.13 Simulation of fish robot velocity and moving distance
with respect to time by using Eq. (3.5).

42

Fig. 3.14 Simulation of propulsion force and drag force of fish robot by using Eq. (3.5). 42
Fig. 3.15 Simulation of fish robot links oscillation and
their angular velocities using Eq. (3.6)

44

Fig. 3.16 Simulation of fish robot velocity and moving distance
with respect to time by using Eq. (3.6).

44

Fig. 3.17 Simulation of propulsion force and drag force of fish robot by using Eq. (3.6). 45
Fig. 3.18 The internal structure of fish robot.

46

Fig. 3.19 The exterior shape of fish robot.

46

Fig. 3.20 Simulation of fish robot links oscillation and
their angular velocities using Eq. (3.7)

47

Fig. 3.21 The comparison of moving distances of two
optimal cases and one non-optimal case.
Fig. 3.22 The comparison of power consumption in three cases.

48
49

Fig. 3.23 The experiment of fish robot in
different frequencies case (Optimal parameters).

49

Fig. 3.24 The experiment of fish robot in same frequencies case (Optimal parameters). 49
Fig. 3.25 The experiment of fish robot in Non-optimal case (Non-optimal parameters). 49

vii
12


Fig. 3.26 The general algorithm of smooth gait optimization process.

51

Fig. 3.27 Fitness function of GA and HCA in smooth gait problem.

52

Fig. 3.28 HCA diagram in smooth gait optimization.

53

Fig. 3.29 Simulation of fish robot links oscillation and
their angular velocities using Eq. (3.9)

55

Fig. 3.30 The position of the fish robot’s links
during each 1-second sampling period as determined by Eq. (3.9).

56

Fig. 3.31 Fish robot link positions for each sampling time
of 1s by applying Eq. (3.9) and the normalization method.

57

Fig. 3.32 Simulation of fish robot links oscillation and
their angular velocities using Eq. (3.11)

59

Fig. 3.33 The position of the fish robot’s links
during each 1-second sampling period as determined by Eq. (3.11).

60

Fig. 3.34 Fish robot link positions for each sampling time
of 1s by applying Eq. (3.11) and the normalization method.

60

Fig. 3.35 Comparison of the links’ positions in the normalization mode
between the optimal case and the arbitrary case for one cycle.

61

Fig. 3.36 Simulation of fish robot links oscillation and
their angular velocities using Eq. (3.12)

62

Fig. 3.37 The position of the fish robot’s links
during each 1-second sampling period as determined by Eq. (3.12).

63

Fig. 3.38 Fish robot link positions for each sampling time
of 1s by applying Eq. (3.12) and the normalization method.

64

Fig. 3.39 Simulation of fish robot links oscillation and
their angular velocities using Eq. (3.13)

65

Fig. 3.40 The position of the fish robot’s links
during each 1-second sampling period as determined by Eq. (3.13).

66

Fig. 3.41 Fish robot link positions for each sampling time
of 1s by applying Eq. (3.13) and the normalization method.

66

Fig. 3.42 Comparison of the links’ positions in the normalization mode
between the optimal case and the arbitrary case for one cycle.

viii
13

67


Fig. 3.43 The diagram of fish robot experiment on gait motion.

69

Fig. 3.44 Joint motions of experiment and Simulink
simulation for the optimal condition.

70

Fig. 3.45 Joint motions of experiment and Simulink

simulation for a non-optimal condition.

70

Fig. 3.46 Experimental joint motions of optimal and a non-optimal condition.

71

Fig. 3.47 Joint motions of experiment and Simulink
simulation for the optimal condition.

72

Fig. 3.48 Joint motions of experiment and Simulink
simulation for a non-optimal condition.

72

Fig. 3.49 Experimental joint motions of optimal and a non-optimal condition.

73

Fig 4.1 Digital compass sensor – HMR3300

75

Fig 4.2 Yaw angle motion of fish robot.

76

Fig. 4.3 Turning motion of a fish robot in counter
clockwise (CCW) and clockwise (CW) directions.

76

Fig. 4.4 The principle of Fuzzy-PID controller.

79

Fig. 4.5 The membership functions of two inputs of FLC.

80

Fig. 4.6 The membership functions of three outputs of FLC.

80

Fig. 4.7 a. Direction control result by fuzzy-PID controller (desired heading angle = 300).
b. Applied flow disturbance.

82

Fig. 4.8 a. Direction control result by fuzzy-PID controller (desired heading angle = 600).
b. Applied flow disturbance.

83

Fig. 4.9 a. Turning control performance of fuzzy-PID controller (desired turning angle = 300).
b. Applied flow disturbance.

83

Fig. 4.10 a. Turning control performance of fuzzy-FPID controller (desired turning angle=600).
b. Applied flow disturbance.

84

Fig. 4.11 Single neuron.

85

Fig. 4.12 The block diagram of the HNNPID control system.

87
0

Fig. 4.13 a. Direction control result by HNNPID controller (desired heading angle = 30 ).
b. Applied flow disturbance.

88

Fig. 4.14 a. Direction control result by HNNPID controller (desired heading angle = 600).

ix
14


b. Applied flow disturbance.

89

Fig. 4.15 a. Turning control performance of the HNNPID controller (desired turning
angle=300).
b. Applied flow disturbance.

89

Fig. 4.16 a. Turning control performance of the HNNPID controller (desired turning
angle=600).
b. Applied flow disturbance.

90

Fig. 4.17 The principle of a conventional PID controller.

90

Fig. 4.18 a. Direction control result by PID controller (desired heading angle = 300).
b. Applied flow disturbance.

91

Fig. 4.19 a. Direction control result by PID controller (desired heading angle = 600).
b. Applied flow disturbance.

92

Fig. 4.20 a. Turning control performance of the PID controller (desired turning angle=300).
b. Applied flow disturbance.

92

Fig. 4.21 a. Turning control performance of the PID controller (desired turning angle=600).
b. Applied flow disturbance.

93

Fig. 4.22 Comparison of sum square error among
controllers with desired heading angle 300.

94

Fig. 4.23 Comparison of sum square error among
controllers with desired heading angle 600.

95

Fig. 4.24 Comparison of turning time among
controllers with desired turning angles of 300.

96

Fig. 4.25 Comparison of turning time among
controllers with desired turning angles of 600.

96

Fig. 4.26 Comparison of steady state error among
controllers with desired heading angle 300.

97

Fig. 4.27 Comparison of steady state error among
controllers with desired heading angle 600.

97

Fig. 5.1 The principle of SMC.

100

Fig. 5.2 The stability boundary of the SMC.

101

Fig. 5.3 The principle of SMC.

101

x
15


Fig. 5.4 a. Direction control result by SMC controller (desired heading angle = 300).
b. Applied flow disturbance.

104

Fig. 5.5 a. Direction control result by SMC controller (desired heading angle = 600).
b. Applied flow disturbance.

104

Fig. 5.6 a. Turning control performance of the SMC controller (desired turning angle=300).
b. Applied flow disturbance.

105

Fig. 5.7 a. Turning control performance of the SMC controller (desired turning angle=600).
b. Applied flow disturbance.

105

Fig. 5.8 The principle of FSMC controller.

106

Fig. 5.9 The membership functions of two inputs of FLC.

107

Fig. 5.10 The membership functions of one output of FLC.

108
0

Fig. 5.11 a. Direction control result by FSMC controller (desired heading angle = 30 ).
b. Applied flow disturbance.

109

Fig. 5.12 a. Direction control result by FSMC controller (desired heading angle = 600).
b. Applied flow disturbance.

109

Fig. 5.13 a. Turning control performance of the FSMC controller (desired turning angle=300).
b. Applied flow disturbance.

110

Fig. 5.14 a. Turning control performance of the FSMC controller (desired turning angle=600).
b. Applied flow disturbance.

111

Fig. 5.15 Comparison of turning time of SMC
and FSMC with desired turning angles of 300.

113

Fig. 5.16 Comparison of turning time of SMC
and FSMC with desired turning angles of 600.

xi
16

113

List of tables
Table 1.1 Difference of BCF swimming modes [28]

10

Table 3.1 Optimal values of parameters with different frequencies

40

Table 3.2 Optimal values of parameters with same frequencies

43

Table 3.3 Fish robot physical parameters

46

Table 3.4 Non optimal parameters set

47

Table 3.5 Optimal value of parameters in the different frequencies case

54

Table 3.6 Non-optimal value of parameters set in the different frequencies case

58

Table 3.7 Optimal value of parameters in the same-frequencies case

62

Table 3.8 Values of the set of arbitrary parameters in the same-frequencies case

64

Table 4.1 Fuzzy rules of FPID controller

81

Table 5.1 Fuzzy rules of FSMC controller

108

Table 5.2 Sum square error in straight motion

112

Table 5.3 Steady state error after turning motion

114

xii
17


Chapter 1
Introduction
1.1 General Researches about Fish Robot in Recent Years & Some Analysis

about Carangiform Fish Robot
1.1.1 General Introduction
Many researchers’ studies about underwater propulsion depend on the use of
propellers or thrusters or buoyancy to generate the motion for objects in the underwater
environment. Thruster is mostly used in ROVs (Remotely Operated Vehicles) or AUVs
(Autonomous Underwater Vehicles) to create the propulsion force for the robot’s
movement. Yeow Cheng Sun and Chien Chern Cheah [1] used thruster to generate the
propulsion force for their AUV. Besides, David McFarland and Ian Gilhespy et al. [2]
employed the buoyancy control inspired by sperm whales to operate their biomimetic
underwater robot. However, most natural methods for underwater propulsion employ the
changes of an object’s body shape to create movement. This changing shape generates the
propulsion force necessary to make the object move forward effectively. The Carangiform
fish is one example of changing its body shape to move itself in an underwater environment.
This kind of Carangiform fish is mostly concerned in developing a biomimetic underwater
robot called fish robot.
One of the most famous initial projects about fish robot is the RoboTuna project. This
project began in 1993 at the Department of Ocean Engineering of MIT by Professor
Michael S. Triantafyllou as the team leader. This project’s goal was to develop a better
propulsion system for autonomous underwater vehicles. Then, in 1995, the fish robot
named Charlie I was build by designer David Barrett for his PhD thesis. Tests were
performed on with the robot to access the swimming capabilities of the fish. RoboTuna
(Charlie I) was designed to mimic the shape and motion of a small tuna. It is controlled by
six powerful servomotors (rated at 2 horsepower each) and has force sensors placed at
various locations along the path of the controlling tendons. The carriage is towed through

1


the water and swimming efficiency is calculated from the forces on the tendons [3]. This
fish robot is introduced in Fig.1.1

Fig. 1.1 Fish robot Charlie I developed at MIT.
Five years later, a new generation of Charlie I called RoboTuna II was born. The new
generation of fish robot was designed by David Beal and Michael Sachinis. This fish robor
is founded on a cable-pulley system, like the original robot, but with several significant
modifications [3].

Fig. 1.2 Fish robot RoboTuna II developed at MIT.

2


Fig. 1.3 Fish robot RoboPike developed at MIT.
Then, another version of fish robot developed at MIT is Robo-Pike. This fish robot
was developed by John Muir Kumph. The characteristics of fish robot are that the authors
want to express the demonstration of a very quick turning and fast acceleration from a stop
[4].
RoboTuna also triggered a renewed interest in biomimetic locomotion in term of fish
robot design, locomotion analysis and control methods. Some other Carangiform fish
robots developed at Caltech were designed using fewer degrees of freedom [5], [6].

Fig. 1.4 Fish robot developed at Caltech.
Besides, huge types of fish robot were also researched and designed by Koichi Hirata
et al. [7]. They proposed the turning three turning modes for fish robot. These three turning
modes are now used as a standard to control the turning motion for most of Carangiform

3


fish robot. Koichi Hirata also designed and developed some other type of fish robot and

researched about their behavior or their motion’s characteristics [8].

Fig. 1.5 Fish robot’s turning mode proposed by Koichi HIRATA et al. [7].
Fig. 1.6 below introduces some fish robots developed by Koichi HIRATA and his
colleagues.

Fig. 1.6 Fish robots developed by Koichi HIRATA et al..
Some other researches about fish robot are also carried out and developed in Korea.
Tuong Quan Vo, Hyoung Seok Kim et al. [9], [10], [11], [12] researched about the
dynamic structure, optimization method of a 3-joint Carangiform fish robot. Some other
researchers use new and small material as the actuator to generate the motion for their fish
robot. For example, Teddy Wiguna, Seok Heo et al. [13] used the piezoelectric actuator to
design for their fish robot. Quang Sang Nguyen, Seok Heo et al. [14], [15] used the
Piezoceramic actuator to drive their small fish robot. The applications of fish robot are
quite large and especially in the works that related to underwater environment such as

4


underwater environment exploration, underwater animals’ life research, and etc. Daejung
Shin, Seung You Na et al. [16], [17] used their fish robot to monitor the water pollution
problem. And, the control method which related to networking system was also applied to
control for their robot.

Fig. 1.7 Fish robot with new actuator developed by [14], [15].
Furthermore, some typical types of fish robot are also researched and developed in
United Kingdom. Jingdon Liu, Huosheng Hu et al. [18], [19] built the simulation
environment to optimize the control parameters for their fish robot. Also, the fish robot’s
behaviors were also paid much attention to be focus on by using the reinforcement learning
algorithm.

Fig. 1.8 Fish robot developed in Essex University, UK.
Jenhwa Guo developed a measurement strategy for a biomimetic autonomous
underwater vehicle (BAUV) to reduce positioning uncertainties while the BAUV was
controlled to reach a target efficiently. The BAUV had the role as a target tracker. It can
swing the pectoral fins when searching for a target and oscillates its tail fin to move
forward [20], [21].

5


Fig. 1.9 Fish robot proposed by Guo Jenhwa, National Taiwan University.
In China, Junzhi Yu et al. [22] also proposed many researched problem about fish
robot such as: fish robot’s dynamic modeling, simplified propulsive model of biomimetic
robot fish and its realization [23], [24]. Tianmiao Wang, Jianhong Liang et al. [25] focused
on the stabilization based design method to reduce the yawing, rolling and pitching of the
fish robot and to improve the efficiency of tail fin propulsion. Guangmin Wang, Lincheng
Shen et al. [26] investigated about the long-based undulatory fin of an Amiiform fish “G.
niloticus”. They paid attention to the kinematic modeling and dynamic analysis about this
type of fish robot [26].

Fig. 1.10 Fishes robot developed by Junzhil et al., and Tianmiao Wang et al..
In addition, in Singapore, fish robot also researched and developed by K. H. Low,
Chunlin Zhou et al. [27], [28]. They focused on the analyzing the motion of parallel
structure mechanism fish robot and parallel structure mechanism fish robot, the motion
planning and control of fish robot [27], [28].

6

Fig. 1.11 Fish robots (cuttlefish and knifefish) researched by K. H. Low et al. [28].

1.1.2 Analysis of Some Current Researches about Carangiform Fish Robot
J. Edward Colgate and Kevin M. Lynch [29] made a review of the artificial swimmer
design taken from fish robot physiology and the control problems that must be solved by a
robot fish. Lauder G.V. and Drucker, E.G [30] made the thorough surveys and analyses of
motion mechanisms of fish fin in order to develop such underwater robot systems. M. J.
Lighthill [31] surveyed the hydromechanics of aquatic animal propulsion because of the
motion mechanisms of many underwater animals were evolved through many generations
to adapt to the harsh underwater environment. Iman Borazjani and Fotis Sotiropoulos [32]
introduced a numerical investigation of hydrodynamics of carangiform swimming in the
transitional and flow regimes. They investigated the influences of relative magnitude of
viscous and inertial forces, i.e. the Reynol number (Re), the tail-beat frequency, i.e. the
Strouhal number (St) on the hydrodynamics of carangiform locomotion. Jehnwa Guo [21]
developed a measurement strategy for a bionic AUV (BAUV) to reduce positioning
uncertainties while the BAUV was controlled to reach a target efficientcy. K. H. Low et al.
[28] discussed a mechanism design and gait of biomimetic fish robot in two major forms:
planar serial chain mechanism and parallel mechanism. Besides, gait functions for two
forms of biomimetic fish robots were also discussed [28]. Thomas D. Jorgensen and
Charlotte C. F. Norlund [33] researched about the mechanical design and develop a robot
herrings. Two different types of mechanical accumulators, which are the bucked spine
accumulator and the bent beam accumulator, were implemented and tested as a spine on the
robot [33].

7


Some other researches have carried out parameter analysis and optimization. Qin Yan,
Zhen Han et al. [34] investigated the influences of the characteristic parameters, including
frequency, amplitude, wave length, phase difference and coefficient of wave amplitude

envelope on forward velocity of fish robot by experiments. Junzhi Yu and Long Wang [23]
calculated the optimal link ratio of a 4-link fish robot by using computer simulations, and
they compared the moving speed of two types of models. One model used in this
comparison utilized the optimal link ratio, and the other model was considered without the
optimal link ratio. Another solution for optimization was proposed by KeehongSeo,
Richard Murray et al. [35]. They used numerical optimization software (NLPP – Non
Linear Path Planning – Tool Box for Matlab) to optimize the control parameters for a
simplified planar model of a Carangiform fish robot.
Besides, Chao Zhou, Min Tan et al. [36] focused on the design and implementation
problem of fish robot. Junzhi Yu et al. [22] also carried on the dynamic modeling of fish
robot. Seunghee Lee and Jounghyun et al. [37] researched about the kinematic and dynamic
parameters of a mackerel-mimicking robot and proposed the optimal control method for
this fish robot in which energy efficient in trajectory tracking problem. Junzhil Yu and
Shuo Wang et al. [24] made an improved approach to design a robot fish based on a
simplified kinematics model that related frequency to speed and joint angle bias to turn,
where geometric reduction was employed and complex hydrodynamic analysis was
avoided. However, all the studies discussed above are based on quite simplified kinematic
models, simplified dynamic models or on experimental approaches.
On the other hand, some other authors paid attention to the optimal design of fish
robot’s mechanical structure. Yu JunZhil et al. [38, 39] used the numerical method to find
the optimal link-length-ratio for their fish robot. Dimitrios Tzeranis et al. [40] listed some
desired standards or requirements for their underwater autonomous robot fish and then they
designed their fish robot to adapt to these standards. Yeffry Handoko et al. [41] researched
about the design of fish robot’s head and body shape and also fish robot’s tail fin
mechanism in order to get the optimal thrust speed. The fish robot’s caudal fin geometry,

8