1-s2.0-S1672652914600986-main

Corresponding authors: Jindong Liu, Young-Sun Ryuh E-mail: [email protected], [email protected]

Journal of Bionic Engineering 12 (2015) 3746

A School of Robotic Fish for Mariculture Monitoring in the Sea Coast

Young-Sun Ryuh1, Gi-Hun Yang1, Jindong Liu2, Huosheng Hu3 1. Korea Institute of Industrial Technology, Ansan, 426-910, Korea

2. The Hamlyn Centre, Imperial College London, London, SW7 2AZ, United Kingdom 3. School of Computer Science and Electronic Engineering, University of Essex, Colchester, CO4 3SQ, United Kingdom

Abstract This paper presents a multi-agent robotic fish system used for mariculture monitoring. Autonomous robotic fish is designed

to swim underwater to collect marine information such as water temperature and pollution level. Each robotic fish has 5 degrees of freedom for controlling its depth and speed by mimicking a sea carp. Its bionic body design enables it to have high swimming efficiency and less disturbance to the surrounding sea lives. Several onboard sensors are equipped for autonomous 3D naviga-tion tasks such as path planning, obstacle avoidance and depth maintenance. A robotic buoy floating on the water surface is deployed as a control hub to communicate with individual robots, which in turn form a multi-agent system to monitor and cover a large scale sea coast cooperatively. Both laboratory experiments and field testing have been conducted to verify the feasibilityand performance of the proposed multi-agent system.

Keywords: robotic fish, mariculture monitoring, multi-agent system, pollution detection, sea coast Copyright 2015, Jilin University. Published by Elsevier Limited and Science Press. All rights reserved. doi: 10.1016/S1672-6529(14)60098-6

1 Introduction

Mariculture is a specialized branch of aquaculture involving the cultivation of marine organisms for food and other products in the open ocean, an enclosed sec-tion of the ocean, or in tanks, ponds or reserves that are filled with seawater. An example is the farming of ma-rine fish, including finfish, shellfish and seaweed in saltwater ponds. Fig. 1 demonstrates the concept of mariculture proposed by East-sea Fisheries Research Institute.

Fig. 1 Concept of mariculture (East-sea Fisheries Research Institute).

Monitoring mariculture around sea coast is an im-portant and challenging task due to the complex geog-raphy of sea bed, ocean currents and human activities. Difficulties in information exchange, management, and energy supply could happen as well. The classical way of using floating sensor devices has the problems of uncontrollable locations and sparse sensing points. Re-cently, multi-robot systems started to be applied to solve these problems. A typical multi-robot system consists of a number of self-controlled robots which can collect data independently within a dedicated area. A central control unit can combine the data from individual robots to cover a large area which otherwise cannot be accom-plished by a single robot.

Multi-robot systems have been deployed in many real world applications such as manufacturing, as well as search and rescue operations[1]. It has been proved that multi-robot systems can reinforce and extend the robot ability in space, time and functionality under uncertain and dynamic environments[2]. However, mariculture environments are extremely difficult for any robots to conduct efficient and consistent jobs since the robots have to be well waterproofed and strong enough against

Journal of Bionic Engineering (2015) Vol.12 No.1 38

huge water pressure. In addition, the distribution of jobs among multiple robots to accomplish should take the function of each robot into a careful consideration[3].

Up to now, unmanned underwater robots have been widely used for the purpose of the marine exploration, surveillance, and environmental monitoring. However, their driving efficiencies are limited by rotary-propellers, i.e. below 70%, and they are very noisy and not friendly to sea animals. Moreover, their turning radii are rela-tively big and evading speeds are considerably slow. These weakness are the major obstacles for traditional underwater vehicles to perform mariculture monitor-ing[4,5]. In contrast to rotary propellers, robot fish applies an undulatory tail movement for propulsion and ma-neuvering[6]. The undulatory tail motion is less noisy, more efficient, and more maneuverable (turning radius is 1/10 of Body Length (BL) contrast to 10 times of body length of a conventional ship[7]). So robotic fish is a good solution for this problem, which can be effectively de-ployed in the mariculture monitoring operations and other similar applications[810].

Many robotic fishes have been developed world-wide for the last two decades. Streitlien et al. compared the performance between a tuna-like robot tail fin and a traditional propeller in 1994, and found that the fins are more efficient method to proceed in the water[11]. The fins of robotic fish could repulse the water more than the square of propellers blade, as a result its energy effi-ciency reached 87% (the propeller is 70%)[12]. Jalbert et al. applied Shape Memory Alloy (SMA) to produce a robot lamprey[13] which aimed to provide mine coun-termeasures. Guo et al. developed a micro robotic fish using ICPF actuators[14]. Yu et al. developed a biomi-metic robotic fish and a motion control algorithm for a group of robotic fish using overhead vision system[15]. Liu et al. produced three energy-efficient multi-joint fish robots to mimic carp fish, which were daily operated at London Aquarium to the public for over 2 years[16]. Kato et al. produced a robot Blackbass[17] in order to study the propulsion characteristics of pectoral fins. Recently, Chen et al. developed an ionic polymer-metal composite caudal fin for robotic fish[18]. Klein et al. controlled a group of robotic fish via a underwater wireless net-work[19]. Polverino et al. deployed a robotic fish to lead live fish away from oil spills[20]. Liao et al. proposed a wire-driven flapping propulsor which can have a maximal cruise speed of 0.288 BLs1 [21].

In this paper, we propose a complete new solution of a multi-robot system for monitoring mariculture. A versatile robotic fish, Ichthus V5.5, has been designed for high efficient swimming and high maneuverability. A number of sensors are equipped for monitoring water temperature, electric conductivity and pH (hydrogen ion concentration) value of the water. The developed robotic fish has the abilities to control its posture, navigate autonomously in a 3D space and to monitor the water quality. A robotic buoy floating on the water surface is deployed as a control hub to communicate with indi-vidual robotic fish, which in turn form a multi-agent system to monitor and cover a large scale sea coast co-operatively. Petri net theory is applied for the multi-agent control.

The rest of the paper is organized as follows. Sec-tion 2 introduces the design of the robotic fish, including its kinematics and onboard sensors. In section 3, a multi-agent control system is proposed for coordinating a school of robotic fish in mariculture monitoring and management. Some experimental results are presented in section 4 to show the feasibility and performance of the proposed system. Finally, a brief conclusion and future work are given in section 5.

2 Robotic fish design and its kinematics

2.1 Bio-inspired underwater robot Ichthus V5.5 The fish robot deployed in this research is named as

Ichthus-V5.5 (Ichthus in short), which is developed at Bio-inspired Robot Engineering Laboratory in Korea Institute of Technology (KITECH). Its shape looks like a trout fish, with a dimension of 500 mm 146 mm 170 mm (length width height). Its weight is about 4.7 kg (Including buoyancy adjustment weights).

As shown in Fig. 2, Ichthus has three joints in the body and the tail to mimic the swimming movements of a real carp fish. It can be characterized with its sufficient underwater DOFs and durability resulted from its high swimming speed and quick turning. It has a low center of gravity, with underwater specific gravity close to 1, in order to maintain a stable posture in the water. It can realize flexible up and down movements using its side fins. The platform is in a modular structure, i.e, a drive part, a sensor part, a communication part, a control part. Therefore, it is easy to assemble, repair and replacement.

On each joint, the servo motors are connected to the body frame of the robot. Therefore, by using the servo

Ryuh et al.: A School of Robotic Fish for Mariculture Monitoring in the Sea Coast 39

motors, we are able to control the frequency and am-plitude of the robot. Table 1 and Fig. 3 show the specifications of Ichthus and actual implementation of the robotic fish. 2.2 Sensors

Ichthus has embedded several kinds of sensors for autonomous navigation and quality detection of water. Sensors used for autonomous navigation include infra-red (IR) sensors, ultrasound range sensors, Global Posi-tioning System (GPS) sensor, water pressure sensor, Inertial Navigation System (INS) and Ultra Short Base-line (USBL). Basically, the GPS sensor is used for lo-calization when the robot is on the surface of water. In case that the robotic fish is located in the underwater environment, GPS data is not accessible. Therefore the robot needs to use water pressure sensor for depth de-tection and INS for localization. To compensate the error of INS, USBL location data is transmitted through an ultrasound modem. With these features, the robotic fish can move autonomously in the underwater envi-ronment as well as on the surface of water. The water quality detection sensors are used for measuring tem-perature, Electric Conductivity (EC), pH (hydrogen ion concentration) of the water. 2.3 Kinematic model of Ichthus

To model a real fish swimming motion to our multi-joint robotic fish Ichthus, we followed the method in Ref. [16]. We assume that all the swimming activities of carangiform fish can be divided into basic fish swimming motion patterns, namely swim patterns, according to their swimming morphologies such as cruise straight, cruise in turning, and C-shaped turn, etc. The undulation motion of fish body in each swim pattern can be represented by an independent function.

From biologists observations of the fish motion, the body undulation function can be obtained. However, such function cannot be used directly for the robotic fish tail control. It is because the body functions are the full body motion while the robotic fish needs a tail motion function relative to the fish head. In order to deduce a tail motion function from a body motion function, two coordinate systems are defined[16], (i) a world coordi-nate system R (x, y) in which the origin is fixed at the connection point B between the fish head and tail, and x-axis is aligned to the swimming direction; (ii) a

Fig. 2 Multi buoy and multi-agent robots.

Table 1 Specifications of Ichthus Ver. 5.5

Ichthus Ver. 5.5

Body size 500 mm 146 mm 170 mm (Length Width Height)

Weight 4.7 kg DOF 5 (Tail 3, Side fin 2)

Battery 14.8 V Li-Io Battery

Fig. 3 Actual implementation of robotic fish Ichthus V5.5.

system Rh (xh, yh) in which it has the same origin point B as R, but the x-axis is aligned to the fish head rather than the swimming direction. The body motion function can been described as fB (x, t) in the R coordinate. The fish tail motion is denoted as fT (xh, t) in the Rh coordinate. From now on, fB (x, t) and fT (xh, t) are simplified as fB (x, t) and fT (x, t).

According to Ref. [16], the deduction from fB (x, t) to fT (x, t) can be expressed in Eq. (1). Given a general form of fB(x, t) = (c1x +c2x2) sin(kx + t) of cruise straight swim pattern, where c1, c2, k, are parame-ters[16], the tail motion function is fT (x, t) = (c1x + c2x ) sin(kx + t) c1xsin (t). Please see the details of de-duction and the swimming function for other motions in Ref. [16].

| 0( , ) ' ( , ) 0( , ) .0 0

B B xT

f x t xf x t xf x t

x= =

Journal of Bionic Engineering (2015) Vol.12 No.1 40

The reproducing fish swim pattern on a robotic fish is to adopt multi-joint tail to generate the same tail mo-tion as a real fish does. Due to the difficulty to deduce the kinematic function of each joint from fT (x, t), the digital approximation method in Ref. [16] is applied here. In this way, fish swimming motion can be mimicked by approximating a number of tail postures with rigid linkages between joints. The approximation results are a serial of relative angles between the joints to its prior. A lookup table is built to contain all angles. So the servo motor in each joint can be controlled by setting the target to the value in the lookup table.

3 Multi-robot control system

3.1 System architecture In this paper, multiple small and low-cost fish ro-

bots with distributed miniaturized sensors and a short distance communication module are deployed. To coor-dinate these robotic fish for the mariculture monitoring task, we propose a multi-robot control system that con-sists of three layers, the GPS, a number of buoy robots and multiple robotic fishes, as shown in Fig. 4. Each buoy robot on surface can manage multiple clients (fish robots) via the acoustic communication. It can receive the measurements from clients and distribute mission assignments to the clients. At the same time, it can communicate with an off-shore control center for mis-sion control and data collection via a real-time satellite communication. A buoy robot is equipped with a GPS receiver and therefore it is able to locate its own position by itself. Fig. 5 shows one of the buoy robots used in our field test in sea. A buoy robot is also called a docking station. It monitors the status of each robotic fish by sending pe-riodic polling signals and assigning proper jobs to them as the main processing unit does. Since radio frequency signals cannot function well in the water due to the dif-fusion of radio wave, it is impossible for a large amount of data transfer, such as images or moving pictures, to be communicated underwater among fish robots.

Instead, fish robots can store collected information on their internal memory up to a certain point, and then these data will be collected by buoy robots and later transfer them to the control center via higher bandwidth communication interfaces such as USB or Ethernet. The buoy robots are also used for charging, cleaning and storing robotic fish. Fig. 6 illustrates a control console on

Global GPS system

Buoy robot

Fish robot 3

Fish robot 2

Fish robot 1

Fig. 4 Control of multi-agent robot (clients) with buoy robot (supervisor).

Fig. 5 An example of buoy robot, aka docking station. One fish robot (white) in the bottom-left corner is returning to the buoy robot.

Fig. 6 A control console for a buoy robot.

a buoy robot.

Within such structure, buoy robots are the servers and fish robots are the clients. The geographical location of each robotic fish can be measured by combining its-

Ryuh et al.: A School of Robotic Fish for Mariculture Monitoring in the Sea Coast 41

relative position data from inertial sensors and the ab-solute position data from buoy robots. Then, underwater exploration can be conducted. Fig. 7 illustrates the functional relationship between buoy robots and robotic fishes. In this paper, an improved Petri net modeling is deployed on controlling multiple robots (buoy robots and fish robots).

3.2 Petri net model for mariculture management

The theory of Petri Nets (PN)[22] provides a more convenient means than that of the general graph theory in expressing asynchronous distributed systems. A PN is a formal method that enables information gathering in dynamic environments or systems. The simplest form of PN model consists of conditions and events. Commands for operating robots can be expressed as mutually de-pendent events and conditions. PNs are well known for modeling the concurrent behaviors of distributed sys-tems because multiple tokens may be present anywhere in the net.

As shown in Fig. 8, the basic components in PN are place, arc, transition and token. Transitions are the events that may occur in the system such as robot returns to charging station (T19 in Fig. 10). Places are the con-ditions which can lead to an event, e.g. true if recall a robot back to charging station (P16 in Fig. 10). Places and transitions are directly connected to each other via arcs (Fig. 8). The existence of token in a place indicates

Fig. 7 Functional relationship between buoy robots and robotic fish.

Token

Arc Arc

Place Transition Place

Fig. 8 The basic components in Petri net model and their rela-tionship.

that the condition of the corresponding place is true, i.e. marking on a place distinguishes whether the given condition is true or false in PN. When all input places of a transition are marked, the given transition is to be fired. In other words, the token on each input place is removed and is moved to each output place when an event occurs.

The generalized form of such a simple condition/ event (C/E) PN is the place/transition (P/T) PN. Instead of simple conditions such as true or false, the number of available spaces is used within an input place or buffer. The case with capacity > 0 for each place should be defined even though it is not applied to this case. In addition, the weight on an arc interconnecting a transi-tion and a place indicates adding or subtracting multiple tokens after firing the transition. The colored PN is gen-erally used in modeling a large scale system in which various different layers of the target are then expressed in different colors.

Here, transitions change their colors as defined by a color function. Such a colored PN can be transformed into a C/E PN. Fig. 9 and Fig. 10 show the PN models for a buoy robot and a robotic fish respectively. The places and transitions are listed below these figures. The two PNs are connected together via the places of P10, P15 and P16 and the transitions of T4, T5 and T7. The de-tailed explanation can be seen in Ref. [23].

4 Experimental results

4.1 Experiment for multiple robotic fish control Mariculture management described in this section

is a laboratory implementation via PN modeling. The multiple small low-cost fish robots are linked with a supervisor robot to achieve stable control and manage-ment. The supervisor robot is run on the main computer, and assigns a mission to fish robot, e.g. to collect in-formation within a complex underwater environment. Each of the three fish robots consists of a communication module, a motion control module and a sensor module. The supervisor robot consists of a known map for navigation, a vision system for locating clients, and a communication module for issuing commands or col-lecting information. Fig. 11 shows the laboratory setup of such a multiple robot control framework.

The multi-robot system is implemented in two phases. First, the main system or the supervisor locates each fish robot, assigns it a mission and delivers the information collected by fish robots to the main control

Journal of Bionic Engineering (2015) Vol.12 No.1 42

Place 0: Stan by

Place 1: Initial global localization

Place 3: Start packet communicationPlace 2: Check dock

Place 4: Check packetPlace 5: Back trackingPlace 6: Stop mission

Transition 0: System on

Transition 1: Start global localization

Transition 3: Start monitoring and data acquisitionTransition 2: Dock open & release

Transition 4: Normal packet communicationTransition 5: Packet receiving errorTransition 6: Recall to charging stationTransition 7: Dock close

P0 P1 P2

P3P4

P5

P6

T0 T1

T2

T3

T4

T5

T6

T7

T6

Start navigation

Recall to charging station

Mission stop

Fig. 9 PN model of buoy robot.

Buo

y ro

bot

Fish

robo

t

Fig. 10 PN model of fish robot.

Ryuh et al.: A School of Robotic Fish for Mariculture Monitoring in the Sea Coast 43

center. The supervisor maintains a communication with all fish robots via an ultrasonic device in order to locate each of them and to collect information from them, and provides a charging station for the client robots in a low battery state. In addition, the supervisor collects infor-mation, prepares for a possible loss of fish robots, cor-rects the positioning information of each fish robot, and assigns the corresponding missions through packet transmission and reception.

As shown in Fig. 5, the buoy robot requests the information for global localization from the GPS module while a token moves from P0 to P1 (refer to Fig. 9) at the initial stage. The GPS module generates the event T1 and secures its own positional information, and then releases the fish robot and transmits the global positional information to each client. The buoy robot can also check the battery level of each client and navigate close to the client if necessary. Each fish robot returns to the station after transmitting a comeback signal to the buoy robot when it completes its own mission, needs to be charged, or is in emergency/danger situation.

Figs. 9 and 10 show the control sequence of dock-ing station and fish robot respectively in the multi-robot system. The experimental implement is mainly focused on exposing the feasibility of controlling multi-robots based on PN modeling in a simplified environment. The main computer is equipped with a vision camera for locating fish robots, i.e. playing a role of a stationary buoy robot. Fig. 12 depicts the snapshot of the main program when a fish robot starts to move (P10 in Fig. 10) for the given mission with the buoy robot in its normal states P4 and T4 in Fig. 9. Fig. 13 demonstrates a process for a fish robot to return to its initial location according to the command issued by the controller. 4.2 Navigation in the field

In order to test our system in a field, we imple-mented a remote control and monitoring program of the robotic fish, i.e. WaveWork 2. It can set the waypoint navigation of the robotic fish through Google Earth. Fig. 14b shows the appointed waypoints and their posi-tions in Google Earth. The functions of WaveWork 2 also include commanding the robot to start or stop, set-ting the maximal swimming velocity and controlling the azimuth heading error.

The docking station was linked with robotic fish via USBL, Radio Frequency (RF), ultrasonic modem

Fig. 11 Laboratory setup for multiple robot control experiments.

Fig. 12 Main control program: Start navigation.

Fig. 13 (a) Normal condition; (b) packet error; (c) recall to charging station; (d) return to C/S [turn & swim].

and GPS module. Tritech MicroNAV USBL positioning system was applied to localize the robotic fish with an

Journal of Bionic Engineering (2015) Vol.12 No.1 44

accuracy of 0.2 m. The USBL is able to track fish ro-bots within a range of 500 m in horizontal and 150 m in vertical. To increase the accuracy of depth measuring, a pressure sensor was equipped on the robotic fish, which improved the accuracy by 0.01 m. A wireless charging system was deployed for the fish robots in the laboratory environment. The wired charging from a docking station was used to the robotic fish in the sea field experiments. The docking station conveys path information to robotic fish by RF and the robot performs a command. Also, the docking station displays a navigation and environment information obtained from the fish robot on the map. The robotic fish can be controlled directly using either RF or ultrasonic modem.

Fig. 14 shows the navigation result of one fish robot with 4 waypoints in a lake near Suin-Ro, South Korea in July 2012. The start position was at longitude 126.893503 and latitude 37.322359. The grid in Fig. 14a is 4 m. The docking station commanded the fish robot to start the navigation, i.e. heading to waypoint 1 first. It was frequently checking the fish robot conditions fol-lowing the PN models of both the buoy robot and the fish robot shown in Figs. 9 and 10. When the fish robot reached the waypoint 1, the docking station commanded the fish robot to head to the waypoint 2 and so on. The total distance between the waypoints was 110 m. The real navigation path of the fish robot was 152 m. The average RMS error of its navigation is 1.4 m.

Fig. 15 shows the field test results of three fish ro-bots with 8 waypoints in the Han River of South Korea in August 2013. The start position was at longitude 127.312878 and latitude 37.541446. In the figure, four events show how the docking station monitors the fish robot and corrects its trajectory. Referring to Fig. 10, event a is a normal navigation state right after the start position. It corresponds to P13 in the PN model in Fig. 10. In event b, the buoy robot found that the fish robot drifted away from the appointed waypoints 1 and 2 (P14 in Fig. 10) so that it stopped the fish robot and recalled it (event c) to the next nearest waypoint, i.e. the waypoint 2. In event d, the fish robot corrected its trajectory and started to navigate to the waypoint 2. The average RMS error of navigation is 1.5 m.

It should be noticed that the waypoint 4 is a singular point, i.e. the fish robot is appointed to return to the waypoint 3 after visiting the waypoint 4 and then con-tinues its journey to the waypoint 5. The results clearly

3

1

4

24 m

(a)

(b)

Start point

Fig. 14 Field test result of one robotic fish with 4 waypoints. (a) Appointed waypoints (boxed numbers) and real navigation path (purple points); (b) the corresponding navigation path shown on the Google Map.

Fig. 15 Field test results of three robotic fish with 8 waypoints. The appointed waypoints are indicated by circles labeled with boxed numbers and real navigation paths are illustrated as color dots. Four red circles give an example of path correction events in the Petri net model.

Ryuh et al.: A School of Robotic Fish for Mariculture Monitoring in the Sea Coast 45

show that our PN model can control multiple fish robots to follow appointed waypoints within the real field. When the fish robot went off course, the PN model can help the docking station to detect the situation and cor-rect it on time.

5 Conclusion

In this paper, a novel multi-agent robotic fish sys-tem is proposed for mariculture monitoring, in which multiple autonomous robotic fish are deployed to swim underwater and collect marine information such as water temperature and pollution level. The mathematical model of the robotic fish with a tail fin of 3-DOF link mechanism has been derived. The developed robotic fish has many kinds of embedded sensors to navigate autonomously in the real environment such as a lake or a river, as well as to measure temperature, electric con-ductivity and pH of water so that the quality of water can be monitored in real time.

Since the change in the underwater environment is very irregular, underwater robots must be able to handle these changes robustly. The proposed exploration method is based on the absolute location obtained by vision and no GPS information is available. The control of multiple underwater robots is experimented based on 2D data. PN models are deployed to provide visible effects and easy modeling. The problems associated with the loss of control parameters, logical conflict or over-lapping avoidance are completely solved. The experi-mental results demonstrate relatively stable collabora-tion and timely gathering exploration data within a specific area.

In the future, we will further investigate the effi-cient collective control and collaboration mechanism of multiple robotic fish in various and complex operations with each fish robot equipped with inertial sensors and acoustic communication devices.

References

[1] Hormann A. A petri net based control architecture for a multi-robot system. Proceedings of IEEE International Symposium on Intelligent Control, Albany, NY, US, 1989, 493498.

[2] Agah A, Doyle B, Drees M, Froehlich C, Kuok K. Robot soccer for the study of learning and coordination issues in multi-agent systems. Proceedings of IEEE International Conference on Systems, Man, and Cybernetics, San Diego,

US, 1998, 4, 35203525. [3] Hiraishi K. A petri-net-based model for the mathematical

analysis of multiagent systems. IEICE Transactions on Fundamentals of Electronics, Communications and Com-puter Sciences, 2001, 84, 28292837.

[4] Hirata K, Takimoto T, Tamura K. Study on turning per-formance of a fish robot. First International Symposium on Aqua Bio-Mechanisms, Mitaka, Tokyo, Japan, 2000, 287292.

[5] Cho K, Park H, Kim S, Yang H, Park Y. Development of robot mimicking propulsion of a fish. Korean Society of Mechanical Engineers, 2007, 4045.

[6] Oyekan J, Gu D, Hu H. Visual imaging of invisible haz-ardous substances using bacterial inspiration. IEEE Trans-actions on Systems, Man and Cybernetics: Systems, 2013, 43, 11051115.

[7] Ryuh Y. Development of swimming mechanism and algo-rithm of fish-like underwater robot. Korea Robotics Society, 2009, 4348.

[8] Oyekan J, Hu H. A novel bio-controller for localising a pollution source in medium peclet environments. Journal of Bionic Engineering, 2010, 7, 345353.

[9] Oyeken J, Lu B, Hu H. A creative computing approach to 3D robotic simulator for water pollution monitoring. Interna-tional Journal of Creative Computing, 2013, 1, 92119.

[10] Wang S, Chen L, Hu H, Xue Z, Pan W. Underwater local-isation and environmental mapping using wireless robots. Journal of Wireless Personal Communications, 2013, 70, 11471170.

[11] Streitlien K, Triantafyllou G, Triantafyllou M. Efficient foil propulsion through vortex control. AIAA Journal, 1996, 34, 23152319.

[12] Kim Y. Robotic fish, the prince of ocean. Dong-A Science, 2005, 8, 5459.

[13] Jalbert J, Kashin S, Ayers J. A biologically-based undulatory lamprey-like auv. Proceedings of the Autonomous Vehicles in Mine Countermeasures Symposium. Montery, US, 1995, 3952.

[14] Guo S, Fukuda T, Kato N, Oguro K. Development of un-derwater micro-robot using ICPF actuator. Proceedings of IEEE International Conference on Robotics and Automation, Leuven, Belgium, 1998, 2, 18291834.

[15] Yu J, Tan M, Wang S, Chen E. Development of a biomimetic robotic fish and its control algorithm. IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics, 2004, 34, 17981810.

[16] Liu J, Hu H. Biological inspiration: From carangiform fish to multi-joint robotic fish. Journal of Bionic Engineering, 2010,

Journal of Bionic Engineering (2015) Vol.12 No.1 46

7, 3548. [17] Kato N. Control performance in the horizontal plane of a fish

robot with mechanical pectoral fins. IEEE Journal of Oce-anic Engineering, 2000, 25, 121129.

[18] Chen Z, Shatara S, Tan X. Modeling of biomimetic robotic fish propelled by an ionic polymer-metal composite caudal fin. IEEE/ASME Transactions on Mechatronics, 2010, 15, 448459.

[19] Klein D, Gupta V, Morgansen K. Coordinated control of robotic fish using an underwater wireless network. Ma-zumder S K (Ed.), Wireless Networking Based Control, Springer, New York, US, 2011, 323339.

[20] Polverino G, Abaid N, Kopman V, Macr S, Porfiri M.

Zebrafish response to robotic fish: Preference experiments on isolated individuals and small shoals. Bioinspiration & Biomimetics, 2012, 7, 036019.

[21] Liao B, Li Z, Du R. Robot tadpole with a novel biomimetic wire-driven propulsor. IEEE International Conference on Robotics and Biomimetics (ROBIO), Guangzhou, China, 2012, 557562.

[22] Reisig W. Petri nets and algebraic specifications. Theoreti-cal Computer Science, 1991, 80, 134.

[23] Ryuh Y, Moon J. Multi-agent control and implementation of bio-inspired underwater robots for mariculture monitoring and control. IEEE International Conference on Robotics and Biomimetics (ROBIO), Guangzhou, China, 2012, 777783.