An AUV Based Experimental SystemFor The Underwater Technology Education

Keisuke WatanabeAssociate Professor

Department of naval architecture and ocean engineeringSchool of marine science and technology, TOKAI UNIVERSITY

3-20-1 Orido Shimizu-kuShizuoka-shi, Shizuoka, 424-8610, JAPAN

keisukegscc.u-tokai.ac.jp

Abstract - In this paper, I introduce a series of AUV relatedbasin experiment platforms now we are developing as a set ofunderwater technology educational tool. Existing AUV platformsare very expensive for general students or schools to purchase as __-_an educational tool so the popularity of AUV technology amongthe general people is not still so high in Japan. On the other hand,land based robotic contest is very popular in Japan and its Aleducational effect seems good for the wide range of students. I gSafti rthink it is important for us to develop a set of handy and low cost 00*0i

tools oriented to the underwater technology education and AUVrelated platforms which can use easily in a small basin will be oneof the best choices. In this paper I will introduce our concept ofthe AUV based experimental systems as an underwatertechnology educational tool. The system is composed of two types "Me ~ SUUM IMof AUV test-beds, precise ultrasound ranging system, LBLF Conceptualsketchofdeapseaconstructoncoordinate detection system and 2FSK ultrasound communicationsystem. The hardware and software design is introduced in thispaper. The AUV test beds are designed considering a convenienceof basin experiment using a 7inches-diameter acrylic sphere or17inches-diameter glass sphere with low-cost embedded systemwith DOS/V PC104 based subsystems. We constructed a set of PIhh'W .

iMHz carrier frequency ultrasound ranging system and l00kHzLBL coordinate detection system. A 2FSK ultrasoundcommunication system for basin experiment was tested. In ourcase, the basin is 3m(W)c3m(D) 6m(L). So we designed both AUTVtestbeds as are enough for the handling and mobility in our basinexperiment. The production cost must be suppressed while it hasenough equipment for our purpose. The hull is transparent so wecan implement optical communication using sets of infrared LEDand phototransistor without using any penetrate connectors forthe data communication between the inside equipment and Fig.2 Conceptual sketch of distributed monitoringoutside equipment like an ultrasound module or a land PC. Theunderwater connector is relatively expensive compared to other Production Veseslequipment so we used it for power supply only and reduced thefabrication cost. On smaller one, the thruster is composed of a4.5W coreless motor with planetary gear and a four-bladespropeller. We designed and fabricated the magnet coupling forwatertight seal of the thruster to reduce the production cost. Thesmaller type?s total weight in air is about 3kgf. As the total size is

T

small (0.4mILI_0.4mIBI_0.2mIHI) we can handle it easily and 4 Transportation Robot Agentmany robots can swim in our basin. The hull is designed to bearthe pressure at least l0m water depth. We selected Lithium ion Manganesebattery for power supply and the total capacity is 450Wh. By Oft Nodules

power consumption calculation we may use it about 4 hours with A"

one full charge of the batteries. The main PC is PC104 module(Advantech PCM3350) of which MPU is NS-Geode 300MHz and64MB RAM. I implemented the RT-Linux based program systemin 1GB compact flash card from which the system boots. Itcontrols the total system through PIC based subsystems. The RS-232 serial communication is used for gaining the ranged distance Fig.3 Conceptual sketch of collection of deepwater resourcesfrom the ultrasound controller of which MPU is PIC16F876. ThisRS-232 signal is also transmitted from the inside MPU to the

1-4244-01 38-0/06/$20.00 ©2006 IEEE 1

outside PC by infrared data communication. So we canmonitor the information inside the pressure hull without opening

...~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~............it. The program is implemented through Ethernet. The frequency LAND PCof the sensing and actuation is at most lHz this time, the timing iscontrolled by the RT-task mentioned later. An electric compassTCM2 is used to detect the heading angle and roll, pitch. I wasafraid that the magnet compass is affected by the thruster orinner electric current, but at present, it doesn't matter. Eachdevice like electric compass has its own PIC and all serialcommunication line is multi-dropped using 12C through the Ul-ut Soundcommunication interface PIC. The 6 underwater connector areattached for ultrasound transducers, thrusters, power supply andethernet. The magnet switch is used for power on and off of 2 DC-DC, which corresponds to 12V for thruster and 5V for computers.The system generates maximum 4 PWM signals for the propeller Aspeed control with motor driver ICs. As we mounted a wirelessCCD camera inside, we can see the underwater image as far as I B

the robot is near water surface. We have just finished the basicfunction test in the basin by free swimming. We have notimplemented modern controller, we will try path tracking Fig.4 Conceptual view of basin experiment setupexperiment in the next step..

basket. If the transportation basket is full then the basketI. INTRODUCTION releases its weight to go up to the surface where the cellection

vessel is waiting for baskets. As the underwater robots areThe underwater multi-agent technology has great potential staying for long period around the seabed, the power supply

for the various activities in the near future ocean development station is needed. The suggested method before now for theor investigation c,2,3,4]. Fig.o 1 illustrates an imaginary picture development of deepsea manganese nodules consists of a veryof deep ocean construction as one example of those activities, large vessel with massive riser pipe and an underwater suctionBecause we can utilize buoyancy force actively to reduce the unit to absorb the mineral resources from the bottom to theself-weight of the structural members in the water and the vessel[8]. The design of the riser system is complicatedexternal force is relatively small compared with near free because of its flexiblity so the total cost of the system issurface in the deep water, there is a new possibility of expensive. The multi agent method may be one of theunderwater structure to be developed if only we can develop improved plans because of its general versatility and systemthe method of automated underwater construction using many flexibility.autonomous underwater robots[5,6]. Considering about the AUV control algorithm designFig.2 shows the schematic view of underwater monitoring procedures, it is important for the designer to synthesize thesystem now we are planning to develop as another possibility balance between autonomy and cooperation in the AUVof underwater multi-agent. We are aiming to realize the character or intelligence. To determine which characteristic iscontinuous monitoring system in time and space where many more important depends on the mission. For example, thecompact AUV groups distributed in target coastal sea area priority is on the autonomy side for the Fig.2 case because thefrom the surface to the bottom. The target monitoring area is interactive situation is limited for communication betweensliced planar hierarchically toward the depth as monitoring each AUV. On the other hand, in the Fig. 1 case, the priority islayers, and a group of AUV arranged on each layer tracks the on the cooperation side because the interaction is not only ongiven trajectory and gathers various environmental information the communication but also dynamical coupling exists in thiscontinuously communicating each other in the layer or between case. And also, in Fig.3 case, we need to implement a kind ofother layers. There are also autonomous moving buoys on the emergent characteristic to optimize the total behavior of thesurface that have the role not only of the observation around team. To extract and solve such problems in the developmentsurface environment but also of the communication interfaces of AUV multi-agent technologies, we need to enlargebetween underwater groups and facilities on land, satellites etc. educational opportunities for students of dealing with variousThe target area also marked off one lot from another and each basic AUV technologies by their own hands. So we shoulddivision has an AUV group communicating other groups to build up a series of basin experiment system which we canorganize information networks from land base through free experience various problems arising from the real system tosurface to the bottom of the sea. We call this system as verify the result of simulated algorithms. The conceptualPADOMS abbreviating Portable and Autonomous Distributed sketch of the experimental system is shown in Fig.4. TheOcean Monitoring System[7]. It will improve the defects of the essential parts of the experimental system are AUV platforms,present ocean monitoring method, that is, the monitoring ultrasound ranging system and ultrasound data communicationdepends on the spotted water sampling so that continuous in- system. The AUV platforms consist of a floating commucationsitu monitoring is difficult. interface between underwater robots and land base PC, and

Fig.3 shows a concept of development of deepsea mineral underwater tiny AUV test beds. In this paper I will introduceresources like manganese nodules by underwater multi agent two types of AUV test-bed platforms and ultrasound systemsrobots. Each distributed robot has manipulators to grab the now we are developing. Although the underwater multi-agentnodules and they gather the nodules into atransportation system will include not only the AUVs but also other non-

vehicle type robotic agents, in this research we restricted ourexperimental task using AUV only. More complicated task thatneeds non-vehicle type underwater robot is the next step of thisresearch.

11. ULTRASOUND RANGING AND LBL SYSTEM _

We are building a 1Mhz ultrasound ranging system that is _

detection. And also we are building a set of 40kHz ultrasoundpinger/transponders to calculate the underwater coordinate byFi5Ulrsudanngm ueLBL with reference to the basin fixed coordinate system[7].Although it depends on the acitivity range of the multi agents,we will need both long distance ranging system and short |1800 = R1aDist.Elz] McsrdEmz] Elr[l]m25distance ranging system in the application of underwater multi 1 600agent system. For example, when each agent works in open 1400 20water it uses long rang system but if it ought to move into 1200 * 15some complicated space like coral reef, cave or underwater 1000 .10artificial structure around, it swithes the ranging system to 800more precise mode. The switching algorithm is important and 600 m5we want to realize the situation in the basin experiment that's 400 0

200_=why we are now developing a precise ranging system using 0 -51MHz carrier frequency. 0 500 1000 1500

We designed and fabricated the ranging circuit as an Real Distance[mm]independent module because of the maintainance andextensibility. The ultrasound module has its own pressure tightFi6Cabrtoreutf1Mzanngmdl

the module. The phototransistor and LED is used for thecommunication between the module and inner main PC which ll i1activates the ranging. Fig.6 shows the calibration result. The ll 11 Ei_.,.. f; 'calibration is executed like Fig.5 we measured the distance i Xusing the reflected ultrasound by the basin wall. Pink dots 1indicate real distance and blue dots indicate the measured | z lr4

reflected carrier is decayed. However, when the distance is less Ithan tin, the errors are within 5mm, which is enough for our I* __7747711 M3m[W] X 6m[L] X 3m[D] basin. A PIG in the main hull is used |*_for the interface between the main PC in the hull and the_ranging module.

Fig.7 shows an experiment setup of 100kHz 2-dimensionalLBL systemand anexperiment resul. Th,lain oo Fig.7 Two dimensional LBL experiment setup and a result

transmits 100kHz carrier and at first the transponder A calls ..

back 1lOmsec after it receives the transmitted carrier signal dere ietvt oaodtemlipt aig i. hwfrom the robot. The transponder B calls back after 20msec the test condition ofthisFSK development.

whenit ecevesthecarierand he obo cacultesits The communication words which we are planning now iS likecoordinate using measured distances. The right part of the as follows- abmtxlOy20z30.......................... \nFig.ure shows the detected coordinate. Now we are building amxo2zo \.. . . . , , . ,, The first character means the communication starter's ID innondiretuidional100khz ultranspondersnt ctethe bt's .this case the AUV iS "a". And the next character is thecoordinate in 3 dimension. destination AUV's ID in this case "b". A space iS used to

separate the meaning block. The next block is an actionIII. ULTRASOUND COMMUNICATION SYSTEM command or a message from the starter to the destination robot.

In this case "mt" means "move to directed position". EachUltrasound communication system.is essential for our AUV has a database of command words and this received

basi1n experiment. We are bu lding a binary FSK serial data command blci. .smcsedtomm iatoidon stem.e tllision dsire for muti-orwagen ........command list.The next block means the data or following wordcommunication that the ultrasound oscillators are non-directive, wis Uth reous can dmormes

LBL ~~~ ~ ~ ~ ~ ~ ~ ~ ~ ~~hc dependreernc the theubasinnfixe coordinat systm7e

But in the narrow basin the multi path fading occurs so in this m detais M nasinstepwofltheeexperimentnwe usedathe oscgiator which handsh e r nt n thei 2examl fr ollow the courdiate

dstanep raninsysepetwem inetheapplication ofcunerate multi .

some~~~ ~ ~ ~ ~ ~ ~ ~ ~~exe1etcoplcae spacelikelcoralreef,wcave oroordinate10

(10,20,30) where the robot "b" ordered to go by the robot "a".The end of the communication is "\n".

In this binary FSK system, the signal 0 corresponds to thecarrier 35kHz and the signal 1 corresponds to the carrier 38kHz.We selected each carrier frequency closely so as to use onetransducer and receiver. A character data is translated to 8 bitsbinary code in the PIG program and integer data is translated to Transduce16 bits binary code. Fig.9 shows the received binary signals. for~ eeved charActerThe yellow line is the raw signal received. The received binary communii At on.signals are reprocessed by the PIC and reconstructed into thecharacters or integers. The blue line shows the demodulated Fig.8 Binary FSK check resultsignal from the carrier 35kHz into 0 signal. The pink lineshows the demodulated signal from the carrier 38kHz into Isignal.

The reconstructed received data is shown in Fig.8. In thisexperiment we transmitted "abxl0" as 40bits binary code. Re666 4d zigraAlthough the basin in this experiment is a kind of storage tank bfso it is very small as shown in Fig.9, we can communicate withno multi path fading. As it is inevitable for underwater robotsthat the communication rate is very low so we must develop sii

the efficient communication protocol and well balancedautonomy/cooperation algorithm using these real system.

Although there are many problems in the development ofthis communication system, the first problem to be solved is _how to share the position information of each agent that is, ........................

how to make and share the mapping information of each agent,which changes continuously and unpredictably. It affects theway of a generation or making algorithm of the communication Fig.9 Received binary signal ofFSKwords. Especially, if the agents must work together in thecooperative mission like underwater construction, we easilyimagine the situation that an agent needs help by its neighborsbut how can it know who is the nearest or nearer one? So we

u V1must develop underwater position information sharingmethodology in the development of underwater multi agenttechnology as well as the development of efficientcommunication protocol and procedures.

IV. A TINY AUV TEST BED SYSTEM

For the development of underwater robotic multi agentthecnology, we need many (at least 3) AUV or non-AUV type Fig 10 Inner configuration oftiny test bedtest beds. So the testbed must be as low cost as possible andalso as small as possible. In our case, the basin is

whic istaapearnouhfrte hnlnand mobilityckviein our basin. i.11BteisadTM ops ipa

hasFenoug.13showseq nth forsou pu..rp The presrTe hull is

compospaedt sof two 7ainches imente acryica semmui-sphere which epniecmaedt te qipen ow sdioaret cofnnected byDand0rn seootal.sigtor sihowsthuing ner oe upyol adrdcdtdfbrcto ot h

configuraticonno f othedevelope tinyAUV testbed. Fig 11

showhstenobattery modueshandlnLCd doislaty.iThe displayed

thprienttalapeaane anodusytemncheckusviewi ouprebasedwin.Fgl1eatresadTM2cmastipaFig13ou showeup ntheosyste diaramse ofe it.ssThe hull is

tranospaent sof twe ca imples imenteoptyica commuicatione using epniecmaedtctehqipen ow sdio

inside equipment and outside equipment like an ultrasound geare and a four-blades propeller. We designed and fabricatedmodule or aland PC. The underwater connector is relatively the magnet coupling for watertight seal of the thruster to

reduce the production cost as shown in Fig.11. The total weight

datain his aseis te TM2 cmpas ouput Fig12 how

in air is about 3kgf. As the total size is small (0.4m[L] XO.4m[B] X O.2m[H]) we can handle it easily and many robotscan swim in our basin. The hull is designed to bear the pressureat least 1 Om water depth. We selected Lithium ion battery forpower supply and the total capacity is 45OWh. By powerconsumption calculation we may use it about 4 hours with onefull charge of the batteries. The main PC is PC104 module(Advantech PCM3350) of which MPU is NS-Geode 300MHzand 64MB RAM as shown in Fig.13. 1 implemented the RT-Linux based program system in 1GB compact flash card fromwhich the system boots. It controls the total system throughPIC based subsystems. The RS-232 serial communication isused for gaining the ranged distance from the ultrasoundcontroller of which MPU is PIC16F876. This RS-232 signal isalso transmitted from the inside MPU to the outside PC byinfrared data communication. So we can monitor theinformation inside the pressure hull without opening it. Theprogram is implemented through ethernet. The frequency ofthe sensing and actuation is at most 1Hz this time, the timing iscontrolled by the RT-task mentioned later. An electric compassTCM2 is used to detect the heading angle and roll, pitch. I wasafraid that the magnet compass is affected by the thruster orinner electric current, but at present, it doesn't matter. Each I gdevice like electric compass has its own PIC and all serialcommunication line is multi-dropped using 12C through the __:_communication interface PIC. The 6 underwater connector areattached for ultrasound transducers, thrusters, power supplyand ethernet. The magnet switch is used for power on and offof 2 DC-DC, which corresponds to 12V for thruster and 5Vfor computers. The system generates maximum 4 PWM signalsfor the propeller speed control with motor driver ICs. As we hmounted a wireless CCD camera inside, we can see theunderwater image as far as the robot is near water surface. Wehave just finished the basic function test in the basin by freeswimming. We have not implemented modern controller, wewill try path tracking experiment in the next step. Fig.13 System diagram of tiny test bed

V. A LARGER AUV TEST BED SYSTEM from the view point of the brittleness of the glass, wide deadspace due to its shape. And the open/close operation is really

Wentionedmadve.tw otyp risdesiofAU tor Tetone platfor ise troublesome because the crude rubber tape is used for watertight sealing. So we are now planning to develop the pressurenot only in the basin but also in the calm sea. We decided to hull for our convenience in the next step of this research.

develop the platform as reduced in size and weight as possible FigiS shows the appearance of the large AUV.because, on the field experiment, we experienced the handling A thruster has a 70W motor with tacho-generator for theof the AUV is very hard if the AUV is massive. From the local feedback control of the thruster revolution speed. Weaspect of cost reduction in both fabrication and operation, we prepared 2 propellers for basin experiment and calm sea fieldaimed that we can handle the platform by two men at shore and experiment whose diameters are 9cm, 15cm respectively. Thewe can put it into the water or out from the water using a 15cm diameter propeller's bollard maximum thrust is aboutgeneral fishing boat with no extra facilities. So we selected 2kgf. Each motor has a motor driver which generates PWMBENTHOS 17inch pressure tight sphere glass which is widelyused for underwater measurement instrumentation. Fig.14 power signal with tacho-generator local feedback. The thrustshows the hardware system of the large AUV. The total weight sinls ffr D/a ard to thendrive. AnrAuV may hve4

0.9m[]X.5m[]. Tis~ thrusters for planar and vertical motion control, but this timein air is about 45kgf, 0.7m[L] s09[]X05[]T isszein'airsabout45kgf7mL 0.9mB] X05mH. This sizewe equipped only 2 thrusters for horizontal motion direction.is the limit that we can handle it by two men in land and put it The total battery power inside is 700Wh in this case and weinto or pull it up from our basin by hand. Regardless this size

w can put 4 more batteries so the maximum power capacity iswe will need small winch when we make an exDeriment in the , -,- -1100OWh in this system. The calculated maximum speed withreal. Although the glass pressure hull is not so expensive and 15cm propeller is about 0.8kt.widely used for ocean research, the 6500m depth spec is too W eetdteP14cmue ytmadtemihigh for our purpose and it is not necessarily convenient for the CPU is NSGeode300MHz with 64M RAM and a 20G HDD.non- professional users like students to handle it, The PCMCIA card for wireless LAN and A/D, D/A cards are

equipped. The wireless LAN is used for the remote login _access from an external computer for program download or _______sensor data acquisition. One of the important functions of thelarge AUV is the interface function between the land basedcomputer and the underwater multi-agents so the wireless LANismpotan asthe communication device. The attitude sensorUSGOE30

is TCM2 which senses the heading, pitch, roll angles and the Isensing data is collected by RS232(COM1). The power.consumption of PC system is less than 25W, and we estimateeach 4 thrustersm- averaged power consumption is about 30W

t wbf Mt m

without disturbance like current, so that we can continue one iexperiment about 4 to 6 hours. It may be enough at this stage, I I

S ~~~ ~ ~ ~ ~ ~~k 'Pg Am.l Pofw wtg M&,iftthrM 11

but we need further good ideas to realize longer navigation 11I____W______||________A_t_C_____time.

The ultrasound ranging/LBL module is separated from themain pressure hull to avoid PWM noise generated by the motor Fig. 14 Hardware system configuration of larger AUVdrivers inside. Because the glass pressure hull is transparent,we used a set of an infrared LED and a phototransistor for theinfrared serial data communication between main PC in themain hull and PICs for ultrasound ranging module outside.Generally speaking, the underwater electrical connector cost isvery high, so we designed to limit the use of the connector onlyfor the power transmission. We reduced the use of underwaterconnector by using the infrared data communication. Once thewatertight seal is equipped we don't want to open the hulleasily so the main power switch is controlled externally usingphoto sensors because the magnet switch could not workbecause the glass wall was too thick.

VI. SOFTWARE SYSTEM

I selected RT-Linux V3.lpre2 as the operating system ofthe both types because it is free and is possible to get hard real- Fig. 15 Total appearance of larger AUVtime operation. The VxWorks is widely used as the operatingsystem of many AUV platforms. But in our experiment RT-Linux is enough. Each operations such as the thruster drive or USER PROGRAMRS232 for ultrasound ranging control is executed by the START/STOPperiodic thread whose period and priority can be varied xHzindependently arranging each operation's necessity. The lower TASK HANDLERlevel time consuming tasks such as ultrasound ranging orsensor data acquisition are executed separately from the main Kernel Modulesthread and processed by PIC programs so the correspondingthread is a kind of a watchdog of the PIC to synchronize the riontYbwhole threads within the desired sampling time. Because thesethreads are implemented as the kernel module it is not easy for Control forcea user to control the threads externally, so we used a user Threadprocess to activate or stop the kernel module through RT-FIFOas commonly used. The acquired data is also transmitted Ultrasoundthrough RT-FIFO and the user process handles all time Commun Cconsumptive file accesses such as acquired data recording tothe compact flash card. The command and data communication | Ultrsounbetween the main PC and each subsystem PICs mainly depend ..l...Ranging COM2on RS-232(COM1, COM2) in this system. This situation is notso effecttive not only in the operation but also its fabrication or LL Threadprogramming, so we are now planning to develop a PIOcommunication using pararell port and PC104 bus. The RT- I < RTFIFOThreadLinux based software development system is very useful for__students to learn about hard real time task and threads basedprogramming. Fig.16 RT-Linux software configuration

VII. SUMMARY

Underwater multi-agent technology is expected to broadenthe chance of underwater activities such as underwaterconstruction, resource development or ocean environmentmonitoring. We must experience many situations and problemsto be solved for the development of its elemental technologiesby using real systems as well as our computer simulations. Andalso we should make effort to broaden the popularity ofunderwater technology. Educational platforms that we are ableto have students see and touch easily are very effective for thatpurpose. In this paper I introduced our concept of theexperiment in the near future and the hardware/software designof our tiny AUV testbed and ultrasound ..................

ranging/communication system for that experiment. Two typesof testbeds are designed for the basin experiment with RT-linux based compact flash boot embedded system with NS-Geode PClO4 CPU card and PIG based multi CPU subsystems.We also made an 1MHz ultrasound ranging system which canmeasure the distance to an obstacle within the 2cm accuracy.The 100kHz LBL system was fabricated and checked, and100kHz non directional LBL system is now in progress. Wemade a binary FSK ultrasound data communication systemwhich we succeeded to transfer 3 characters and 1 integersuccessive data. The undergraduate students who engaged in

making parts of this experimental system all enjoyed to buildNup and create them. And I believe their academic skills were

improved extensively through this development. As the basiccomponent of the basin experiment was developed, we mustassemble the total system to execute some mission for groupbehavior experiment in our next step of this research.

Fig. 17 Fabrication scene by students

ACKNOWLEDGMENT [5] H. Suzuki, T. Qi, K. Watanabe, "Learning Tracking Contr

The author gratefully acknowledge the contribution of ollers under Unknown Disturbances for the Installation ofundergraduate students Hiromasa Miura, Kenichi Ishise in my Rigid and Flexible Structures", SNAJ, Journal ofMarinlaboratory, Saori Nakada, Masako Fukazawa, Ayano Ishihara e Science and Technology, Vol .4 No.4, pp.187-199, 1999and Makiko Fujita in Tokai Junior College. This work was [6] K. Watanabe, H. Suzuki, K. Yoshida, "Active Control forsupported by the Grant-in-Aid for Young Scientists (B) Installation of Underwater Structures", SNAJ, Journal of16760654. Marine Science and Technology, Vo 1.3 No.3, pp.15 1-160,

1998REFERENCES [7] K. Watanabe, K. Sugiyama, A. Nakamura, "A Concept of

Distributed Compact AUV Groups for the Continuous M[1] J. Guo, H.Y. Wei, F.C. Chiu, SW. Chen, "A MaximumE onitoring of Coastal Sea Environment", Proc. Oceans Tec

ntropy Method for Multi-AUV Grouping", Proc. Oceans hno-ocean 04, pp.511-516, 2004Techno-ocean 04, pp.R325s36, 2004 [8] N. Hatta, "Hydrodynamic Examinations and Problems on

[2] P. Didao, E. Batlle, D. Ribas, M. Carreras, "NEPTUNE: Lifting Characteristics of Manganese Nodules by Air-LifA HIL Simulator for Multiple UUVs", Proc. Oceans Tech t Pump System", Energy Resources, Vol.50, No.1, pp.50-no-ocean 04, pp.524-5J31, 2004 56, 1996

[3] A. Okamoto, J.J. Feeley, D.B. Edwards, R.W Wall, Rob [9] J. T. Mian, T. Ishimatsu, and Y. Nagashima, "Compact Uust Control of a Platoon of Underwater Autonomous Vehi nderwater Vehicle with Acoustic Link for Communicatiocles", Proc. Oceans Techno-ocean 04pp.505-510, 2004 n and Positioning,", Proc. ofthe 2000 International Symp

[4] C.A. Reeder, D.L. Odell, A. Okamoto, M.J. Anderson, D. osium on Underwater Technology, pp. 237-241, 2000.B. Edwards, "Two-Hydrophone Heading and Range Sens [10] S. Yu, "Multi-agent based AUV System", Journal of thor Applied to Formation-Flying forAUVs", Proc. Oceans

Techno-ocea 04, pp.5 7-523, 200 e Robotics Society ofJapan, pp. 718-722, 2004

7