research and development of matsya 4.0, autonomous

Research and Development of Matsya 4.0, Autonomous

Underwater Vehicle

Nilesh Kulkarni, Mohit Chachada, Hardik Godara, Sant Kumar, Kunal Tyagi, Tushar

Sharma, Anshuman Kumar, Sachin Garg, Devyesh Tandon, Varun Mittal, Kavya

Kavi, Sandeep Dhakad, Rizwan Ahmed, Nithin Murali, Varad Pingale, Akash

Kamble, Dheeraj Kotagiri, Mishal Asif, Jay Prakrash, Arpit Dangi,

Abhinav Kumar, Somesh Kulkarni

Faculty Advisors: Prof. Leena Vachhani, Prof. Hemendra Arya, Prof. V. Kartik

Indian Institute of Technology, Bombay

Abstract— Matsya, stands for a fishin Sanskrit, is a series of AutonomousUnderwater Vehicles (AUVs) being de-veloped at Indian Institute of Technol-ogy(IIT) Bombay with the aim of deliv-ering a research platform in the field ofunderwater robotics and to promote au-tonomous systems. AUV-IITB is a fo-rum for students and faculty to pursuetheir interests in underwater robotics.AUV-IITB team has developed four ve-hicles each one more advanced as wellas more capable than it’s predecessor.Major architectural changes have beenmade to the subsystems by designingthem from the perspective to handletasks in real time. Although, the latestvehicle, Matsya 4.0 has its design philos-ophy similar to Matsya 3.0, it has ma-jorly improved upon the previous ver-sions in terms of weight optimization,reliability, endurance, speed, aestheticsand cognition.

1 IntroductionMatsya 4.0 is an AUV developed by a multi-disciplinary student-faculty group at IIT Bom-bay to facilitate research and development inUnderwater Robotics as well as to partici-pate in the International Robosub Competi-tion. With integration of a robust AcousticLocalization System and improved control sys-tem, this year’s vehicle is capable of performingmajority of the tasks and addressing the chal-lenges defined by the competition.AUV-IITB is a group of 25 students fromdifferent specializations having a strong mo-tivation to explore the field of UnderwaterRobotics. It has three divisions namely Me-chanical, Electronics, and Software. Matsya

Figure 1: Matsya 4.0

4.0 has seen a year-long development cycle withmajority of components, underwater connec-tors, software stack and electronics boards de-signed in-house by the team members.

2 MechanicalMatsya’s mechanical subsystem consists ofthree specialized sub-divisions namely: actu-ators, frame and hull. Every sub-division un-dergoes predefined design flow (Figure 2) to de-liver reliable and optimized components.

Figure 2: Design Flow

2.1 HullHull is a pressure chamber meant to providea waterproof enclosure to place electronics andother water sensitive components of Matsya.To incorporate various needs like accessibility

AUV-IITB, Indian Institute of Technology Bombay 1

to inner component(s), field of sensing etc.,Matsya houses four types of hulls namely: bat-tery hull, camera hull, main hull and sensorhull. The hull sub-division also designs custommade connectors, penetrators and underwaterswitches.

Battery Hull This year Matsya has two bat-tery hulls (Figure 4(a)) of same design. The de-sign objectives for the battery hull include easeof opening & closing while ensuring robust wa-terproofing. The hull is designed to allow barehand operation and have a life of 600 cycles ofopening & closing.

Camera Hull Because of modular design phi-losophy, Matsya 4.0 reuses same hull for cam-era as Matsya 3.0 (Figure 3(a)). Accuratemounting and enough field of view for cam-eras are some of the primary objectives thatthe hull must meet. Similar to battery hulls,the camera hulls are also designed to allow wa-terproofing operation and opening and closingto be as intuitive as handling a water-bottle.

Sensor Hull Doppler Velocity Log (DVL) hulland Inertial Measurement Unit (IMU) hull fallunder this category. For early detection of wa-ter leaks, these hulls have been chosen to allowmaximum visual from their inner parts. Ac-curate mounting and electromagnetic shieldingare some of the other important design aspects.

Main Hull As the name suggests, the mainhull houses almost all the electronics of Mat-sya. It is not only an amalgamation of all theabove design objectives but also includes fea-tures like layered electronic stacking and goodheat dissipation qualities. The main hull com-prises of four parts: base-plate, lateral walls,flange (Figure 3(c)) and end-cap (Figure 3(b)).To reduce cost, side walls are bent from a sheetof aluminum to form two C-shaped structureswhich are then welded together and thus aremade up of more ductile aluminum alloy. Thebase-plate is made up of a more rigid aluminumalloy since the main hull also provides struc-tural strength to Matsya. The flange is de-signed to avoid accumulation of water on thetop surface and to ensure that no water seepsinside during disassembly. An acrylic end-cap

at the top provides a transparent interface forvisual detection of water seepage as well as elec-tronic displays and indicators. For ensuringrugged waterproofing, the hull has undergonevacuum impregnation - an industrial porositysealing technique.

(a) Camera hull & as-sembly

(b) End cap

(c) Flange

Figure 3

Connectors The team has fabricated in-house dry mateable six pin underwater connec-tors (Figure 4(b)) to route connections amonghulls, thrusters and sensors. The plug and playnature of these connectors makes it convenientto integrate additional components in Matsyaand eases the phase of disassembly and replace-ment. The manufacturing cost for these con-nectors is around $5 which is nearly twentytimes cheaper as compared to the alternativeoption available in market.

(a) (b)

Figure 4: (a)Design of End Cap for BatteryPod; (b)Connectors


Penetrators Penetrators are designed toroute cables out of hulls for direct connectionto the sensors for which simple plug and playconnectors cannot be used as they cause dataloss and impedance matching problems.

Underwater Switches This year’s Matsyahas four switches of same design. The designobjective for these switches include reliabilityand good diver interface.

Latches For sealing main hull, pull actiontoggle latches are fixed on top of the end capto squeeze the O-ring sandwiched between endcap and flange of main hull. They also incor-porate a custom designed lock to prevent acci-dental opening.

2.2 FrameFrame is responsible for providing a rigid struc-ture to Matsya’s peripherals. The position-ing and mounting of peripherals have beendone strategically to develop a bottom-heavyopen-frame structure which exhibits symme-try, modularity and stability. Benefits of anopen frame structure includes easy & fast ac-cessibility and inspection of any component ofMatsya. To make the vehicle dynamically sta-ble, the peripherals are placed so as to alignthe Center of Buoyancy (COB) and the Cen-ter of Mass (COM) vertically, with COM lyingbelow COB.

Modularity is one of the most important as-pects of Matsya’s frame. Every peripheralhas been housed inside a hull and the hull ismounted onto a sub-assembly which is thenmounted with the rest of the frame in sucha way that the peripheral can be replaced orchanged without affecting other parts of Mat-sya. This year, the optimizing process has beendivided into two parts: unit optimization andintegrated optimization. Both kind of opti-mization have been done using Solid Structuralon ANSYS workbench. In unit optimization,only the sub-assembly is optimized on the ba-sis of weight, strength, impact resistance, andspace. While in the integrated optimization,entire assembly of Matsya is optimized on sim-ilar grounds as that for unit optimization. Oneof the key features of integrated optimizationis that it distributes load of components and

stress on frame in a uniform manner.

Figure 5: SolidWorks rendering of CADmodel of Matsya

Thruster Positioning Matsya in total has sixthrusters: four Seabotix BTD150 and twoVideoRay brushed thrusters. Since the Video-Ray thrusters are rated for significantly highervoltages, they are operated at 30V and arechosen for surge direction in order to providehigher surge speed. Since the heading angleis controlled by sway thrusters, their separa-tion has been kept as high as possible in or-der to decrease the thrust for maintaining bal-ancing moment. This particular arrangementallows to cut down power requirement and in-crease the endurance. As a side effect, this ar-rangement forces heave thrusters to get close tothe main hull. Also the fore thruster is morecrowded than that of the aft thruster result-ing in production of different thrusts. In orderto equalize those, ducts are placed around thethrusters to create similar environments.

2.3 ActuatorsThis sub-division handles grippers, torpedolaunchers and marker droppers of Matsya. Allthe three actuators are connected to a cen-tralized pneumatic system which uses a stan-dard paintball CO2 tank as the power source.To control the power, Matsya incorporates sixcontrol valves and a pressure regulator to stepdown the tank’s pressure in order to produce asteady output of 100 psi.

Gripper Matsya has two grippers, one on eachside of the frame. They have been designed insuch a way that they require less ground clear-ance but when actuated, provide large gripping


area. To achieve this, Genetic Algorithm hasbeen used as an optimization tool.

(a) Normal position (b) Actuated position

Figure 6

Torpedo There are two torpedo launchersmounted on either side of Matsya. The tor-pedoes are designed to have twisted fins whichprovide gyroscopic stability. These fins are op-timized using ANSYS Fluent and allow torpe-does to travel 2 meters in a straight line withdeviation of less than 2 centimeters.

Marker Dropper Like torpedo launchers,there are two marker droppers on each side ofMatsya. Though the design of markers is sim-ilar to that of the torpedoes, they are bottomheavy which provide them orientational stabil-ity. They also incorporate twisted fins whichare optimized using ANSYS Fluent. Currentversion of marker travels 2 meters with devia-tion of 10 centimeters or less.

3 ElectronicsWith electronic framework being the core ele-ment for interfacing mechanical hardware withsoftware stack, Matsya’s electronics has beendesigned focussing upon the verticals of sim-plicity, stability and modularity within a dis-tributed architecture (Figure 7). The differ-ent subsystems have been integrated througha GPIO Board (Section 3.2) reducing in-airwiring and enhancing the ease of replaceabil-ity in case of any failure. A separate daugh-ter processing board has been incorporatedin GPIO board to facilitate uninterrupted ex-change of information among different subsys-tems and Single Board Computer (SBC) (Sec-tion 3.1). A very subtle emphasis has beengiven to the power requirements and distribu-tion among the subsystems, thus increasing theendurance manifold.

3.1 Single Board ComputerThe core processing unit of Matsya has beenupgraded in order to suffice the increased com-

Figure 7: Electronics System CommunicationDiagram

putational requirements due to significant en-hancement in the software and acoustic signalprocessing systems. The current SBC consistsof Axiomtek’s mini-ITX motherboard (Figure8) powered by an Intel 4th Generation Corei7 processor running at 2.5GHz along with 8Gigabytes of RAM.

Figure 8: MANO882 mini-ITX Motherboard

3.2 GPIO BoardThe General Purpose Input Output (GPIO)Board works as a shield for the BeagleBoneBlack and provides input and output periph-erals to other electronic subsystems. Beagle-Bone Black is a standalone development boardwith ARM Cortex-A8 processor running at1GHz with 512MB RAM. This processor di-rectly communicates with SBC through eth-ernet. GPIO board has improved the modu-larity of the system significantly as any newsensors or electronic subsystem can be directlyintegrated through this board. Moreover, thisboard also facilitates the required voltage con-version to ensure compatibility of logic levelsamong all subsystems. It provides digital out-put peripherals for pneumatic actuators, PulseWidth Modulated (PWM) output signals for


motor drivers, input peripherals for underwa-ter switches, battery voltage sensor and pres-sure sensor. An array of RGB LED is also pro-vided to check status of Matsya and to debugthe electronic system. As per the set-pointsprovided by the SBC, the GPIO board executesclosed-loop control algorithms providing thedesired PWM outputs to individual thrusters.

Figure 9: Power Distribution Diagram

3.3 Power Management SystemPower Board performs task of regulating poweramong all the sub-systems while incorporatingvarious other features like short circuit protec-tion, reverse battery polarity protection withvoltage measurement. Standard voltages 3.3V,5V and 12V are accessible from a power rail en-suring that the voltage requirements for mostof the standard industrial sensors have beensatisfied. Power management has its own ded-icated microcontroller which performs all theimportant tasks like battery voltage and cur-rent monitoring. In case of low voltage orhigher current than expected, it shuts off powerof all the vital components.

3.4 Pneumatic BoardThe Pneumatic Board provides functionalityfor the independent operation of the pneumaticactuators without affecting the other subsys-tems. The board receives commands from theBeagleBone Black for the purpose of firing tor-pedoes, dropping markers and object manip-ulation using grippers. There are 6 solenoidvalves which can be switched independently us-ing solenoid driver-LMD18400 from Texas In-struments.

3.4.1 Motor Driver Board

The Motor Driver Board integrates all sixSiren-10 motor drivers of Dimension Engineer-

Figure 10: Electronic Hardware Stack

ing LLC on a single custom designed board andprovides ease of replaceability and safety fea-tures. With a power rating of over 150W, eachof these motor drivers are capable of efficientlydriving high-power thrusters. This board canbe switched ON/OFF anytime with the helpof a heavy duty relay which is controlled byBeagleBone Black.

3.5 Acoustic Localization SystemAcoustic localization unit of Matsya uses fourhydrophones to localize with respect to an ac-tive sound source. A custom board is designedto condition the signal coming from the hy-drophones. This conditioned analog signal isconverted to digital signal using DAQ NI 9223from National Instruments and saved & pro-cessed on SBC to estimate the Direction OfArrival (DOA). The entire system execution isdivided into following stages:

Figure 11: Acoustic Localization: Five stageprocess

• Pre-Amplification: This stage involvespre-amplification of raw hydrophone sig-nals that intrinsically have very low peakto peak voltages. The gain is controlled inorder to compensate for voltage compat-ibility with Analog to Digital Converter(ADC) levels and improve sensitivity.

• Filtering: Elliptic band pass filter is usedto remove noise from the amplified signaland passes them to the post amplification


unit. Two second order active elliptic fil-ters are cascaded for each hydrophone inorder to provide high roll-off factor.

• Post-Amplification: The filtered signal isagain passed through a post amplificationphase with a fixed gain to make it com-patible with the ADC’s voltage levels.

• Sampling: Signals from all hydrophonesare sampled using NI 9223 data acquisi-tion board. The digital signals are thentransferred to SBC through ethernet forfurther processing and direction of arrival(DOA) estimation.

• Digital Signal Processing: Digital Signalsof all hydrophones are processed on SBCto estimate position of pinger relative tothe vehicle. After implementing the FastFourier and Inverse Transform the signalsare then pre-processed to discard noiseand to make envelope of the signals uni-form. The DOA is finally estimated fromthe processed signals using hyperbolic po-sitioning [Chan and Ho 1994].

Figure 12: Custom Designed Acoustic board

3.5.1 Batteries

Matsya’s entire system is powered by two 5-cell18.5V 16Ah Lithium Polymer batteries. Onebattery is being used for thrusters & pneumaticactuators while the other one is being usedfor powering electronic system & the acousticlocalization system. With all mission-criticaltasks running along with the onboard peripher-als, these batteries gives an endurance of aboutsix hours to the vehicle.

3.6 SensorsVarious on board sensors are used to get feed-back for control and underwater navigation.

3.6.1 Cameras

Matsya uses two Unibrain Firewire color cam-eras for its machine vision. The vehicle’s bot-tom view camera is a FireWire-i BCL 1394aboard camera while the front view camerais a FireWire-i 1394a board camera. Basedon ICX-625 Sony CCD sensor, these cam-eras provide various features like high resolu-tion color video streaming at adequate framerates, multi-camera synchronization and real-time control of several camera parameters.Lenses of specific focal lengths are used forfront and bottom cameras for getting an ap-propriate underwater Field Of View (FOV).

3.6.2 Inertial Measurement Unit

For low-drift and precise orientation measure-ments, Matsya uses 3DM-GX3-25 AttitudeHeading Reference System (AHRS) from LordMicrostrain Sensing Systems as its primarynavigator. Based on MEMS sensor technol-ogy, this device fuses data from its triaxial ac-celerometer, triaxial gyroscope, triaxial mag-netometer and temperature sensors using anon-board processor and provides very accurateinertial measurements. It is directly interfacedto the SBC via RS-232 protocol.

3.6.3 Pressure Sensor

US381 pressure sensor manufactured by Mea-surement Specialties Inc., is used for the vehi-cle’s depth measurements. With an operatingpressure range of 100 PSI, this sensor outputscurrent in 4-20mA range proportional to thepressure exerted on it’s outer diaphragm.

3.6.4 Doppler Velocity Log

Teledyne RDI’s Explorer Doppler Velocity Log(DVL) has helped significantly in improvis-ing the navigation and localization frameworkof Matsya. By fusing data from an onboardinertial sensor and depth sensor, this deviceprovides high precision velocity data usefulfor real-time underwater navigation using theDoppler effect of sound waves. Rated at depthsup to 1000m, it communicates directly with theSBC using RS-232 serial protocol.

3.6.5 Hydrophones

Four Reson TC 4013 hydrophones are used asan input array to receive the pings of an ac-tive sound source emitting at regular intervals.


Figure 13: Teledyne RDI’s Doppler VelocityLog with its enclosure

With an input sensitivity of -211dB ± 3dB re1V/uPa and usable frequency range of 1 Hz to170 kHz, these hydrophones provide an opti-mal solution for a passive acoustic localizationsystem.

4 SoftwareThe Software Stack of Matsya 4.0 has beendeveloped on top of Robot Operating System(ROS), developed at Willow Garage.

The software system is implemented as onestack with different packages representing var-ious modules like vision, navigation, controls,firmware and mission planning. The designspecifications of the stack ensured that it isextendable, independent and generic. ROShelped in meeting these design goals andkeeping the software modular with differenttasks compartmentalized into various processescalled nodes. ROS handles all the Inter ProcessCommunication between the nodes via mes-sages and services. The software can scale withrespect to the tasks or missions that can beaccomplished. It is also generic enough to beplugged into other robotic frameworks. Thebroad division of the software stack into differ-ent levels is as follows:

• Off-Board Computing: This layer consistsof specialized hardware like filters, ampli-fiers, ADC, GPIO, etc. The communica-tion with SBC is over TCP/IP and UDPprotocols.

• Firmware/Driver Layer: Responsible forall of the Hardware Abstraction, this layerhelps abstract data from the sensors andpresent them as processes to the SBC.Thislayer also consist of abstracted module forhandling GPIOs, PWMs and ADCs. Each

Figure 14: Software Data and Control Flow

hardware peripheral connected to the SBCis abstracted out as a ROS node.

• Middle layer: This layer is responsible forprocessing information from sensors suchas IMU, DVL and Cameras and presentingit to the higher level logic nodes.

• Top layer: This layer uses informationfrom the middle layer to generate actua-tion commands. It also provides a real-time interface to monitor Matsyas perfor-mance and as well as to implement indi-vidual missions.

4.1 LocalizationThe localization module targets clear segmen-tation of upper layers of cognition and lowerlayers which consist of different sensors andcontroller driver modules.

It is responsible for filtering and fusing datafrom inertial and acoustic sensors to providethe top layer with a single source of all relevantlocalization data. At all points of time, Matsyauses the localization data to localize itself inthe belief map.


4.2 Inter Board CommunicationThe number of boards has been reduced by us-ing an on board controller (Beaglebone Black)which has an ARM microprocessor on board.The communication between SBC and onboard controller is handled over TCP/IP andUDP protocol which provide a reliable mecha-nism for transferring of data. A server runningon the controller handles all the incoming com-munications from SBC, where one node actsas the client. A similar architecture exists forcommunication with the Acoustic LocalizationSystem.

4.3 NavigationThe Navigation system is responsible for con-verting motion primitives generated by the Toplayer to set points understandable by the lowerlayers. It handles the generation of set-pointsfor intermediate locations on the trajectorygenerated for motion from point A to pointB. Based on a state machine architecture, itchooses the motion to be performed based onchecks on the localization data with respect tothe desired set point.

4.4 Mission PlannerThe mission planner controls active behaviorof Matsya. Its main goal is to schedule tasksdepending upon the active state of the systemto maximize the score in the competition. Toachieve its objective, it runs a state machinewhich consists of 3 states for each task:

• Transition State: When the currently ac-tive task is not in the range of Matsyassensors, Matsya navigates to the optimumpoint using a user-fed belief map. As soonas Matsya has the task within range of itssensors, it switches to the Scan State.

• Scan State: When the currently activetask is in the range of Matsyas sensors,Matsya tries to search around itself usinga combination of its history and user-fedscan plans. Using its sensors, it tries tocover the most probable region for the cur-rently active tasks till it receives an ac-ceptable amount of sensor feedback. Onreaching an acceptable position to per-form the task, it switches to the ExecutionState. If it fails to find the location of thetask within pre-specified time it aborts the

current mission

• Execution State: When Matsya enters theExecution State from Scan State, there isa very high probability that the task canbe completed. In this state, Matsya per-forms the required motions to completethe task, and moves onto the next taskon successful completion. On losing trackof the task, it switches to scan state so asto position itself properly again.

Figure 14 describes the state machine andthe various transitions.

4.5 InterfacesQuite a few improvements were done to makethe user interfaces robust, modular and intu-itive to use with a few ROS utilities added tothem.

• Debug Interface

This aids in the complete manual navi-gation of Matsya and helps in resolvingboth minor and major errors during test-ing. It also incorporates some task relatedparameters which are essential and form acore part of Matsyas run. The underlyingidea behind developing Matsya 4.0’s De-bug Interface was to provide the user max-imum control over Matsya and to allowtask execution with fewer button clicks,as well as easier modification of parame-ters to be viewed or controlled. This de-manded for a responsive and modular UIwhich was met using the rqt plugins.

Figure 15: Debug Interface• Vision Interface

This interface is used to set various visionrelated parameters which are to be tuneddepending on the environmental condi-tions.

• Map Interface


Provides a drag and drop interface toschedule a list of tasks to perform on anestimated map of the arena. The outputof the map is used by the mission plannerfor navigating from one task to anotherand for real-time tracking of Matsya withrespect to the tasks in the arena.

4.6 VisionThe vision subsystem provides crucial informa-tion about the target object which is used fornavigating Matsya to the target location andperforming the desired mission. Software mod-ules for image processing and computer visionare built using Intel’s OpenCV 2.4 library.

• Framework: The image processing code isimplemented as a library employing ob-ject oriented philosophy for different pro-cessing techniques. Since the other soft-ware elements are entirely based on ROSarchitecture, a single vision node in ROSenvironment is used to handle the com-munication between vision and other soft-ware elements. This helps in reducing theinter-process communication delays whichmight occur in ROS and thus complex al-gorithms can be used in the vision subsys-tem.

• Camera control and pre-processing: Toaccount for the varying environmentalconditions such as illumination variation,brightness artifacts, sunlight reflection,etc., auto exposure and gain control algo-rithms have been implemented which up-date the camera parameters in real-time.Also, several localized and adaptive pre-processing techniques have been developedto handle water color cast, poor contrast,and reduced color quality underwater.

• Processing techniques: The major pro-cessing modules include color detection,shape based validation, region growing,contour analysis and machine learningbased object detection and classification.A modular framework has been developedfor video-based object training and modelgeneration. Each task object is detectedby a combination of aforementioned pro-cessing techniques. The final output ofthe processing module gives generic infor-mation about the target such as its loca-

tion, orientation, dimension, distance-to-camera estimate and some task specific in-formation. This information is communi-cated to other software elements via theVision ROS node.

• Parameter tuning: Some of the image pro-cessing techniques require some tuning ofparameters for better performance. Thus,an easy-to-use GUI based parameter tun-ing interface has been prepared which canbe used for the live update of the param-eters.

Figure 16: Torpedo Cutout Detection

One major addition in Matsya 4.0 is a genericobject tracking algorithm which is able to trackthe target object even if the detection tech-niques fail. Thus, a combination of objecttracking and object detection algorithms hassignificantly improved our task execution ca-pabilities.

4.7 ControlsThis year Matsya’s Motion Control System hasbeen redesigned to provide precise control forfive degrees of freedom, namely, surge, sway,heave, pitch and yaw. The control loop runsat a frequency of 10 Hz, acquiring localizationdata from the DVL and IMU and generatingappropriate PWM signals for each of the sixthrusters. A cascaded control architecture hasbeen developed which comprises of a kinematiccontroller and a dynamics controller. The kine-matic and dynamic controller pair is able toachieve global position and velocity referencetracking.

A 3D motion simulation tool has also been de-veloped to analyze the performance of the mo-tion controller. This tool is also used to test ourmotion planning and navigation techniques.


5 ConclusionWith much rigorous design, analysis and test-ing done by all the three divisions of the team,Matsya 4.0 has the ability to easily adaptand incorporate additional systems integrationwithout major developmental changes. Thiswill help the team to focus more on softwaretesting and sorting out runtime issues effi-ciently in future. Innovative indigenous solu-tions like underwater connectors and switchesprovided unique insights into the team’s de-sign philosophy. With major advancements insoftware and electronics architecture to incor-porate DVL and Acoustic Localization systemin the navigation framework, Matsya 4.0 of-fers a great opportunity to the team to pur-sue further its research in Underwater and Au-tonomous Robotics.

6 AcknowledgementsWe would like to thank the Industrial Researchand Consultancy Centre of IIT Bombay forcontinuous administrative and monetary sup-port during the project. The support of theDean R&D’s office was essentially crucial in thesuccessful execution of the project.We sincerely appreciate the generous supportfrom our sponsors. They played an instrumen-tal role in helping us meet our goals within ourbudget constraints.

ReferencesChan, Y. T., and Ho, K. C. 1994. A simple

and efficient estimator for hyperbolic loca-tion. Signal Processing, IEEE Transactionson 42, 8 (Aug.), 1905–1915.

Iyenger, P. Underwater Image Pro-cessing. http://eternalwandering777.

wordpress.com/.

Quigley, M., Conley, K., Gerkey, B.,

Faust, J., Foote, T. B., Leibs, J.,

Wheeler, R., and Ng, A. Y. 2009. ROS:an open-source robot operating system. InICRA Workshop on Open Source Software.

Tellakula, A. K. 2007. Acoustic source lo-calization using time delay estimation.

Welch, G., and Bishop, G. 1995. An in-troduction to the kalman filter. Tech. rep.,Chapel Hill, NC, USA.

Zhen Cai, Jonathan Mohlenho, C. P.

2009. Acoustic pinger locator (APL) sub-system. Tech. rep.


research and development of matsya 4.0, autonomous

Documents