auv-iitb, indian institute of technology bombay 1 … · santnam bakshi, anuj verma, devansh...

AUV-IITB, INDIAN INSTITUTE OF TECHNOLOGY BOMBAY 1

Research and Development of Matsya 5.0,Autonomous Underwater Vehicle

Varun Mittal, Sandeep Dhakad, Kunal Tyagi, Jay Prakash, Abhinav Kumar, Yash Bhagat, Arpit Singh,Santnam Bakshi, Anuj Verma, Devansh Sharma, Laukik Mujumdar, Athul Arun, Mayank Pandey, HariPrasad V, Anvitha Sudhakar, Viraj Nadkarni, Ninad Kale, Aditya Biniwale, Anant Taraniya, Ananay

Garg, Vipul Ramtekar, R. Sudarsanan, Pranav RaoFaculty Advisor: Leena Vachhani

Abstract—Matsya, stands for a fish in Sanskrit,is a series of Autonomous Underwater Vehicles(AUVs) being developed at Indian Institute ofTechnology(IIT) Bombay with the aim of deliver-ing a research platform in the field of underwaterrobotics and to promote autonomous systems.AUV-IITB is a forum for students and facultyto pursue their interests in underwater robotics.AUV-IITB team has developed five vehicles eachone more advanced as well as more capable thanit’s predecessor. Major architectural changes havebeen made to the subsystems by designing themfrom the perspective to handle tasks in real time.

I. IntroductionMatsya 5 is an AUV developed by a multidis-

ciplinary student-faculty group at IIT Bombay tofacilitate research and development in UnderwaterRobotics as well as to participate in the InternationalRobosub Competition. With integration of a robustAcoustic Localization System and improved controlsystem, this year’s vehicle is capable of performingmajority of the tasks and addressing the challengesdefined by the competition.AUV-IITB is a group of 23 students from differentspecializations having a strong motivation to explorethe field of Underwater Robotics. It has three divi-sions namely Mechanical, Electronics, and Software.Matsya 5 has seen a two year-long development cyclewith majority of components, underwater connec-tors, software stack and electronics boards designedin-house by the team members.

Fig. 1: CAD Rendering of Matsya 5

II. Design StrategyThe designing of Matsya 5 kicked off two years

ago when team decided to build a completely new

vehicle from scratch, starting with a blank sheetof paper. Initially 7 rough sketches of Matsya wereproposed and after thorough discussion, the besttechnologically and aesthetically designed was sin-gled out. Having experience of building 4 vehicles,we listed down the drawbacks and scope of improve-ment that could be made. On the basis of this list,electronics and software sub-divisions were planningand doing experiments that would be implementedin the vehicle. Once the preliminary designs wereready for both electronics and mechanical, consid-erable amount of time was devoted in visualizingmechanical-electronics stack integration and place-ment of connectors on different hulls. This resultedin less cumbersome wiring both inside and outsideof hulls. Whole process was visualized in solid-works.In parallel the software team would test their algo-rithms in simulator to foresee any bug before goingto actual pool testing.

III. Vehicle Design

TABLE I: Matsya 5.0 SpecificationsSpecification Description

Degrees of Freedom 6Length 1750mmBreadth 780mmHeight 640mmEndurance 240 minutesDepth Rating(ANSYS) 150 feetActuation System 5 T200 Thrusters,

3 Videoray BLT,Marker, Torpedo& Gripper System

Power Two 16000mAHLiPo Batteries

Feedback DVL, Visual, Pressure,Acoustic and Inertial

A. Mechanical

Matsya’s mechanical subsystem consists of four spe-cialized sub-divisions namely: Frame, Hulls, Pneu-matic System and Surface. Every sub-division un-dergoes predefined design flow (Figure 2) to deliverreliable and optimized components.


Fig. 2: Design Flow

1) Hulls: Hulls are the pressure chambers of theAUV meant to provide a waterproof region for theelectronics and other water-sensitive components ofMatsya. To incorporate various needs like accessi-bility to inner component(s), field of sensing, pro-vide modularity, etc., Matsya houses four types ofhulls: battery hull, camera hull, main hull and thesensor hulls. For ensuring rugged waterproofing, allthe hulls have undergone vacuum impregnation - anindustrial porosity sealing technique. Apart from thisthe other design objectives include ease of assemblyand disassembly and efficient heat circulation. Eachhull has a fully removable endcap which mechanicallysqueezes a Nitrile O-ring fitted on the flange toachieve waterproofing of the hulls. This is achievedthrough the use of latches in main hull and nuts-boltsin the other hulls.a) Battery Hull: This year Matsya has two batteryhulls each of which contains 2 batteries and a BMSboard. One of the battery hulls is used by Matsya,and the second is used when the batteries in the firsthull is discharged. Easy installation and uninstalla-tion of battery pods allows to change the batteriesin less than a minute with hot swappable powersupply feature to power the vehicle. The designobjective was to achieve robust waterproofing andefficient placing of the 2 batteries and the BMSboard. The battery hull comprises of four sections:endcap, flange, walls and baseplate. The endcap ismade of polycarbonate, is fully removable and isfastened to the hull through a set of 20 nuts andbolts. This helps to get a view inside the hull fordiagnostic purposes. The baseplate and flange aremade of high strength Al-6061-T6 alloy. The wallsare made of the more ductile Al-5052(O) alloy. Toreduce the welded zones the walls are bent from asheet to form two C-shaped structures. This resultsin improved strength and lesser probability of waterseepage.b) Camera Hull: This year Matsya has two camerahulls: Front camera hull and Bottom camera hull.The bottom camera hull is integrated with the mainhull by welding them together. Because of modulardesign philosophy, the camera hull has similar de-sign to that of Matsya 4A. Accurate mounting and

Fig. 3: CAD Rendering of Battery Hull

enough field of view for cameras are the primaryobjectives that the hull must meet. The camera hullcomprises of six sections: endcap, 3 flanges, mainbody and a baseplate. The main body of the camerahull is made of Al-1060 alloy since it is easier tomachine along with providing strength. The endcapis made of polycarbonate for the camera to have clearview. The flanges are made of Al-6061-T6 alloy andused for fastening the endcap and baseplate to thehull. The camera is mounted on an adjustable ABSmount which is press fit inside the hull.

Fig. 4: Exploded View: Camera hull

c) Sensor Hulls:• Doppler Velocity Log (DVL) Hull The design of

the DVL hull is similar to that of the batteryhull. For early detection of water leaks, this hullhas been designed to allow maximum visibilityto the inner parts. The design objectives are ac-curate mounting and electromagnetic shielding.

• Inertial Measurement Unit (IMU) Hull For max-imum visibility and reducing weight, the IMUhull is made entirely of acrylic. Similar to theDVL hull, IMU hull has electromagnetic shield-ing.

d) Main Hull: Main hull houses all the electronics ofMatsya 5 and includes features like layered and back-plane electrical stacking, and good heat dissipationqualities. The height of the main hull is reduced andlength is increased proportionately to accommodateboth, the backplane and the stack. The main hullcomprises of four sections: endcap, flange, walls and


(a) CAD Rendering of DVL Hull

(b) IMU Exploded View

Fig. 5: Sensor Hulls

baseplate. The endcap has two parts - a frame madeof Al-6061-T6 and an acrylic plate. The acrylic plateis press fit into the frame and the whole endcap isfastened to the hull through 10 180◦ latches. Theacrylic plate provides a transparent interface forvisual detection of water seepage as well as electronicdisplays and indicators. Similar to the battery hull,walls of the main hull are made of Al-5052(O) andbent into two C-sections for welding. The flange ismade of Al-6061-T6 and designed to avoid accumula-tion of water on the top surface and to ensure that nowater seeps inside during disassembly. The baseplateis made of Al-6061-T6 which has more strength sincethe main hull also provides structural strength toMatsya.

Fig. 6: Main hull Exploded View

e) Connectors: The team has fabricated in-housedry mate-able two pin underwater connectors toroute connections between main hull and batteryhull. The plug and play nature of these connectorsmakes it convenient to integrate additional compo-nents in Matsya and eases the phase of disassemblyand replacement. Since these connectors are custommade, their manufacturing cost is 20 times cheaperas compared to the alternative option available in

the market.f) Penetrators: Penetrators are designed to routecables out of hulls for direct connection to the sensorsfor which simple plug and play connectors cannot beused as they cause data loss and impedance matchingproblems.g) Underwater Switches: Matsya 5 has 4 underwaterswitches at the back end of the vehicle- two onboth sides. This year rotatory switches have beendesigned for the ease of use by the diver. Theycontain magnets and reed switch for their operation.

Fig. 7: CAD Rendering of Underwater Switch

h) Latches: For sealing main hull, pull action togglelatches are fixed on top of the end cap to squeezethe O-ring sandwiched between end cap and flange ofmain hull. They also incorporate a custom designedlock to prevent accidental opening.2) Frame: The frame of Matsya 5 has been designedto achieve a more hydrodynamic vehicle and to bringmore stability than before. To engineer this, thelength of the vehicle is increased and its height isreduced. A skeleton type internal structure has beendesigned to reduce weight without compromising onthe strength. The cage-like structure is made up ofwater-jetted aluminum 6061 alloy with an externalcoating to prevent its corrosion. The corners andthe angles have been joined by nylock nuts (SS),bolts (SS) & CNC-machined delrin parts acting asa support to prevent relative motion. The structureof frame has been provided with handles, at frontand rear end, to transport the vehicle by hand, anda horizontal rod at the rear end, as a handle for thediver.

Fig. 8: Internal Frame CAD

a) Component Positioning: Every component andhull of Matsya 5 has been positioned to makethe vehicle stable (static) in Roll, Pitch and Yaw.The positively buoyant hulls, Main Hull & DVLHull, have been kept above the negatively buoyant


components, Battery Hull, Pneumatic valves &pressurized CO2 tank. The hydrophones have beenattached to the studs of the vehicle to receive clearsignals.A total of 8 thrusters have been used in Matsyato provide all possible degrees of freedom to thevehicle. The surge thrusters of Matsya 5 have beenkept on the vertices of an equilateral triangle. Thesurge thrusters hence can also be used for pitch andyaw control. The line joining the centroid of theequilateral triangle and CG (Center of Gravity) ofthe vehicle is along the surge axis of the vehicle. Thiswas done to provide maximum stability during surge.

Fig. 9: Surge Mount CAD

In Matsya 5, three heave thrusters have been usedand are kept at the vertices’s of an isosceles triangle,two at the front and one at the rear. This allows tooperate these thrusters for pitch and roll control aswell. The two sway thrusters, one near the front endand the other at the rear end, have ducts aroundthem to ensure symmetric unobstructed flow of thewake. The centroid of the isosceles triangle of heavethrusters, the midpoint of the sway thrusters andthe CG of the vehicle have been made collinear formaximum overall stability.b) Analysis: Each of the structural elements of theframe has undergone static analysis using the FiniteElement Method (FEM) on ANSYS Workbench. Theanalysis was done on the designs before fabrication toensure that the components can handle the realisticsituations like:The frame of the vehicle should be able to withstandthe weight of the hulls and the force of the thrustersin air and underwater. The hulls of the vehicle shouldbe able to withstand the hydrostatic pressure atatleast 50 meter depth. The stresses and deflectionswere considered for each case and a safety factor ofatleast 1.3 was achieved everywhere.3) Actuators: Matsya 5 has several actuatorsmounted on its frame to make it capable of accom-plishing several tasks at RoboSub. All the actuatorshave been connected to pneumatic valves which aredriven by a CO2 gas tank.a) Gripper: Matsya has a gripper mounted on theright side of the frame. They have been designed insuch a way that they require less ground clearance

Fig. 10: ANSYS of DVL Hull at 50m depth

but when actuated, provide large gripping area. Thegripper was designed using Genetic Algorithm as anoptimization tool.

Fig. 11: Gripper CAD

b) Torpedo: Matsya 5 has 2 torpedoes just above thefront camera hull. The torpedoes have twisted finsat the and to provide gyroscopic stability allowingtorpedoes to travel in a straight line for 1 meter withdeviation of 10 cm or less.c) Marker Dropper: Matsya 5 has two marker drop-pers and the actuation system has been optimizedto use only one pneumatic piston as opposed to twoin Matsya 4A. The design principle of the markersis same as that of the torpedoes, with the onlydifference being markers are heavier.

(a) Torpedo

(b) Marker(c) Marker Dropper CAD Render-ing

Fig. 12: Torpedo & Marker System

4) Surface: The previous series of AUVs had openframe structure which considerably hindered thehydrodynamic characteristics of the vehicle. An en-closure was conceptualized, in order to direct the


incoming fluid, and therefore improve the drag char-acteristics of the vehicle.The design objective is to heighten hydrodynamicperformance and aesthetics. The initial design wasdeveloped in collaboration with IDC, IIT Bombay.To permit easier access to the components of thevehicle, the surface is split into 6 parts, each of whichis detachable. The surface is mounted on the vehicleby using a press-fit mechanism, ensuring a fasteropening and closing of the surface.Acrylonitrile Butadiene Styrene (ABS) has beenused for making the surface, due to its relatively goodstrength and impact characteristics (Izod values),among thermoplastics. The manufacturing processinvolved CNC-machining a stack of MDF plywoodsfor the surface mould, heating the ABS sheet to80oC and vacuum suctioning it over positive moulds(Vacuum Forming). Each part was mounted on thevehicle using custom 3D printed mounts.

Fig. 13: Surface CAD

Fig. 14: Surface Exploded View

B. ElectronicsElectronics framework is the main element for inter-face between the mechanical and software systems Italso serves for better endurance and safe operationof the vehicle in real time.

Fig. 15: Electronics System Communication Diagram

1) Architecture: The whole of electronics can bebroadly divided into the 3 parts viz. Main Stack,SBC and the Battery Management system. TheSBC does the higher level processing such as vision,Mission planner and controller. It reads/publishesdata from various sensors(IMU, DVL and Pressuresensor) and electronics stack on serial and CANbus respectively. The Main stack features a customdesigned backplane system for ease of accessibility ,increased modularity and enhanced ease of replace-ability in case of failure in one of the components.The backplane has slots for Power, GPIO and MotorDriver modules with connectors for thrusters, Sen-sors, Power and kill switches placed on the peripheryfor easy accessibility and removal. It serves as theinterface between different boards and the SBC.

Fig. 16: Electronic Hardware Stack

2) Controller Area Network: The communicationover different subsystems occurs over CAN. Higherlevel communication protocols are designed in a wayto have a highly robust system considering the casesof failures and their solution. The CAN bus servesas a single point of I/O for the main processor i.e.SBC from the electronics stack. All the messagesfrom different subsystems are published on CANbus which may be accessed by all other subsys-tems. Unconcerned data can be easily filtered bysubsystems thus reducing the processing time on thesubsystem. The data traffic consists of informationabout speed of thrusters, state of actuators, Pingmessages from different subsystems, Battery SOC aswell as the interrupt messages published when anyof the external switches are used.3) GPIO Board: Integrates the sub-modules for gen-erating PWM, Digital I/O and ADCs. The PWMsignal are generated for thrusters. Digital Outputsare used to control power to actuators. The pressuresensor reading are recorded by 12 bit dedicated ADCwhich is sent to SBC on serial. Different referencevoltage line is used to keep the pressure sensor signalunaffected by noise or fluctuations in GND used forpower lines.4) Power Management System: Takes input powerfrom BMS and converts them into different powerrails needed by sensors, SBC, Data Acquisition Cardand cameras. All the power rails are controlled byAtmega328P which ensures proper powering up se-quence and communicates with SBC via CAN. Inthis way any device’s power can be rebooted in case


needed without powering off the entire system. Asoft kill signal can also be generated for switchingoff thruster upon message from SBC which is sent toBMS for implementation.

Fig. 17: Power Distribution Diagram

5) Battery Management system: It is designed toensure maximum endurance for Matsya. The BMStakes power from two 16000mAh 4 cell lipo batteries,check the battery with higher SOC along with killswitch and software commands to channel power tothrusters. If any of the three criteria fails, power willnot be channeled to thruster ensuring the safety ofbattery, vehicle and diver swimming near the vehicle.The battery voltage is measured at a time whencurrent is not being drawn to get accurate voltage.It is done alternately so that power supplied toelectronics is not affected. It periodically sends theSOC left to mission planner to estimate run timeavailable. If SOC drops below certain threshold, suchmessage is also published on CAN bus to warn themission planner. If SOC further drops below anotherthreshold, BMS shuts of all the power to electronicsto prevent batteries from over discharging.6) Motor Driver Board: The Motor Driver Boardintegrates all 8 BlueESC from BlueRobotics on twoidentical custom designed boards and provides easeof replaceability and safety features. With a currentrating of 30A, each of these motor drivers are capableof efficiently driving both Videoray brushless as wellas Blue Robotics T200 thrusters. The power to thisboard is controlled by MOSFET switches indepen-dently by kill switches as well as command fromSBC.7) Acoustic Localization System: Acoustic localiza-tion unit of Matsya uses four hydrophones to localizewith respect to an active sound source. A customboard is designed to condition the signal coming fromthe hydrophones. This conditioned analog signal isconverted to digital signal using DAQ NI 9223 fromNational Instruments and saved & processed on SBCto estimate the Direction Of Arrival (DOA). Theentire system execution is divided into followingstages:

• Pre-Amplification: This stage involves pre-amplification of raw hydrophone signals that in-trinsically have very low peak to peak voltages.The gain is controlled in order to compensatefor voltage compatibility with Analog to DigitalConverter (ADC) levels and improve sensitivity.

Fig. 18: Acoustic Localization Flow

• Filtering: Elliptic band pass filter is used to re-move noise from the amplified signal and passesthem to the post amplification unit. Two secondorder active elliptic filters are cascaded for eachhydrophone in order to provide high roll-offfactor.

• Post-Amplification: The filtered signal is againpassed through a post amplification phase witha fixed gain to make it compatible with theADC’s voltage levels.

• Sampling: Signals from all hydrophones aresampled using NI 9223 data acquisition board.The digital signals are then transferred to SBCthrough ethernet for further processing and di-rection of arrival (DOA) estimation.

• Digital Signal Processing: Digital Signals of allhydrophones are processed on SBC to estimateposition of pinger relative to the vehicle. Af-ter implementing the Fast Fourier and InverseTransform, the signals are then pre-processed todiscard noise and to make envelope of the signalsuniform. The DOA is finally estimated from theprocessed signals using hyperbolic positioning.

8) Batteries: Matsya’s entire system is powered bytwo 4-cell 16Ah Lithium Polymer batteries. Boththe batteries are merged so as to maximize the en-durance. Special provision is made to charge the bat-teries without having to open the enclosure. Safetymeasures are also taken for the gases leaking outduring charging.9) Sensors: Various on board sensors are used to getfeedback for control and underwater navigation.

• Cameras There is a significant jump in termsof quality of images we used to get in matsya4.0. This year we are using Mako G-234 cameraalong with Kowa LM6HC lens. The camera isoperated at 14 fps and 405x645 resolution. Thecameras are PoE compliant and communicatedirectly to SBC via Ethernet. Power is providedthrough through a PoE injector in intermediatepath.

• Inertial Measurement Unit For low-drift andprecise orientation measurements, Matsyauses GX4 Attitude Heading Reference System(AHRS) from Lord Microstrain Sensing Systemsas its primary navigator. Based on MEMSsensor technology, this device fuses data fromits triaxial accelerometer, triaxial gyroscope,triaxial magnetometer and temperature sensorsusing an on-board processor and provides veryaccurate inertial measurements. It is directly


interfaced to the SBC via RS-232 protocol. Thesensor is mounted so as to be least affected bythe electromagnetic interference caused by highcurrent consuming components.

• Pressure SensorUS300 pressure sensor manufactured by Mea-surement Specialties Inc., is used for the vehi-cle’s depth measurements. It is a voltage basedpressure sensor and is used to calculate depthfrom surface. The analog output of the sensoris fed directly to ADC which is sent to SBC viaserial communication.

• Doppler Velocity Log Teledyne RDI’s ExplorerDoppler Velocity Log (DVL) has helped signifi-cantly in improvising the navigation and local-ization framework of Matsya. By fusing datafrom an onboard inertial sensor and depth sen-sor, this device provides high precision velocitydata useful for real-time underwater navigationusing the Doppler effect of sound waves. Ratedat depths up to 1000m, it communicates directlywith the SBC using RS-232 serial protocol.

• Water Seepage sensor An in-house water seep-age sensor was developed to alert us whenwater seeps inside hulls and becomes dangerto electronics. The current conducted by thesmall amount of water is processed to workwithin TTL level and programmed to generateinterrupt when water seeps inside

• Hydrophones Four Reson TC 4013 hydrophonesare used as an input array to receive the pingsof an active sound source emitting at regularintervals. With an input sensitivity of -211dB ±3dB re 1V/uPa and usable frequency range of1 Hz to 170 kHz, these hydrophones provide anoptimal solution for a passive acoustic localiza-tion system.

Fig. 19: CAD rendering of Thruster Profiling Setup

C. SoftwareThe Software Stack of Matsya 5 has been developedon top of Robot Operating System (ROS), developedat Willow Garage.

Fig. 20: Software Data and Control Flow

The software system is implemented as one stackwith different packages representing various mod-ules like vision, navigation, controls, firmware andmission planning. The design specifications of thestack ensured that it is extendable, independentand generic. ROS helped in meeting these designgoals and keeping the software modular with differ-ent tasks compartmentalized into various processescalled nodes. ROS handles all the Inter ProcessCommunication between the nodes via messages andservices. The software can scale with respect to thetasks or missions that can be accomplished. It isalso generic enough to be plugged into other roboticframeworks. The broad division of the software stackinto different levels is as follows:

• Off-Board Computing: This layer consists ofspecialized hardware like filters, amplifiers,ADC, GPIO, PWM, etc. The communicationwith SBC is over TCP/IP, UDP, as well as CANprotocols.

• Firmware/Driver Layer: Responsible for all ofthe Hardware Abstraction, this layer helps ab-stract data from the sensors and present themas processes to the SBC. This layer also consistof communication modules required for handlingGPIO, PWM and ADC. Each hardware periph-eral connected to the SBC is abstracted out asa ROS node.

• Middle layer: This layer is responsible for pro-cessing information from sensors such as IMU,DVL and Cameras and presenting it to thehigher level logic nodes.

• Top layer: This layer uses information from themiddle layer to generate actuation commands.It also provides a real-time interface to monitorMatsya’s performance and as well as to imple-ment individual missions.


1) Localization: The localization module targetsclear segmentation of upper layers of cognition andlower layers which consist of different sensors andcontroller driver modules.It is responsible for filtering and fusing data frominertial and acoustic sensors to provide the top layerwith a single source of all relevant localization data.At all points of time, Matsya uses the localizationdata to localize itself in the belief map.2) Controls: This year Matsya’s Motion ControlSystem has been redesigned to provide precise con-trol for six degrees of freedom, namely, surge, sway,heave, pitch, roll and yaw. The control loop runs at afrequency of 10 Hz, acquiring localization data fromthe DVL and IMU and generating appropriate PWMsignals for each of the eight thrusters. A single loopmodified PID control architecture has been devel-oped which comprises of a kinematic controller anda dynamics controller. The kinematic and dynamiccontroller pair is able to achieve global position andvelocity reference tracking.A 3D motion simulation tool has also been developedto analyze the performance of the motion controller.This tool is also used to test our motion planningand navigation techniques.3) Navigation: The Navigation system is responsiblefor converting motion primitives generated by theTop layer to set points understandable by the lowerlayers. It handles the generation of set-points forintermediate locations on the trajectory generatedfor motion from point A to point B. Based on a statemachine architecture, it chooses the motion to beperformed based on checks on the localization datawith respect to the desired set point.4) Mission Planner: The mission planner controlsactive behavior of Matsya. Its main goal is to sched-ule tasks depending upon the active state of thesystem to maximize the score in the competition. Toachieve its objective, it runs a state machine whichconsists of 3 states for each task:

• Transition State: When the currently active taskis not in the range of Matsya’s sensors, Matsyanavigates to the optimum point using a user-fed belief map. As soon as Matsya has the taskwithin range of its sensors, it switches to theScan State.

• Scan State: When the currently active task isin the range of Matsya’s sensors, Matsya triesto search around itself using a combination ofits history and user-fed scan plans. Using itssensors, it tries to cover the most probable re-gion for the currently active tasks till it receivesan acceptable amount of sensor feedback. Onreaching an acceptable position to perform thetask, it switches to the Execution State. If itfails to find the location of the task within pre-specified time it aborts the current mission

• Execution State: When Matsya enters the Ex-ecution State from Scan State, there is a very

high probability that the task can be completed.In this state, Matsya performs the required mo-tions to complete the task, and moves onto thenext task on successful completion. On losingtrack of the task, it switches to scan state so asto position itself properly again.Figure 20 describes the state machine and thevarious transitions.

5) Vision: The vision subsystem provides crucialinformation about the target object which is usedfor navigating Matsya to the target location andperforming the desired mission. Software modules forimage processing and computer vision are built usingIntel’s OpenCV 2.4 library.

• Framework: The image processing code is imple-mented as a library employing object orientedphilosophy for different processing techniques.Since the other software elements are entirelybased on ROS architecture, a single vision nodein ROS environment is used to handle thecommunication between vision and other soft-ware elements. This helps in reducing the inter-process communication delays which might oc-cur in ROS and thus complex algorithms can beused in the vision subsystem.

• Camera control and pre-processing: To accountfor the varying environmental conditions suchas illumination variation, brightness artifacts,sunlight reflection, etc., auto exposure andgain control algorithms have been implementedwhich update the camera parameters in real-time. Also, several localized and adaptive pre-processing techniques have been developed tohandle water color cast, poor contrast, and re-duced color quality underwater.

• Processing techniques: The major processingmodules include color detection, shape basedvalidation, region growing, contour analysis andmachine learning based object detection andclassification. A modular framework has beendeveloped for video-based object training andmodel generation. Each task object is detectedby a combination of aforementioned processingtechniques.The final output of the processing module givesgeneric information about the target such asits location, orientation, dimension, distance-to-camera estimate and some task specific informa-tion. This information is communicated to othersoftware elements via the Vision ROS node.

• Parameter tuning: Some of the image processingtechniques require some tuning of parametersfor better performance. Thus, an easy-to-usetext file based parameter tuning interface hasbeen prepared which can be used for the liveupdate of the parameters.

One major addition in Matsya 5 is a generic processarchitecture which enables breaking algorithms inmodular pieces, reducing lines of code as well as


Fig. 21: Image Processing Interface

increasing test coverage.

IV. Experimental resultsAn apparatus was developed for profiling of thrustercurrent and thrust vs PWM values. This would givethe controller data which would help it to optimizeits performance. The output given by controller isin terms of force, so the force to PWM map aidsin optimal control of the vehicle. The setup consistsof Load cell attached to a frame which is rigidlyattached to ground so as to get the most accurateresult. The thruster is attached to load such that thethrust from it produces a torque. The setup is cali-brated and its parameters are calculated from knownvalues of force and used to calculate thrust. Alsoattached in series of current path given to thruster asense resistor to measure current consumed for giventhrust. The profiling results of thruster gave the thenon-linear relation between PWM and thrust. Thisgreatly helped in optimizing controller.As the theoretical testing of controller was beingdone in parallel with mechanical and electronic fabri-cation time, the real testing of controller was reducedto tuning of parameters. This resulted in maintainingfocus on the tasks in problem statement. Matsya 5has observed 300 Hrs of rigorous testing comprisingof testing of acoustic, vision and mission planner. Itis now going to perform integrated testing of all tasksand will be optimized for time.

V. ConclusionWith much rigorous design, analysis and testing doneby all the three subdivisions of the team, Matsya5 has the ability to easily adapt and incorporateadditional systems integration without major devel-opmental changes. This will help the team to focusmore on software testing and sorting out runtimeissues efficiently in future. Innovative indigenoussolutions like underwater connectors and switchesprovided unique insights into the team’s design phi-losophy. Major advancements in software and elec-tronics architecture achieve a reliable interface of

software with mechanical end. Matsya 5 offers agreat opportunity to the team to pursue further itsresearch in Underwater and Autonomous Robotics.

VI. AcknowledgementsWe would like to thank the Industrial Research andConsultancy Centre of IIT Bombay for continuousadministrative and monetary support during theproject. The support of the Dean R&D’s office wasessentially crucial in the successful execution of theproject.We sincerely appreciate the generous support fromour sponsors. They played an instrumental role inhelping us meet our goals within our budget con-straints. Special thanks to our vendors such as Bluerobotics, Teledyne RDI for their support in case oftechnical problems.

References[1] Chan, Y. T. and Ho, K. C., A simple and efficient estima-

tor for hyperbolic location. 1994.[2] Quigley, Morgan and Conley, Ken and Gerkey, Brian and

Faust, Josh and Foote, Tully B. and Leibs, Jeremy andWheeler, Rob and Ng, Andrew Y., ICRA Workshop onOpen Source Software. 2009.

[3] Ashok Kumar Tellakula, Acoustic Source Localization Us-ing Time Delay Estimation. 2007.

[4] Welch, Greg and Bishop, Gary, An Introduction to theKalman Filter. 1995.

[5] Zhen Cai, Jonathan Mohlenho, Cassondra Puklavage,Acoustic Pinger Locator Subsystem. 2009.


APPENDIX:Community outreachThe AUV-IITB Team, each year attends many work-shops and/or exhibitions to reach the community.It is through these exhibits the team encouragesyoung school as well as high school students to takeup robotics. The team demonstrates working of theAUV followed by a detailed seminar and a question-naire session to motivate students, and increase theirknowledge about AUVs and robotics in general.

Fig. 22: Matsya at Exhibition at Techfest

The Team last year participated in events such asTechFest- Asia Largest Technical Festival, organizedby IIT Bombay; Nehru Science Centre Exhibitionand MakerMela. Also, the team itself held manyworkshops in the campus, open for all, to havethought provoking discussions with the students andprofessors about the design strategy and the generalworking concept of the AUV. This not only helps theteam get fresh ideas, but it also helps us ponder ona few details which the team might have missed.Apart from this, the joy it brings to others, espe-cially young enthusiastic school students (for exam-ple students of Witty International School shown inphoto)is a rewarding experience and motivates theteam further to work harder and continue to makemore developments.

Fig. 23: School children at our lab

Fig. 24: Matsya at Tech and RnD Expo

The research done by the team also helped severalstudents in their Masters/ BTech projects on topicslike Control of Overactuated Nonlinear Systems,

Navigation of Unmanned Vehicles, Design of a 2-Linkgripping mechanism, etc. This further fuels the teamto work harder and deliver results.The team also mentors quite a few other teamsfrom India, who are keen on making AUVs, likeIEM Kolkata, Sahyadri College and VIT Pune. Theteam guides them through the overall procedure ofmaking an AUV, importance of communication anddocumentation and the process of acquiring funds formaking AUVs in their respective colleges.

auv-iitb, indian institute of technology bombay 1 … · santnam bakshi, anuj verma, devansh...

Documents