DTU-AUV 1

DESIGN AND DEVELOPMENT OF AN AUTONOMOUS

UNDERWATER VEHICLE - LAPRAS
Mihir Ranjan, Megha Goyal, Dhruv, Harshul Bansal, Siddhant Verma, Gunagya Singh Mamak, Sriram
Chava, Saurav Ranjan, Shivang Dudani, Utkarsh Verma, Shorya Kumar, Kritika Bansal, Rajat, Sundeep
Chand, S. Suchit, Chirag Garg, Samarth Sachdeva, Suryansh Gangwar, Ishpreet Singh Luthra, Utkarsh
Kuchhal, Shivam Rastogi, Saransh Goel, Md Masood Raza, Abhinav Aggrawal, Rakshit Garg

Delhi Technological University, Shahbad Daulatpur, New Delhi-110042, India

ABSTRACT: corresponding props would be easier from the

LAPRAS is the flagship Autonomous vision perspective as they are quite
Underwater Vehicle developed at Delhi distinguishable. The smaller torpedo opening of
Technological University. As first-time
the bootlegger is a trapezium that is easier to
participants at RoboSub, the vehicle was
designed to be compact and robust while aim for than a star. For the buoy task, we would
simultaneously being able to function as a need a rough estimate of the distance to the
reliable AUV to complete all required tasks and buoys, which could be done using a simple
unmanned operations. This entailed a massive vision algorithm since the buoy dimensions are
design overhaul with improving vision known, as once the bot gets close to them,
processing and a comprehensive round of detection would not work. For the torpedo task,
simulations and testing the final design that
to reduce the chances of failure, we decided to
exhibited a clear improvement compared to our
previously created designs. first shoot through the bigger opening and then
through the smaller opening. In the bin task, our
strategy is to use two markers. Once the first
I. COMPETITION STRATEGY
marker is dropped, we would track it and based
To coordinate well in a remote work on its success, we would take the second
environment, we had to introduce some attempt if needed.
significant changes in the software department Once the bot reaches the centre of the octagon
workflow. Using a monorepo on Github to hold through the pinger guidance, we start searching
all the project files helped us collaborate for the bottles located on the table via object
seamlessly in a remote working environment. detection through the front & the bottom
This enabled us to start testing our vision and camera. Once the position of the bottle is
control algorithms early on. localised we pick it up and resurface at the
For enhancing the colours in the videos centre of the octagon. Then we try to find the
captured, we decided to use our image dehazing table having dollar sign, by searching for it,
method[1], on which we started working along the diameter of the octagon and finally we
recently. try to place the bottle on the table after
Coming to our task strategy, we decided to pick localising it. Then we try to repeat the previous
the G-Man side since detecting the
 DTU-AUV 2

steps, by moving to the centre of the octagon, It makes precise cuts to the metal necessary to
till no bottle is left on the platform. manufacture the frame's intricate cavities and
edges. The frame is made up of aluminium
6061-T6 alloy, selected considering the harsh
working environment and operating conditions.
The bottom plate is attached to the base plate
using 2mm stainless steel L-brackets,
manufactured using sheet metal bending
operation. The mounts and fixtures for the hulls
are produced by FDM based 3D printing of
ABS material, chosen as it absorbs and dampens
the structural vibrations to acquire precise
dimensions and strength of the design. The base
plate holds the main hull, front camera hull, 4
heave thrusters, 2 sway thrusters, 4 heave
thrusters, torpedo launcher and gripper. Each
Figure 1.1 - Cad render of Lapras
side plate holds one surge thruster each. The
bottom plate holds the battery hull, bottom
II. DESIGN RATIONALE camera hull and dropper.

A. MECHANICAL DEPARTMENT

FRAME

Figure 2.2 - Mesh of Lapras base plate

Figure 2.1 - Frame Render of Lapras

The frame of Lapras is the main supporting
structure in the assembly. The structure is well
optimised to avoid stress concentration and to
sustain the loads without any considerable
deformation or failure. The frame is designed to
have minimum assembly time, easy access to
the components attached and structural strength
under static and dynamic conditions. Figure 2.3 - Stress analysis of Lapras base plate
The sheet metal parts of the frame are
manufactured using waterjet cutting technology.
 DTU-AUV 3

The structure is smartly designed so that the MANIPULATORS

base plate not only act as the mounting space for
the electronic housing and manipulators but also 1. TORPEDO LAUNCHER
it has shrouds for the heave thrusters, thus the
placement of the base plate was crucial as we
planned to place heave thrusters in the plane of
centre of gravity for higher propulsion
efficiency.

CUBOIDAL HULLS
The AUV has 4 hulls in total attached to its
frame. Except for the front camera hull, all the
hulls are cuboidal in shape. It is easier to store,
access and monitor electronic components in a
cuboidal hull as compared to cylindrical hulls, Figure 2.4 - Design render of Torpedo Launcher
where it was not possible to utilise the complete
volume in a clean way. The torpedo launcher system consists of two
parallel launch tubes, 3-D printed with springs
The main hull is mounted upon the frame with
mounted at their ends. The launch tubes are
the help of 4 wall mounts and 2 latches, here the
placed adjacent to each other to reduce the
latches secure the hull in place.
inaccuracy of the torpedo. The tubes are
The battery hull is made up of aluminium to attached to the frame with the help of 2 snap-fit
dissipate the heat generated by the buck attachments, this makes it easy to load the
converters. Thus eliminating the need for a torpedoes. The torpedoes are kept inside the
dedicated cooling system. launch tubes with the help of a circular
cross-section (obstruction plate) made up of
aluminium (6061-T6 alloy). The obstruction
THRUSTER PLACEMENT plate is controlled by a waterproof servo motor,
AUV has a total of 8 T200 thrusters, which are mounted above the launch tubes with the help of
placed as follows: latches for easy assembly. The servo motor
rotates by an angle of 30 degrees from the mean
● 4 heave thrusters are placed at the base position for a successful launch.
plate that controls the roll and pitch of
the AUV. The front of the torpedo is ellipsoidal. This
● 2 sway thrusters are placed at the bottom reduces the stress at the tip and thus protects
plate to control yaw. from deformations. The tapered cylindrical body
● 2 surge thrusters are placed at the side of the torpedo is ideal for reducing drag and
plates. mass simultaneously. The torpedo has 4 fins that
prevent any unwanted deviation from its
Overall the AUV has 6 degrees of freedom. trajectory.
The torpedo has a slenderness ratio of 14.66
Torpedo Coefficient of Drag:
2
𝐶𝑑*𝐴*ρ*𝑉
Drag Force = 2
2
𝑁𝑒𝑡 𝑇ℎ𝑟𝑢𝑠𝑡*𝑇
Vertical deviation = 2
 DTU-AUV 4

0.3
𝑑𝑥
The dropper for the AUV holds two solid steel
T= ∫ 𝑉 balls (Markers) adjacent to each other, separated
0
by a partition. The markers are dropped by
2 2
2
𝐶𝑑*𝐴*ρ*𝑉 *𝑥 rotating a heart-shaped plate placed
Where, 𝑉 = 𝑉𝑙 − 𝑚 perpendicular to the dropper. The balls then free
0.3 𝐶𝑑*𝐴*ρ*𝑥
fall under the influence of gravity. The dropper
(1 + )𝑑𝑥
T= ∫ 𝑚 body is kept adjacent to the bottom camera for
2
0 𝑉𝑙 better mapping and accurate dropping. The
obstruction plate has a circular cut in the centre,
Where, which is designed to lower the total rotation
Cd = Coefficient of Drag made by the servo motor, which significantly
reduces any chance of failure. The servo motor
A = cross-sectional area of Torpedo has to rotate 30 degrees from the mean position
to release the marker. The markers are placed in
ρ = Density of Water
their respective cylindrical cavities, and a
V = instantaneous velocity of torpedo detachable cover is used to secure them from the
top. The dropper body is made up of ABS
VL = launch velocity of torpedo plastic and is manufactured by the 3-D printing
m = mass of torpedo process. This design is adopted as it takes up a
small amount of space and requires only one
T = total time travelled waterproof servo motor to operate.
Simulating the torpedo for speeds ranging 3. GRIPPER
between 0.25 m/s to 2m/s :
Cd = 0.091
Taking maximum vertical deviation to be 0.5
cm, the launch velocity comes out to be 0.46
m/s. The escape velocity of the torpedo is
around 0.46 m/s in water, this speed is sufficient
for the torpedo to hit the target from a safe
distance without any deviation and without
damaging anything in the environment.
2. DROPPER
Figure 2.6 Design Render of Gripper
For most of the manipulator designs we focussed on
workability and simplicity, thus the gripper for
Lapras is manufactured by 3-D printing, and the
material used is ABS plastic considering its high
strength to weight ratio. The gripper fingers have
toothed gears in contact with each other and are
constrained to rotate together. A servo motor is
mounted above the fingers, which provides torque
to the whole system. The shape of the fingers is
Figure 2.5 Design Render of Dropper triangular, with the internal surface slightly curved
inwards for easier gripping of circular and
 DTU-AUV 5

cylindrical objects. At the end of each finger,

abrasions are made to increase the gripping strength S.no Component Current draw
and make lifting the bin covers easier in the marker
dropping challenge. A cavity is made in each 1 Thruster (T200) 4A

finger, which sheds off excess mass while retaining 2 Servo Motor (HS5646WP) 2.1A
structural rigidity and strength.
3 Intel NUC (NUC7i7BNH) 3.5A
B. ELECTRONICS DEPARTMENT 4 Arduino Mega 0.6A
The main hull houses all the electronic
components, and the lithium polymer batteries 5 Hydrophone (Teledyne 0.028A
have been placed in a separate hull. Multiple TC4014)
PCBs have been designed to reduce the wiring
6 MAX11043ATL 0.5926A
and make the connections more modular.
Individual PCBs have been designed for Power
distribution, Sensor and motor control and for Total current being drawn at all times =
cutting off and switching the batteries. 6 x (4A) + 1 x (2.1A) + (3.5A) + (0.6A) + 3 x
(0.028A) + 0.5926A = 30.8766A
1. Power
The Power Distribution Board (PDB) was custom
Total battery capacity = 10000 + 10000 mAH =
designed and had a buck converter for the servo
motors. The PDB enables us to power all the 20000mAH
electronics onboard. PDB distributes power to the
thrusters, servo motors, Intel NUC through the Ideal operating time of Lapras =
DC-DC NUC, which boosts 12V to 19V and the 20000/30876.6 = 0.6477 hours = 38.86 minutes
hydrophones. The Intel NUC powers the
microprocessor board. The buck converter drops The servo motors will not be working
the 12VDC to 7VDC for the servo motors. A
buck converter is also implemented for 12VDC - simultaneously, only one servo will operate at a
5VDC for the acoustic systems. A DPDT switch time. Similarly a maximum of 6 thrusters will be
located outside the AUV is connected between the operating at once. The runtime has been calculated
DC-DC NUC and the PDB as a kill switch to halt by taking these situations into consideration.
the vehicle’s functioning.
2. Battery monitoring system
2, 4 cell Lithium polymer batteries with a total
capacity of 20000mAH were chosen for the input
power. A custom battery monitoring system PCB
was designed, which continuously monitors the cell
voltage and the total voltage of each battery and
cuts it off from the circuit as soon as it reaches the
threshold minimum voltage (3.1V). This circuit also
switches between the two batteries through the
inrush as soon as one is cut off. Previously, a market
purchased battery management system was used,
But after thorough testing and designing, this PCB
was finally used.
Figure 2.7 Power distribution board
 DTU-AUV 6

Figure 2.8 Battery cutoff/switching circuit

3. Higher-level control board

A microprocessor board for sensor and motor Figure 2.9 Microprocessor board
control was designed by the team. Arduino Mega
2560 was chosen as the microprocessor for 4. Acoustics
higher-level control because of the high number of We have used three Teledyne Reson TC4013
PWM, digital, and analog pinouts. This PCB acts as hydrophones placed in a right isosceles triangle
a shield for the Arduino mega; It has pinouts for the with a side length of 27.2 cm. The acoustic stack
IMU, Pressure sensor, current sensor, servo motors consists of two boards, namely the data acquisition
and ESCs. The processed sensor data from the board and the digital signal processor. We have used
Arduino mega is relayed to the CPU using the Maxim’s MAX11043ATL Evaluation Kit, which
Rosserial protocol. Current sensor ACS712 has along with the simultaneous sampling of ADCs, has
been implemented to monitor the current drawn by an integrated programmable amplifier, anti-aliasing
the gripper servo motor. By monitoring this current filter, and digital filters. The data acquisition
it can be determined whether the gripper has board’s data is sent to the digital signal processor
successfully grabbed the object or not. All the via the SPI communication protocol. We have used
grippers implemented in our previous ROVs and the Texas Instruments LAUNCHXL-F28379D
AUVs did not involve this technique. Using this microcontroller as our DSP to lay off the strain
significantly increases the success rate of the from our main CPU. The coordinates of the pinger
gripper involved tasks. are calculated by TDOA and sent to the Intel NUC
Apart from the current sensors, Voltage sensors and via USB.
temperature sensors can also be attached to this The hydrophones have been carefully placed away
PCB which find their use during the debugging from the thrusters to reduce the effect of noise in the
phase. It was noticed in the previous vehicles that input signal captured by the hydrophones. The
wiring increased debugging periods—developing placement was done keeping in mind the
the microprocessor PCB greatly improved this dimensions of the hydrophones and symmetry. The
situation. arrangement has been kept such as to maximise the
value of receiving time difference which makes the
algorithm easier to compute and estimate values.
 DTU-AUV 7

Figure 2.10 Hydrophone placement

Figure 2.12 Estimating pinger

C. SOFTWARE DEPARTMENT

1. Underwater Image Enhancement

Recently we started working on underwater image
colour enhancement techniques to improve the
image quality for the object detection algorithms.
We implemented an algorithm based on the
Underwater Single Image Dehazing [1] method. To
check the performance of the method used, we used
the video data from the 2018 tasks.
Figure 2.11 Hypothetical arrangement

We have used an approximate model of 2D TDOA

localisation, using 3 hydrophones neglecting the
Z-plane delay, which gives us substantially reduced
complexity in exchange for some loss in accuracy.
In order to compensate for the lost accuracy, we will
be using the camera modules considering the fact
that the pinger tasks include the presence of visually
distinguishable objects present which will be easier
to detect using cameras.
The algorithm computes the time difference of
arrival of multiple nodes wrt 1 node to form a
hyperbolic line of positioning to find the unique
solution of the pinger. This estimation is done
recursively for the auv to move in the direction of Figure 2.13 Result of image enhancement on path
the calculated location. marker
 DTU-AUV 8

2. Detecting the Targets 4. Simulations

We would use YOLO v3 real-time object detection, We used Gazebo simulator along with the UUV
along with colour thresholding (for segmenting package to simulate our AUV in the underwater
parts from images) and contour detection (to detect environment. Gazebo provides plugins for various
borders of objects and localize them) for detecting sensors like camera, IMU, Pressure sensor, etc.
the objects. We calculated the center of the using which we were able to mimic the bot’s
bounding rectangle around the object for alignment. sensors. Having a simulation environment setup in
When rotationally aligned with the target, we used Gazebo helped us to test our controls and gain
the property, the bounding rectangle will be a enough confidence in our algorithms that they will
perfect rectangle instead of a trapezium to control perform well in the competition environment.
the AUV. However, to implement the vision
algorithms in Gazebo, we used OpenCV based
object detection techniques. These techniques
sufficed to test the vehicle’s overall control;
however, in a real-world scenario, using the YOLO
object detector would help us improve our accuracy
to a great extent.

Figure 2.16 Gazebo scene setup for the gate task

5. Mission Planner
We used Smach as a Mission planner to divide a
given task into a structure of subtasks. It allows us
to compose and coordinate complex robot
Figure 2.14 Identifying and classifying the targets capabilities to robust robot applications. The nodes
in torpedo task in each task are operated through messages
published on different topics. The subtask to be
3. Distance Estimation performed next is determined by the message from
In order to accomplish certain tasks, like the buoy the current subtask or the subtask just executed. The
task, it was necessary to determine the distance current task, if executed successfully, publishes
between the camera and the target. So to do this, the messages that lead the robot further to another task
concept of “Triangle Similarity” was used. The in the task structure. In case any of the tasks fails,
performance of the algorithm was tested inside the corresponding defined action is taken.
Gazebo, and it gave us fairly good estimates of the
distance to the target.
III. EXPERIMENTAL RESULTS
Since we had very limited access to the pool
because of this ongoing pandemic situation,
Gazebo based simulations helped us to draw
some insights about our chances of succeeding
in the tasks. We had also planned on using
YOLOv3 based object detection in the real
Figure 2.15 Estimating the distance to the target.
 DTU-AUV 9

situation, however because of the ongoing ● Since we have tested the torpedo in
situation we could not train a custom YOLO underwater environments previously and
detector in a remote environment. But since we based on that we have improved the
have used this algorithm in some of our torpedo design, we are predicting a
previous vehicles, we are pretty confident about success rate of 70% while shooting
its performance. So for testing in simulations we through the bigger opening and a success
have used OpenCV based object detection. rate of 60% while shooting through the
In order to make some valid predictions about smaller opening since we haven't tested
our success rates it was necessary to model the our improve torpedo design underwater.
vehicle as well as the environment as accurately ● We have tested the gripper mechanism
as possible. To test the robustness of the previously, but since the margin of error
algorithms in real time scenarios it was in the task is significant, our confidence
important to perform simulations in as different of successfully completing the task is
environments as possible. Also, simulating the about 40%.
sensor noises was a crucial part of this stress
testing process. Dropper Testing:
The team successfully tested the detect and drop
Based on the simulations and various tests code for the vehicle using a prototype model of
performed, as well as from our previous the dropper in the initial phase of testing.
experiences, we can conclude that our chances
of successfully completing the tasks is as
follows:
● For the gate task along with style it
would be approximately 95%,
● For the buoy task as well as the path
markers, we are estimating a success rate
of 100%, since we have performed these
tasks previously and have simulated Figure 3.1 Successful dropper testing
them also.
● We have also tested our marker droppers
in underwater environments during past Power Connector for Underwater
competitions as well as simulated it on Application
Gazebo, hence we are confident about Because of the high hydrostatic pressure and
completing the task without lifting the corrosive nature of the marine environment, it is
lid with 80% success rate. difficult to secure and reliable electrical
● We are quite confident with our connections. IP67 and IP68 connectors have to
acoustics navigation system and predict be used for transmitting electrical signal and
an 80% chance of successfully detecting power. Therefore, the reliability of the whole
the pinger to a sufficient distance such system during deployment, lifetime and
that the image recognition system can maintenance becomes dependent on its
take over. connector. For quick testing, we designed an
 DTU-AUV 10

underwater connector which allowed us to make

waterproof connections for quickly
disconnecting the vehicle.

Fabrication. Metal pins or the bullet connectors

were soldered on a custom-designed PCB and
assembled with the rest of the body.
Marine-grade epoxy resin was used to seal the
bullet connectors inside the plastic body. The
issue with 3D printing a waterproof connector is Figure 3.3: Prototype claw mechanism testing
that since the part is produced by a continuous
stacking of polymer layers, the resultant parts
IV. ACKNOWLEDGEMENTS
are highly porous. The connector body was
post-processed to tackle the issue by treating it DTU AUV would like to thank and express
gratitude towards everyone who has come together
with Acetone (C3H6O) vapours.
and helped us to make the project a huge success.
We’d like to express our sincerest thanks to our
faculty advisor Dr Ajeet Kumar and his guidance on
the design considerations. Moreover, the team
wouldn’t have been able to function without its cold
sponsors who generously funded us and helped us
acquire all the resources that were required during
the many stages of the project. Finally, we would
like to thank the knowledgeable and resourceful
network of DTU AUV alumni who made
themselves available for advising and sharing their
Figure 3.2 Watertight connector design and valuable experiences and helping us circumvent
analysis multiple roadblocks that one faces while designing
a vehicle.
Gripper testing:

The successful testing of the claw mechanism

was done, later on we optimised the mechanism
for better gripping design and reduced the claw
weight, thus improving the overall performance.
We plan to test the force sensing using the
mapped force values with the current values in
the coming days.
 DTU-AUV 11

V. REFERENCES
[1] N. Carlevaris-Bianco, A. Mohan, and R. M.
Eustice. Initial results in underwater single image
dehazing. In Proc. MTS/IEEE OCEANS, 2010.
[2] Natu, Aditya, Vipul Garg, Purnyatre Gaur,
Piyush Raj, Dhruv, Shivam Sain, Upasana Biswas,
Deepanshu Bansal et al. "Design and Development
of an Autonomous Underwater Vehicle VARUNA
2.0."
[3] Wang, G., Zhou, J.Q., Zhao, Y.T., Feng, J. and
Sun, Q., 2021. Performance Degradation Modeling
and Reliability Analysis of Aviation Connectors in
Marine Environment. In IOP Conference Series:
Materials Science and Engineering (Vol. 1043, No.
3, p. 032029). IOP Publishing.
[4] M. M. M. Manh ̃aes and S. A. Scherer and M.
Voss and L. R. Douat and T. Rauschenbach, UUV
Simulator: A
Gazebo-based package for underwater intervention
and multi-robot simulation 2016.
 DTU-AUV 12

APPENDIX A: COMPONENT SPECIFICATIONS

Component Vendor Model/Type Specs Cost

Buoyancy Control NA NA NA NA

Frame In House Plane truss structure 4mm Aluminium 6061 T6 sheets $150
3D printed attachments 3D Printed ABS $81.20

Waterproof Housing Hammond 1554YA2GYSL1554C2 IP68 rated, Polycarbonate $108

Hammond GYCL IP68 rated, Polycarbonate $27
Polycase AN-07F-01 IP68 rated, Aluminium $33.69

Waterproof Connectors Blue robotics Potted Cable penetrator Quantity: 20,M-10 thread, 6mm $100
cable

Thrusters Blue robotics T200 Quantity: 8,Operating $1432

Voltage:7-20V, Full Throttle
Current(@20V):32A

Motor Control Blue ESC for T200 Quantity:8,7-26 volts (2-6S) $216
Robotics

High-Level Control Arduino Mega 2560 6-20V,54 digital I/0 pins,16 analog $24.30
pins, Clock speed:16MHz

Actuators Hitec Servo HS5646WP IP67, Torque:11.3-12.9 sponsored

Kg-cm^2/s^2

Battery Tattu Tattu 14.8V 25C 4S 14.8V 25C 4S 10000mAh $350

10000mAh Lipo
Battery Pack

Regulator I Robu LM2596S Quantity:2,Input $1.60

Voltage:3-40V,Output
Voltage:1.5-35V(Adjustable)

Regulator II Mini-box DC-DC NUC 12V or 19V selectable output, 6A $59.5

peak

CPU Intel NUC7i7BNH 12-19 VDC,3.50 GHz,1.21Kg $300

Internal Comm Network N/A N/A N/A N/A

External Comm Interface N/A N/A N/A N/A

Inertial Measurement Unit Bosch BNO055 9 Axis, supply voltage:2.4-3.6V, $10.70

(IMU) Range:125-2000 deg/s

Hydrophones Teledyne Reson TC-4013 Quantity:3,1 Hz-180KHz $538x3

,depth:700m

Acoustics I Maxim MAX11043ATL 5V, 4-channel, 16-bit, $157.50

simultaneous-sampling ADCs
 DTU-AUV 13

Acoustics II Texas LAUNCHXL-F28379D 200 MHz dual C28xCPUs and dual $33.80
Instruments CLAs

Vision Logitech C930E Full 1080p HD video at 30 frames $ 130

per second

Pressure sensor Blue BAR30 300m depth $72

Robotics

Current sensor Robu ACS712 Measure Current Range: -30A ~ $1.90

30A

Voltage sensor Robu Voltage detection sensor Voltage detection range: 0.02445V $1.10
module - 25V DC

Manipulator In House Gripper 3D Printed ABS $24.33

Developed Dropper $13.10
Torpedo Launcher $15.73

Algorithms: vision N/A OpenCV, YOLOv3 Underwater Single Image N/A

Dehazing, YOLO Object
Detection, Distance Estimation
using triangulation, ORB

Algorithms: acoustics N/A TDOA Hyperbolic positioning N/A

Algorithms: localization N/A OpenCV Localisation using vision N/A

and algorithms
mapping

Algorithms: autonomy N/A Smach Finite State Machines N/A

Open-source software OpenCV, YOLOv3, Robot Operating System, Smach, Gazebo, Python, C++

Team Size 29
(number of people)

Expertise ratio 19/10

(hardware vs. software)

Testing time: simulation 170 hours

Testing time: in-water 4 hours

Inter-vehicle N/A
communication

Programming Language(s) C++, Python

 DTU-AUV 14

basics of modern AI, as well as some of the

representative applications of Artificial
APPENDIX B : OUTREACH ACTIVITIES Intelligence, were covered.

Each year DTU AUV attends and organizes many The team believes in spreading awareness about
workshops to reach out to the community. Through what advancements are taking place in the fields of
these exhibitions, DTU-AUV tries to reach young science and technology. Every year we conduct a
and insightful minds to take up robotics. number of school workshops to promote such
activities and also give school students a sight of
exactly what kind of environment they face when
they go for higher studies.

Presenting our vehicle to Hon’ble Chief

Minister:- Arvind Kejriwal

The team organised a Machine Learning and

Artificial Intelligence Workshop. The program
implemented a well-researched plan to spread AI
Workshop organised for school students
talent at global tech hubs for applications in
practical decision-making. In this workshop, the

ML and AI workshop organised by DTU-AUV

Tricky circuit competition organised by DTU

AUV
 DTU-AUV 15

We have organised workshops in collaboration

with NGOs to provide an opportunity to students
with limited means to develop a knack for science,
enhance their problem solving abilities, analytical
thinking and practical skills to support their
intellectual growth.

Workshop organised for students with limited

means

Workshop organised in RP public school , Rohini