IITB Mars Rover Team

System Acceptance Review

University Rover Challenge - 2024

Team Leaders
Khush Jain
[email protected]
+91 93284 41633
Harsh Shah
[email protected]
+91 94094 75801
Gaurav Rathod
[email protected]
+91 86697 23750
 I. Robotic Arm II. Wheels and Chassis
Apart from the gripper jaw motion, the robotic The load-bearing capacity, traction, ability to
arm boasts six degrees of freedom. The end effector overcome obstacles of over 20cm height, and shock
workspace spans a height of 1.5m above the ground, absorption were considered for wheel and grouser
and the base rotation enables a full 360◦ rotation design. The stainless steel wheel rim and spoke hub
about the vertical axis. The robotic arm has been were laser-cut, rolled, and welded with the hub to
form the core load-bearing structure of wheel. The
spoke’s design ensure compliance with shock ab-
sorption in vertical direction, and stiffness in lateral
and longitudinal direction. A planetary gearbox in
introduced this year in the motor-hub which serves
the purpose of reducing the stress on motor shaft
and simultaneously increasing motor torque.

(a) Robotic Arm (b) Laser gripper with

lead screw mechanism
Fig. 1: Robotic Arm

enhanced this year by incorporating a customized

cycloidal gearbox 2 into the base rotation, providing
a higher reduction ratio of 56:1 in a compact space
(a) Wheel Hub (b) Convex (c) Wavy Grousers
while minimizing backlash. The newly developed
Grousers
gripper, utilizing a lead screw mechanism enables Fig. 3: Wheels Assembly
application of substantial clamping force without
backdrive. Its precision claws facilitate meticulous
movements, particularly in tasks like equipment
servicing, and simplify interactions with panels fea-
turing knobs, switches, and buttons. Furthermore,
we’ve integrated a mini linear actuator with a laser
pointer to ensure accurate keyboard typing and bolt
fastening. Fig. 4: Motor Hub
The arm’s shoulder consists of parallel aluminum In this iteration, rubber is replaced with wave-
plates, folded at the edges to enhance stiffness. For shaped TPU grousers in order to reduce wear and
lifting heavy objects, the elbow is crafted from a provide more shock absorption. The wave pattern
carbon fiber tube, ensuring robust strength. The helps in reducing deviations in axis of rotation
components were simulated using ANSYS Static during skid steering. Further, the grouser shape
Structural which allowed the team to choose the is being re-iterated to a convex shape to improve
ideal materials for the arm while keeping in mind straight line drive performance and overcome small
the required properties. With a lifting capacity obstacles while skid steering by reducing lateral
exceeding 7kg and an 8cm wide gripping span, the stress. The radius, thickness and number of grousers
arm effortlessly handles heavy loads and securely was optimized to improve traction, and suspension
grasps objects during delivery missions. kinematics.
The chassis framework employs a novel layered
design with sectioned regions for the various subsys-
tems. The primary structural members involved are
square aluminum alloy pipes. The base structure of
the chassis consists of the main frame with provisions
for anterior coupling to the bio-frame and the arm-
frame. The central part of the main frame houses the
electrical box. The updated Elec-Box in our rover is
fully concealed to avoid dust inflow, with materials
Fig. 2: Customized cycloidal gearbox - exploded view mitigating Faraday caging. Quick-remove panels at

1
the front and top end have been added to aid with have been done to the areas of weakness. The rocker-
accessibility. ANSYS static structural simulations bogie assembly has a dynamic load factor of 2 and a
were performed on the chassis frame to optimize safety factor of 1.5 under extreme load conditions.
the dimensions before final fabrication. Maximum
deformation and von-Mises stress were analyzed to IV. Bio-assembly
have safe values of 10µm and 5MPa, respectively. 1) Surface soil collection
Utilizing a three-claw gripper with integrated
III. Suspension
sensors, this setup ensures soil sample purity (by
With a reduced span of 1m (end to end) and a eliminating cross-contamination) during deposition
ground clearance of 300mm, the six-wheeled rocker- into the biobox. A transportation unit divides sam-
bogie suspension assembly has been fabricated with ples into test tubes via a funnel. Within the testing
structural members of hollow 200gsm bi-directional module, designated reagent-filled test tubes await
carbon fibre square tubes for optimal stress distribu- analysis, supported by continuous visual data from
tion. The suspension system’s weight was effectively USB cameras. To maintain a sample temperature of
reduced while preserving strength by transitioning at least 100°C for the Diphenylamine test, Peltier
from hollow aluminum pipes to carbon fiber since, modules and temperature sensors regulate a minia-
carbon fiber offers 3 times more strength than an ture water bath.
aluminum alloy for same weight. The novel four-
bar linkage mechanism implements a differential
between two rockers which minimizes the tilt of
the mainframe during obstacle climbing and allows
transverse slope traversal. The complete suspension
weighs 4 kg.

(a) Three-claw gripper (b) Drill

Fig. 6: Surface & Sub-Surface Collection Mechanism
2) Sub-surface soil collection
A novel drill mechanism featuring a high-torque
gearbox has been designed, enabling drilling to
depths of more than 10cm into the soil. This mech-
anism achieves constant translation and rotational
motion through a driver motor and lead screw
Fig. 5: Full Rover Assembly arrangement. Once the desired depth is attained,
The chassis and electrical box designs are com- a servo motor facilitates precise rotation of the
pliant with the suspension to set the center of inner collecting cylindrical unit which allows for the
mass approximately at the center of the rover, with storage of uncontaminated soil sample. Currently
the center of mass having a sufficient clearance of the setup has been fabricated with PLA, but are
300mm from the ground. Efficient mass distribution in process to manufacture it with aluminum sheet
of the subsystem components in the chassis allows metal.
the rover to climb and descend a maximum of 3) Visual Rock Analysis
54◦ slope, and traverse a slope of up to 38.10◦ The visual analysis pipeline includes super seg-
sideways. With the diameter of the wheels being mentation of rock images using LSC (Linear Spec-
200mm, the obstacle climbing capacity of the rover tral Clustering, [1]), for close inspection of specific
is around 300mm which is equal to the bogie link clusters. The Region-of-Interest (ROI) in the image
length. The suspension parameters were determined is passed through neural super resolution (CARN,
after repeated motion analysis on various extreme [2]) for enhancement of the textures. The textures
terrains. The kinematics and compliance of the are then compared with a database curated for map-
simulation were compared with the field trials to ping the textures to possible stratigraphic profiles.
verify the suspension model. Critical locations that For documenting the site, a wide angle panorama
lead to failure were correctly identified using ANSYS is taken using the mounted Pan-Tilt-Zoom (PTZ)
and MATLAB simulations, and design modifications cameras and the cardinal directions are fetched from
the INS sensor.
2
 V. Electronics and Controls This year, Resolved-rate Motion Control (RRMC)
has been implemented for providing the rover oper-
ator an alternate control of the arm, allowing the
controller to vary the end-effector in cartesian space
rather than joint space.
4) Base-station Control
A ROS-interfaced dual-shock analog joystick [5]
and a mono-analog joystick [6] control are used
for controlling the rover. Additionally a Django-
based GUI is developed to integrate video streams
from 3 Pan-Tilt-Zoom (PTZ) camera. The GUI also
provides visualization of data from bio-sensors for
soil analysis.

B. Communications
Presently, the communication system operates
within the 5 GHz frequency band, enabling signifi-
cantly higher data transfer rates and hence allowing
minimal response latency. Our setup utilizes Mi-
mosa c5c point-to-point bridges in conjunction with
Ubiquiti omnidirectional antenna on the rover to
establish a robust connection between the rover and
the base station. We have also integrated a sector
Fig. 7: Electrical Components Flowchart antenna on the base station, which tracks the rover,
ensuring a consistently solid and secure connection
The Electronics & Controls subsystem has focused (compliant with FCC UNII-1 regulations) over a
on developing a compact and reliable electrical stack. non-line-of-sight range of 950 throughout our oper-
Improving the communication stack is imperative ations. The most suitable frequency is decided band
to ensure uninterrupted rover operability in each by employing a link metric considering RSSI and
mission. throughput values. Additionally, the C5c supports
a wide frequency range (4.9-6.4 GHz), which allows
A. Controls and Locomotion avoidance of crowded 5 GHz spectrum bands [7]. It
1) Hardware Present also restricts channel widths to 20MHz.
The rover utilizes a combination of DC motors, For rover maneuverability, three IMOU cameras
servos, & linear actuators. To interface DC motors & (PTZ cameras), Intel Realsense D435 [8] and web-
linear actuators, custom-made PCBs interfaced with cam for gripper are used to create multiple video
RoboClaw [3] motor drivers, which have high current streams. IMOU cameras provide wide angle view
ratings (45 A), built-in encoder & potentiometer with 0-355° pan & -5-80°tilt [9]. The overall network
support are employed while an ESP32 is used to is bifurcated with an onboard LAN and a point-
control the servo. We use the Intel NUC [4] to control to-point network for long-range communication (Fig
all the components onboard via ROS. 7). A Netgear Router [10] provides easy scalability
2) Rover Mobility in onboard cameras, enabling multiple views for ease
Each of the 6 Brushed DC motors are equipped of operation.
with encoders, which are used for closed-loop velocity C. Power Distribution & BMS
control and provide odometry feedback to the base The rover can operate for 2̃ hours on a single
station for improved operability. charge with the support of four 12 Volt Li-Ion
3) Arm mobility battery packs, each boasting a maximum current of
The robotic arm is actuated via three linear 105A and an energy capacity of 35Ah, all equipped
actuators, three DC motors, & one servo motor. The with built-in BMS. The power system relies on
problem of non-linear control due to combination regulated DC-DC USB NUC Converter [11] and
of different actuators is circumvented by placing a DC-DC buck/boost converters. To ensure efficient
linear limit on the movement of these joints. wiring, we utilize high-quality silicone wires [12] in

3
12, 14, and 18 AWG standards, along with aviation- mounted on a rotating platform, controlled by PID,
grade GX16 connectors and T-Tap terminals. expands the rover’s field of vision while minimizing
D. Bio-Assembly turning while performing search. Additionally, veloc-
A Raspberry Pi oversees a diverse array of de- ities provided by the planners undergo smoothing to
vices, including relays, temperature sensors, heating maintain stable motion and avoid abrupt changes.
modules, and lasers [13]. Additionally, an ESP32
has been incorporated to manage stepper and servo
motor controls. Currently, the drilling mechanism,
which collects soil from a depth of 10 cm, is driven
by a stepper motor and a DC motor for smooth
operation. We have used a relay in addition to the
GX16 pin for efficient power distribution to bio-
setup, including heating elements. Webcams inside (a) Stereo camera (b) Quad Search for AR
bio-assembly provide video feed to observe ongoing on rotating tag
experiments and test results. platform
VI. Autonomous Navigation Fig. 8: Search for Objects/AR Tag

For precise autonomous navigation, we utilize B. Object Detection

various sensors/devices, including the YDLiDAR X4 To detect the specified objects (plastic bottles and
Pro for 2-D occupancy grid, the Intel Realsense D435 mallets), the navigation pipeline has been augmented
stereo camera, the SBG Ellipse-D RTK INS sensor with yolov8 object detection algorithm for real-
for IMU and GPS, and Wheel Odometry. The col- time operation. Before passing the live video feed
lected data undergoes noise reduction and sensor fu- to the algorithm, the images are passed through
sion using Extended Kalman Filter, enhancing accu- an anomaly detector ( [15]) for checking outliers.
racy for localization. The sensor fusion enhances the This is done to avoid false prediction on Out-Of-
alignment of point cloud data, resulting in reduced Distribution inference images, hence reducing false
error rates and mitigating visual odometry drift. positive rates.
Real-Time Appearance-Based Mapping (RTABMap) VII. STEM Outreach
aids in loop closure detection for SLAM, while Visual The business division of the team participates
Odometry from Intel RealSense further refines local- in international exhibitions, and also hosts educa-
ization. The resulting map supports path planning, tional outreach campaigns, geared towards spreading
navigation, obstacle avoidance, and optimized rover space tech awareness among high school students,
traversal. especially in girl students, and those belonging to
A. Path Planning underprivileged background.
We utilize the Dynamic Window Approach (DWA)
for local path planning and the A* algorithm for
global path planning. DWA considers parameters
such as the goal location, rover velocity, and obstacle
distance, restricting velocities to ensure safe stops
within reachable intervals. This approach strikes a
balance between speed towards the goal and obstacle
avoidance, thereby ensuring robust collision preven-
tion. Goals are strategically assigned to facilitate Fig. 9: Educational Outreach Campaigns
efficient tag search, involving a rotating maneuver VIII. Budget and Project Management
around GPS coordinates following a quad-spinning
strategy. The system is designed to manage various The project has been funded by the institute as
failure scenarios. In case of stereo camera failure, a Student Technical Project (STP) with aprooved
we have implemented monocular SLAM using DSO budget of 18101.52 USD. Apart from this, the funds
(Direct-Sparse Odometry) [14], as a fallback mech- from Dean ACR are in process and will be sanctioned
anism. In case of LiDAR sensor failure, we obtain post-SAR Results for travel and logistics purposes.
2D Laser-scan using stereo camera data. To enhance Table I contains the breakdown of the expenses of
energy efficiency and save time, a stereo camera the funds for technical and logistics purposes.

4
 IX. Project Plan & Logistics

Fig. 10: Gantt Chart (Sept’23 - May’24)

TABLE I: Detailed Breakdown of Funding and Expenditure (All amounts in USD)

INCOME EXPENSES
Income Sources Amount Project Expense Category Amount
1. Industrial Research & Consultancy Center 18101.52 Core Rover Development 11603.88
Amount Received Till Date 12103.88 Rover onboard computer & Regulator 1110.23
Motors, actuators, & drivers 2232.52
Prototype & Fabrication : 3D printing 820.60
Prototype & Fabrication : Metal 4327.07
Chemical reagents and apparatus 241.35
Cameras & Lidar 1580.87
PCB, Comms, & Batteries 1291.24
Amount Anticipated 5997.64 Inventory Equipments & Spares 1248.20
Misc. electrical instruments 422.37
Mechanical tools 181.02
Spare parts & Equipments 500
Lander and test stand 144.81
Travel & Logistics 16864.75
2. Dean ACR & Institute of Eminence Cell 12067.68 Rover shipping to USA 3620.30
Amount Received Till Date 0 Batteries purchase in USA 1206.77
Amount Anticipated 12067.68 Travel to and from USA 12037.68
TOTAL INCOME 30169.2 TOTAL EXPENSE 29716.83
In-Kind Sponsorships Funds Surplus 452.37
Solidworks & Ansys Software License
Network-Orbiter & Ruckus Wireless Communications Module
SBG Systems Navigation Module

5
 X. Science Mission Assembly B. Bio-sensors
In the assay of biosensors, we have used SHT10
A. Experiments Performed temperature and humidity. Further, we are in the
process of integrating MQ-4 methane sensor with the
In this iteration we have build upon our pre- new tri-claw gripper. As a indicator to assess pres-
vious science mission experiment assays and have ence of specific microbes, temperature and humidity
enhanced the simplicity, accuracy, and robustness sensor is measured. Proteins’ temperature prefer-
of tests. In order to detect any extinct or extant ences help identify potential microbial life forms:
life forms and elements suggesting a favourable Psychrophiles (10-15°C), Psychrotrophs (15-30°C),
environment for life, colorimetric tests are used. Mesophiles (30-45°C), or Thermophiles (50-85°C and
1) Diphenylamine DNA estimation higher). In the situation of detection of methane
The diphenylamine test provides a reliable means on Mars, there are two possible arguments can be
of detecting the presence of DNA and RNA, serving made. At first, climate change causes a release of
as an indicator of life. This method capitalizes on frozen methane remnants from extinct microbes into
the nitrogen-rich composition of DNA’s nitrogenous the atmosphere. Alternatively, live microbial activ-
bases. When DNA’s deoxyribose reacts with acid, ity remains plausible. Geological (abiotic) methane
it generates β -hydroxylevulinaldehyde, which sub- sources result from hydrogen reacting with carbon,
sequently reacts with diphenylamine, resulting in a which are facilitated by specific inorganic catalysts,
blue hue. The procedure involves mixing the sample forming methane. This can be attributed to the
with the reagent, briefly heating it to a minimum presence of methane to subsurface hydrothermal
of 100°C, and then allowing it to cool to room activity.
temperature for observation. The presence of DNA C. Mineral testing
is signaled by a change to blue coloration, whereas
RNA produces a green solution due to the creation We are working on a image-based mineral clas-
of non-deoxy pentose species. [16] [17] sification study. We have generated the dataset
for rock classification and are currently in process
2) Bradford method for protein estimation of developing Convolutional Neural Network based
Proteins, often considered vital components of life, architecture for the classification tasks.
present a promising avenue for life detection and
can be evaluated using Bradford’s method. This D. Results
approach relies on the ability of Coomassie Brilliant Together, data from the mineral/rock analysis and
Blue G-250 (Bradford’s Reagent) to interact with visual feed from the bio-analysis allow us to estimate
the hydrophobic segments of proteins, causing a the viability and presence of life of the collected
change in their absorbance spectrum. This alteration samples.
converts the initially light brown, unbound dye into
a blue-colored complex bound to the protein. Results
are obtained within minutes of the reaction. [18]
3) XTT assay
To analyze metabolic activity within living cells,
the XTT assay, a type of colorimetric analysis
is used. In this test, the yellow tetrazolium
salt XTT (known as 2,3-bis-(2-methoxy-4-nitro-
5-sulfophenyl)-5-[(phenylamino)carbonyl]-2H-
tetrazolium hydroxide) undergoes reduction to
produce an orange-colored formazan product when
exposed to an active electron transport chain in
viable cells. This color change can be observed
with naked eye. To conduct the assay, a sample
from the extraterrestrial environment is incubated
with XTT, and the resulting alteration in formazan
production is observed. [19] [20]
Fig. 11: Bio Testing Pipeline

6
 References
[1] Z. Li and J. Chen, “Superpixel segmentation using
linear spectral clustering,” in 2015 IEEE Conference
on Computer Vision and Pattern Recognition (CVPR),
pp. 1356–1363, 2015.
[2] N. Ahn, B. Kang, and K.-A. Sohn, “Fast, accurate,
and lightweight super-resolution with cascading residual
network,” arXiv preprint arXiv:1803.08664, 2018.
[3] “RoboClaw.” https://www.pololu.com/category/124/
roboclaw-motor-controllers.
[4] “InterNUC.” https://www.intel.in/content/www/in/
en/products/details/nuc.html.
[5] “Logitech F310.” https://www.logitechg.com/en-in/
products/gamepads/f310-gamepad.940-000112.html.
[6] “Logitech Extreme Pro 3D.” https://www.logitechg.
com/en-in/products/space/extreme-3d-pro-joystick.
942-000008.html.
[7]
[8] “Intel Realsense D435.” https://www.intelrealsense.
com/depth-camera-d435.
[9]
[10] “Netgear Nighthawk RAX10 AX1800.” https://www.
netgear.com/in/home/wifi/routers/rax10/.
[11] “DCDC Converter.” https://www.mini-box.com/
DCDC-NUC.
[12] “Silicone wires.” https://robu.in/product/
high-quality-12awg-silicone-wire-1m-red-1m-black/.
[13] “RaspberryPi4B.” https://www.raspberrypi.com/
products/raspberry-pi-4-model-b/.
[14] “Direct sparse odometry.” https://cvg.cit.tum.de/
research/vslam/dso?redirect=1.
[15] I. García-Aguilar, R. M. Luque-Baena, E. Domínguez,
and E. López-Rubio, “Small-scale urban object anomaly
detection using convolutional neural networks with prob-
ability estimation,” Sensors, vol. 23, no. 16, 2023.
[16] K. Burton, “A study of the conditions and mechanism
of the diphenylamine reaction for the colorimetric esti-
mation of deoxyribonucleic acid.,” Biochemical Journal,
vol. 62, pp. 314–323, 1956.
[17] J. Jayaram, Laboratory Manual in Biochemistry. New
Delhi: New Age Internation Ltd., 1981.
[18] M. Bradford, “A rapid and sensitive for the quantitation
of microgram quantitites of protein utilizing the princi-
ple of protein-dye binding.,” Analytical Biochemistry,
vol. 72, pp. 248–254, 1976.
[19] N. Roehm, G. Rodgers, S. Hatfield, and A. Glasebrook,
“An improved colorimetric assay for cell proliferation
and viability utilizing the tetrazolium salt XTT.,” Jour-
nal of Immunological Methods, vol. 142(2), pp. 257–265,
1991.
[20] R. Crawford, A. Paszczynski, and Q. Lang, “Mea-
surement of microbial activity in soil by colorimetric
observation of in situ dye reduction: an approach to
detection of extraterrestrial life.,” BMC Microbiology,
vol. 2(22), 2002.