LAGARI UAV TEAM 2019

Technical Design Paper for AUVSI SUAS Competition

Yildiz Technical University
Faculty of Mechanical Engineering
Mechatronics Engineering Department

Abstract
This paper presents the design and tests processes of the ZD-02 UAV for the AUVSI SUAS 2019 Competition.
Lagari UAV Team has been founded at Yildiz Technical University in 2013. In the past years, Lagari UAV Team
created multiple innovative aeronautical & aviation projects and won various awards. This year will be the third
year in the AUVSI SUAS competition for the team. Lagari UAV Team consists of undergraduate students from
the department of Mechatronics Engineering of Yildiz Technical University. ZD-02 was designed, manufactured,
improved and tested by Lagari UAV Team. The aircraft was manufactured with using advanced composite
technologies to provide maximum reliability and safety during the flight. The avionics were renewed and
integrated to the ZD-02 to complete all missions on the competition autonomously. The team realized that the
future of UAS is guidance, control and navigation systems. With this goal, the team efforts to develop the UAS
with high performance and hi-tech. The system has capabilities such as high-performance image processing,
customized user interface, long distance wireless communication, advanced composite manufacturing based on
ANSYS CFD analysis and customized algorithms for ODCL, Obstacle Avoidance and Air Delivery tasks.

Lagari UAV Team Technical Design Paper - 2019 1/15

 The UAS must avoid obstacles, identify the potential
1 System Engineering Approach drop location and drop the package to a safe location
ZD-02 was designed, developed and tested to and then carry the payload autonomously to the
accomplish all the tasks of AUVSI-SUAS 2019.This customer’s location.
Unmanned Aerial Vehicle(UAV) can fly
With the experiences and the knowledge gained
autonomously, avoid obstacles, detect objects and
from last years, the team focused on the inspection
drop a payload safely, accurately and efficiently.
of the weaknesses of 2018’s performance and gave
This section includes the mission requirements, the importance to adjust them. The team also spent time
design rationale, and a programmatic risk analysis. to analyze carefully the tasks importance,
requirements and the action to be executed to
1.1. Mission Requirements Analysis accomplish them successfully. Mission focused sub-
systems were designed to perform efficiently. Table
This year the competition simulates a Unmanned 1 shows our mission strategy.
Aerial System(UAS) tasked by a delivery company.

Used Sub-
Mission Explanation Mission Execution Requirements Outcomes
System
• The team train to
assemble the aircraft
• Complete this stage in minimum • Finish the mission as fast as
in less than 10
time (80%) possible
minutes
Timeline (%10) • Needs to be completed under 40 • Well prepared and organized team UAS
• Fully mission is
minutes • Simulate the full mission as much
scheduled to be
• Avoid from take a penalty (20%) as possible before the competition
completed into 20
minutes
• Fly autonomously over • High maneuverability Control • A new CB was
waypoints (40%) • Well-calibrated equipment (Air Box(CB) & designed and
Autonomous Flight • Refrain from safety takeovers Speed sensor, Compass, Ground manufactured
(%20) (10%) Accelerometer) Control • More sensitive
• Capture the waypoints as close • Investigate the log data after each Station equipment was
as possible (50%) flight (GCS) installed
• The algorithm was
• Developing a new algorithm for CB developed depending
Obstacle Avoidance • Avoid stationary, cylindrical
obstacles and test it & the UAV’s
(%20) obstacles
• Have a good maneuverability GCS maneuverability
competence
• A new PB was
• Object characteristics must be • Have a better angle of vision designed and
determined correctly (20%) • Use a good imaging system Payload manufactured
Object Detection, • Localization of detected object • Developing a deep-learning Box(PB), • The Imaging
Classification, (30%) algorithm for detecting the objects Imaging system was
Localization (%20) • Sending files during flight time autonomously System & redesigned including
(30%) • A new algorithm for geolocation GCS a gimball with a new
• Autonomous detection (20%) calculation more competent
camera
• Drop safely a Unmanned Ground
• UGV was
Vehicle(UGV) to the drop • Well-qualified UGV and
Air manufactured from
location accurately as possible mechanism
Delivery carbon-fiber
Airdrop (%20) (50%) • Decrease the falling speed of the
mechanism • A sufficient
• Drive the UGV with his 8 oz UGV with a parachute
& UGV parachute was
water bottle to a specific • Give attention to calculation
manufactured
destination (50%)
• The team uses
Asana[1] to give
• Well-defined personal Check List
• Professionalism, communication tasks
Operational responsibilities &
between members, reaction to • During test flights
Excellence (%10) • Usage of walkie-talkie between Communic
system failures the sub-teams
the GCS and assembly team ation
communicate with
each other
Table 1: Mission Table

Lagari UAV Team Technical Design Paper - 2019 2/15

1.2 Design Rationale problems. Studies were practiced on a safer
autonomous take-off and landing.
In the entire design and development process of ZD-
02; all equipment’s were selected to maximize the Obstacle avoidance, the algorithm was tested on
task performance considering the team’s budget, Software in the loop(SITL) and on flights.
capability and past accumulations. Since the team
ODCL, a new gimbal system and plexiglass dome
needs sponsors throughout the pre-competition
was designed to get a wider area coverage without
process. Budget is an important factor that limits us
any problems of vibration. A new camera was
during the creation of the system.
bought according to the calculations in Section
Due to limitations of the sponsors in this year, the 2.4.1. More practices were given on detections and
team decided to go with the same UAV frame from tests were conducted.
last year. This had several advantages and
According to the new air delivery mission, a new
disadvantages.
design was made. Also some tweaks were
Advantages: implemented on the drop algorithm.

• More time was spent on algorithms

2 System Design
• The budget was shifted to new systems
• Manufacturing new systems had more 2.1 Aircraft
priority
This section contains subsections of the design and
Disadvantages: development on the airframe, surfaces and the
propulsion system.
• The UAS system was taken into
maintenance 2.1.1 Airframe and Surfaces
• Changes on new systems were harder to • Fuselage
implement
• Shorter flight time due to extra weight The fuselage is a significant part of the plane, its
shape is designed in accordance with the missions of
The team worked to find solutions to the identified the competition.
problems within the possibilities. The team’s
capabilities, budget and acquired experience played The size of the fuselage was determined by taking
an important role in bringing these solutions. Fig. 1 into consideration that it must contain all the avionic
shows the decisions that were followed during the boxes and batteries. All the components, including
design processes the wings, avionic boxes and the batteries, were
mounted in a key-lock match. As a result of which,
the components remain stable during flight, while at
the same time it is easier to detach and mount them.

The tendency to use carbon fiber in aerospace

industry results from its high-tension strength,
chemical resistance and stiffness also low thermal
expansion and weight properties. Because the
fuselage of ZD-01, which was previous year’s UAV,
already meets the requirement and also the cost of
manufacturing was high. Last years fuselage was
used.
Figure 1: Flow diagram of decisions
• Wings
Firstly, last year’s problems were identified and
solutions were done according to the tasks required. Wings are the parts of the plane that enable flight by
providing the lift.
For timeline, same configuration of screw heads was
implemented on the overall UAV to accomplish a Extruded polystyrene(XPS) foam was preferred as
faster mount and dismount time rate. the main material of the wings structure because it is
cheaper than other foams and more ductile than balsa
For autonomous flight, a new Pixhawk was bought
wood. The wing was manufactured from carbon
because of the older one causing us some technical
fiber using vacuum method to obtain the final shape

Lagari UAV Team Technical Design Paper - 2019 3/15

of the wings. The wings were supported by carbon Among many different combinations, the tricycle
fiber pipes as spars extending into the fuselage to landing gear was chosen due to the advantages
increase the strength and durability. The 16% comparing to other configurations[4].
increase in weight this year has caused an increase in
wing loading. The main Landing gear was manufactured from
carbon fiber. Also, two layers of Kevlar were added
A wing loading test was carried out in order to check to make it more elastic and two Rohacells were used
and verify the endurance of the wings and control to make it more durable. This part was analyzed in
surfaces. To achieve this, Schrenk approximation[2] ANSYS with the new weight as shown in Fig. 4.
was used. As a result of which, weights were spread
on the wing surface as shown in Fig. 2.

Figure 4: Landing Gear Stress Analysis

The landing gears are one of the components which
are highly affected by the increasing weight. A 2 G
force test was conducted according to the
FAR23(Federal Aviation Regulation) is shown in
Fig. 5. The component did undergo no plastic
Figure 2: Wing Loading Test deformation as a result.

According to the flight maneuvers, the loads on the

wings were calculated with an 1.2 Factor of
Safety[3]. Tests were concluded up to 2 G
deformation. The graph of force-strain is linear.This
can be regarded as proof of that the result was within
the resilience limits (Fig. 3).

2G Test

Figure 5: 2 G Force Test

One of the significant problems from last year was
related to the nose landing gear. The montage was
difficult and endurance was questionable. For these
reasons, a new design was adopted this year. It is
more durable and whose assembly is easier despite
being heavier Table 2.
Figure 3: Line Chart of Moment-Deformation
Nose
Weight [g] Demount Damper
Gear
• Landing Gear New x2 M3 Up to
193
Design Setscrew 4[kgF]
Landing gear is important for a successful take-off, Previous M5 Nut +
landing and taxi. The landing gear components must 124 Lacking
Design M8 Nut
bear high stress.
Table 2: Comparison of new and previous design

Lagari UAV Team Technical Design Paper - 2019 4/15

A aluminum nose landing keeping the AR between 8-10. UGV and some extra
gear with a suspension (1) payloads hava increased the weight and drag of ZD-
was too short and for this 02. These amendments let the aerodynamic team to
reason, a component (2) test the UAV from scratch.
was manufactured from
Al-6061 which is a According to the feedbacks from the safety pilot, the
lightweight material LAElev [5]distance was reduced from 1250 [mm] to
widely used in aviation. 1150 [mm] to get a more effective tail.
Knife steel (DIN- Lagari has performed a full plane ANSYS fluent
115CrV3) was used to simulation by using Navier-Stokes equation[6] to
assemble the main parts. It check how the velocity and pressure values are
is not only used to prevent distributed around the aircraft (Fig. 7). Total CL/CD
components breaking off (Lift Coefficient to Drag Coeffiecient) ratio was
from sharp edges but also calculated in XFLR5 (Fig. 8). Also, a similar
resistant to wear. As a simulation was made on SimScale to confirm our
result of this process, a results. The following values were obtained:
component which is 50%
lighter than the Figure 6: Nose • Cruise speed has changed from 15,41 to
completely-steel version Gear Design 17.91 [m/s] and stall speed from 11,22 to
was chosen. With the help 12.59 [m/s].
of a 3D printed component (3), one side of this part • Total CL/CD = 21.6
is square and the other side is compatible with head • Taper ratio is 1.5
of the servo. A turning movement is transferred from
motor to the nose landing gear. The whole
mechanism is fixed to the housing with a roller
bearing (4) Fig. 6.

2.1.2 Aerodynamics
Lagari used asymmetric Eppler 214 airfoil for main
wing to maintain necessary lift. NACA 0012 was
used for horizontal and vertical stabilizer. To
improve maneuverability, the Aspect Ratio(AR) was
determined in accordance with the optimum AR
range of sailplane aircraft models. The tapered Figure 7: General destitution of velocity
trapezoidal wings platform was preferred for

Figure 8: XFLR5 VLM analysis

Lagari UAV Team Technical Design Paper - 2019 5/15

 Figure 9: Top and Side view of ZD-01

GENERAL CHARACTERISTICS PERFORMANCE MAIN WING

24.47 [Kts] (12.59
Length 1.95 [m] Stall Speed Airfoil E214
m/s)
34.81 [Kts] (17.91
Total Wingspan 2.80 [m] Cruise Speed Span 2.80 [m]
m/s)
Empty Weight 3.50 [kg] Max Speed 48 [Kts] (25 m/s) Area 0.81 [m²]
Wing Aspect
MTOW 11 [kg] 13.686 [kg/m²] 8.9
Loading Ratio
HORIZONTAL
Rate of Climb 250 [m/min] VERTICAL STABILIZER
STABILIZER
Minimal Turn NACA
20 [m] Airfoil NACA 0012 Airfoil
Radius 0012
Maximum Flight Approximately
Span 0.30 [m] Span 0.40 [m]
Autonomously 18 [minutes]
Operational
4 [km] Area 0.07 [m²] Area 0.20 [m²]
Range
Operational Aspect
450 [m] Aspect Ratio 2.6 3.2
Ceiling Ratio
Table 3: General Features

2.1.3 Power and Propulsion System ZD-01. This system provided cruise speed with
This year, Lagari team checked previous year’s DC approximately 55% throttle in last year. According
brushless Motor AXI 5320/18, whether it is enough to AIRCRAFT DESIGN: A Conceptual Approach
book[4], for ZD-02 type plane, trust to weight ratio
to handle the new design. This motor was selected
[T/W] value must be in between 3.25 and 3.75. To
last year by using manufacturer datasheet which is
find this ratio, thrust test was conducted on AXI
shown in Table 4. 5320/18. Results of the test is shown in Fig. 10.
Weight Motor consume approximately 38 [A] for 55%
Ampere(max)
KV Propeller of throttle. According to Fig. 10, this ampere value is
Battery
Model corresponding to 3.261[kg]. As a result of this test,
370 55A/6S Li-Po 18"x10" 9000 [g] [T/W] value was 3.174 for previous year. This year,
Table 4: Properties of Motor according to system gained 1.6 [kg] and [T/W] ratio rose to 3.44
manufacturer for new system. This value is still in acceptable
interval.
Manufacturer supplies this information by making
educated guess based on the knowledge about
aircraft models. This motor was enough to handle

Lagari UAV Team Technical Design Paper - 2019 6/15

 CB has been designed to prevent the cable
complexity and easy debugging. Flight control
system is combined in CB for more stable and simple
interior design. In this year, aluminium 661 series
was used for CB instead of plexiglass shown in Fig.
11. Despite of weight increase, aluminium provides
more solid stance and composes Faraday Cage effect
(protect IMU from magnetic field). In addition, fan
was removed from CB because it was drifting the
barometer. The CB includes:

• Pixhawk 2.1 Cube

• RFD900+ Telemetry
Figure 10: Thrust test • PPM Encoder
• Power • Radio Reciever

Same power distribution was used on the ZD-02.

Because similar system was used.
There are three main power units made by Li-Po
cells which each of them has 2280[mAh] and 25[C].
Power sources and properties are shown in Table 5.

Battery Connected to [mAh]

6S/6P Motor 13680
2S/2P Control Surfaces 4560
3S/2P Avionic Boxes 4560
Table 5: Power Supplies
Figure 11: Control Box
In order to determine the flight time, average ampere ArduPlane has been used on Pixhawk because of its
drain in system was used in Equation 1. This useful properties such as Proportional Integral
average value was taken from flight logs. Derivative (PID) controller, L1 flight controller and
6[# 𝑜𝑓 𝑝𝑎𝑟𝑎𝑙𝑙𝑒𝑙 𝑢𝑛𝑖𝑡𝑠] ∗ 2,28[𝐴ℎ] ∗ 60[ 𝑠 𝑡𝑜 𝑚𝑖𝑛] EKF. First few flights were conducted for tune PID
parameters in Autotune flight mode. PID controller
45[𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑠𝑢𝑝𝑝𝑙𝑦 [𝐴] 𝑡𝑜 𝑚𝑜𝑡𝑜𝑟]
provides smooth and stable autonomous flight. PID
Equation 1: Calculations of flight time controller commands and actual values are nearly
According to calculation, our flight time is same after tuning as shown in Fig. 12. L1 controller
18.24[min]. is a guidance algorithm that was used for path
control. Unfortunately, outputs of sensors have high
2.2 Autopilot noise. In this point, EKF is get involved and
eliminate errors for optimal estimation As a result of
In the selection of flight controller, elements such as
that, team achieved 88% accuracy of waypoint
autonomous flight capability, reliability, stability
capture with a maximum 13.5 [m] offset in 5 flight
and existince of reliable documents were taken into
which each of them lasted 15 [min].
consideration. Pixhawk 2.1 Cube can fulfil desired
requirements. Other key features for choosing
Pixhawk 2.1 Cube are:

• Fail-safe backup controller

• Redundant power supply
• Powerful development capability
• Redundant Inertial Measurement Unit
(IMU)
• Redundant internal and external sensors
• Support of two Global Navigation Satellite
System(GNSS) module
• Extended Kalman Filter(EKF)
Figure 12: Autopilot Performance

Lagari UAV Team Technical Design Paper - 2019 7/15

UGV can be reach desired location after dropping. the UAV. With respect to yaw angle of the UAV,
Lagari choose Pixracer(Fig. 13) as a UGV’s two alongside corners can be selected. Through these
autopilot because of its smallness and capabilities. corners, more appropriate path for UAV can be
STM32F4 processor, 6 PWM output, internal created. An example of the mission and the flight
sensors are some of important features of reason for path are shown in Fig. 15 and Fig. 16. In addition
choosing Pixracer. Mission Planer (MP) is using as when algorithm creates new waypoints, they are
a GCS. Connection between UGV and GCS is initialized in 50 meters radius in order to decrease
established via XBee S2B which is small and high- the sharp turning. On the other hand, mission
range telemetry. waypoints radiuses will stay at one meter for the
purpose of maximum waypoint accuracy.
Overshoots as shown in Fig. 16 are accepted because
of the high waypoint capture accuracy. Finally, for
the emergency cases and riskly overshoots, control
mechanism was added to manage waypoints’ order
easily. Test results are shown in Table 6.

Number of SITL and ZD-02

12 (approx. 160
Test flights
[min])
Stationary obstacles 73
Figure 13:Pixracer Wiring Stationary obstacle
68
avoidance
Table 6: Test Results of Algorithm
MP is used as GCS that connects to Pixhawk via
RFD900+ telemetry shown in Fig. 14. This interface
is a full-featured software to control flight
dynamically. Lagari team made some modifications
on MP to reduce mission operator’s workload such
as obstacle avoidance, airdop mission and
communicate with interop server.

Figure 15: Waypoints (Green) and Obstacles (Red)

Figure 14: Mission Planer Interface

2.3 Obstacle Avoidance
This year, obstacle avoidance algorithm was
completely changed because its mission is getting
harder due to the increasing number of obstacles and
former approach was not suitable for large quantities
of obstacles. The basis of this system is tracking
distance between obstacles and UAV. When the
UAV is approaching obstacles, a new path is created
around related obstacles. The algorithm is treating
obstacles as squares and calculates the distance from
each corner of this square with respect to location of Figure 16: Flight path of the ZD-02

Lagari UAV Team Technical Design Paper - 2019 8/15

2.4 Imaging System Fig. 17 shows the two-axis gimbal system ,which is
controlled with Alexmos control board, used to fix
2.4.1 Camera the camera angle, reduce the vibration. The system
This year, Sony Alpha A5100 is chosen due to was designed to be lightweight and easy to assembly.
competition’s primary objective. As compared to the Two brushless DC motors and 3D printed parts were
last year’s camera, Sony A5100 has a larger sensor used.
size, lower weight and takes 105 more frames with
single charge. Both cameras have nearly 76 Pixel per 2.5 Object Detection, Classification
Meter [PPM] from 50 meters. Minimum resolution and Localization
to detect object’s shape and alphanumeric is
determined as 50 [PPM] from 50 meters altitude. For this mission, PB was produced (Fig 18), which
Cruise speed of ZD-02 is approximately 19 [m/s] and is include Odroid-XU4, router, M5 Bullet and
Sony A5100 ‘s capture rate is 2 [s/image]. According gimbal
to Equation 2 and informations shown in Table 7, controller.
the camera can capture 50 meters vertically at every Odroid used for
image and search the area without any blank. take photo and
determine
Camera Weight Sensor Angle of View timestamp of
Size (VxH) taken photo.
Sony 283[g] 23.5x 54° x 77° For connection
A5100 15.6[mm] M5 bullet and
Canon 302[g] 22.3x 52° x 73° Odroid-XU4,
EOS 14.9[mm] router is used. Figure 18: Payload Box
M100
Table 7: Camera Specs
A ratio between the ground and photo pixels are
𝑆𝑐𝑎𝑛𝑛𝑒𝑑𝐴𝑟𝑒𝑎 = 2 ∗ 𝑎𝑙𝑡𝑖𝑡𝑢𝑑𝑒 ∗ 𝑡𝑎𝑛(𝐴𝑂𝑉) established because the camera is parallel to the
Equation 2: Scanned Area Calculation ground with the gimbal. The pixel coordinates of
objects are obtained in meters in the body frame axis
2.4.2 Gimbal System of the plane according to Equation 3 and Fig. 19.
• Design
The main problem in the ODCL mission as the
gimbal camera angle. The plexiglass dome was not
semi-spherical so the shape was preventing the
camera from covering a wide angle. For this reason,
both the gimbal and the plexiglass dome were
changed. In the new gimbal design, the plexiglass
dome placed lower than last year’s model. This has
cause to an increase in drag surface. However, our
viewing angle has consequently increased 23%. Figure 19: Relation Between Ground
And Pixels
Plexiglass is chosen because of its high level of
endurance in varying atmospheric conditions. 1 [𝑚] (𝑡𝑎𝑛𝛼) x (𝐴𝐺𝐿)
=
Plexiglass is easily scratched as opposed to regular 𝑝𝑖x𝑒𝑙 𝑃𝑖x𝑒𝑙 𝑜𝑓 𝑝ℎ𝑜𝑡𝑜 of 𝑥 𝑎𝑥𝑖𝑠
glass but it is 60 % lighter and less brittle.
Equation 3: Relation Between Ground And Pixels

The body frame is

converted to the
East North Up
(ENU) coordinate
system (Fig. 20),
by rotating the old
coordinates
according to the
Figure 17: Design of two-axis gimbal Figure 20: Earth-Centered Earth-
angle of yaw. Fixed (ECEF) and ENU Coordinate
Frames and Euler Angles
Lagari UAV Team Technical Design Paper - 2019 9/15
The ECEF coordinates are obtained by applying the 2.5.2 Autonomous Detection
specified transformation matrix in Equation 4 to the The autonomous detection system consists of 5
ENU coordinate system. Finally, ECEF coordinate different parts that are compatible with the image
of the object are converted to geodetic coordinate processing interface (Fig. 21) developed for the
system. ODCL task.
−sin 𝜆 − sin 𝜑 cos 𝜆 cos 𝜑 cos 𝜆
𝑅 = ൥ cos 𝜆 − sin 𝜑 sin 𝜆 cos 𝜑 sin 𝜆 ൩
0 cos 𝜑 sin 𝜑

Equation 4: Transformation Matrix ECEF to ENU Find region of

Coordinates
2.5.1 Manual Detection interest point
Manuel Detection is performed by using Imaging
Console Interface. System can download all images
from the Odroid-XU4 at 0.25 [s/image]. System
assists users by changing saturation, brightness and
gamma values for identifying the objects easier.
Furthermore, it can zoom up to 10 times to see Crop the
objects clearer. After user finds the objects, objects image
will be cropped and new form pop-up that includes
every information needed to identify objects
(Geolocation is calculated by system and showed in Shape
this pop-up). Users selects the properties of the segmentaion
objects from cropped-images and send them to Alphanumeric
interop server. Additionally, system saves the segmentation
objects and its information to database at the GCS and rotate
for safety.

Binarization

Shape Shape color Alphanumeric Letter

detection detection detection and Color
algorithm algorithm Oriantation detection

Figure 22 : Autonomous Detection Algorithm

Figure 21: Imaging Console

Shape Letter
Image Shape Letter Comment
Color Color

Colors of the shape and letter are

75% 65% 48% 59%
similar.

Shape and it’s color is easily decetable

72% 90% 44% 39%
but letter is close to edge and slim.

Resoution of the image is low but all of

91% 85% 78% 88%
the necessary information can be found.

Table 8: Autonomous Detection Table

Lagari UAV Team Technical Design Paper - 2019 10/15

 • Find Interest Point 2.6 Communication
The first thing that needs to be done for the ODCL Three wireless communication links between the
task is to identify the points of interest in the picture aircraft and GCS are established as shown in the
as shown in Fig. 22. Firstly, Hue Saturation Fig.23.
Value(HSV) conversion, Canny edge detection,
Otsu's binary and contour detection were used for
this process. This process takes about 15 [ms] for a
single image. Noise in the image can cause wrong
evaluation. To solve this problem, confidence rate
in the Tesseract library[7] is used. Figure 23: Block Diagram of Communication
• Shape Classification
2.6.1 Safety Pilot Communication
To classify the shape of the target, Convolutional Futaba 10J was chosen because it was enough for our
Neural Network(CNN) was implemented in system. Also, Futaba has high quality in this market,
structure by using Keras library[8]. The network was and it has high range capacity (3.95 [km]). Device
trained with synthetic samples taken from kaggle[9]. works on 2,4 [GHz] bandwidth.
Dataset include 8K samples consist of one of the
thirteen possible shapes(circle,trapezoid,etc.). 2.6.2 Telemetry Communication
This link is used for telemetry connection between
• Color Detection the Pixhawk and GCS. RFD900+ was selected for
Before the color detection, interest point this part of the communication because its high range
segmentation was applied. Morphological capacity (>40 [km] depending on selected antenna).
transformation, threshold and bitwise logical Devices work on 915 [MHz] bandwidth.
operations were used for segmentation. Then, k- In telecommunications, Received Signal Strength
means algorithm was used for the color detection. K- Indicator(RSSI) is a measurement of the power
means algorithm works well despite working slowly. present in a received radio signal. Minimum signal
strength for reliable packet delivery is -70 [dBm].
• Alphanumeric and Orientation
When flying within a radius of 1.5 [km], the lowest
Detection
RSSI value -60 [dBm] was observed as shown in
The Tesseract library was used for the determination Fig. 24.
of alphanumeric detection. There will be a wrong Just in case, team took precautions about this subject
reading due to the letter orientation. In order to solve by doubled communication link using the payload
this problem, the image was rotated from certain link. Pixhawk and Odroid XU-4 are connected as
angles and the highest confidence rate was selected. serial. If any trouble occurs in telemetry channel,
In doing so, orientation was found. these devices communicate by using MAVLink
protocol. Pixhawk sends the telemetry data to Odroid
• False Positive Elimination
XU-4 and it sends this data to GCS via M5 Bullet.
To avoid extra-object penalties, the noise found as
interest points is eliminated. Various elimination
methods are used for this purpose. Several filtering
processes are applied to minimize noise, for example
bilateral filter. If the Tesseract library returns a low
confidence rate then that object is not reported. If
interest points have too few pixels to give a shape, it
is considered as noise and eliminated. Results of all
system is shown in Table 8.

Figure 24: RSSI values

Lagari UAV Team Technical Design Paper - 2019 11/15

2.6.3 Payload Communication The DC motors were mounted with brackets and 3D
Through the last years’ experience, UBIQUITI printed Acylonitrile Butadiene Styrene(ABS) parts
antennas shown in Table 9 are enough for our that included inside a damping system. According to
system. 5.8 [GHz] bandwidth is enough in image Hook’s Law, a absorbiton of 7.35 [J] was achieved
transferring. which is approximatly 14 % of the impact energy.

UBIQUITI Nanost. M5 It will also improve the driveability in not flat

Dimensions 294x31x80 [mm] terrains. Fig. 30 shows the full design of the UGV.
Weight 400 [g]
Frequency 5,8 [GHz] • Trigger Mechanism
Power Gain 14,6-16,1 [dBi]
To drop the UGV, a trigger system was designed for
UBIQUITI Bullet M5
our needs. The mechanism was installed in the CG
Dimensions 15,2x3,7x3,1 [mm]
of both the aircraft and the UGV to preserve stability
Weight 180 [g]
before and after the drop. It had to be rigid enough to
Frequency 5,8 [GHz]
withstand any turbulence during the flight, easy
Power
0,6 [W] detachable and release the payload without any
Consumption
Table 9: Properties of antennas delay.

2.6.4 Antenna Tracker

The team developed an
antenna-tracker Fig. 25 for
efficient telemetry and payload
communication.

The algorithm embedded in

calculates a vector between
aircraft and GCS using GNSS
data of the aircraft that is
constantly received from the Figure 26: Drop Simulation of UGV
ZD-02 and GCS’s location. Figure 25: Antenna-
Turning is achived by using tracker A two step separation configuration was planned.
servo motors in azimuth and elevation directions to This system consists of two identical subsystems.
ensure that it is directed to the vector. First, the servo on the fuselage will be activated to
release the UGV with the parachute. The second step
2.7 Air Delivery is to trigger the second servo on the UGVto release
the parachute. In order to make the circular
2.7.1 Design movement of servos linear, a wire that can turn
The objectives of the mechanical design is to around a base was placed between the servo handle
assembly a trigger system to drop the UGV, provide and the fiberglass pin.
a safe and precise landing to the given point. Also,
properly drive the UGV to the given destination. For In relation with the plane's maneuvers, the
these tasks our UGV had two major limitations. intermediary parts in the system will be exposed to
First, we already had a functional UAV so size of the stress factors such as bending, torsion, and tensile.
UGV must be compatible with fuselage. Second For this reason, these parts were manufactured from
regarding the competition, we had a weight limit of aluminum using the CNC instead of 3D Part which
1,3 [kg] that should not be exceeded. showed breakability, despite the heavy weight and

Firstly, analysis and simulations were made for the

drop of the system. Solidworks Simulations was
used mainly for Explicit Dynamics. This analysis
was ran with the condition of a 4 [m/s] touchdown
speed. According to the Drop simulation illustrated
in Fig. 26, it will get a maximum stress of 166,1
[MPa]. This value doesn’t exceed of the mechanical
properties of the chosen Carbon Fiber. Drop tests for
the real UGV are planned for further flights.
Figure 28: Trigger System

Lagari UAV Team Technical Design Paper - 2019 12/15

cost. The openings which will contain the parts were 2.7.2 Drop Algorithm
manufactured from 3D printer because of less stress With the experience and formulas of last year, team
they will undergo. System is shown in Fig. 28. made a look-up table to detect the drop distance and
time. Cd of the drop was calculated by the SimScale
• Recovery System
program, surface area was calculated by
To land the UGV and its Reliability SolidWorks, air density was taken at sea level and
payload, Lagari team Repeatability 35° [C]. Combine this information with altitude and
decided to use a velocity of the plane, look-up table was created using
Low Volume
parachute to slow down by Equation 7. The claculated trajectory shown in
Low Weight
the fall speed to a Fig. 31 This year’s air delivery has 1.1 [kg], 0.015
Simple Design [m2] surface area and 0.56 Cd.
reasonable value of 4
Simple
[m/s]. The parachute is a
Manufacturing
popular solution which
Few Maintenance
has proved its efficiency
Indifference to
over the years. Table 10 damage
describes the reasons of
why parachute solution Table 10: Advantage
was chosen. of a parachute

Equation 5 gives us force which the ropes of the

parachute must handle when the parachute opens,
and Equation 6 gives us the area depending by the
desired fall speed.
FD = Drag Force Figure 30: Design of the UGV
Equation 5: Drag Force ρ = Fluid Density
𝐹𝐷 = ½. 𝜌. 𝐴. 𝐶𝐷 . 𝑉 2 A = Parachute Area
CD = Drag
Equation 6: Area Coefficient 𝐵𝑎𝑙𝑙𝑖𝑠𝑡𝑖𝑐 = 𝐶𝑢𝑟𝑣𝑒(𝑉0 ∗ 𝑡 ∗ 𝑐𝑜𝑠(𝜃) ∗ 𝑉0 ∗ 𝑡
V = Velocity
𝐴 =
2.𝑚.𝑔
m = Weight ∗ 𝑠𝑖𝑛(𝜃) − 0.5 ∗ 𝑔 ∗ 𝑡 2
𝜌.𝐶𝐷 .𝑉 2
g = gravitational + 𝑦(𝐻𝑒𝑖𝑔ℎ𝑡) ∗ 𝑡 ∗ 0.5)
acceleration
Equation 7: Parametric Curve for ballistic aplicaitons

As a result of convincing tests,

team decided to manufacture a
bigger parachute by using a
brand-named spinnaker.
Spinnaker has a high drag
coefficient and is constructed of
lightweight fabric, usually nylon
and generally used in sail.
Considering the parachute’s
material, the shape hemispheric, Figure 31: Graphical illustration of air drop
Figure 29:
UAS’s cruise speed 19 [m/s], and trajectory (Red line: with air resistance,
Spinnaker Blue Line: without air resistance)
UGV targeted fall speed 4 [m/s]
Parachute
our calculation indicates that
ropes must resist against 289.21
[N] and the area need to be 1.09 [m2]. Fig. 29 show 2.8 Cyber Security
the parachute which was manufactured according to
In our unmanned aerial vehicle, many security
the calculations.
precautions and communication protocols are used
in order to provide security in communication
system as shown in Table 11.

Lagari UAV Team Technical Design Paper - 2019 13/15

COMPONENT RISKS MITIGATION FALLBACK PLAN
A second telemetry communication over 5.8 [GHz] Return to home plan of Pixhawk starts. If no
Telemetry Jamming
is designed. connection in 3 mins. Failsafe landing.

Wi-Fi link Spoofing WPA/WPA2 security and encryption None.

If any monitoring is detected Database locks

Wi-Fi link Monitoring WPA/WPA2 security and encryption
itself and secures the data.

RC transmitter and receiver are paired before Change the flight mode over the Mission
RC receiver Hacking competition, and FHSS prevents others from Planner and continue the flight.
hijacking the link without knowing the initial seed.
RC receiver is placed on the bottom of the airplane Change the flight mode over the Mission
RC receiver Jamming
to guarantee the connection. Planner and continue the flight

Two GPS modules are placed on different places Airplane waits 3 mins. for connection. After
GPS Jamming
to make connection safe. that failsafe mode starts.

Table 11: Cyber Security

3.Safety, Risks & Mitigations 3.1. Developmental Risks &

The teams can encounter various risks and
Mitigations
unexpected problems during the process of There might be many risks about safety, budget and
multidisciplinary engineering projects. Since Lagari timing in the process of UAS development. In order
Team is an old-established and well experienced to overcome unexpected or expected problems, risks
team in aviation, the risks were listed that may be must be analyzed in advance and measures should be
encountered in management, production and test taken to these risks. Table 12 demonstrates the
process based on old experiences. The Lagari Team analysis of the projected risks and mitigation.
created a mitigation plans to prevent being affected
by them. For Table 12 and Table 13, likelihood
factors are rated as frequent, rare and infrequent.
3.2 Mission Risks and Mitigation
Severity is rated as high, medium and low
respectively numbering as 3, 2, 1. Following All mission risks and mitigations are shown in Table
calculation is manipulated (Risk Factor = Likelihood 13.
x Severity). If risk factor is less than or equal 3, it is
acceptable for risk management. Otherwise it is
unacceptable, and a specific mitigation is needed.
Risk
Risk Explanation Likelihood Severity Mitigation
Factor
-First aid kit and safety equipment (gas mask,
Injury resulting
gloves, glasses etc.) was supplied to the
from
workshop.
developmental Staff may be injured
1 3 3 -Some senior team member is guarding for
and during operation.
safety during manufacturing.
manufacturing
- All team members attend occupational
process.
health training at the beginning of each year.
Budget shortage -New sub-team was established that is called
Budget
adversely affects the 2 3 6 “Sponsorship Sub-team”
Constraint
development process. - A web site is opened for advertising.
Work plan may be
Delay on
lagged due to air -Asana is an interactive workspace that is
manufacturing
condition, time 2 3 6 used for flow of work.
and testing
constraint, lack of
process.
material etc.
- The back-up of system components was
Failure of System components
provided. (for landing and drop tests, landing
system may be suffering on 2 2 4
gear and UGV components were supplied
components tests.
more than one.)

Table 12: Development Risks and Mitigations

Lagari UAV Team Technical Design Paper - 2019 14/15

 Risk
Type of Risk Risk Explanation Likelihood Severity Mitigation
Factor

-Loss of signals can

-High power antennas are preferred.
cause the aircraft crash
-Receivers and transmitters were placed
Loss of data or failure of the image
1 3 3 out of fuselage.
link transfer.
Competition Mission

- RTL mode initiated in case of 1 minute

-Carbon fiber creates
signal loss.
faraday cage effect.

-The parachute is used while dropping.

-When releasing the
The breaking -The UGV is fabricated from carbon
UGV from air, it may 3 3 9
of UGV fiber.
be broken.
-Suspension system was used.

The failure of -Autonomous detection

autonomous algorithm has not 100 3 3 9 -False elimination method was used.
detection % accuracy.
Triple redundant power to autopilot by
-Complete loss of
Loss of power power module, Battery Eliminaton
control of aircraft is 1 3 3
to autopilot Circuit(BEC) to servo and USB from
caused to crash.
Payload Box.
Autonomous Flight

Failure of -It causes failure of - Two avionic boxes were redesigned

avionics ODCL missions, even 2 3 6 and manufactured by aluminum to
components causes any crash. reduce any failure.
-Flight logs are analyzed in detail and
-It can cause damage on
Failure of parameters are compared different flights
landing gears or
autonomous by Autonomous Subteam.
fuselage due to flight 2 3 6
take-off and -Extensive Checklists are examined
controller parameter
landing. every flight.(Before&After Flight
error.
Checklists)
- Covered Li-Po Batteries with a hard-
Damage to Li- -It may cause fire in the plastic material to make them rigid.
2 3 6
Po Batteries UAV. - Insulated cells with a nonflammable
rubber.
-A cable fault that can
Cable -Connector and cable assemblies fitted in
Testing

occur during flight may

connectivity 3 3 9 the fuselage.
cause the aircraft to
error -Sockets used at the ends of the cable.
crash.

Vortex wind -It can cause damage on -Using strengthened and spare landing
1 3 3
during landing landing gear or fuselage gear and fuselage.

Table 13: Mission Risks and Mitigations

4. Conclusion 5. References
This paper summarizes a work done by students
[1] https://asana.com/
from the Mechatronic Engineering Department of
the Yildiz Technical University in preparation for
[2] T. Dalgleish et al.J. Exp. Psychol. Gen.
the AUVSI SUAS 2019. Throughout the year, the
Lagari team designed, improved and tested the ZD- [3] C. T. Modlin and J. J. Zipay 2014.
02 and UGV system with an improved software for
payload control and an enhanced hardware for flight [4] D. P. Raymer, “Aircraft Design” 1992.
control. Advanced algorithms were written to
accomplish competition tasks. Numerous ground [5] “About stability analysis using XFLR5.”
and flight tests were conducted to confirm the UAS
capabilities and assured its excellence in the SUAS [6] C. L. Fefferman, Navier Strokes eq.
competition.
[7] https://pypi.org/project/pytesseract/

[8] https://keras.io/

[9] https://www.kaggle.com/

Lagari UAV Team Technical Design Paper - 2019 15/15