Вы находитесь на странице: 1из 21

MS Ramaiah Institute of Technology

Journal Paper

Abstract
Edhitha, a team formed in 2011, has undergone exponential growth. Edhitha has delved into
diverse and challenging endeavors, ranging from systems integration to design and
development of airframes using advanced composite technology. Last year, Edhithas
venture into custom built airframes yielded a valuable experience in understanding the
demands and resource limitations of developing a fully functional Unmanned Aerial
System. For SUAS 2015, Edhitha believed in the vision of delivering a high performance
UAS and strived to meet the ever increasing demand for pioneering research in the field of
Unmanned Aviation. Edhitha realized that in order to maintain a degree of reliability from
the UAS, an extensive optimization process was necessary which aimed at an in-depth
testing of all components of the UAS. Consequently, Edhitha shifted its focus transitioning
into its most extensive R&D phase yet, clocking the largest number of test-flights in the four
years since conception. In each successive mission Edhitha has improved the efficacy of the
system through a systematic identification and removal of inconsistencies and
inefficiencies. Safety has remained a key element in the design and development of the
UAS.

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

Contents
1. Mission Requirement Analysis .. 3
2. System Design Ra:onale .. 3
2.1 Airframe ..4
2.2 Naviga:on and Autonomy ..5
2.3 Power System ..5
2.4 Payload System ..5
3. Expected Task Performance . 6
4. Programma:c Risks and Mi:ga:on Methods . 7
5. Descrip:on of UAS . 8

5.1 Airframe 8

5.2 Power System 8

5.3 Method of Autonomy .. 9

5.4 Data link .. 10

5.5 Payload System .. 10

6. Ground Control Sta:on 11


6.1 Naviga:on Unit ..11
6.2 Imagery Unit .11
7. Data Processing 13
8. Mission Planning . 14
9. Payload System Tes:ng and Performance . 14

9.1 Airdrop 14

9.2 Infrared Tes:ng .. 15

9.3 SRIC .. 15

9.4 Imagery Tes:ng 15

9.5 On-board Computer Performance Tes:ng . 17

9.6 Interoperability . 17

10. Naviga:on Tes:ng and Performance . 19


11. Safety Considera:on and Approach .. 20

11.1 Safety and Risk Mi:ga:on Approach in UAS Design 20

12.2 Opera:onal Safety 20

12. Conclusion. 20
13. Team Members Edhitha ..21
14. Acknowledgement . 21

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

1. Mission Requirements and Analysis


After thorough analysis of the mission as specified in the SUAS rulebook, Edhitha has identified
the following as the tasks to be completed by the UAS:
1. Autonomous flight: Autonomous take off, waypoint navigation, search and landing.
2. Image acquisition: Visible spectrum and Infrared (in realtime)
3. Data processing: Extraction of image characteristics i.e. shape, background color, letter
orientation, alphanumeric, alphanumeric color, and location
4. Data exchange: Data transfer to Ground Control Station (Imagery and Telemetry; Data
exchange with interoperability and SRIC servers)
5. Search Detect and Avoid: Algorithm enabling the UAS to avoid virtual objects
6. Autonomous Air-drop: Specialized payload designed to perform automated drop of relief
package
Having identified the tasks to be performed, the UAS essentially translated into the following:
1. Stable airframe capable of accommodating required avionics, and subsystems
2. Integrated autopilot system interfaced with the Ground Control Station
3. Image acquisition payloads i.e. High definition visible spectrum camera and Infra Red
camera
4. On-board computer for processing data
5. RF telecommunications unit capable of transferring data to and from the UAS
6. Air-drop mechanism
Edhitha also theorized the following broad processes that need to be functional during mission:
1. Data processing algorithms for detection and extraction of target characteristics
2. Autopilot Interfacing with GCS for navigation control and real-time mission planning
3. Automated background scripts to relay data during the Interoperability, SDA and SRIC tasks

2. System Design Rationale


For SUAS 2015, Edhithas approach from which the rationale and sequence of development of
the UAS was need based. The idea was to extract from the UAS exactly what was needed for the
mission with minimal resource input and to engineer a UAS system capable of providing
optimal mission performance. In order to do this, Edhitha indulged in a conscious effort with an
open mind for experimentation and relied greatly on test flight results to evaluate the
operational performance of the UAS. Hence the UAS was optimized accordingly. Safety was one
of the primary factors influencing the design of the system. Plausible causes of system failure
were eliminated through a thorough analysis of each individual component and its
functionality. Edhitha also kept a constant focus on ensuring modularity throughout design of
the system, providing sufficient avenue for redundancy. Therefore Edhitha successfully obtained
the set standard of reliability in the UAS.

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

Fig 1-System Design Procedure

2.1 Airframe
Keeping the motto for the year 2015 as Go lean,Edhitha acquired an airframe which was
lighter, easy to handle and deploy. After extensive testing on a custom built twin boom airframe,
a similar configuration was favored. This configuration offered larger usable fuselage volume
compared to other configurations having the same weight. This selection also ensured greater
accessibility to components while enabling easy isolation of failure. From having used a
gasoline engine in the past, it was seen that undamped vibrations adversely affected the
functionality of the sensitive electronics and the structural integrity of the airframe over a
period of time. Hence, Edhitha decided to acquire an off-the-shelf airframe powered by an
electric propulsion unit. Airframes were compared with features such as cruise speed, All Up
Weight (AUW), payload volume, deployment methods, et al. With a payload capacity of 4Kg
and a low cruise speed of 18m/s, the Anaconda was an ideal slow flying plane to support the
requisite systems and payloads. Thus the Anaconda best suited the teams requirements.

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

Table 1Airframe Comparison

2.2 Naviga:on and Autonomy


The UAS required an autopilot that was reliable and tested. Using a familiar system meant
saving time that would otherwise be used on experimentation. It was also desirable to use an
open source autopilot in order to have large community support. Edhitha needed to ensure
quick replaceability of the autopilot module in instances of malfunction or damage during test
flights. This meant that the autopilot would have to be both low cost and easily acquirable.
Realizing that processing capabilities would be pushed to the limits during execution of tasks as
demanding those of SUAS, less powerful options like the APM 2.6 were dismissed. Having
arrived at the above considerations, Edhitha decided to use the 3DR Pixhawk.

2.3 Power System

An uninterrupted and consistent power source had to be chosen to power all components of the
UAS. The use of an electric propulsion system mandated the separation of power distribution
over the UAS. Therefore the designed power system consisted of 2 batteries in tandem-one to
power the propulsion and avionics and the second to power the OBC and data link systems.
This arrangement ensured a lower current draw from a single battery with even power
distribution across all systems, and increased the overall endurance of the system.

2.4 Payload Systems


2.4.1 Airdrop: Simplicity and safety in operation were prioritized during the design of the air
drop system. The goal was to create an autonomously operable mechanism that could release
the air-drop payload with the actuation of a servo. The mechanism was required to be robust in
order to avoid accidental release of the payload. For additional safety, the operator was always
given control of the mechanisms actuation, thus keeping the design in compliance with the
mission requirements.

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

2.4.2 Imagery, On-board Computer & Data link

Table 2Imagery, On-board Computer & Data link Design

3. Expected Task Performance


Throughout testing the UAS exhibited robustness in performance. It delivered an average of 27
minutes of continuous flight over 10 test flights with all systems onboard. The UAS was
successful in completing multiple mock SUAS mission flights during the final stages of testing.
The image acquisition system was able to capture all targets and relay data to the GCS at an
average speed of 40mbps. The data processing algorithms detected all six characteristics of the
targets in real time, thus meeting the objective requirements of ADLC and actionable
intelligence. Further improvement in accuracy is expected in the detection of alphanumeric
character and GPS location of the target. The IR camera was able to detect the secondary IR
target during test flights, delivering high quality images. The package drop mechanism was
successfully integrated with the autopilot system over 11 test flights during which the threshold
requirement was met. With further testing, complete autonomy is expected in the airdrop task.
From having completed the SRIC task successfully for 2 years in a row, Edhitha had an in-depth
understanding of the demands of the SRIC task. Edhitha managed to successfully exchange data
with a remote server simulating the SRIC antennae. Attempting the Interoperability and SDA
tasks for the first time this year, the UAS was able to transmit and receive the desired data at an
excess of 10 Hz. In all of the tests carried out for interoperability, the UAS was able to
successfully sending data to a mock external server. With further testing, Edhitha expects to meet
the threshold requirement for SDA task. From the evaluation of the performance as seen in
mock missions and test-flight, the UAS is expected to complete all tasks meeting the objective
requirements in most of them.

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

4. Programmatic Risks and Mitigation Methods


During the course of development of the UAS, Edhitha encountered, analyzed and resolved
several programmatic risks. During the testing phase of the project, Edhitha acquired a
comprehensive understanding of the various risks and possible failures likely to occur during
flight. The likelihood of failure was methodically removed after each subsequent flight test.
Likewise, all systems have been acquired and designed to maximize reliability and safety
during flight. In areas of high risk and unexpected failure of components, a sufficient budget
has been allocated for replacement with spares.

Table 3ProgrammaEc Risks & MiEgaEon Strategies

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

5. Description of UAS

Fig 2Block Diagram, DescripEon of UAS

5.1 Airframe
The Anaconda features a twin boom, inverted V-tail configuration with a wingspan of 2.06m
constructed with high density EPP foam. The airframe, with a dry weight of 2.4Kg and the
ability to carry a payload of up to 4Kg, provides sufficient volume to install the payload in a
modular manner. Compared to previously used airframes like flying wings, the Anaconda
provides control in all three axes, giving an additional degree of maneuverability and stability
along yaw. The airframe is powered by a 1KW, 880 kv brushless out-runner motor which draws
power from a 16000mAh 4S LiPo battery.
The airframe was customized extensively to suit Edhithas requirement. The wings were
reinforced with carbon fiber tubes to reduce load induced wing flexure which significantly
improved the flight characteristics. The stock landing gear of the UAS was replaced with
aftermarket gears to avoid failure during hard landings. To ensure a better utilization of space,
custom made mounts were designed to accommodate all systems with an organized wiring
layout.

5.2 Power System


Main Power System: The 16000mAh 4S LiPo battery powers the motor, navigation system, and
all servos giving a total endurance of 30 minutes. The maximum current draw was measured to
be 64A, of which the motor drew 60A at full throttle, while the 7 servos and the navigation
system drew a maximum of 4A. To distribute power over the system, a 100A ESC with a 5A
BEC was used.

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

Auxiliary Power System: This system was used to power the OBC which had a power
requirement of 2A at 5V and the 5.8GHz data link requiring 1.75A at 14.8 V. However, in certain
unanticipated conditions, it was seen that the power drawn by the OBC exceeded 2A.
Anticipating a maximum current draw of 5A and keeping a factor of safety as 2, a 10A UBEC
was used. For the above power requirement and an endurance of 30 minutes, a 2200mAh 4S
LiPo battery was used to supply power.

Fig 3-Auxiliary Power System

5.3 Method of Autonomy

Fig 4NavigaEon System


The 3DR Pixhawk autopilot system, integrated with an array of specialized sensors listed below,
was used for achieving navigational autonomy.
A sensor suite consisting of a 3 axis 16 bit gyroscope, 3 axis accelerometer to maintain
orientation.
Barometer to maintain altitude.
A differential digital airspeed sensor coupled with a pitot tube to obtain the airspeed.
Ublox LEA 6H GPS along with a compass module used as a peripheral unit to determine
GPS location as well as heading.
An SD card to store flight data logs for post flight analysis.

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

The autopilot was mounted as close as possible to the Center of Gravity of the UAS after
carefully factoring in the safety considerations and ergonomics of the autopilot system.
Sufficient vibration damping was added as an additional measure of safety. A long range RC
receiver was coupled to the Pixhawk to facilitate manual override of the autopilot as a necessary
safety measure. Waypoint tracking was accomplished by writing GPS coordinates to the
autopilot from the GCS in real time through a dedicated telemetry data link as described below.

5.4 Data Link


Communication with the GCS was established over two independent RF links:
Two 915 MHz telemetry modems were used to relay navigation data between the autopilot
system and the GCS. The telemetry link uses the Mavlink Protocol to pack C-structs over serial
channels with high efficiency and send these packets to and from the ground control station.
A 5.8GHz data link channel was used to control the OBC to monitor imagery systems and
transfer the images captured to the GCS. The RF signals were relayed to and from the GCS
using a Ubiquiti Bullet M5 modem mounted on board the UAS. The data link operates as a Line
of Sight communication channel over a range of 5Km.

5.5 Payload System


5.5.1 On Board Computer: For its exceptional processing capabilities, the ODROID U3 was
chosen as the On-board computer(OBC). The OBC was interfaced with the bullet to provide
remote access from the GCS using the 5.8GHz data link channel. Armed with a 2GB RAM,
1.75GHz quad core processor, the OBC interfaced with payload systems used for image capture,
and SRIC operations. The 2.4GHz Realtek Wi-Fi module connected to the OBC was used to
connect to the SRIC router.

Fig 5Onboard Systems

5.5.2 Camera: The UAS was equipped with a Nikon D3300 DSLR to capture images of the
ground targets. The camera was interfaced with the OBC via USB and was remotely accessible
from the GCS. The gPhoto2 package running on the OBC was used to control camera
parameters and trigger the camera. In order to cover a larger ground area and optimize image
overlap, the camera was operated in burst mode at 2 fps. A large sensor size and a resolution of
24.2 MP ensured high image quality.

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

10

5.5.3 IR Camera: The SJCAM SJ4000 action


camera was integrated onto the UAS for
active IR imagery. The IR sensitivity of the
camera was increased by replacing the IR
filter with a longpass filter which blocked
light in the visible spectrum up to a
wavelength of 750nm (Fig 6). A Standard
USB connection was used to interface the
OBC with the camera. The captured images
were then sent to the GCS via the 5.8GHz
data link.
5.5.4 Air Drop: The canister for the air drop
task was custom designed and 3D printed to
obtain an aerodynamic profile. The design of
the canister included a flange which was
installed in flush with the fuselage. A pocket
was provided for a snug fit of the actuating
servo. The canister was then attached to the
bottom of the fuselage directly below the
Center of Gravity of the airframe to prevent
any change of weight balance after the
airdrop. At the target waypoint the servo is
actuated by the autopilot thereby opening
the door of the cylinder to release the egg. To
prevent accidental release, a safety switch
was configured on the transmitter which
activated the airdrop mechanism.

Fig 6-Wavelength RecepEon of Longpass Filter

Fig 7- Air Drop Mechanism

6. Ground Control Station


The GCS constitutes systems that control and co-ordinate the functions of the on-board systems
to elicit the desired task performances. The system was designed to optimize the ergonomics of
UAS handling and monitoring. The GCS comprises of two independent units that function in
parallel during the flight.

6.1 Naviga:on Unit


The navigation team uses the GUI based Mission Planner software to communicate with the
Pixhawk. The 915 MHz data link is used for both monitoring of flight characteristics and
dynamic mission re-tasking. The autopilot operator constantly monitors the flight
characteristics and navigational accuracy of UAS. The Waypoint Operator computes and
controls the flight path for various mission tasks. Navigation unit is interfaced with the
interoperability server to publish and receive data.

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

11

6.2 Imagery Unit


Operated by a four member team, the imagery unit consists of the following:
1. Communication system: The Ubiquiti NanoStation M5 at the GCS was paired with the
Ubiquiti Bullet M5 on board the UAS to establish the 5.8GHz data link. This link allowed the
GCS systems to remotely control the OBC (ODROID U3) and extend this control over the onboard imagery systems interfaced to it.
2. Processing unit: A network of four systems was forged for the purpose of processing the
data being forwarded to this unit by the GCS communication system. This data is filtered and
processed by System 0 which in turn generates an output that is processed by subsequent
systems 1, 2, and 3 as depicted in fig 8. This network of systems employed a model that was
aimed at obviating the effect of system failures on the completion of mission objectives. The
model allowed each system to be a functional duplicate of every other system providing the
highest degree of functional redundancy.

Fig 8- GCS Layout

7. Data Processing
Edhithas approach towards data processing was aimed at carrying out real time processing
while maximizing autonomy and ensuring redundancy with a manual alternative. The OBC
on the UAS was used as the data acquisition centre for still images with the GCS systems
acting as the data processing centers. Image acquisition was carried out by means of an open
source library running on the OBC, which offers an API to trigger and download images
from a DSLR. Acquired images were routed to the GCS systems by means of the 5.8GHz
data link. Automatic Detection, classification and localization (ADLC) was carried out by the
algorithms mentioned below.

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

12

7.1 Target Isolation


Images that were received at the GCS underwent a filtration process to isolate images with
potential targets in them. These filtered images were processed to identify regions of interest
(ROI) by eliminating background noise. Noise elimination was carried out by an optimized
algorithm that works on histogram back projection on the hue channel of the image. Extracted
ROIs were further filtered to segregate ROIs containing targets, QR targets and emergent targets
from false positives.

7.2 Procedure to localize targets


Each of the ROIs containing targets, QR targets and emergent targets were localized by
translating the ROI centers in the images into the geographic coordinate system. The
localization algorithm utilized the GPS and compass heading extracted from the metadata on
these images and carried out Euclidean transformations to localize the position of the ROIs on
the original images. Orientation of these ROIs was estimated from the compass heading
extracted from the metadata on these images.

7.3 Procedure to determine color shape and alphanumeric character


Shape classification of targets utilized an ensemble of classifiers using shape approximation
algorithms and template matching algorithms to classify the shape of the targets. Alphanumeral classification was executed by using the Tesseract Optical Character Recognition (OCR)
engine with a series of pre-processing algorithms being executed to improve the accuracy of the
output.
Color classification of targets was carried out by employing a K-means clustering algorithm to
determine the color of the alpha-numeral and shape. The RGB channels of each target were
partitioned into 5 clusters by the algorithm with each cluster centre representing a color. These
clusters were sorted by the frequency of occurrence in the descending order with the largest
cluster being the color of the shape and the second largest cluster being the color of the alphanumeral.

7.4 QR code
QR targets were fed into an open source QR decoder. The accuracy of the QR decoder was
restricted by the orientation the QR image. To overcome this limitation, the QR targets are fed
into the decoder with multiple rotations.

7.5 Redundancy
Operators overlooked the above ADLC algorithms running on the GCS systems and were
provided with an option to intervene and modify the output of these algorithms if they were
not satisfied with the outputs. Additionally, emergent target classification was carried out by the
GCS operators with the locational attributes being extracted automatically.
All the image attributes were automatically logged into CSV files on each of the GCS systems
with the final CSV file containing the processed data in the required format, being logged into
the USB stick that would be handed over to the judges.

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

13

8. Mission Planning
After identification of tests to be carried out during the mission, the agenda of flight is
announced at least 12 hours prior to the mission. Two sets of procedures are then undertaken,
the first of which involves the planning of the sequential execution of mission objectives and
accordingly, the design of the flight path and the second involves a ground-check of all
components to be tested during the mission while setting up the UAS & GCS .
Mission waypoint planning: With objectives of the mission and the required tests, clearly
outlined, the flight path is planned strategically in order to ensure successful completion of each
task according to its relative priority. The guidelines for mission planning differ according to the
type of tasks to be completed. For the autonomous navigation task, the waypoint sequence is
made progressively complex in order to observe the flight characteristics and accuracy of
waypoint tracking through increasingly difficult turns. During mock missions, the team
employs similar flight paths as were used in the previous years of the competition. In order to
optimize the coverage of imagery payloads, the team devised a technique of grid tracking that
involves a partial coverage of the search area along the circumference followed by tracking
along a set of closed spaced parallel paths.
On-field set-up: A thorough check is conducted of all components to be tested during the
mission including a preliminary check of airframe health. All components and items required
for on field testing are packed into separate storage units according to a series of checklists. At
the airfield, the team members function in parallel to minimize the time taken for UAS and GCS
set-up. The navigation and imagery operators set up their respective systems. Once the systems
are set-up the data link between the UAS and GCS is established. The imagery team conducts a
set of tests to verify system connectivity while the navigation team checks the flight readiness of
the autopilot. The flight path is then written onto the autopilot. After a final verification of the
functionality of all systems, the safety pilot is handed complete control over the UAS.

9. Payload Testing and Performance


9.1 Airdrop
A trial and error method was adopted to determine the GPS location, speed and heading
required for the release of the package. A series of manual drops were conducted to estimate the
offset distance at which the airdrop was to be initiated. The offset distance was measured along
the direction of UAS heading, parallel to the ground. The cruise speed and the altitude of the
aircraft were kept as predetermined variables. Over 11 trials were conducted for testing the
accuracy of airdrop across which the average offset distance for the servo actuation was found
to be 20ft at an altitude of 400ft. Therefore, the autopilot was programmed to actuate the servo
at this offset distance in order to hit the bulls eye. Fig. 9 describes the distance of the package
drop from the point of the reference for each air-drop trial. With each test, the team aimed to
further centralize the offset distance.

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

14

Fig 9-Air Drop Accuracy

9.2 Infrared Tes:ng


During test flights, an array of Infrared LEDs emitting a wavelength of 850nm was used to test
the modified camera. Testing was carried out in both daylight and low light conditions. In low
light, the modified Near Infrared(NIR) camera was able to detect the target from over 100m
away. The same tests were carried out in broad daylight conditions and the range of detection
was reduced to 60m. The removal of the Infrared filter from the camera improved its nearinfrared sensitivity, but the poor contrast led the team to look into the use of bandpass and
longpass filters to block a certain spectrum of the visible light. Subsequently, the images
captured were found to have the requisite characteristics for further processing.

Fig 10-IR TesEng W/O Longpass lter

EDHITHA

Fig 11-IR TesEng with Longpass Filter

MS RAMAIAH INSTITUTE OF TECHNOLOGY

15

9.3 SRIC
The radius within which the UAS was allowed to loiter to establish connectivity with the SRIC
router was calculated using the given antenna beam-width of 60 degrees and the altitude of the
UAS. Using FTP, two files were uploaded and downloaded to and from the server respectively.

9.4 Imagery Tes:ng


Performance of Automatic Detection and Extraction of targets: The algorithms for detection
and extraction of the targets were tested during flight tests and on the ground. In case of
exceptions like images containing anomalies like cement pavement occupying more than 50% of
the image, a failsafe was introduced.
Color Characteristic Testing: The issues in clustering more than 255 colors into 16 easily
recognizable colors were resolved by choosing optimal thresholds. The percentage of the color
of shape in some images was less than the noise which led to erroneous results. These erroneous
results were corrected by a manual mechanism which allowed a keyboard input of the results.
Shape Characteristic Testing: Targets containing shapes such as circles, squares, stars were used
as inputs to the algorithm. Constraints like the angles, aspect ratio and number of vertices were
analyzed and added to the algorithm through testing. All the regular shapes extracted from the
test data were successfully identified by the algorithm. Elliptical shapes were not identified
accurately and the algorithm is being improved to correct the inaccuracy.

Table 4-Results of Imagery TesEng

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

16

GPS accuracy testing: The GPS algorithm was tested during the initial flights and verified
against the GPS values given by the Google APIs. The error rate of 40% was identified due to
improper calibrations of the Geo-tagger and the algorithm. Upon careful calibration and
refinement of algorithm, the error decreased to 12% in the second test flight and a constant 5%
in all subsequent test flights.

Fig 12-Results of GPS TesEng

QR Code accuracy testing: The QR code was tested with images of size lesser than 10 KB and
resolutions lower than 100x100 pixels with varying orientations up to 50 degrees from the
normal. The output of the code depends on the focus of the image and the visibility of the three
positions
Algorithm Performance and Reliability Test: The false alarm rates of each algorithm were
improved significantly by including a scope for manual intervention. This step depreciated the
error percentages to 5% or further below, exceptions being the localization and QRC algorithms.
The remainder of the errors were attributed to human errors. The algorithms were first tested
upon individual images and then extrapolated to a batch of images. The processing times of
subsequent systems were decreased due to the simultaneous processing across different systems
as depicted in fig 8. This reduced data processing time from 10 minutes to less than 4 minutes
for a batch of 50 images.

Table 5-Individual Error rates and Processing rates of Algorithms for a batch

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

17

9.5 On-board Computer Performance Tes:ng


The OBCs maximum current draw and CPU usage were put to test by powering it up and
connecting the DSLR, SJ4000 & Wi-Fi module so as to simulate the final payload configuration.
Inference:
Max Current Draw of peripherals from OBC was 602 mA
Threshold Current Draw: 2000 mA
Maximum CPU usage at any time was within 50%

Table 6-Current Draw by Peripherals of OBC

Table 7-CPU Usage- OBC

9.6 Interoperability
In the initial stages of testing, the Interoperability server was setup using information from the
Github repository and duplicate GPS data was obtained. Upon investigation it was observed
that the GPS module on the UAS had a limitation of sending data at 5 Hz to Mission Planner, but
the data rate to the server was higher than 10 Hz. Analysis of the uploaded data on the server
revealed constraints of the GPS module. In the first set of flights, a simple curl approach was
employed and a data rate of 10 Hz was achieved, which met the given mission task objective.
On further testing an optimized code was developed and a data rate of 14 Hz was
achieved ,which was faster than the previous approach.

Table 8-Interoperability Test results

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

18

10. Navigation Testing and Performance


The final goal for the navigation team was to ensure good accuracy in tracking waypoints while
keeping the airframe within its operational safety limits. Prior to testing navigation systems, a
number of manual flights were conducted with the airframe in fully loaded condition to
understand flight characteristics. With an idea of basic flight characteristics the navigation team
carried out testing over two phases described in the table below:

Table 9- NavigaEon TesEng Results

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

19

11. Safety Consideration and Approach


11.1 Safety and Risk Mi:ga:on Approach in UAS Design
With the experience of past years competitions as valuable lessons in hand, Edhitha has made a
conscious effort to incorporate factor of safety and risk mitigation techniques for the airframe
and the systems integrated in it.

Temperature control: In order to prevent component overheating during flights, Edhitha


improvised by providing additional air-ducts for ventilation in the fuselage. The ESC was
placed external to the fuselage to negate any possibility of failure due to overheating.

Voltage spike negation: Unregulated and unchecked voltage spikes can leave the autopilot
module susceptible to frequent lock ups. Edhitha has eliminated unchecked voltage spikes by
connecting a Zener-diode in parallel with the output rails of the autopilot module.

Isolating stray signals: Edhitha paid keen attention towards isolating stray noise which can
be picked up by servo wirings causing faulty functioning of servos. Any unavoidable long wires
have been twisted and passed through toroid rings canceling out any RF noise.

11.2 Opera:onal Safety


Edhitha has incorporated two layers of operational safety during each mission-

Pre-ight checks:

Component checks for wiring, motor, motor mount, propeller, airframe health.
Ensuring power source voltages are up to the required levels.
Telemetry signal data check between RF modules and autopilot sensor health check.

In-ight safety approach:


Return-to-launch: The autopilot has been programmed to return to the GCS in the event of a
signal loss ensuring safety of UAS, operators and spectators.
Monitoring Telemetry Data: Active personnel stationed at GCS to monitor telemetry data of
the UAS ensuring a pre-failure detection of impending threats.
Externally accessible arm/disarm switch: provide visual indication of the status o systems on
board and transfers control to the autopilot only when all pre-flight checks are complete.
Voice updates: Navigation GCS provides continuous voice updates about autopilot status,
signal strength and health of on-board sensors to the operator.
Go/No Go Indicators: Critical checklist items linked to operations are fitted with visual
indicators to indicate the status of the item.

12. Conclusion
Edhitha has experienced a remarkable year, undertaking a clinical and exhaustive approach
towards optimizing the performance of the UAS. Almost all mission tasks specified by SUAS
were achieved with outstanding accuracy even meeting the objective requirement. Having been
conscientious in its efforts to achieve full mission task performance capability, Edhitha is
confident of putting on a tremendous performance at SUAS 2015.

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

20

13. Team Members Edhitha

Table 10- Team Members, Edhitha

14. Acknowledgements and Sponsors


1. 3D Robotics
2. Management, MS Ramaiah Institute of Technology
3. Solmeta
4. Mr. Ananth Nairy
5. Former Members, Team Edhitha (Sriraghav S, Gandhar K)
6. Team Mentors and Test Pilots (Vishnu BN, Vishwanatahan R)

EDHITHA

MS RAMAIAH INSTITUTE OF TECHNOLOGY

21

Вам также может понравиться