Upload
others
View
6
Download
0
Embed Size (px)
Citation preview
Delhi Technological University
1
Delhi Technological University Student Unmanned Aerial Systems
Competition 2013 Journal Paper
Figure 1: Aarush-M
Abstract Aarush-M is a twin-boom, inverted V-tail UAS designed for delivering situational awareness in a disaster struck area. Command and control over UAS is done via a 2.4 GHz radio link, while the intelligence gathered is transmitted over a 5 GHz link. A highly modular and portable system, Aarush-M can be flight ready in less than 30 minutes, providing an endurance of 20 minutes. This paper presents requirements analysis of the UAS followed by the design description. Flight testing and evaluation results are also presented which validate the performance parameters .The final section elaborates the safety measures adopted by the team to ensure safety of the personnel and UAS at all times. Having successfully conducted two dry-runs of the competition mission, team UAS DTU is confident that Aarush-M will be able to support the US Marines in their humanitarian relief and security mission in the earthquake struck Caribbean island.
Delhi Technological University
2
Table of Contents Abstract .................................................................................................................................................. 1
1. Introduction .................................................................................................................................... 3
2. Systems Engineering Approach ........................................................................................................ 3
2.1. Mission Analysis and Requirements Specification ..................................................................... 3
2.2. Design Rationale ...................................................................................................................... 3
2.2.1. Intelligence Gathering Payload ......................................................................................... 4
2.2.2. Guidance and Navigation .................................................................................................. 5
2.2.3. Airframe ........................................................................................................................... 5
3. UAS Design Description.................................................................................................................... 6
3.1. Air Vehicle .................................................................................................................................... 7
3.1.1. Conceptual & Preliminary Design ...................................................................................... 7
3.1.2. Wing analysis and configuration selection ........................................................................ 7
3.1.3. Fabrication and Developmental Tests ............................................................................... 8
3.1.4. Propulsion System ............................................................................................................ 9
3.1.5. Power System Design and Layout of Avionics.................................................................. 10
3.2. Payload .................................................................................................................................. 11
3.2.1. Imagery System Payload ................................................................................................. 11
3.2.2. SRIC System Payload ....................................................................................................... 11
3.3. Data Processing ...................................................................................................................... 11
3.3.1 Image Processing ............................................................................................................ 11
3.3.2. SRIC Data Acquisition...................................................................................................... 14
3.4. Communications .................................................................................................................... 14
3.5. Ground Control Station .......................................................................................................... 14
3.6. Mission Planning .................................................................................................................... 15
4. Flight Testing and Evaluation Results ............................................................................................. 15
4.1. Navigation Performance ............................................................................................................. 15
4.2. Payload Performance ............................................................................................................. 16
4.2.1. Imagery .......................................................................................................................... 16
4.2.2. SRIC ................................................................................................................................ 18
5. Safety ............................................................................................................................................ 18
5.3. Safety in design of the UAS..................................................................................................... 18
5.4. Safety in mission execution/operation the UAS ...................................................................... 19
5.5. Failure Mode Effect Analysis .................................................................................................. 20
6. Acknowledgements ………………….…………………………………….……………………………………………………………..21
Delhi Technological University
3
1. Introduction
The Unmanned Aerial Systems team at Delhi Technological University is proud to present Aarush-M to compete at the eleventh annual SUAS competition. The team has been participating in the Student Unmanned Aerial Systems SUAS Competition since 2009 and has gained valuable experience in design, development, and operation of a UAS. The team comprises of undergraduate students from diverse engineering backgrounds. The preparation for the competition started in October 2012 and after 4.5 hours of flight testing till May 25, the team is confident that Aarush-M will complete the mission safely and successfully.
2. Systems Engineering Approach
Taking cue from the judge’s comments from last year, this year, the team’s approach to the competition has been mission-oriented, rather than technology-oriented. This section describes the systems engineering paradigm followed by the team.
2.1. Mission Analysis and Requirements Specification
The mission clearly entails objectives and thresholds for Key Performance Parameters. To score maximum points, the performance with regard to each parameter should be close to the objectives. Therefore, the team defined a set of KPPs given in Table 1 below with modified thresholds and objectives based on their importance and likelihood. The expected performance during competition is highlighted in green color in the table.
S. No. Parameter Threshold Objective
1. Navigation Dynamic Waypoint Dynamic Search Area
2. Launch/Recovery Automatic Launch Automatic Launch and Recovery
3. Imagery All Characteristics (3 Autonomously )
All Characteristics (All Autonomously)
4. Target Location Within 250 feet Within 50 feet
5. SRIC Data Acquisition Manually Autonomously
6. Mission Completion <=30 minutes <= 20 minutes
*SRIC Simulated Remote Intelligence Center ** Green Color Expected Performance Table 1: Key Performance Parameters
The analysis of the mission KPPs defined the initial driving inputs of the overall UAS design. These factors along with performance parameters generated the derived requirements of the sub-systems which have been described in detail later in this report.
The team also deduced certain operational requirements which contribute to the overall robustness of the UAS operation. They were:
1. Wind Tolerance > 15 knots and Gust Tolerance ~ 20 knots. 2. Setup time should be less than 30 minutes, which would require a better portable ground
station, efficient mission planning and more test runs.
2.2. Design Rationale
Once the overall system requirements were generated, different sub-teams were tasked with dissecting the UAS requirements’ document to generate quantitative subsystem requirements. These low level requirements guided the design process that followed. The team identified four design elements which
Delhi Technological University
4
dictated the performance of the overall system: 1. Intelligence Gathering Payload (IGP) 2. Guidance , Navigation and Control System (GNC) 3. Communications 4. Airframe
For each of the design elements, anomalies in the previous year’s competition entry were analyzed and the goal of the development process was to mitigate these in the new UAS. A variety of options for each design element were considered, and the option that fulfilled its own requirements, while not conflicting with the requirements of the other elements was chosen.
2.2.1. Intelligence Gathering Payload
This design element is considered the most important from the competition’s perspective. Gathering situational awareness is the foremost requirement of the competition, ergo the team locked down this subsystem in the early stages of the design. This system also dictated the “payload fraction” and “payload aperture” requirements of the airframe design team, thereby making the IGP the most critical design element. The IGP consists of two functionally different elements:
Imagery System Our previous year’s imagery system comprised of a Canon G10 for capturing images, which after rigorous testing, was found incapable of giving the required image quality, especially for character recognition. After analysing the imagery system requirements, a DSLR camera was sought as a suitable upgrade. All DSLR cameras supported by libgphoto2 library were listed and further shortlisting was done based on camera properties, weight, camera availability and budget. Various tests inside laboratory and during test flights revealed that Canon 500D was able to capture images at a higher rate as compared to Canon G10. This ensured that UAS covers entire search area by providing significant overlap of land between consecutive images.
Parameter Canon G10
Canon EOS 500D
Nikon D5100
Comments
Resolution (MP)
14.7 15.1 16.2 12-16 MP was found to be optimal to meet requirements.
Sensor size (mm)
7.44 X 5.58
22.3 X 14.9 23.6 X 15.7 Larger sensor captures more light and gives better quality image.
Shutter Speed(s)
1/4000 to 15
1/4000 to 30 1/4000 to 30
Fast shutter speed is preferable.
ISO 80 - 1600 100–3200 100 – 6400 High ISO crucial in low light conditions
Weight 14.15 oz. 38.51 oz. 41.58 oz. Canon 500D is lighter among other DSLRs
Price (USD) 509/- 691/- 619/- Prices nearly equal.
Availability Low High Moderate In Indian markets.
(* green color highly favorable *yellow color moderately favorable * red color not favorable)
Table 2: Comparison between Canon G10, Canon EOS 500D and Nikon D5100
A B: B derives requirements from A Airframe
GNC IGP
Communications
Figure 2: Relationship between the requirements of sub-systems
Delhi Technological University
5
SRIC It was decided that gathering data from the SRIC would be accomplished using a wireless network adapter because of the simplicity of the approach and minimal weight addition. The selection of the network adapter was done by limiting the physical size and weight of the adapter and connectivity, as shown in Table 3:
Table 3: SRIC Requirements
Previous year’s UAS failed to connect to the SRIC because of the large turning radius of the vehicle. To improve the probability of data acquisition from SRIC unit, the UAS was deemed to have a lower orbit radius and higher communication range.
2.2.2. Guidance and Navigation
The single component affecting the performance of the navigation system is the autopilot. The team surveyed different COTS (Commercial-Off-The-Shelf) and open source autopilots. The team’s previous year’s UAS used Piccolo II autopilot from Cloud Cap Technology, which gave satisfactory performance. Piccolo II, initially served as a reference autopilot used to compare other, low-cost alternatives. The main contender to Piccolo II in the open source domain was ArduPilot Mega 2.5, which was the team’s initial choice. However, preliminary flight testing revealed serious design limitations in the APM 2.5. The performance of the APM, even after hours of tuning, was nowhere near the ballpark required for the competition and hence the APM 2.5 was replaced by the Piccolo II as the GNC unit.
2.2.3. Airframe The team had previously been using a commercial, off the shelf airframe - Sig Rascal 110 that supported a payload of 2 lbs..However, Sig Rascal was deemed unsuitable for this year’s mission due to the increased payload weight, low wind tolerance and smaller turn radius that was demanded. The team thus decided to develop a custom airframe to meet the unique requirements of the system.
2.2.3.1. Design Objectives
According to the requirements analysis of the aircraft, a Statement of Objectives (SOO) was made which set the basis of the aerodynamic and mechanical design. Table 4 lists the SOO for Aarush-M. The complete design description of the airframe is given in UAS Design section.
Parameter Objective Threshold
Gross Take-off Weight (GTOW) < 25 lbs. < 55 lbs.
Endurance > 150 mins >30 mins
Payload >13.2 lbs. >10 lbs.
Take off Distance <80 ft. <150 ft.
Landing Distance <80 ft. <150 ft.
Min Control Speed 26.2 ft./s 49.2 ft./s
Turning Radius < 98 ft. < 164 ft.
Wind Tolerance >20 knots >15 knots
Table 4: Statement of Objectives
Parameter Limiting value
Weight <2.5 oz.
Form Factor 3.5” X 3.14” X 0.8”
Connectivity Ethernet
Frequency of operation 2.4Ghz
Delhi Technological University
6
3. UAS Design Description
Figure 3: System Overview
2.4 GHz 5.8 GHz
2.4 GHz
AARUSH-M
Delhi Technological University
7
3.1. Air Vehicle
Team UAS DTU chose to develop a custom airframe with a wingspan of 122” and having an empty weight of 32 lbs.. The airframe has been designed and fabricated by undergraduate mechanical engineering students of the team. It is powered by Hacker A80-8 electric motor with a 22x8 propeller. A single axis gimbal for the camera has been integrated with the system, to provide +/- 45 degree roll compensation.
Considering the time constraints, the prototyping of the airframe was broken into three phases for rapid development:
i) Conceptual and Preliminary Design ii) Fabrication and Developmental Tests iii) Flight Testing and Evaluation
3.1.1. Conceptual & Preliminary Design
A preliminary weight estimate was deduced from the statement of objectives and data gathered about other UAS belonging to the same class as Aarush-M. MATLAB was used for all the theoretical analysis. Various design specifications of different small class UAS (10 – 100 lbs.) were studied and this statistical data was used to estimate the weight using regression analysis. This gave a good initial estimation of 35 lbs.. Once the weight was determined, the SOO and four parameters namely stall speed, take-off distance, landing distance, turning radius were used to construct a constraint analysis graph. A design space for the airframe which would meet all the threshold requirements was obtained as can be seen in the Figure 4. The power loading vs. wing loading plot depicts infinite number of points which satisfy the design requirements; ergo it was difficult to choose an optimum value. Few points for low power loading and wing loading were selected and compared on the basis of overall scoring with weightage assigned to different parameters. The highest scoring point in the design space was selected.
Table 5: Parameters Weightage
Since practical outcome and performance always deviate from the theoretical analysis and estimation, sensitivity charts for GTOW vs. Payload, Endurance were prepared. This gave an opportunity to assess the changes in performance with the changes in design value.
Output of Conceptual Design – Specifications of the Air Vehicle
3.1.2. Wing analysis and configuration selection
A variety of low speed, high lift airfoils were analyzed with the help of XFLR5 and their drag polars and
GTOW Endurance Payload Take off Run Landing Distance Min. Control Speed
35% 20% 25% 5% 5% 10%
GTOW 35 lbs. Wing Span 122 inches
Power required 3.8 hp Wing Area 11.02 sq ft.
Take-off Distance 110 ft. Payload 7 lbs.
Landing Distance 100 ft. Endurance 20 mins
Figure 4: Power Loading vs. Wing Loading
Delhi Technological University
8
lift curves were studied. Simulation for wing analysis was performed based on various computational models (Lifting Line Theory, Vortex Lattice Method). The result thus obtained concluded the wing design.
Propulsion configurations such as pusher, tractor, twin-engine were considered as prospects for the design. Propeller efficiency, vibration isolation, flexibility in tail, manufacturability & weight were the key parameters which were used to discern these configurations. Pusher configuration with twin boom
inverted V-tail was chosen for being lighter than H-tail. Besides being lighter, inverted V-tail also gives an advantage of proverse yaw which increases the wind and gust tolerance of the UAS. The pusher configuration provides a larger field of view for the camera and better vibration isolation. Such a configuration also allowed the avionics system to be easily accessible. The wings have been designed for high turning rate and tolerate a structural load of 6 Gs. Figure 6 shows the finalized assembly of the airframe
Simulations were done to diagnose and fix the problems in flight characteristics and dynamic stability of the airframe. The stability and control analyses were performed in AVL and a full 6 DoF simulation was done in X-Plane. The two simulations concurred, with a static margin of +10% giving a satisfactory result, which was chosen for the airframe.
3.1.3. Fabrication and Developmental Tests
One of the ancillary objectives of the team was to develop a robust aerial platform fit for indigenous research, besides performing in SUAS ‘13. The fabrication process for the new airframe was carried out completely in the UAS-DTU lab at DTU. The fuselage features a monocoque shell design composed of carbon fiber/epoxy sandwiching balsa sheet for additional stiffness. Sandwiched laminates of carbon fiber, glass fiber and balsa sheet were made for testing and experimentally determining their strengths which would be further used for wing and tail skins. As a result 200 GSM glass fiber (45 degrees) and balsa sheet were used. The 45 degree orientation provides much greater load transfer and shear strength. Wing spar was constructed according to the structural calculations which gave the load and the bending moment along the spar. Unidirectional CF strip, 400 GSM CF and balsa wood was used to build the spar. It was subjected to cantilever destructive test failing at 168 lbs. where as it was designed for 124 lbs. with a factor of safety 2, thereby passing the test with a good margin. The design features twin CF booms and solid spring CF landing gear. Landing gear, wing & tail skins were manufactured using CNC cut medium density fiber molds. The wing assembly consists of two
Figure 6: Final CAD Figure 5: Wing Analysis
Figure 7: Load vs. Wingspan Plot
Figure 8: Wing Spar destructive test
Delhi Technological University
9
Figure 12: Hacker A80-8 Brushless DC motor Figure 13: Aarush-M at Karnal Airport
outboard sections and one mid-section. To reduce the time in assembly and easy replacement standard bolts of 4 and 5 mm have been used.
3.1.4. Propulsion System The team studied two options – a two stroke engine and an electric motor, to meet the power requirement of 3.5 HP. A comparison chart was prepared, based on prior experience, as shown in Table 6, where green color indicates a favorable condition and red indicates an unfavorable condition.
A two stroke 50 cc DA engine was tested, but encountered several mid-air engine failures leading to emergency landings. The reliability of the engine was not satisfactory, especially at elevated temperatures. Operational factors such as maintenance, troubleshooting etc. deemed the engine unfit for operation with Aarush-M. However, these risks were mitigated by an electric motor which provided equivalent thrust and higher reliability, low acoustic signature and almost no maintenance.
MotoCalc was used to compare different motors and propellers. A brushless DC motor – Hacker A80-8
powered by three 10S 5000mah Lithium Polymer batteries with a propeller of 22x8 was selected. This
propulsion system provided an endurance of 20 minutes under static conditions. The result was in
accord with the time required to complete the mission.
Table 6: Propulsion System Selection
Parameter Two Stroke Engine
Electric Motor
Reliability
Vibration
Endurance
Maintenance
Weight
Figure 9: Fuselage Bottom Skin Figure 10: Mold Pattern Figure 11: Left Wing
Delhi Technological University
10
3.1.5. Power System Design and Layout of Avionics
The power sources were selected on the basis of their size, weight and energy density. Lithium Polymer chemistry was chosen because of its high discharge capacity and high energy density. The battery
capacity was optimized recursively with flight tests so as to save weight. The power of the control surfaces actuators was kept separate from the avionics system. A switch board was placed under the avionics hatch to allow selective powering of components during ground testing.
Table 7 shows the power requirements chart which was prepared for battery selection once the on-board components were specified:
Avionics Component Mission Ampere Hours (30 minute flight time)
Power Source
Factor Of Safety
Piccolo II Autopilot System 600 mAh
14.8V 3900 mAh Li-Poly
1.9 PandaBoard 350 mAh
DLink Network Switch 200 mAh
SRIC Wi-Fi Router 300 mAh
5 Ghz Wireless Router 600 mAh
Canon EOD 500 DSLR 600 mAh 7.4 Wh 1000 mAh Li-ion 1.67
Actuator servos 1350 mAh 7.4 V 2600mAh Li-Poly 2 Table 7: Power Requirement Chart
Separating the control surface power and avionics power sources had twofold advantages:
Increased reliability of aircraft control: In case of avionics power failure, the aircraft control systems shall remain active. This increases the reliability of the UAS as a whole.
Eliminating loading effects at servos: Isolating the power at servos precludes the dropping of voltage at their input below their operating point i.e. 4.8 V. It was been empirically ascertained that running the avionics and control servos simultaneously from the same 5V source resulted in loading effects which may lead to terminal voltage at servos dropping below their operating point i.e. 4.8 V.
Figure 14 Layout of avionics on-board Aarush-M
Delhi Technological University
11
Figure 15: Onboard Imagery Peripherals
3.2. Payload
3.2.1. Imagery System Payload
The team opted for a user-centred approach for the effective operation of imagery system during mission. The imagery system is designed to require minimum human intervention and deliver minimum false positives. Several use cases were designed to test out the individual features of the software and
hardware system, and various test flights were conducted to simulate the mission which helped in identifying bugs and bottlenecks. It was observed during flight tests that aircraft banked as much as 40 degrees. Hence, the Canon EOS 500D is housed inside a gimbal which is roll compensated up to +/-45 degrees. The competition objectives also require imagery system to be capable of analysing the off-centre target which could be up to 250ft. cross-range. The gimbal is capable of being controlled via a joystick at the ground station when put into manual mode, to accomplish this objective.
3.2.2. SRIC System Payload
Data from SRIC is accessed via network adapter that transmits the file data via the Imagery link itself. The Imagery router is used for the SRIC data downlink because the amount of data transferred from the SRIC is a) intermittent and b) small enough to not hinder any pending Imagery data transfer for more than a few seconds. These assumptions were well justified when the setup was tested in the lab and during flights. The test procedures are described under the Testing and Evaluation section of this paper.
3.3. Data Processing
3.3.1 Image Processing
The mission objectives require the imagery system to perform Automatic Detection/Cueing,
Figure 16: SRIC Information Flow Diagram
Delhi Technological University
12
Classification, or Identification (ADCCI) on acquired aerial images in real time. The data processing unit was, therefore designed to be reliable, efficient and fast. Rigorous testing during test flights revealed more than eighty five per cent success rate of the entire sub-system.
a. Image Acquisition: The on-board computer runs a headless version of Ubuntu 11.04. It controls the camera parameters such as aperture, shutter speed, focus, and image quality etc. using libgphoto C library. The code running on Pandaboard on-board computer captures the images every three seconds to provide optimum overlap. Excess overlap is avoided to reduce computational overhead. As soon as the image is captured, the GPS information is stored in the image metadata as exif tags. These images are simultaneously transmitted to the Ground Station using a secured Wi-Fi link created by Groove Routers. Image transfer takes about 2-3 seconds which is equal to the time required to capture one image. This time interval has proven to be sufficient for real-time image processing within given mission time.
b. Graphical User Interface (GUI): The GUI was developed in C++ using QT library. The primary objective while designing the GUI was to reduce mission execution time by making the compilation of target data sheet easy for the imagery operator. The need to increase the speed of the GUI was catered by running few small processes that can run independently on separate threads. The visible components on the GUI are divided such that all data being processed is displayed on one screen while all processed target data is displayed on the other. This separates the active target-related data from diagnostic information which is not used during normal operation. The GUI also lets the administrator communicate directly with the on-board computer. It stores all processed targets in a SQL database common to all users and is capable of generating a text file for submission in accordance with the competition’s requirement. Screenshots of GUI are shown in Figure 17:
c. Image Analysis:
The image processing code for autonomous target classification and identification was written in C++ using OpenCV, an open source image processing library. To process about three hundred images in the allotted mission time, it is imperative for the image processing software to be fast and accurate. Thus, a laptop with NVIDIA Graphics Processing Unit is used to improve the image processing rate. The image processing technique has been described in the flow chart shown in Figure 18:
Screen 2: Contains data being processed
or the active data Screen 1: Contains Processed Target Data and
diagnostic information Figure 17: Image Processing
GUI
Delhi Technological University
13
Segmentation: Segmentation was found to be a crucial part for autonomous processing. A frequency tuned approach of segmentation to extract salient objects is used to segment targets from the images. The result of segmentation is a gray scale image in which salient objects (targets in our case) appear whiter. A graph cut based technique is then used to extract targets from its background.
Color Recognition: To recognize the colors, a histogram of colors for the target is generated using the Hue values from HSV color space. The highest peak of this histogram gives the shape color while the second highest peak gives the color of alphabet.
Shape Recognition: Previous year’s system used ray tracing technique to identify shapes of target. The technique was found to be highly reliable and was improved for distorted shapes also. Once correctly segmented, the unit can now recognize various polygonal and non-
Original Image Texture of ground flattened
using mean shift filtering
Saliency Map of Image
Extracted Target Using Mask Mask created using Graph
Cut Segmentation
Histogram of color distribution in
extracted target for color
recognition
Distance Theta curve for
Shape Recognition Letter extracted from Image for
recognition based on Eigen Space
Figure 18: Imagery Analysis Flowchart
Delhi Technological University
14
polygonal images with 78% accuracy.
Letter Recognition: Letters are first segmented from the target using graph cut approach. This is done using the method of scale and rotation invariant letter recognition based on Eigen spaces. This letter recognition module was trained by an artificial dataset. The implementation resulted in a success rate of less than fifty percent, with certain characters being recognized only if they were well defined in the image.
3.3.2. SRIC Data Acquisition
Data acquisition from the SRIC required manual intervention from the payload operator to access the file. This was one of the reasons why the team couldn’t access the file last year. After selecting the payload required for this task, it was decided to automate the process by writing scripts for both Windows (batch file) and Linux (shell script), which allowed us to perform this task from either the Mission Control Centre (MCC), or the Information Gathering Station (IGS). The script, when given the required parameters of the SRIC’s remote laptop and router e.g. IP Address, Username, Password etc., tries persistently to connect to the FTP server running on the remote laptop. Once a connection is established, it navigates to the specified directory, extracts the text file, and stops execution upon successful file transfer.
3.4. Communications
There are three communication channels between the air vehicle and the ground station:
Manual R/C control
Telemetry downlink and uplink
Imagery downlink
The manual R/C control is the most critical link and utilizes a 2.4 GHz frequency hopping spread spectrum transmitter receiver to ensure a robust link, and allow manual override at any time. Such a modulation technique provides superior noise immunity as compared to FM/AM transmitters.
The telemetry downlink and uplink is done via 2.4 GHz Microhard transceivers that are part of the autopilot package. The frequency was chosen because the other option of utilizing the 900MHz band is not possible without licensing in India.
The payload connectivity requires large bandwidth to keep latencies to a minimum. As a result, a 5.8GHz wireless router is used to communicate with the air vehicle. The TCP/IP protocol of the router ensures that the transmitted packets are delivered at the ground station.
3.5. Ground Control Station
The PGS setup makes the gathered intelligence easily accessible to the judges and the operators. A portable ground station has reduced our setup time by a factor of 10 and thus allows Aarush-M to be ready for deployment in less than 30 minutes, as stipulated by the competition rules.
The ground station supports a maximum of three payload operators: one administrator and two other users. Each user runs an independent Graphical User Interface that shares same database over wired network. A separate MCC operator laptop displays the telemetry data from the UAV.
The PGS requires an 110V AC power source and provides the user ample control over the power of various system components, thereby allowing the controller to switch off / reset them when needed. Spare power outputs are also given to allow future expansion. It also provides an umbilical power cord
Delhi Technological University
15
to power the RC transmitter separately which precludes battery drain in ground testing. A charging port for transmitter battery is also incorporated to charge the same during long flight hours.
3.6. Mission Planning
At the onset of preparation for the competition, the mission plan was based on retrospection of previous year’s performance, this plan, was evolved with each flight test to adapt to the new system’s capabilities.
Figure 19: General Flight Plan Profile
It was observed that the initial waypoint navigation, including the take-off took about 3 minutes, while one traversal of the search area in progressive wave pattern took about 4.5 minutes. Since sufficient overlap is maintained between consecutive aerial images clicked, only one round of the search area was sufficient. Extraction of text file from SRIC required less than 30 seconds. The pop-search area was also tried and it was found out that its traversal took about a minute. These time estimates helped us develop a general flight plan, which would be used for mission during competition.
4. Flight Testing and Evaluation Results
The test flights were methodically scheduled to test and tune the performance of each major sub-system rigorously, with minimum risk, to ensure that the system gets enough flight time to be reliable and worthy of a competition entry.
Subsystem F1 F2 F3 F4 F5 F6 F7 F8 F9 F10 F11 F12
Autopilot R/C Lateral Tuning Longitudinal Tuning
Waypoint Navigation
Autonomous Takeoff & Landing
Buffer Flights
Imagery Camera Parameters Selection
Altitude Optimization
Competition rehearsal – real-time code run; sweep pattern analysis
SRIC Static Testing Altitude Variation
Orbit Radius Variation
Transferred File Changes
Competition rehearsal
Table 8: Flight Testing Schedule
4.1. Navigation Performance
Improving the navigation system performance is central to the success of the mission, since it directly impacts the performance of imagery and SRIC data acquisition. Prior to the flight testing phase, the dynamic model of the aerial-vehicle was developed using AVL, R/C test flight results and vehicle design parameters. This dynamic model was used in Piccolo II’s proprietary simulator to carry out control law tuning in software-in-the-loop simulation environment.
The optimum gains found during simulation were then used as initial points for in-flight gain tuning. The gains were tuned iteratively, until the desired performance of the various command loops was attained.
Delhi Technological University
16
Canon G-10
Canon EOS 500D (DSLR)
Figure 21: Target Image Quality Comparison
Following two images show the waypoint navigation performance during normal day and a windy day.
Figure 20: Clockwise from top-left: Waypoint navigation on normal day (wind = 2 m/s); Waypoint navigation on windy day (wind = 7 m/s); Bank angle strip chart during orbit
4.2. Payload Performance
4.2.1. Imagery
4.2.1.1 Target characteristics
Accuracy of the imagery system was validated by various tests inside the laboratory and in test flights. Images clicked by Canon EOS 500D showed significant improvements in image quality in comparison with the images clicked by Canon G10. For example, target acquired from images captured by Canon G10 and by Canon EOD 500D are shown in Figure 21.
Data analysis unit was rigorously tested for different target shapes, colors and letters. It was found that
S. No. Parameter Result
1. Bank angle tolerance 2 deg
2. Altitude tolerance 13 ft.
3. Waypoint tracking tolerance 10 ft.
4. Airspeed Tolerance 4.9 ft./s
5. Auto-takeoff 5 successful attempts
6. Auto-landing 1 successful attempt Table 9: Autopilot performance specifications
Delhi Technological University
17
processing an image took about 4 seconds, which makes it reliable for real time applications. The current image processing software can segment targets autonomously with 72% accuracy. This number would increase on a ground with lesser pattern variation which gives lesser false positives. The shapes that can be identified autonomously include star, cross, circle, semi-circle, triangle, square, rectangle, arc, trapezium and rhombus with 78% accuracy. The accuracy for character recognition was found to be less than 30%. This number is low, because of the size of the character in images, noise and other complexities involved in recognition. The Table 10 shows the results obtained during test flights with few of the target types.
(* green color correct * red color incorrect *grey color Not Analyzed) Table 10: Imagery Test results for a given set of inputs
4.2.1.2 Target Location and letter orientation Determination of the target location requires vehicle’s GPS coordinates, altitude and heading information. These values are tagged by the on-board computer in each of the image. If the plane’s GPS location at the time of capturing is considered as the actual target location, the maximum error was found to be within the threshold value of 250ft. It can be calculated as follows(refer to Figure 22):
Max error will occur when target is present at corner of the image.
√(
)
(
)
Where, VFOV is Vertical field of view = 42.2°
HFOV is Horizontal field of view = 63.3°
To improve the target location estimation a set of mathematical equations were used which utilize the latitude, longitude, altitude, heading and camera field of view to transform the target’s pixel coordinates into the actual GPS coordinates. This code, however, gave poor results in test flights. It was later
S. No.
Cropped Image
After Segmentation
Shape Shape Color Letter Letter Color
1
Semi-Circle Red Not analyzed White
2
Square Blue P Yellow
3
Star Sea Blue A Sea Blue
4
Semi-Circle Sea Blue Not analyzed Red
5
Triangle Pink Not analyzed Not analyzed
6
Cross Red Not analyzed Red
7
Circle Yellow T Grey
8
Semi-Circle Pink Z Blue
9
Triangle Sea Blue D Pink
10
Rhombus Yellow T Grey
Vehicle’s
GPS
Figure 22: Max GPS Error
Delhi Technological University
18
0
50
100
150
200
0 100 200 300 400
Tim
e Ta
ken
(se
con
ds)
Orbit Radius (feet)
Transfer of 1 MB of data at different altitudes and orbit radii
200 ft
300 ft
400 ft
500 ft
600 ft
observed that the GPS heading information was being updated with a delay of 5-10 seconds. This problem is being corrected as of this writing by using a better GPS unit.
4.2.2. SRIC Having missed out on extracting data from the Simulated Remote Information Center (SRIC) in the 2012 SUAS Competition, the team was keen to perform rigorous testing for our SRIC setup to successfully execute it in the competition. Testing of the SRIC was done in three phases:
1. Lab testing: The test environment was set up in the lab to verify the functioning of all systems. This testing phase was used to debug and tune the scripts for automated data extraction.
2. Altitude and Orbit Radius variations: The plane’s altitude and orbit radius above the SRIC was varied and the access time was recorded, if the file was received at all. The result from these tests showed that for best results, the plane should fly somewhere between 200 – 300 feet above ground level, with
orbit radius 150 feet. 3. Changes in transferred files: Further tests were conducted to figure out the maximum amount
of data that could be acquired from the SRIC in under a minute. To do this, the size of file(s) to be acquired was increased, starting from a simple text file, all the way to a video. The results showed that at least 3 MB of data could be accessed in under a minute with optimum flight conditions. (Note that every trial required around 5 seconds making the initial connection. The Time Taken value does not take this into account)
Type of File Size (MB) Time Taken (seconds)
Text File 0.1 0
JPEG Image 0.5 2
PDF File 2.2 13
MP4 Video 26.4 204
5. Safety
The competition demands special attention to safety of personnel and the UAS. The team approached the development of each design element keeping in mind these crucial criteria. There are two levels of safety measures adopted by the team:
5.1 Safety in design of the UAS
Safety in design of all the sub-systems was accounted for at the onset of development. The degree of safety was quantified by a number called “factor of safety” (FOS) which provided a safety margin for the design of all critical elements. A higher FOS implied greater the safety margin and hence, more reliability. The FOS also accounted for theoretical and fabrication inaccuracies which were revealed only
Figure 23: Time taken vs. Orbit Radius for different altitude
Table 11: Time taken for acquisition of different files
Delhi Technological University
19
in the later stages of the development.
The foremost safety requirement from the UAS design was the accessibility of manual override in case of autopilot control failure. Such a failure could happen due to autopilot power loss, or accidental faults in the autopilot output lines. The team mitigated these sources of error by separating the actuator and avionics power, and incorporating an RX MUX that allowed manual override any time during the flight. These corrective measures ensured that the actuators would be powered in the case of avionics power loss, and that the safety pilot will be able to take control of the plane.
Most of the faults that have compromised the mission have, in the team’s experience, been traced to electrical sources. The team has taken extra measure of precaution, by using conduits to protect wires from physical damage and extend their “mean time between failures”. The selection of wire gauges and insulation types was made keeping in mind the required ampacity, operating temperature and a conservative factor of safety.
Booster extensions have been used at the output of the RX MUX, in order to prevent loading of its outputs and signal loss at the actuator. The connectors between the different components of the UAS were identified as major failure points, and all of them have either been substituted by “positive locking type” Molex connectors or, encapsulated by servo extension locks.
The airframe has been designed to handle load factors in excess of 4 which is much higher than the normal flight envelop. The camera is one of the most critical components on board the UAS; ergo, it has been recessed inside the fuselage to ensure its integrity even in the case of a main landing gear collapse.
5.2 Safety in mission execution/operation the UAS
While measures were taken in design process to ensure a safe build, the operation of the UAS ensures that a safe design stays safe for long periods of time. The team tackled this problem by characterising the “Mean Time between Failures” (MTBF) for critical components. This number allowed the team to anticipate fatigue of components and thus replacements were made accordingly.
The safety of the UAS during competition and flight testing is ensured by conducting rigorous pre-flight checks. These checks have evolved with the team’s experience with failures and ensure that the UAS is fit to fly prior to mission execution.
Efforts have been made to minimise human errors during mission execution by establishing a communication protocol amongst the ground crew. The ground crew now follows a chain of command to ensure that only the right people are making the relevant choices, thus avoiding errors. A pre-determined mission protocol is in place to ensure that there is no panic in the case of any crisis.
The safety officer is accoutred with a master checklist and is in-charge of overall safety of the mission. He is second in command, after the flight director to call off the flight in the case of any aberration in system performance.
A rigorous range check is performed to ensure that the airplane stays under manual control for the desired airspace boundary and there are no glitches in control surface actuation.
Delhi Technological University
20
5.3 Failure Mode-Effect Analysis
The team used the flight testing experience to identify risks and develop multiple contingencies for each
fault. A detailed Failure Mode-Effect Analysis was carried, out and the mission status after each failure
mode is classified as follows:
Failure Mode Indication Effect Primary Response Secondary Response
Telemetry Link Loss
‘Link’ Indicator turns red at MCC terminal, unusual
navigational response
Link between 60% to 80%
Observe Autopilot telemetry for link
improvement.
Observe Autopilot telemetry for link
improvement.
Link less than 60%
Observe Autopilot telemetry for 15 seconds for link improvement.
Switch to manual and troubleshoot
communication link.
Image Acquisition System Failure
Image synchronisation
fails or is unresponsive
Image processing
possible but slow
Observe link for 2 minutes for
improvement
Reset router power and observe link
Image processing not
possible
Reset router power and observe link again
Emergency landing to troubleshoot imagery
subsystem
Mission Control Centre computer
crashes
Command Centre hangs or Shuts
down
Autopilot Navigation
Affected
Shift to R/C, meanwhile backup
computer brought in
Resume mission after setting up backup Autopilot terminal
Imagery Terminal Crashes
No output on screen
Image Processing
affected
Terminal restarted, backup image
processing terminal brought in
N/A
Avionics or Propulsion
Battery level unsafe
Indicated on PCC plugin
Flight Endurance Affected
Emergency landing within three minutes
Swift battery replacement and take-off
Motor cut-off Continuously falling Airspeed and/or Altitude
Flight Stability affected
Shift to R/C and emergency landing
engaged
Swift battery replacement and take-off
Component Disintegration
Falling debris, erratic behaviour
Aircraft integrity affected
Shift to R/C and emergency landing
engaged
Quick ground assessment and take-off if feasible
Unable to hold altitude/Enters
no fly zone
Altitude or position error
observed on MCC
Autonomous navigation accuracy affected
Switch to manual, mission continues;
Adjust the control law gains
Switch to autopilot and observe
Code Blue: Mission Continues, Fully Autonomous
Code Yellow: Mission Continues, Manual Override
Code Red: Mission Haults, Emergency Landing
Table 12: Failure Mode Effect Analysis
Delhi Technological University
21
6 Acknowledgements
The Team extends its gratitude towards University Vice Chancellor Prof. P B Sharma for his constant
support and encouragement to the project. The team is indebted to its project advisor Prof.N S Raghava
for his timely guidance and motivation during the course of the project.
Team UAS-DTU would like to immensely thank Lockheed Martin Aeronautics Company for their
mentorship and financial support in the project. The team also acclaims the support of the former team
members who helped the team in preparing flight plans and execution of mission.
The team is grateful for the efforts of Mr. Jasvinder Singh, who was the safety pilot for developmental
test flights of Aarush-M.