LAGARI 2017 Technical Design Paper for AUVSI Student UAS Competition Yildiz Technical University Faculty of Mechanical Engineering Mechatronics Engineering Department Figure 1: Huma Mini UAV Preface The Lagari UAV Team has been founded at Yildiz Technical University in 2013. In the past years, Lagari created multiple innovative aeronautical & aviation project and obtained various awards. This year will be the first year in the AUVSI SUAS competition for the team. The Lagari Team consist of five undergraduate students from the department of Mechatronics Engineering of Yildiz Technical University. Abstract This paper presents the design and tests processes of the Huma Mini UAV for the AUVSI Student UAS Competition 2017. Huma is designed, developed, manufactured, operated and tested by Lagari Team. The aircraft was manufactured with using advanced composite technologies to provide maximum reliability and safety during the flight. The full autonomous avionics was integrated to the Huma in order to complete the all missions on the competition autonomously. Lagari 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 COMSOL CFD analysis and customized algorithms for ODCL, Obstacle Avoidance and Air Delivery tasks.
Contents 1. System Engineering Approach... 3 1.1. Mission Requirements Analysis... 3 1.2. Design Rationale... 3 1.3. Programmatic Risks and Mitigation Methods... 4 2. System Design... 5 2.1. Aircraft... 5 2.1.1. Aerodynamic... 5 2.1.2. Airframe... 5 2.1.3. Wing Design... 6 2.1.4. Landing Gear... 6 2.1.5. Propulsion & Power System... 7 2.2. Payload... 8 2.2.1. Autopilot... 9 2.2.2. Imaging System... 9 2.2.3. Communications... 11 2.3. Ground Control Station... 12 2.3.1. Mission Control Station... 12 2.3.2. Imaging Console... 12 2.4. Object Detection, Classification, Localiation (ODCL)... 12 2.4.1. Off Axis... 13 2.5. Obstacle Avoidance... 14 2.6. Air Delivery... 14 2.7. Interoperability... 15 2.8. Cyber Security... 15 3. Tests & Evaluation Plan... 15 3.1. Developmental Testing... 15 3.1.1. Wing Loading Test... 15 3.1.2. Wing Flow Test... 15 3.1.3. Propulsion System Trust Test... 16 3.1.4. Landing Gear Loading Test... 16 3.2. Individual Component Testing... 16 3.2.1. Autopilot System Tests... 16 3.2.2. Imaging System Tests... 17 3.2.3. Data Link Tests... 17 3.2.4. Obstacle Avoidance... 18 3.2.5. Air Delivery... 18 3.2.6. Interoperability System... 18 3.3. Mission Testing Plan... 18 4. Safety, Risks, & Mitigations... 19 4.1. Developmental Risks & Mitigations... 19 4.2. Mission Risks & Mitigations... 19 4.3. Operational Risks & Mitigations... 20 4.3.1. Autopilot Failsafe Safety Functions... 20 4.3.2. Pre-flight & Post-flight operational Checklist... 20 5. Conclusion... 20 Lagari UAV Team Technical Design Paper - 2017 2/20
1. System Engineering Approach 1.1. Mission Requirements Analysis Lagari Team designed and developed Huma to accomplish all of SUAS competition tasks with paying attention rules that based on AUVSI 2017 Rules for SUAS Competition document despite it being their first participation on SUAS competition. Lagari team classified tasks as primary and secondary according to team budget, time and grade. Primary tasks which are autonomous flight, search area, ODCL, actionable intelligence, emergent target, airdrop, interoperability were developed and tested. Secondary tasks which are stationary obstacle avoidance, moving obstacle avoidance, off-axis were in various development stages. Lagari team is going to attempt these mission, shown in Figure 2. -Take off -Waypoint Navigation -Landing AUTONOMOUS FLIGHT OBSTACLE AVOIDANCE -Stationary Obstacle -Moving Obstacle -Search Area&Off-Axis -Actionable Intelligence -Interoperability ODCL AIR DELIVERY -Air Drop Figure 2: Mission Table 1.2. Design Rationale In the entire design and production stages of Huma; equipments were selected according to the team budget and team's past experience. While creating Huma, specific focus points were set to achieve the best performance: - Reduce the weight of the aircraft without decreasing its strength - Make the aerodynamically most stable aircraft - Perform image processing correctly and effectively - Use reliable control card and parameters for autonomous flight - Provide high speed internet connection over long distances - Prepare a system that can do airdrops in a practical and accurate way Some choices were made to reach the specified focus points: - Carbon fiber composite material was preferred for the structure of fuselage and wings in order to provide lightness without reducing the strength. - The wings were positioned higher than the tail part and the dihedral angles were given at wingtips to achieve better stability. - High-performance on-board computer was selected for fast image processing and transfering. - A camera with has high resolution and auto focus was preferred. - M5 bullet was selected to provide desired link performance. - Pixhawk autopilot was selected for autonomously take off, flight and landing. - The airdrop mechanism was designed to be lightweight and with little friction. - In all of the design and selection process, safety was our most important criteria. All improvements were made with ensuring the reliability of our solution. Lagari UAV Team Technical Design Paper - 2017 3/20
1.3. Programmatic Risks and Mitigation Methods The teams can encounter with various risks and unexpected problems during the process of multidisciplinary engineering projects. Since the Lagari is a old-established and well-experienced team with a lot of experience on aviation, the risks were listed that may be encountered in management, production and test process based on our old experiences. The team created a mitigation plans to prevent being affected by them. The Lagari team adopted the agile systems engineering approach in order to facilitate the team management and to make the follow-up the team work plan within comfortable. The management and work plan of the team are carried out through the interactive Asana workspace. It s easy to manage multiple projects, tasks, and subtasks in Asana. That workspace represents a wonderful platform to facilitate collaboration. Asana is a significant upgrade from using excel or emails to manage projects and tasks. Table 1 shows the risks and mitigation methods during designing and building a UAS system. Figure 3: Asana workspace Risk Cause Likelihood Impact Migration Plan Delay on manufacture and test process Damage on the airframe structure Incapability of team members to design UAV for SUAS Financial inadequacy of Team - Air conditions, - Lack of spare component and workshop -Vibration, -Crash during flight test, -Hard landing - Inexperience at the SUAS competition - Lack of sponsorship MEDIUM (%34-66) HIGH HIGH LOW (% 1-33) MAJOR MAJOR MINOR (%1-33) MAJOR -Gantt chart plan was prepared for test process. -Components were backed up during the purchasing. -Composite workshop was founded. -Damper was used with a fittings elements. -Carbon fiber material was selected as main structure and three aircraft were produced for tests. -The Journal Papers of SUAS 2016 were investigated by paying more attention. -Sponsorship agreements were made with companies working on the aerospace industry. Lack of collaboration and communication between the team members Selection of electronic equipment that can not accomplish all tasks - Students from various faculties - Classic project management methodology - Incorrect selection of avionics MEDIUM (%34-66) LOW (% 1-33) MODERATE (%34-66) MAJOR -The team consist of department of Mechatronics Engineering. -Agile Scrum was adopted as a project management methodology. -Comprehensive literature review was done before the selection of the components. Table 1: Risk Management Plan Lagari UAV Team Technical Design Paper - 2017 4/20
2. System Design 2.1. Aircraft 2.1.1. Aerodynamic While Lagari Team is planning their strategy for AUVSI SUAS 2017 competitions, the team realized that the aircraft must have advanced aerodynamic stability and high maneuverability to obtain maximum points from missions. The aerodynamic sub-team did comprehensive literature review in parallel with this purpose. Asymmetric Eppler 214 airfoil with high camber ratio was used for improved stall characteristic of wings and to get more lift force at zero degree angle of attack(aoa). In order to improve maneuverability, the AR was chosen in accordance with the optimum aspect ratio range of aerobic aircraft models. The tapered trapezoidal wings platform preferred for keeping the Aspect ratio in this range. The wings have five degree dihedral. The dihedral angle provide to keep spiral dive stability characteristics on the lateral axis so the aircraft can be at controllable levels with Pixhawk autopilot unit. Also, the main wings generate more lift force with the dihedral. The most important factors in aircraft design processes such as airfoil selection, Center of Gravity (CG) localization, wings-tail sizing and localization were analyzed in low Reynolds numbers with COMSOL Multiphysics and XFLR5 CFD analysis programs which were based on VLM (Vortex Lattice Method). Maximum aerodynamic efficiency was obtained with comparing the analysis results. As the result of the analysis, the system was designed at low stall speed, high cl /cd ratio (aerodynamic efficiency), stable in all AoA under the stall angle on the longitudinal axis. As control surface, Flaps were not found necessary on wings because of the high camber ratio of airfoil, advanced aerodynamic cl/cd ratio and low stall speed characteristics. Ailerons were designed as length about % 60 of half wingspan, and width about %25 of root chord. This sizing provided to the UAS greater maneuverability. Figure 4: XFLR5 VLM Analysis 2.1.2. Airframe Since the fuselage of the aircraft hosted all of our electronic equipment and fitting elements, more attention was paid to its volume and strength. It was aimed to be easily accessible to all avionic components in the fuselage during the 3d cad design process. It is possible to produce high strength and lightweight material by advanced composite technology. For this reason, the manufacturing team has decided to produce by using process of composite laying on the wooden mold. Completely carbon fiber fuselage was obtained from wooden mold by vacuum method. According to result of the strength and durability analysis, an extensive lamination plan was optimized and fuselage durability was increased by using different types and amounts of carbon fiber on areas where the loading was most exposed during the flight. Carbon fiber reinforced balsa spars were used into the fuselage in order to eliminate lateral buckling on the fuselage and mount the payload. Angle of incidence was determined with respect to result of aerodynamic analysis. Figure 5: Composite manufacturing of fuselage Lagari UAV Team Technical Design Paper - 2017 5/20
Three degree wing slots were created on the fuselage with the purpose of easily mounting the main wings with this angle of incidence. Carbon Fiber pipes were used as spar to increase the strength and prevent the vibration for the assembly of the wings with the fuselage. In order to decrease fuselage weight and improve durability at same time, carbon fiber pipe was used between the fuselage and the tail part. 2.1.3. Wing Design The wingspan length is 2700 mm. Wings taper off from the root to the tip so that chord length at the wing roots are 320 mm and the chord length at the wing tips are 200 mm. Also, the main wings have three degrees of angle of incidence. The cl / cd ratio reached maximum level with this incidence angle in CFD analysis. Symmetric airfoil was used on the tail part. The elevator was designed widely for providing more lift during takeoff and the fin was also designed with the purpose to improve the maneuverability of Huma. Lightweight extruded polystyrene (XPS) foam was preferred as the main material of the wings structure because it is more ductile than balsa wood. In the production process of main wings and the tail part from xps foams, our own cutting mechanism was used by hot NiCrom wire. Foam molds of wings was prepared with this production technique. The wing was covered with carbon fiber using vacuum method to obtain the final shape of the wings. The wings were supported by carbon fiber pipes as spars extending into the fuselage to increase the strength and durability. Carbon fiber ribs were inserted into the foam in the production to prevent vibration of the carbon pipes during flight. The foam weight was decreased without damaging the durability and strength by using lattice structure. For easy transportability, wings were designed and manufactured to be detachable from the fuselage.(figure 6) 2.1.4. Landing Gear The design, analysis and production of the Huma s landing gear were made to provide the desired balance and durability. Tri-cycle landing gear consisting of main and nose parts was chosen due to the advantage of take-off/landing run and stability during taxi. The wheelbase of main landing gear was determined to provide the necessary force absorption during landing and stability during rotation. Localization of the main and nose landing gear was designed with calculating the weight balance to be 80% to 20% with reference to the CG. Figure 6: Comsol Wing CFD Analysis Figure 7: Landing Gear Ansys Loading Analysis The height of the landing gear was calculated so that the tail and the propeller would not hit the ground. Using carbon fiber on wooden mold, landing gear was produced by vacuum technique. The areas exposed to the highest load according to the ANSYS analysis, were supported by extra carbon fiber. No Single main Bicycle Tailgear Nosegear Quadricycle Multibogey Human leg 1 Cost 9 7 6 4 2 1 10 2 Aircraft weight 3 4 6 7 9 10 1 3 Manufacturability 3 4 5 7 9 1 10 4 Take-off/landing run 3 4 6 10 5 8 2 5 Stability on the ground 1 2 7 9 10 8 5 6 Stability during taxi 2 3 1 8 10 9 - Table 2: A comparison among various landing gear configurations (10: Best, 1: Worst) Lagari UAV Team Technical Design Paper - 2017 6/20
Figure 8: Huma s Dimensions(in mm) GENERAL CHARACTERISTICS PERFORMANCE MAIN WING Length 2.10 m Stall Speed 23 Kts (12 m/s) Airfoil E214 Total Wingspan 2.70 m Cruise Speed 31 Kts (16 m/s) Span 2.70 m Empty Weight 3.50 kg Max Speed 48 Kts (25 m/s) Area 0.66 m² MTOW 8 kg Wing Loading 12 kg/m² Aspact Ratio Rate Of Climb 250 m/min VERTICAL STABILIZER HORIZONTAL STABILIZER Minimal Turn Radius 20 m Airfoil NACA 0012 Airfoil NACA 0012 Maximum Flight Autonomously 45 minute Span 0.30 m Span 0.41 m Operational Range 4 km Area 0.07 m² Area Operational Ceiling 450 m Aspect Ratio 1.9 Table 3: General Features Aspect Ratio 8.7 0.20 m² 2.1.5. Propulsion & Power System The motor has to provide sufficient thrust throughout the whole flight. DC brushless electric motor has been chosen due to its light weight, efficiency and ease of use. We chose Axi 5320/18 which is brushless outrunner motor with respect to the result of CFD analysis in order to provide about 55 N thrust force. It has been used for producing thrust with suitable propeller and power system, shown in Table 3. MOTOR PROPELLER BATTERY RPM/V 370 Diameter (in) 18 Voltage(V) 22.2 No load current[a] 1.4 Pitch (in) 10 Capacity(mAh) 9120 Maximum current[a] 78 Weight(g) 109 Weight(g) 2000 Table 4: Propulsion & Power system specifications. 3.6 Lagari UAV Team Technical Design Paper - 2017 7/20
Power system has two components; Three 6S LiPos are used for powering servo rails with 5V voltage regulator, Pixhawk power module and motor with ESC. 4S LiPo is used for powering Jetson TK1 and Bullet M5 with 12V voltage regulator. 6S LiPo; Servos Pixhawk Motor 4S LiPo; Jetson TK1 Bullet M5 These LiPos were packed by the eletronic sub-team and they provide 45 minutes flight time. Also, the system has two separate power systems for image processing and flight control systems. Therefore, each system is protected at any failure for safety flight. 2.2. Payload Figure 9: Payload System The electronic systems which required to accomplish all tasks are shown in Figure 9. The power system at the bottom left of the table feeds the autopilot system and the power system on the top left also supports image processing system. Pixhawk flight control card has been used in autopilot unit. Image processing is performed by Jetson TK1 with Sony A6000. There are separate two screens for autonomous flight and image processing in the ground station. Communication is provided between Huma and the ground station via three diffirent data links. The main system consists of three subsystems as follows: 1. Autopilot(AP) & Flight Control System 2. Imaging System 3. Communications Lagari UAV Team Technical Design Paper - 2017 8/20
2.2.1. Autopilot Flight Control System has satisfied fundamental five requirements. These requirements: 1. Send data about aircraft s altitude, location and airspeed to the ground station. 2. Autonomous Airspeed; Must not exceed 70 knots, Should be up to 23 knots which is greater than stall speed. 3. Navigate to pass current waypoints autonomously. 4. Navigate without leaving the flight boundaries during the all missions. 5. Support with manual Radio Controlled (RC) flight. There are many flight controller such as Pixhawk 2, APM, Flymaple, Multiwii etc. The Pixhawk has been used as the flight controller because it satisfies subsystem requirements with many advantages. Firstly, Pixhawk has the faster microprocessor than other s microprocessors. The faster CPU obtains additional features and better performance during the flight. Secondly, Pixhawk has high sensitive sensors and provides more accurate data compared to other autopilots, therefore more accurate flight data is transmitted to the ground station. Receiving right data, which especially aircraft s height, velocity, is very important for airdrop, auto landing and auto take off, so Pixhawk is used with externally mounted GPS and Extended Kalman Filter (EKF) which is an algorithm to estimate aircraft position, velocity and angular orientation based on rate gyroscopes, accelerometer, compass, GPS, airspeed and barometric pressure measurements to reject measurements with significant errors. In addition, Pixhawk s failsafe function is a robust solution that ensures the safety of the entire system and support pilot manual override. It is a useful feature to prevent many failures. L.AILERON SERVO R.AILERON SERVO R.ELEVATOR SERVO L.ELEVATOR SERVO RUDDER SERVO AIRDROP SERVO MOTOR ESC BATTERY PPM RECEIVER TELEMETRY PIXHAWK LANDING SERVO PITOT SAFETY BUTTON POWER MODULE JETSON TK1 GPS+COMPASS x2 Figure 10: Flight Control System Architecture 2.2.2. Imaging System The imaging system consist of a On Board Computer (OBC), a camera, gimbal and 5.8G communication system. The imaging system takes and processes the image, calculates the location of targets and send all data to ground control station. Sony a6000 was selected as camera because of its fast shutter speed, high resolution and low weight. Sony a6000 has a 24-megapixel APS-C CMOS sensor and the weight of the camera is 344g.The camera was connected to Jetson TK1 OBC with an USB cable. Jetson TK1 is a multi-core GPU based computer which is developed for image processing. Simple CPU based implementation achieved approximately 4.5 frames per second (FPS) for processing time of the search area. To increase the speed of the process, Compute Unified Device Architecture(CUDA) parallel programming using a graphic processing unit(gpu) on an NVIDIA Jetson TK1 development kit was used. An approximate 10 times improvement in processing speed was achieved compared to simple CPU processing. Lagari UAV Team Technical Design Paper - 2017 9/20
Thus, the image processing algorithm can be operated in realtime with Jetson TK1. The OBC also communicates with the autopilot to take telemetry data. There is also a 2 axis gimbal with storm32 control board to fix camera angle and reduce vibration. Two brushless DC motors and 3d printed parts were used in Gimbal mechanism. Gimbal also allows to change the camera angle for off axis mission. After image processing and calculation of target points, all the data is sent to ground control station via m5 bullet 5.8G wi-fi radio. Camera Selection: Sony A6000 SeeCAM_CU130 Canon Eos Rebel SL1 Shutter Speed 30-1/4000 20-1/2000 30-1/4000 Resolution 24,3 MP 6000 x 4000 13MP (4224x3156) 18 MP Auto Focus YES NO YES Weight 285 g 19 g 370 g Dimensions 120 x 67 x 45 mm 29 x 29 x 30 mm 117 x 91 x 69 mm Price $450 $120 $499 Table 5: Camera s Parameter Due to the ease of image transfer and connectivity with the OBC, firstly the USB 3.0 connected cameras were investigated. As a result of investigation See3CAM CU130 was selected because of its lightweight and high resolution. (Table 5). Although there were some doupts about its lens and shutter speed, this camera was tried because of its low price. CAMERA GIMBAL BATTERY Figure 11: Gimbal System NVIDIA JETSON TK1 BULLET M5 Unfortunately, our doupts turned out to be right and the camera worked unsuccessfully for the ODCL tasks. (Figure13b). Cameras were listed to solve these problems based on the previous extensive investigation about weight, size and price/performance ratio. As a result, Sony A6000 was preferred. Sony has the capacity that can accomplish all the tasks successfully with a suitable gimbal (Figure 11). Sony A6000 can take the sufficient image for ODCL mission as shown in Figure 13a. POE PIXHAWK Figure 12: Image Process Architecture Figure 13a: The photo taken with Sony A6000 Figure 13b: The photo taken with See3CAM_CU130 Lagari UAV Team Technical Design Paper - 2017 10/20
2.2.3. Communications Undoubtedly, one of the most important issue in unmanned aerial system is the data transfering. If the data transfer is not ensured properly, the aircraft cannot perform manual or autonomous control and missions. Therefore, an extensive research was carried out in the selection of the communication equipments. According to the result of the price / performance analysis, the most suitable materials were selected. Problems that could be encountered were identified and the solutions were lined up after the election step. To give a few examples from these; Possibilities; The carbon body reflects the signals Excessive data file size Connection failed due to (long) distance Figure 14: Communucation Architecture Solutions; Attention to the position of antennas in the body Using high bandwidth Using powerful aerials and omni antenna in the aircraft, directional antenna in GCS The system has three different data link between GCS and Huma UAV. So these data links as followed: 1. 2.4 GHz Link: The 2.4 GHz link ensure manual control owing to AT10II radio controller for safety pilot. AT10II was chosen at the end of the optimization process because it has high range (up to 2.48 miles (4 km)), 3 milliseconds response time, low energy consumption (under 105 ma) and lower price than other radio controllers which at the same class. 915MHz 2.4GHz 5.8GHz 2. 915 MHz Link: The 915 MHz link which used for telemetry connected between Pixhawk and GCS. Figure 15: Antenna placement The link is most important link because it is about safety home function. If the data link cannot establish the communication for a period of 30 seconds, the failsafe mode will be activated. When this occur, Huma immediately returns to the predefined home position. 3. 5 GHz Link: The 5 GHz link is the main link for downloading picture during the image processing and interoperability task hence it must be high speed and with superior bandwidth. To serve this requirement, Team decided to use Ubiquiti Network especially M5 series such as bullet M5 for Huma and Nano station M5 for GCS. Omnidirection antenna which radiates radio wave uniformly in all directions, was used on Huma and directional antenna which also radiates in specific directions was used at ground control station. The cloverleaf antenna was chosen as the omni antenna because the antenna has an angle of 360 horizontal and greater than 90 vertical polarization and suitable for bullet M5. In this way if the antenna is positioned perpendicular to the ground, it has a very wide viewing angle. Also, the cloverleaf antenna has a λ / 4 wavelength. In this case, the size of the cloverleaf antenna is half the length of the same powerful λ / 2 wave length antennas. Figure 16: Cloverleaf Antenna On the other hand, the GCS antenna must be always be located relative to the aircraft position. The position of the antenna was controlled by manually because of team budget. Lagari UAV Team Technical Design Paper - 2017 11/20
2.3. Ground Control Station Figure 17: Ubiquiti Nano Station M5 and Ubiquiti Bullet M5 Specifications. 2.3.1. Mission Control Station Ground control station shall actuate 915 MHz telemetry link to communicate with the Pixhawk and send telemetry data to the interop server. There are many ground control station softwares such as Skywiev, APM Planner, Q-Ground Planner, Mission Planner etc. Ground control station has communicated with Pixhawk via Mission Planner. Mission Planner has been customized according to mission requirements which is air drop, off axis target, searching area and interoperability. It is open source that allows significant modification for these tasks. It also enables to provide real time waypoint achievement and control system tuning in flight borders that are actived by GeoFence. However, it can always indicate all significant aircraft data such as aircraft altitude, location, airspeed, climb rate etc. in the simplified interface while recording in log file. 2.3.2. Imaging Console Lagari team designed an imaging console as a part of ground control stations. İmaging console contains a specific software that have a dedicated user interface. The software that developed with C++ on Windows OS has some critical features: Show real time camera image. Allow user to see and edit detected targets data Send target data by connecting to interop server Control the camera and mission Show geo-location data Figure 18: Mission Control Station Figure 19: Imaging Console If there are objects that cannot be detected autonomously, the user can manually identify the objects using this interface. 2.4. Object Detection, Classification, Localiation (ODCL) Software prepared for ODCL mission is written in C++ using the OpenCV library. The high-resolution images taken from the camera for the ODCL mission are too large to be processed at real-time speed. In order to be processed without delay, some of the unnecessary parts are eliminated by some filters. The remaining picture is ready to be transferred to the classification steps. The classification and calculation algorithm basically uses edge detection and deep learning algorithms. This system consists of three main stages: Lagari UAV Team Technical Design Paper - 2017 12/20
Shape Classification Regions that may be the target are detected using the edge detection algorithm in high resolution images. Two methods are used to elimination the non-targets from the detected areas: - Using the altitude of the plane, the area in which the objects can be located is calculated in pixels - Objects whose aspect ratio is greater than 2.5 are eliminated The regions that can pass through these stages can be called targets. These targets are classified according to their shape by the K Nearest Neighborhood (KNN) algorithm. Figure 16: Background Subtraction Character Recognition START After the classification process, the alphanumeric character in the figure is continuously rotated by 15 degrees to determine the letter. In each cycle, the character on the target is compared with the taught character samples. The alphanumeric data of the character on the target is decided according to the highest similarity rate. Take photo Catch the objects that can be target with edge detection methods Detect target by the previously trained deep learning algorithm Geo-location Calculation To determine the position of the detected and classified targets, a calculation is made using the data such as the position of the target in the image, the gimbal position, the direction and position of the aircraft. Theoretically derived functions are combined with experimental results and algorithms are prepared to determine the object position. After these data have been collected, the object is informed to the interop server together with all the properties. NO Subtract the background to reduce image size Are there any object that is not processing? YES Determine color and location of target Determine the Alphanumeric character Upload all data to server Figure 17: Image Process Algorithm 2.4.1. Off Axis To perform the off axis task successfully, an algorithm is used to control the Storm32 card associated with PixHawk. The location of the off-axis target is taken from the interop server. According to this information, new waypoints are drawn to approach the target without leaving the flight boundary. The algorithm calculates the gimbal rotation angle using the altitude of the aircraft and the distance to the off-axis target. After calculation, the ground control station orients the gimbal via Pixhawk. Received images are processed and reported to interop server. Lagari UAV Team Technical Design Paper - 2017 13/20
2.5. Obstacle Avoidance A customized algorithm based on Rapidly- exploring Random Tree (RRT) was used to draw a fast and effective new path plan on the task of obstacle avoidance. A RRT is an algorithm designed to efficiently search nonconvex, highdimensional spaces by randomly building a spacefilling tree. The tree is constructed incrementally from samples drawn randomly from the search space and is inherently biased to grow towards large unsearched areas of the problem. This algorithm easily handles problems with obstacles and differential constraints (nonholonomic and kinodynamic) and have been widely used in autonomous robotic motion planning. As shown in Figure 18, a path is drawn around the obstacles to reach the blue point from the red point. Afterwards,periodic waypoints are added on this path to obtain flight plan with avoiding obstacles. These new waypoints are transmitted to the flight controller via the MavLink. Autonomous flight is carried out in accordance with the new route. 2.6. Air Delivery Lagari team analyzed the system requirements in order to accomplish air delivery task and listed the obtained results; - The Airdrop system should be easily detachable from the fuselage. - The release mechanism should be durable to about 3.5 G shock load during flight. - When the Airdrop system reached to lost hiker, the water bottle must be undamaged and the water must be sealed in the bottle. - To increase the accuracy of the airdrop position, a rapid and high quality algorithm should be created. Considering these conditions, the air drop mechanism was designed in SolidWorks 3D CAD design software. The system was located outside of the fuselage so it could be easily detached in the airdrop position. A rocket-shaped aerodynamic design was made to reduce the drag effect under the fuselage. The airdrop was exactly located on the CG in order to prevent variation of CG. Airdrop Design Figure 18: Simulation of RRT Algorithm Figure 19: Airdrop Draft The protective rocket case was produced from carbon fiber material for high strength and light weight. The water bottle is located into a rocket box with using a special damping element to absorb the shock of collision and prevent the spread out of water in the bottle. The mission planner interface is optimized for the airdrop task. Figure 20: Location of Airdrop When the airdrop mission is activated on the interop server, the mission planner starts to calculate the distance between the UAS and the airdrop position. This calculation is done with taking into consideration the UAS speed, wind speed, UAS altitude and the system delays. The separation is accomplished by the servo mechanism which in the fuselage when the UAS reached the airdrop position. Lagari UAV Team Technical Design Paper - 2017 14/20
2.7. Interoperability The interoperability system has three main tasks: receiving task information, making sense of it and sending UAV status to the server continuously. The software developed by the Lagari Team, uses TCP/IP for communication with the server using post and get methods. 2.8. Cyber Security The imaging system and console uses TCP/IP protocol to send and receive the data. To prevent TCP/IP hacks, the imaging console and on-board computer have static IP address and a matching password. To provide the security, the system checks the information of the sources that try to connect the server. Bullet M5 and Nano Station M5 are locked to each other via mac addresses. Any external connection is not allowed, even if the same product is in the communication network, since it will not keep the MAC address. At the same time, SSID name and password are used to prevent illegal connection to the network. The 915 MHz telemetry line communicates from point to point. For this reason, the data system is protected against any other connection. 3. Tests & Evaluation Plan 3.1. Developmental Testing 3.1.1. Wing Loading Test Aircraft will be exposed to many forces because of various reasons, including altitude changes and turns during the flight. The endurance of the wings is crucial for flight safety. Structural wing loading tests were performed in order to be durable for these forces. The bending and deformation of the wing were examined. In the first flight test, it was observed that there was too much jolts and vibrations in the wings. The design of the wing joints was modified because excessive vibration would affect adversely the strength of the wings and the flight performance. It was experienced that balsa spars were not enough in the connections. Carbon fiber reinforced balsa spars were used to reduce vibrations and jolts. Figure 21: Wing Loading Test Figure 22: Eppler 214 charectrists at low reynolds 3.1.2. Wing Flow Test The aerodynamic team focused on airfoils which wellknown with aerodynamic characteristics at low reynolds because there is no facility for wind turbine testing. Thus Eppler 214 was selected as an airfoil. This airfoil has high lift coefficient (cl) and improved stall angle at low reynolds. (Figure 22) Lagari UAV Team Technical Design Paper - 2017 15/20
3.1.3. Propulsion System Trust Test Propulsion system equipments were selected according to the UAV weight envisaged prior to the production process. The engine search was made according to the C L / C D (it means thrust/weight) ratio. Considering the safety factor which is 1.8, propeller calculations were done to lift about 13 kilograms. The theoretical calculations were simulated with thrust test programs. The propulsion system had to be tested to determine that also satisfy the system requirements in flight conditions. Thrust, voltage and current tests were performed to reach the desired system properties in the test mechanism that we prepared. It was seen that the engine had about 11.5 kilograms thrust with an 18x10 propeller (Figure 23). For this reason, the test results almost confirmed the theoretical calculations and simulations done. 3.1.4. Landing Gear Loading Test The landing gear's strength and durability analysis were done with Ansys Fluent CFD simulation program. Material was selected based on this analysis.the Huma s landing gear was produced from carbon fiber with using vacuum method. According to the lamination plan created to withstand 3.5G-Force during the collision with ground, landing gear was produced and tested under load. Landing gear was damaged when the load reached 3G (Figure 24). That is why the lamination plan has been revised to get better test results. The amount and variety of carbon fiber was improved. The reconstructed test with the changes made gave the desired results. Figure 23: Thrust Test System 3.2. Individual Component Testing Figure 24: Landing Gear Loading Test 3.2.1. Autopilot System Tests Autonomous Take off & Landing Automatic takeoff function sets the throttle and raise the aircraft until a specified altitude. Take off command has two parameters; the minimum pitch angle and the takeoff altitude. These parameters were set properly for Huma design using Mission Planner interface. Figure 25 shown takeoff glide slope. Automatic landing command has many parameters. Flare point is the most important parameter. The flare is the final stage of the landing when Pixhawk decreases the throttle level and raises the pitch angle. As a result of this, drag increases and the aircraft decelerates. The appropriate time depends on the Huma s design to flare. All of landing parameters were set properly for design by protecting glide slope that is 10%. Figure 25: Take off glide slope Lagari UAV Team Technical Design Paper - 2017 16/20
Waypoint capture and accuracy Waypoint capture and accuracy are significant parts in autonomous flight tasks to receive maximum points. Therefore, two GPS were used on Pixhawk. They were separated from each other to prevent signal interference and EKF is used for improving waypoint accuracy. In addition, the radius of each waypoint was determined by the design so waypoint radius must exceed minimal turn radius and calibrated airspeed sensor. It enables to fly with much better control over its airspeed. Autonomous waypoint navigation was tested. Also, autonomous takeoff and landing were performed. Waypoints were followed rightly with 4-5 m(%10) error. Figure 26 shows waypoint navigation. 3.2.2. Imaging System Tests Images were recorded to use in the imaging system test during test flights. Therefore, the performance of the imaging system could be constantly tested when the test flight was not done. Objects were prepared in different colors, letters and sizes for image processing test. With the prepared objects, the task of search area was simulated to ensure successful results of the objects in different terrain (grass, asphalt, soil etc.). After the experiments, edge based object detection algorithm was used instead of color based object detection algorithm. The edge-based target detection algorithm yielded 72% better results than the color-based target detection algorithm on different surfaces. The average test results for some targets are given in Table 6: Target Shape Color Letter Letter Color 75% 82% 56% 54% 80% 85% 36% 47% Comment The letter identify is difficult because the letter was near the edge lines. Difficult to identify characters because of the thin letters 78% 89% 62% 67% Overall good results 32% 40% 55% 39% 81% 85% 12% 10% Table 6: Imaging System Test Results Figure 26: Waypoint Navigation Difficult to detect because of its small size and black color. Letter reading rate is too low due to sunlight shine 3.2.3. Data Link Tests All the data transfer links were tested in the vicinity of the high-noisy city where there are a lot of radio frequencies. In the speed test, data transmission was measured from close distance to 135 mbps. It has been observed that data can be transferred with 75% efficiency at a distance of 2 km. (Figure 27). This data link is able to send two images per second which are reduced in size with a special algorithm. Figure 27: Main data link distance test (Left side about 2 km, right side close distance) Lagari UAV Team Technical Design Paper - 2017 17/20
Distance test was performed on the line used for autonomous flight. It has been found that up to 4 km of data transfer is provided. In addition, all of the communication tests were performed successfully. Speed tests were performed for various angles and distances using different omni antennas. The most productive cloverleaf antenna type was chosen as the omni antenna for the Bullet M5. 3.2.4. Obstacle Avoidance The RRT obstacle avoidance algorithm that was created with the python language was simulated on the mission planner control interface. The Jetson TK1 connects to the interop server and downloads the obstacles informations, than creates a new flight plan by addig the received data into the RRT algorithm. The created new path is transmitted to autopilot via Mavlink to create new waypoints. Obstacle avoidance task has not yet experienced in flight practically. A test plan was developed to complete the tests of this task until the competition. 3.2.5. Air Delivery The Airdrop mission was tested at all flights because it was a relatively easy task. In the first tests, it was observed that the water bottle could be damaged and the water could leak out of the bottle. For this reason, a protective system was designed for the water bottle. It was seen that 100% of the water in the bottle reached the target. Due to the delay in the separation algorithm, the distance from the target was 113 feet, the algorithm was updated with the test results and the best score was achieved with a distance of 13 feet. 3.2.6. Interoperability System The interoperability system connection has been tested in short-range and long-range under load and unloaded data transmission networks. Also, faulty stations have been tested to fix faults in the fastest way possible and establish the connection again. The tests results show that the interop server connection was found to be sufficient in all conditions. 3.3. Mission Testing Plan Tasks Success Rate Threshold Objective Backup Strategies Autonomous Flight; Take off 90% 80% Waypoint Navigation 95% 90% Autopilot calibrations will improve. Landing 80% 60% Obstacle Avoidance; Stationary Obstacle Avoidance 60% - The algorithm worked on simulation, it will be tested during Moving Obstacle Avoidance 45% - the flight. Object Detection, Classification, Localization; Figure 28: Graphs of various speed tests Search Area & Off-Axis 90% 75% Gimbal controller will be reprogrammed for off-axis task. ODCL 80% 75% Image filters will improve with Emergent Target 90% 75% proper threshold limits. Actionable Intelligence 100% 100% - Interoperability 100% 100% - Air Delivery; Airdrop 95% 95% - Table 7: Test results of missions and backup strategies Lagari UAV Team Technical Design Paper - 2017 18/20
4. Safety, Risks & Mitigations The Lagari team focused on the system security factor as the first priority in the design, development, production and testing process of the Huma. Numerous risk factors are present at UAV systems such as Huma, which are emerging as applications of Multidisciplinary Aerospace System. Since the our first priority was system safety during the preparation for the AUVSI SUAS 2017 competition, the Lagari team created mitigation methods and solution strategies taking into account all these risks. As a result of the systematic application of these mitigation plans, maximum safety was achieved during the production, development and testing processes. This section of the paper explains these plans that are created and implemented by the team in detail under three main headings such as development, mission, operational risks and mitigations. 4.1. Developmental Risks & Mitigations System requirements have been analyzed in order to minimize potential risks that could be met during the design and development process. Table 8 shows the risk&mitigation methods. Risk Bad aerodynamic stability Structural insufficient of fasteners between wingstails and fuselage Low durability of landing gears on landing collision impact Vibration during flight Foreign object damage Fire Failure of avionic equipments Mitigations Methods Motor was mounted in front of the fuselage. There is a five degree dihedral angle at wings to increase lateral axial stability. Horizontal stabilizer surfaces were designed broadly to increase longitudinal axial stability. The wings were located close to the upper part of fuselage. The carbon fiber spars in the fuselage and the wing s carbon pipes were assembled with aluminum fasteners. The tail fasteners were supported with chemical resin adhesive. Stainless screws, nuts and washers were used. The landing gear was made entirely of carbon fiber. The connection of the landing gear to the fuselage was made using special damping elements. Motor was mounted with flexible polyurethane plastic damping washer. Damper elements were used to the gimbal and control card. Carbon fiber pipes and plates were used as ribs and spars in the wing. Motor was mounted in the fuselage and the propellers are used with the spinner. LiPo batteries were covered with a static bag and during LiPo packing, the cells were insulated with a nonflammable felt. AWG high current cables were used in the main power line. The fuselage was earthed to prevent current leakage. A zener diode was used to prevent the servo-rail voltage drop in the autopilot. Capacitors were used in the regulators in order to prevent the noises in the equipment. Table 8: Developmental Risk & Mitigations 4.2. Mission Risks & Mitigations Mission risks and requirements were entirely analyzed in order to accomplish competition missions successfully. In the direction of the obtained results, some design parameters were created with the goal of providing maximum safety during mission demonstration. UAS subsystems such as mechanical design, selection of avionics equipment and creation of mission algorithms were optimized according to these parameters. The Lagari team has designed and produced the UAS system that can eliminate all of the risks posed by competition missions with these analysis and optimization processes. Table 9 shows the risk and mitigations methods used in optimization and design processes. Lagari UAV Team Technical Design Paper - 2017 19/20
Risk Mission Likelihood Loss of data links Servo failure Inaccuracy of Autopilot data s Blurry photo due to camera shake Failure during the separation Poor avoidance algorithm All missions All missions All missions ODCL Air Delivery Obstacle Avoidance LOW (%1-34) LOW (%1-34) LOW (%1-34) MEDIUM (%34-66) MEDIUM (%34-66) MEDIUM (%34-66) Impact on The Mission MAJOR MAJOR MAJOR MAJOR MAJOR MODERATE (%34-66) Mitigations Strategy -Antennas were located out of fuselage because Carbon fiber reflect the signals. - RF receivers were fixed on fuselage spars. - High power antennas were preferred. -When the servos was selected, the load safety factor determined as 2. -Servo connecting cables were isolated by using cable heat shrink tubes. -Double GPS were used in control unit. -Extended Kalman Filter (EKF) were used. -Ublox M8N supports Galileo GPS system. - Pixhawk 2 was selected as autopilot with improved sensor properties. -2 axis gimbal was used to prevent vibration. -Camera that has high shutter speed was selected. - The airdrop system that looks a rocket was located on out of fuselage. -RRT algorithm was preferred to obtain the best avoidance strategy. Table 9: Mission Risk and Mitigation Methods 4.3. Operational Risks & Mitigations 4.3.1. Autopilot Failsafe Safety Functions There are many failsafe functions on Huma s autopilot unit that provide aircraft and personal safety. The UAS will return to home position (RTH) if there are 30 seconds loss of data link The UAS will terminate the flight if there are 3 minutes loss of data link If the battery level drops below the specified voltage or mah, the system will generate a warning signal to GCS and UAS will return the land (RTL). 4.3.2. Pre-flight & Post-flight operational Checklist After completing the first flight test, Lagari team realized that there could be confusion during assembly of UAS because of a lot of system equipment. Therefore, numerous precautions were taken: Cable labels used to prevent cable clutter All team members were entrusted with mounting a part of system in all tests throughout the year. A detailed checklist containing all of the assembly procedures has been prepared. The main points of attention in the operational concept are: - Durability of fasteners connection and whether there is a break in the cables checked. - Center of Gravity (CG) location is checked. - Data link communication, servos in manual and auto mode, GPS and compass health are checked. - After ensuring that all connections are made, the motors are allowed to operate via the safety button. 5. Conclusion This paper summarizes a work done by students from the Mechatronic Engineering Department of the Yildiz Technical Universy in preparation for the AUVSI SUAS 2017. Throughout the year, the Lagari team designed, developed and tested the Huma system with an improved software for payload control and an enhanced hardware for flight control. Advanced algorithms were written to accomplish competition tasks. Numerous ground and flight tests were conducted to confirm the UAS capabilities and assured its excellence in the SUAS competition. Lagari UAV Team Technical Design Paper - 2017 20/20