Anadolu University
Transcription
Anadolu University
ABSTRACT Anadolu Suas team that will participate in AUVSI SUAS competition for the first time aims to fulfill the duties such as automatic flight, navigation, surveillance, real time actionable intelligence. Hexacopter which performed its design and production is used in competition. Hexacopter made the mechanical and aerodynamic tests in wind tunnels is equipped with navigation and imagery features. The 900 MHz frequency band system is used in order to communicate with the ground control station. Real-time situation, position, location of hexacopter and other sensor datas can be monitored on the ground control station. Pictures taken with GoPro camera at least 9 Megapixel resolution are transmitted to the ground control station with 5.8 GHz frequency band. Imagery task is fulfilled automatically thanks to these two? camera. Ardupilot Mega flight control card is used as autopilot system. Hexacopter can manually be controlled with HITEC Aurora 9 controller which communicates at 2.4 GHz frequency band in every phase of the flight in every phase of stage. In addition, GPS informations can be transmitted to the ground station with Hitec 2.4 GHz radio telemetry system in real time. Energy of hexacopter is met with two 7000 MAh, 14.8 V li polymer battery. Team Lead : Gökhan GÖL Team Members : Tuna Ayan, Kemal Mert Makinacı, Seçkin Yener, Furkan Üzmez, Gözde Yaşar, Emrullah Aslankaya, Ezgi Tetik, Diler Öz, Özge Ermak, Cansel Öğretmenoğlu Team Pilot : Zafer Öznalbant Faculty Advisor : Dr. Tansu FİLİK 1 1. INTRODUCTION 1.1. Mission Requirements Analysis We expected to be made in the contest task is related to the waypoint navigation and image recognition. Part of waypoint navigation is made automatically in all phases of flight provide that taking point as objective. Part of image recognition included search area from the area where video or image transfer. The purpose that identifies the same characteristic of an unknown number of targets in search area. These features include target shape, background color, alphanumeric character, alphanumeric color, orientation and location. Maximum time is given to perform command of contest is 40 minutes. These commands are doable with all platform that heavier than air. Contest commands are firstly analyzed with engineering approach. So Key Performance Parameters (KPP) is identified. Worked on tasks by teams are divided into four separate groups. This group included that airframe design, communication, navigation and imagery. Key Performance Parameters (KPP) chart was prepared, that was used by the team to identify goals on which further work was to be done once threshold were complete: Characteristics Autonomy Threshold Waypoint mission Objective All phases of flight Waypoint Tracking Static Waypoints Dynamically changeable waypoints Transmitting video or image During all flight Only detection target Imagery Acquisition and visual identification Mission Time Within 40 minutes (imagery recognition and location done at the end of competition) Manual Autonomous color and shape recognition, GPS and orientation mapping. 25 minutes (Imagery, recognition and location done in real time in flight). Automatic and manual Control Safety Manual control in all phases of flight, Loiter mode in signal breaking Manual control in all phases of flight, Loiter mode in signal breaking and fail safe, Automatic landing in getting out of limit of battery Table 1. Identify goals on which further work 2 1.2. Systems Engineering Approach Team is divided by 4 four groups as air frame design, communication, navigation, imagery. Carbon-fiber, fiber-glass and aluminum are selected for used for structure by regard confidence, lightness and endurance. As a result of tests, the load carrying capacity of aluminum profile hexacopter was found to be better than the response windy wind values than others in flight. Communication frequency defined 2.4 GHz for manual control, 900 MHz for system of telemetry, 5.8 GHz for transferring of video. Communications equipment to support safe flight was given to be selected. Firstly it is started with fixed-wing aircraft in order to fulfill the competition missions. However rotary wing aircraft is chosen because of the fact that fruitful results cannot be gotten (figure 1). The reason for this, speed of fixed-wing aircraft is faster than rotary wing aircraft. Over speed makes difficult the detection of targets in ground. Additionaly, some problems are seen in fixed wing aircraft related to navigation. At automatic detection with hexacopter, it can hold its altitude when it detects target. In this way, you can also achieve all the features of the target. It is intended that ardupilot communicates with imagery computer as created software is loaded to ardupilot flight control card. In this way hexacopter will stay loiter mode in the parts of the target and it is able to take a picture of target. Because of the fact that automatic detection cannot be done in targeted features in this year, prepared autopilot system constitutes preparation for the next year. Figure 1. Hexacopter 3 2. Airframe System Design Hexacopter is defined as unmanned aerial vehicle which is capable of vertical take off and landing, is capable of hanging on the air, has a high maneuverability, a simple structure, six rotor. Also, it is an unmanned aerial vehicle with rotary wing which create transportation force via its propeller by taking advantage of moving force produced by engine. Air vehicle is designed in a way that distance between motors is 79 cm and height from the ground is 21 cm. Materials we used in making wings is aluminium and fiber glass in making body. Aluminium is durable and light at the same time, which is why we used it in making wings. 4s1p 7000 mAh Lipo battery is used in our system. As it is known lipo batteries have bigger capacity of storing energy in small volumes, which is why they are most used battery type in air vehicles. Locating body produced according to the modified electronic board, system that prevent oscillation greatly increases the stability of the vehicle is used. Equilibrium point is set to be in the center of the vehicle. 2.1. Design Approach Multicopter systems have not too much type of design. Because there are certain limits. Hexacopter all equipment's design to ensure proper placement is made. Sensor kits are mounted to the body vibration inhibitors to protect from vibration. Flight control board is mounted with anti-vibration to be at the center of the vehicle. The GPS sensor is mounted on the body to be away from the electronic device. Batteries are placed in the suitable slots that under of body to ensure that the x-axis under. Holes were opened to parts on the body strength to some minor to lighten the weight of the hexacopter. Motor cables are passed through aluminum profiles to protect against shock. 2.2. Manufacturing The primary is hexacopter is made mainly of carbon fiber. 2 composite body was produced but the engine thrust levers to reduce the impact has been made thick for strength. Produced single mold body plus the resistance caused too much vibration. So other hexacopter was manufactured using aluminum and fiber glass materials. Strength statically tested and it was found that 14 kg of withstanding loads. 2.3. Power System With a payload take-off weight of 2.9 kg, hexacopter must be equipped with a power system that can provide enough thrust to keep the hexacopter aloft for at least 20 minutes with ample 4 reserve power for emergency situations. Additionally, gimbal and FPV (flight person view) systems energy supplied by external battery. 6 pieces Tiger MT2814-10 770 Kv brushless electric motor was used in hexacopter. This engine was chosen based on its ability to deliver 132 Watts per kg of nominally in cruise, as well as 147 Watts per kg of burst, power reserve. As a result of static tests of motors using 4S 14.8 V battery with 13 x 47 propeller efficiency was found to be high. Figure 2 is shown in. So 3 CW, 3 CCW 6 pieces of 13 x 47 carbon fiber propellers are used. 12*45 Propeller-Motor Static Test 1500 Thrust (g) 1000 500 0 0 10 12 13 14,3 14,6 14,9 16 18 20 17,2 19,1 Current (A) 13*47 Propeller-Motor Static Test 1500 Thrust (g) 1000 500 0 0 9,2 11 12 13,4 13,6 13,9 15,1 Current (A) Figure 2. Propeller – Motor Static Tests As a drive brushless motor 50 A electronic speed controller (ESC) is used. It supports high voltage up to 29.6 V 8S ESC. BEC (battery elimantor circuit) circuit output is 5V/3A thus receiver is supplied. 5 In Hexacopter two units 14.8V 7000mAh 45C 4S batteries are used. Batteries connected in parallel to each other thanks to this strength times were increased .Power module is used for making battery voltage control (Figure 3). Power module measures battery voltage and current drawn and then transmits to Ardupilot flight control board. Also thanks to the voltage regulator that is on it provides the energy of Ardupilot. Figure 3. APM Power Module Lion polymer batteries’ that are used for hexacopter system feeding times are calculated as follows: Current capacity = 2*700 =14000 mAh ( it may generate 14 A in 1 hour) Discharge Coefficient = 45C In normal state capacity of current generate in 1 hour = 14A*45 = 630A According to this calculation batteries that are used can supply 630A continuously for 1 hour. Continuous current values that are drawn by motors are 27 and 6 motors are used. The current drawn by the motor = 6* 27A =162A External 3s 1500 mAh Lipo battery is used for Gimbal and FPV (First Person View ) systems. This battery's voltage and current information is transmitted to the ground control station as a real time by telemetry system. 2.4. Navigation System Autopilot flight occurs almost in every missions in the SUAS competition. If it is performed, there will be positive effect in the scoring. Ardupilot flight control board is used for autopilot system of hexacopter. This flight control board that has Atmel 2560 processor is open source. For the competition, autopilot code is improved and all flights during the competition will be aimed with the autopilot. Waypoints are marked for autopilot over the Google Map in the Ground Control Station and loaded to the system (Figure 4). 6 Figure 4. Mission Planner Drawing Waypoint There are 3-axis accelerometer sensor, 3-axis Gyroscope, 3-axis compass and barometric pressure sensor is located on Ardupilot flight control board. The data received from these sensors is provided through control of hexacopter. In addition to autopilot, flight global positioning system (GPS) sensor is used. Hexacopter follows coordinates that is loaded to flight control board thanks to signals that are gotten from GPS. Many waypoints are occurred to scan all search area in shortest time. There is an example in Figure 5. Figure 5. Search area waypoint 7 2.5. Camera / Gimbal System 2.5.1. Camera Selection The primary mentions of competition recognize the targets and identify the following specifications: background color, shape, orientation, alpha-numeric, alpha-numeric color. When we thought these specifications and our hexacopter, our camera requirements has to be below: high resolution camera between 100 and 750 ft., small size, lightweight, take images and video In the end, we compared GoPro-Black Edition Hero 3, GoPro-White Edition Hero 3 and Canon 600D. Weight Image Quality X4 GoPro Black Edition 3 Lens Distortion X4 3 3 4 Weight X3 4 4 1 Size X3 4 4 1 Ease of Interfacing Cost X2 3 3 3 X1 3 4 1 57 54 45 Total GoPro White Resolution 2 Canon 600D 4 Table 2: selection table of cameras Finally, we decided to use a GoPro Black Edition Hero 3 camera for competition (Figure 6). The camera stream is fast video stream that is used by the camera operator to control the operation of the camera gimbal. 8 Figure 6. GoPro Black Edition hero 3 camera 2.5.2. Gimbal Frame After chosen to GoPro Black Edition Hero 3 camera, we determined to use Walkera 2d gimbal (Figure6). Because this gimbal is specifically designed for our camera, adoping aluminum alloy CNC frame, brushless motor, high accuracy electronic system. Also, when hexacopter altering the flight or shaking happened, the camera can still be horizontal and stable. High quality CNC precision machining, high performance brushless driven, more precise and longer life. Control range are -45 / 45 roll and – 135 /90 Tilt.2 Dimension gimball is suitable for our system, high precise and stable frame design to make sure the camera control accurately during high speed flying. During the flight our camera focuses on targets on the ground. If it is necessary, the gimbal is controlled manually through a joystick operated from the ground station. Other important thing is lightness for UAV G –2D gimbal system weight is 120 g. Figure 7. Gimbal System 9 3. Data Link 3.1. Radio Module Communication between the air vehicle and the ground station is one of the most important parts of this project. Thanks to telemetry system ground station, user can obtain data about final situation of air vehicle. Communication between the air vehicle and the ground are provided with three wireless links. These links are 2.4 GHz radio control transmitter and 5.8 GHz imagery ground station, the 900 MHz autopilot ground station. A 900 MHz data link is used by autopilot to communicate telemetry and aircraft information to the ground station (Figure 8). For 900 MHz data link Xbee Pro 900HP wireless transceiver module is provided. We used this module because; XBee-PRO 900HP embedded module provides best-in-class range wireless connectivity to devices. It takes advantage of the DigiMesh networking protocol, featuring dense network operation and support for sleeping routers, and is also available in a proprietary point-to-multipoint configuration. Supporting outdoor RF line-of-sight ranges up to 28 miles with high-gain antennas and outdoor RF line-of-sight range up to 6 miles (9.6 km) with dipole antenna and data rates of up to 200 Kbps, the module is ideal for extended-range applications requiring increased data throughput. The XBee-PRO 900HP requires no programming and can be configured easily using Digi’s free X-CTU software or via a simplified AT command set. XBee modules are pre-certified for use in multiple countries, further reducing development costs and time to market. This modules are easy to use and are available in a variety of different protocols, enabling users to substitute one XBee for another with minimal development time and risk. Two Xbee modules are used one of them on UAV and the other one is on ground station. Using this module Flight informations (speed,position,angle..etc.) that is about UAV are sent or received by user interface that in groun station pc and are provided for user. Using this xbee destination points, control parameters and another flight information is sent to UAV. Figure 8. Xbee Pro 900HP A 5.8 GHz data link is used for transmission for video from the camera that is on aircraft to the image processing software at the ground station. Thanks to 5.8 GHz power of frequency high quality video transmission is provided. Furthermore, 5.8 GHz frequency power barriers frequency distortion that is happened due to modems, ADSL, other electronic devices. 10 A 2.4 GHz data link is used to communicate remote control signals from controller to the aircraft. When autopilot becomes in out off control, receiver is connected to aircraft using this data link. 3.2. Antennas 3.2.1. Xbee SMA Connector Antenna 900MHz 5dbi SMA antenna is used to communicate between autopilot and ground station. This antenna works in the frequency band of 900Mhz. dBi is a representation of relative power on the other hand this means gain, higher dBi is usually better. According to this comparison we decided to use 5dbi 900 MHz antenna. Furthermore, this antenna has flexible antenna features a tilt-andswivel SMA connector, allowing them to be used vertically, at a right angle, or any angle inbetween. Figure 9. SMA Connector Antenna 3.2.2. Cloverleaf Antenna 5.8GHz 9dbi Clover-leaf antenna is used to transmit video to imagery ground station. This antenna works in the frequency band of 5.8GHz ISM. It can be used in the 5.8GHz wireless video module) and also other transmitter and receiver working in the frequency band of 5.8GHz. The antenna has low profile, light weight, high reliability and wide communication range. We made test with patch antennas that have same frequency but we chose cloverleaf antenna design because it has one of the best antenna designs for aeromodeling activities, especially the FPV flying. Also it has the characteristic of wide wave band range and circular polarization to ensure the stable transmitting and receiving when the aircraft is in any attitude and position. So it can provide reliable wireless video signals during a flight. 11 Figure 10. Cloverleaf antenna 4. Ground Station In our system, all flight operations are controlled through the ground control station. For example, we can program the search area, and emergency target coordinates before the takeoff. The main purpose of the ground station is communication with the unmanned vehicle system. Ground station consists of a network of computers that are used to process flow of images, display telemetry datas from received the UAS and detect the target manually. There will be three computers in the ground station. One of these computers is used to display telemetry data. Ardupilot 2.6 is used in the system. Therefore, Mission Planner is using in order to display telemetry data. Mission Planner is a fullfeatured ground station application for the APM open source autopilot project. Mission Planner can be used as a configuration utility or as a dynamic control supplement for your autonomous vehicle. It helps us about the vehicle’s motion and situation. Mission planner supplies to display hexacopter attitude, speed, altitude, location, wind direction and speed, battery voltage, waypoints which are passed or which will pass, etc... 12 Here some of Mission Planner properties: Point-and-click waypoint entry, using Google Maps. Select mission commands from drop-down menus Download mission log files and analyze them Configure APM settings for your airframe Interface with a PC flight simulator to create a full hardware-in-the-loop UAV simulator. Second computer is used to flow of imagery that received from unmanned system. If it is need to detect target manually, this computer can be used. In this situation print screen button will used. Third and last computer is used to run image processing algorithms. OpenCV libraries are used for this purpose. All of the programs are constructed with using OpenCV libraries. OpenCV (Open Source Computer Vision) is a library of programming functions mainly aimed at realtime computer vision. Received video from the unmanned vehicle is processed in real time. Targets are detected with using the color properties. Then, this target is cut from the video and saved in a folder as a jpeg. Finally, another program is executed over these detected targets to detect other properties. 5. Target Recognition and Analysis Our target recognition and analysis are based on some operations that on original and refined images for the purpose of identification of target properties. Our system has two parts. One of them is on-board system which perceives the target, takes a photo on JPEG format and sends it to 13 ground station. The other part is ground station system which analyses the image according to shape and alphabetic character. 5.1. On-Board System On-board imaging system is based on recognize different colors from background and take images. Our real-time algorithm which is written C language with OPENCV tries to detect according to targets’ colors. First of all, real-time imaging is taken inside of the code. These imaging are RGB (Red, Green, Blue) format. After that, this format is transformed HSV (Hue, Saturation, Value) format. Targets’ colors are tried to perceive on images of HSV and our algorithm do color detection according to our specify color interval. For this reason, we need to introduce HSV color interval for using every color. Therefore, a color filter is designed. When doing this, we thought that the targets’ colors are basic colors like red, green, blue and yellow. Example targets are created these fundamental colors according to SUAS competition rules specifications and ANADOLU BORAN is tested on these targets. Images are saved from taking GoPro Black Edition camera. These, which are included targets, are separated for later use. These frames are transformed from RGB format to HSV format with designed color filter. Pictures of HSV format have three channels, so those are converted binary format. HSV values, that can change manually, are varied until target is detected on figures of binary format via designed a simple interface. When the optimum color intervals are found for every basic color, our real color ranges are formed. Different ranges of hues, saturations and values have been determined for each basic color (upper hue, upper saturation, upper value, lower hue, lower saturation, and lower value). Also, some other morphologic operations (erosion, dilation) are done on images to destroy noises. Thus, non-target small areas have been prevented. 14 Figure 11. Designed interface and operations Our target recognizing algorithm goes through the same process with the color algorithm. Realtime frames are taken inside of the code and all frames are processed separately in the code. Again, images of RGB format are converted HSV format in different functions for each color and those also are transformed binary format according to our HSV range color. As using the OpenCV’s moment functions, area, which is appropriate for HSV ranges, are calculated number of pixels. Besides of that, center (x,y) coordinates of area are found as pixel values. If the calculated pixel amount is in the possible range (the range was calculated by considering the space that targets will cover in the competition flying height rule), it is thought that the target is in the color of the function that works in the image. To be sure, if there is really a target in the zone, based on the center of the zone, a smaller image is obtained by cropping the photo of a wider zone that covers it. This frame is run in the same function. If enough space in the same color can be detected, then it is accepted that there is really a target in that photo and saved in a folder. 15 Figure 12. Image 5.2. Ground Imagery 5.2.1. Target Recognition Target recognition duty is realized on ground station via MATLAB. When the target is detected, camera takes a frame by Panda Board then the image is sent to ground station for the purpose of processing with Matlab. Actually, processing HD image with matlab is so slow. So many methods have been tried to make quick processing with matlab. The target recognition code in Matlab consists of two main parts. First part is binary image processing part and second part is RGB image processing part. While first part provides recognition of the shape and alphanumeric, background and alphanumeric colors are specified by second part. In the first part, after binary image is obtained, morphological operations are applied for reducing noises. If the noises are not eliminated, essential object cannot be distinguished from irrelevant objects, also cannot be labeled. The other purpose of morphological operations is clarifying the alphanumeric which is destroyed by binarization operation. Morphological operations result is shown below; 16 Figure 13. Original (RGB) Image Figure 15. Opening Operation Figure 17.Label 1/shape Figure 14. Binary Image Figure 16.Dilation Operation Figure 18. Label2/Alphanumeric Binary image is available for labeling operation at the end of the morphological operations. Labeling operation scans the image pixel by pixel and examines connectivity of them in terms of 4 or 8 neighbourhoods, so categorizes the connected components. Resultant connected components after labeling operation are shown below; As a result, shape and alphanumeric are separated from each other so, different algorithms can be applied for shape and alphanumeric individually. While correlation algorithm is applied for 17 recognizing the shape of the target, OCR (Optical Character Recognition) is applied for identifying the alphanumeric. However, if the object orientation is not corrected, these algorithms cannot work effectively. So, orientation of the objects must be fixed before applying these processes. Correlation algorithm is based on matching original image and template image. Correlation estimations have the values between [0,1]. If the correlation estimation is equal to 1 or greater than 0.8, similarity can be happened among the matched images matrixes. In the OCR algorithm, features are extracted from each individual character. This features extraction operation is required for constitution a statistical model for each character class. The particular classes that the characters belong to are labeled. Each character class tends to cluster together. For instance, in the figure below, it can be seen that the 9’s have a mean Orientation of 118 and HPSkewness of 0.035. In the second part of the program,RGB image is processed for detecting background and alphanumeric colors. In accordance with this purpose, K-means algorithm is applied to original (RGB) image. objects in cluster 1 objects in cluster 2 18 5.2.2. Target Recognition Flowchart 6. Flight Safety and Risk Management Safety is a priority for Anadolu SUAS. The derived risks were based on previous flight test experiences and listed based from historical research on common UAV flight risk operations. Each risk was evaluated on its impact to the flight mission and likelihood of the issue to occur. Potential risks and damaging situations and previous failures discussed. After these discussions, team indicates risk definitions and according to them team has a risk management plan. 6.1. Flight Safety 19 Anadolu SUAS has an autonomous hexacopter. In case of unexpected flight operation failure system has some specifications. Air vehicle system has a three control modes which are manual, loiter and auto mode and additionally fail safe mode which is feature of control system. For appropriate situations, appropriate mode can be select. For instance in autonomous flight unexpected route changing, manual pilot will select immediately. On the other hand, when manual pilot controlled the air vehicle, pilot may loss of his/her control then auto-pilot can take control immediately. In case of loss of communication signal, vehicle engine reduced the throttle 50% and haxacopter stay on the air in Loiter mode. When the battery cell voltage drops below 3, 6 V, hexacopter starts to automatic landing. In the case of battery power loss hexacopter starts to automatic landing and can be controlled manually. In the case of vehicle loss GPS satellite, hexacopter stays on the air 30 sec then switchces to Loiter mode and wait control signal from ground station. Failure Hexacopter Response Battery Power Loss Hexacopter starts to automatic landing and it can be controlled manually. Loss of Communication Link Hexacopter stays on the air in Loiter Mode until battary cell voltage drops below 3.6 V. Then it treats as in the case of Battery Power Loss failure. Hexacopter stays on the air 30 sec and switches to Loiter Mode and it waits control signal from ground station. Loss of GPS Table 3. Risk Management 6.2. Physical Safety Removable parts and components, safely fastened and secondary fastener attached. Every cable connections, isolated from environment. When choosing cables, their current capabilities and heat specifications are noticed. Manufacturer advised original cables have been used with appropriate lengths. Cable, antenna, battery and other electrical circuit connections has strength to avoid splitting from each in case of shaking, vibration and strong wind. And also these connections have been isolated electrically and also for liquid contact. In case of fire theam? has two different fire extinguishers which are involve all type of fires (ABC type) and second extinguisher is especially for electrical fires. On hexacopter chasis is aliminium so fire risk may occur just because of electricity. Team has knowledge about how to use fire extinguisher and which extinguisher is used for what kind of fire. Preflight Go, No-Go checklist to ensure optimum operational status of the hexecopter. 20 Team has a precautions check table. o Before the every flight operations, dedicated and trained team members checks the safety precautions. o During the competition, these members will use check tables to avoid any mistakes Team has a some safety practices o Fire extinguishing o First aid in case of small injuries o Handling with explosive and damageable parts on UAV 7. Conclusions This journal paper has shown that Anadolu SUAS team has a solid understanding of mission requirements and rules for the 2014 AUVSI Student UAS Competition. Anadolu SUAS team worked over the past nine months to set up the Anadolu SUAS hexacopter. The system was developed by a team of ten undergraduates and one graduate students split into team. Our aims fulfill the duties such as automatic flight, navigation, surveillance, real time actionable intelligence. Anadolu SUAS team looks forward to the opportunity to prove our abilities to succeed in the competition. 21