Unmanned Aerial Vehicle Competition Team Fixed Wing

Transcription

1 Unmanned Aerial Vehicle Competition Team Fixed Wing May 31, 2007 Faculty Advisor: Dr. Charles Reinholtz Graduate Teaching Assistants: Ben Dingus Rick Bis Shane Barnett Undergraduate Team Members: David Garber Ryan Haac John Humphreys Justin Macdonald Brendan Mulhare Eric Personne Stefan Toussaint Reuben Walton Jeff Ward Curtis Wilkes

2 Abstract This Autonomous Aerial Vehicle Team (AAVT) of Virginia Tech has developed an Autonomous Reconnaissance System designed specifically to meet the demanding mission requirements of the 2007 Student Unmanned Aerial Vehicle (UAV) Competition mission. The 2007 competition places increased mission demands on the system, particularly in the areas of side-looking cameras and automated target recognition. To meet these demands, the Virginia Tech team has developed a new, innovative vehicle platform, integrated a gyro-stabilized, retractable gimbaled camera system, upgraded our flight controller and ground station hardware and created a new suite of mission specific software. This new system is capable of fully autonomous operation in every phase of the mission, including take-off and landing. GPS waypoints tracking to reach the search area, automated search patterns, computer-vision based target recognition, gimbaled camera target tracking, and automated logging of the number, orientation, and location of targets. Wherever possible, our system is based on commercial, off-the-shelf hardware and electronics to leverage existing technology and to ensure system reliability. Our focus was on developing custom mission software, system integration, testing and safety. In event of an in-flight autopilot failure, two separate backup systems can be used to regain control before a hard e-stop is invoked. We believe our system is capable of meeting every requirement of the 2007 UAV competition. 2

3 Table of Contents Introduction 4 Mission and Objectives 5 Scope and Assumptions 5 Approach and Methods used in the Design Process 5 Results and Discussion 6 Airplane Subgroup 6 Navigation Subgroup 8 Vision and Communication Subgroup 10 Summary and Conclusions 14 References 15 Appendix A 16 Appendix B 18 Appendix C 18 3

4 Introduction The Student Unmanned Aerial Vehicle (UAV) Competition is organized by the Association for Unmanned Vehicle Systems International (AUVSI). The purpose of the competition is to construct an aerial reconnaissance vehicle capable of meeting physical criteria and mission objectives as outlined in the rules provided by the competition organizers. To accomplish the UAV Competition goals, the team must design and construct a UAV that meets several vehicle requirements and mission objectives. The competition is judged on three components: a journal paper, a static presentation, and the mission performance. The tasks the UAV should be able to complete for the mission performance portion include autonomous control of the vehicle and the detection of targets on the ground. Control of the aircraft will be assessed several ways. First, the vehicle must fly autonomously once in the air without aid from a human pilot. It bonus points are awarded for autonomous takeoff and landing. Once in the air, the vehicle must follow predefined GPS waypoints while avoiding given no-fly zones. These waypoints, similar to airspeed and altitude, are dynamic and must be able to be changed during the mission via the ground station. The second part of the mission performance component to the competition is that the UAV must find and identify several targets in a given search area. Once a target is identified, the UAV should be able to transmit the position, shape, compass orientation, background color, alphanumeric symbol, and alphanumeric color of the target to the ground station in real time. Also, the UAV must be able to survey an area 60 degrees in all directions from directly below the vehicle, identify a known target from an altitude of 500 feet, and find a target 250 feet from the centerline of the vehicle s path while flying at 200 feet. To be allowed to participate in the competition, the UAV must first meet a set of physical and safety guidelines that are outlined in the competition rules. The physical aspect of the competition is primarily a limitation on the weight and type of the vehicle. The rules state the vehicle must be less than 55 pounds and also cannot be of a lighter-than-air variety, such as a blimp. The UAV must be able to operate in wind gusts of up to 20 knots and temperatures at or below 110 degrees Fahrenheit. The team is limited to one vehicle in the air at any time and the system cannot use any ground based sensors for target detection. In preparation for the 2007 competition, several intermediate goals were established with deadlines to be met throughout the year. An early step the team took was breaking up into three subgroups to focus on different components of the vehicle. These subgroups are airplane, navigation, and vision/communication. The airplane team is responsible for vehicle selection, modifications to the airframe, maintenance of the vehicle and the physical layout of system components. The navigation group is focused on learning the autopilot system for autonomous flight, obtaining GPS coordinates, and developing autonomous take off and landing strategies. Finally, the vision/communication team will take the pictures and video taken from the plane, input the information into a computer program they develop to interpret the data, and relay it the ground station personnel. 4

5 The remainder of this report discusses the team mission and objectives, the scope of the project, our design methodology and our solution. Mission and Objectives The overall objective for the team is to create a competitive entry in the Student UAV competition. This objective will be obtained through the design, testing, and refinement of an autonomous aerial reconnaissance vehicle. The components for the vehicle were mainly chosen from off-the-shelf products combined. The new team was determined to push the envelope in this competition by meeting the new, more demanding mission requirements. We began the year by trying to learn from last year s design; we conducted test missions based on the rules of the previous competition and we sought input from members of last year s team. Scope and Assumption In approaching this project, the team knew that funding would be limited. We also knew that the autonomous reconnaissance task required an unmanned aerial vehicle that is able to fly with a preset path, have a visual image of the ground around the UAV, and relay information back and forth to a ground station. Approach and Methods used in Design Process This section of the report will discuss the team s approach to solving the problem. The team held numerous brainstorming sessions to generate ideas that could be further developed and tested. With this information, ideas were reduced to realistic solutions to the problem. Next, the customer needs and performance criteria were identified so that the decision process toward a final solution could begin. Customer Needs. The primary customers identified were the competition judges, the team s faculty advisor, and potential customers such as the United States Marine Corps. A majority of the customer needs were derived from the AUVSI rules, since these expectations defined the required vehicle performance. The resulting customer needs stated the vehicle had to be capable of seeing targets not solely below the airplane, the UAV has to be able to identify target characteristics, and the team must be able to show dynamic retasking from the ground station while operating autonomously. Performance Measures. Once the customer needs were determined, performance measures were established that provided target specifications for the vehicle. The target specifications require that the vehicle must be able to see and take a photograph 60 degrees from vertical in all directions, and the target parameters of shape, background color, alphanumeric symbol, alphanumeric color, orientation, and location must be identified by the vehicle and transmitted in real time to the ground station. The vehicle s airspeed, altitude, and heading must be able to be changed via the ground station while in flight. In addition, while at an altitude of 200 feet the UAV must be able to identify the target parameters of a target 250 feet away from the vehicles flight path. Finally the vehicle must be capable of autonomous take off, flight and landing. 5

6 Results and Discussion This section discusses the results for each of the subgroup s designs for the unmanned aerial vehicle team. This section also includes a discussion of the critical issue of safety. Airplane Subgroup. The airplane design group had the task of developing a reliable airframe that could meet performance requirements and handle the payloads developed by the vision/communication and navigation groups. Also, choosing a suitable engine to propel the airplane and payload is important as well. In order to achieve these objectives, the group investigated various remote-control airplanes and engines. From this research, the group determined that the Sig Rascal, shown in Figure 1, best fulfilled the objectives. The group chose this platform over many others, including last year s Sig Cadet Senior. A primary consideration was the much larger payload capacity of the Rascal. Table 1 show that the Rascal is a larger plane with a greater payload capacity. The larger fuselage is also allows the insertion and extraction of larger, more sophiticated payloads. Figure 1. Sig Rascal after construction 6

7 Table 1. Product Specifications for the Sig Kadet Senior and Rascal 110 RC Airplanes. Airplane Wing Span (in) Wing Area (in^2) Flying Weight (No Payload) (lb) Sig Kadet Senior to 6.5 Sig Rascal to 13 Payload Capacity (lb) Wing Max Payload Price ($) Engine Size (glow) Airplane Sig Kadet Senior to 1.00 Sig Rascal to 1.60 The purchase of a larger airplane meant upgrading other components as well. First, we determined we needed a larger engine, not only for the larger plane, but for the increased payload capacity. Based on the Rascal s weight and the payload, the OS 1.60 engine was a reasonable choice. It has ample power to meet the needs of the competition. Table 2 shows that the OS 1.60 is the most powerful engine among those considered in our research. A larger engine and payload, combined with the required endurance time, called for a larger fuel tank. Instead of the stock 16 ounce tank, a 24 ounce tank provides the ability to fly for an estimated 25 minutes based on preliminary fuel consumption data. We determined that an 18 x10 propeller, though significantly larger than previous propeller, would provide the best performance for competition. Table 2. Product Specifications for Remote Control Airplane Engines. Engine Power (bhp) Weight (oz) Max RPM OS OS OS GMS Engine Fuel Consumption (oz/min) Price ($) OS 0.61 ~ OS OS GMS 1.20 ~ During preliminary construction of the Sig Rascal, we encountered a few problems that also need to be solved. First, we needed to alter the interior design of the fuselage to house the sensor payloads. This required moving the servo mounting plates for the rudder, elevator, ailerons, and throttle. Second, since the engine we chose to handle our large payload was not the typical engine used on the Sig Rascal, and since stability is a very large issue when it comes to taking aerial photographs, we deemed it necessary to install vibration damping engine mounts. The engine dampers reduce vibration seen by the camera allowing for better images during flight. Finally, since the new Cloud Cap camera system is sensitive and expensive, the aircraft was 7

8 equipped with a innovative scissor-lift device that lowers and then retracts the camera into and out of the plane. This allows the camera to be secured inside the plane and not be susceptible to immediate damage from the ground during takeoff and landing. When the lift is fully extended, the gimbaled camera has its full field of view during operation. This custom scissor lift is shown in Figure 2. Figure 2. CAD drawing of scissor lift to retract gimbaled camera. Navigation Subgroup. The Navigation team of the UAV project is primarily focused on improving the accuracy of the vehicle s mission execution. These requirements were addressed in several ways. The most obvious goals was to ensure that the airplane flies smoothly and 8

9 reliably. The navigation subgroup decided that the best avionics package available to meet this objective was CloudCap Tech s Piccolo LT and ground station. The Piccolo LT is an avionics package that is optimized for small aerial vehicles. The ground station is necessary to manage the wireless link between the Piccolo LT and ground station computer. It also provides differential GPS corrections and a convenient interface viewable from a supporting PC. Figure 3. Cloudcap Technologies Piccolo LT When the autopilot was first tested with stock settings from the manufacturer, the flight dynamics were unstable. This made flying difficult and it made autonomous landing and takeoff virtually impossible. The team spent many hours in the time-consuming process of tuning the PID loops and gains the autopilot uses to control the plane. The aim was to maximize the vehicle s stability by teaching it how to react to a given input. In the end, the airplane was able to fly smooth lines or arcs, allowing it to stay close to the designated path assigned by the judges or defined by our reconnaissance plan. This increased stability will also enable us undertake autonomous takeoff and landing sequences, fulfilling some of the desirable mission requirements. The navigation subgroup has done extensive software-in-loop and hardware-in-loop bench testing. We have successfully compiled our own simulator file, which is composed of the various important parameters that define the dynamics of the Sig Rascal. We have been able to find geo-reference map files of Virginia Tech s Kentland Farm and Webster Field, the two locations that the UAV will be flown. We have uploaded these map files, duplicated the previous teams flight plans, and successfully taken off, flown and landed all autonomously these flight plans. We have done an extensive amount of these simulations with Kentland Farm and Webster Field. 9

10 Figure 4. Cloudcap Tech s Ground Station After performing these bench tests in the software-in-loop mode we integrated and installed the Piccolo LT into the plane, and attached all the servos, power, and communication ports correctly and did hardware-in-loop testing. The navigation group setup the same flight plans in the Piccolo User Interface and watched the plane, which was stationary in the lab, controlled the various surfaces depending on the path of the aircraft. The plane responded well to the hardware in the loop testing. We expect to achieve target position accuracy within 15 meters. To be able to do this, we expect GPS accuracy within 5 meters. We should be able to easily program updated flight paths during flight as requested by the judges. Our team will also ensure that fully autonomous flight, including takeoff and landing, will be a reliable process. A list of components to be used and their specifications is outlined in Appendices B and C. Vision and Communication Subgroup. The Vision and Communications sub-group had the tasks of creating a reliable communication link between the plane and ground, providing realtime video to the ground during flight, and acquiring images of targets from the air. Our design focuses on the use of the Cloud Cap Technology s Piccolo LT autopilot and TASE gimbaled camera. These two components are mounted inside the aircraft and communicate wirelessly with our ground station. Two lithium ion batteries are used to power the payloads. One 11.1V 4400 mah battery powers the TASE gimbal and the 1W AV transmitter, shown in Appendix A1. The Piccolo LT is independently powered by its own 11.1V 2200 mah battery for safety. In addition to these components, we have installed an RxMUX switch to provide an extra failsafe to prevent loss of control of the aircraft. We know of previous flight 10

11 failures when the autopilot would not give manual control back to the pilot. The RxMUX provides a reliable toggle between RC and autopilot control by integrating a second RC receiver to bypass the autopilot. A 4.8 V, 2100 mah battery will be mounted to provide power to the RxMUX which powers the control servos. The video sent to our ground receiver is split three ways as illustrated in Appendix A2. The video signal is sent to our automatic target recognition computer, our gimbal control computer, and a minidv camcorder. The use of the automatic target recognition and gimbal control computers will be discussed below. The minidv camcorder is used to store video for later study. The Piccolo telemetry and GPS data sent to a separate receiver is delivered to Cloud Cap Technology s ground station. This data is then sent to the gimbal control computer and the Piccolo interface computer via serial connections. Finally, an ethernet switch provides communication between the automatic target recognition and gimbal control computers. TASE Gimbal Camera Our team was fortunate to be able to purchase Cloud Cap Technology s TASE gimbaled camera, shown in Figure 5. The TASE gimbaled camera was chosen because it provides inertial stabilization, the ability to view targets on either side of the plane, and a high zoom capability. The inertial stabilization is important so that the camera is not bouncing around aiding in a clear picture. The gimbals ability to pan left and right gives us the ability to find the targets that will be placed outside of the fly zone. The camera s high zoom capabilities will allow us to clearly read targets from high altitudes. Another feature of the TASE gimbal and camera is its selfcontained GPS and transmitter. This allows the TASE to work independently of the autopilot. Figure 5. Cloud Cap Technology s TASE gimbal and camera Communication An important metric to successful automatic target detection is the resolution of the image being processed. To ensure high-resolution real-time video, a 1W AV transmitter was selected from Black Widow A/V. This transmitter broadcasts at the unlicensed 2.4 GHz band. To facilitate a noise free transmission, a 6 dbi omni-directional antenna will be used on the onboard 11

12 transmitter and a high gain directional antenna along with a high gain omni-directional antenna will be used on the ground diversity receiver. The coax video cable from the receiver will enter a video splitter and then sent to the two computers and video recorder. The Piccolo LT will broadcast at 900 MHz and the RC Controller and RxMUX Safety Device will broadcast at 72 MHz. Vision and Programming After we acquired the TASE, our team created several programs capable of performing all of the vision requirements set forth by competition rules using the TASE s video stream. The entire task was split into two main components, control of the TASE and acquiring of images. Using National Instrument s LabVIEW software, we used a frame grabber to convert the video stream into individual picture files. With changes in this year competition rules encouraging autonomous target detection, we utilized LabVIEW s Vision tools to detect the targets by color. The image is first run through a color threshold to pick out individual colors. These colors are black, white, red, blue, orange, yellow, and green. The color threshold is run for each of the seven colors in a stacked sequence for each picture frame. The program then runs the image though a particle filter which removes extraneous pixels. Anything under or over the specified range is removed from the image. This process can be seen visually in Figure 6. If there is a target remaining, the program saves the original image with an overlay of what it found as the target. A pop-up window of the target appears while it is being saved to file. The program is currently complete but is in need of fine tuning. To make the program more robust, the color thresholds of the seven colors can be changed directly on the front panel and ideally can be changed to include a color it might have skipped on its first pass. Lastly, the program automatically returns a pixel X, Y coordinate along with a timestamp. The purpose of the coordinates is to find the corrected GPS location of the target. The timestamp assures that the correct GPS data is used. The application and reasoning behind this is discussed later. Figure 6. Original image from TASE video camera and same image after processed by the LabVIEW program. 12

13 Also created is a manual LabVIEW program that fulfills all competition goals as a backup to the automatic target recognition program. This program displays streaming video from the TASE for an operator to see. If a target comes into view, the operator clicks Target which then prompts the operator the click the target in a pop-up window. The prompt is to specify the pixel X, Y coordinate of the target. Unlike the automatic program, both target recognition and X, Y coordinate selection have to be specified by the operator. Lastly, the program overlays the corrected GPS location and saves the image to file. Software It is mission-critical to be able to control the hardware onboard the plane from the ground station. The software we use affords us this ability to relay commands and receive status information from the UAV. The purpose of the software is to provide logical communication to the Piccolo LT and the TASE system, and perform calculations on the data that we receive. The Piccolo LT and the TASE system send information in the form of packets, but the packets that propagate from the systems must be decoded by our software in order to understand the data. Cloud Cap Technologies provides a CommSDK as a wrapper for the low level packet communication. Our software utilizes this library to simplify the process of receiving and understanding data from the plane. Once the data is received, the software then uses the data to perform the necessary transformations to georeference the video received from the TASE system. The software is able to receive an X, Y coordinate from the video screen and transform that screen point to a Lat/Long point on the surface of the earth. We also use this system to provide a real-time visualization of the calculations throughout the mission. This visualization shows us the location of camera, the field of view, and the area we've covered during flight. This enables us to be able to have a sense of what is actually happening during the mission and provides a platform for debugging our software. Figure 7 is a screen shot of the software. Figure 7. Screen shot of the software visualizing the location of the camera and the field of view. 13

14 This software informs the user of the location and the orientation of the camera. Using this information we can calculate the field of view of the camera and georeference the video received from the TASE system. This information is used in accurately locating targets during our mission. Summary and Conclusions We have developed a new fixed-wing unmanned aerial vehicle and the supporting hardware and software that are collectively capable of meeting the mission requirements specified for the 2007 Student UAV competition. Our system includes a collection of commercially available components that were specifically selected for optimal safety and performance. These components include a Sig Rascal airframe, a Cloudcap Piccolo LT controller and ground station computer, a gyro-stabilized TASE gimbaled camera system and an RxMUX switch to provide failsafe control of the aircraft. In addition, we have developed an extensive collection of custom mission-specific software and key electromechanical subsystems such as the novel scissor-lift mechanism for stowing the gimbaled camera. We look forward to presenting our complete solution at the competition. 14

15 References Association for Unmanned Vehicle Systems International- Student Competitions October 15, 2006, Gregg Vonder Reith, Ken Meidenbauer, Imraan Faruque, Chris Sharkey Jared Cooper, Shane Barnett, Dr. Charles Reinholtz, 2005, Development of an Autonomous Aerial Reconnaissance Platform at Virginia Tech, Virginia Tech Department of Mechanical Engineering, 2006 Appendices: Appendix A1: Plane Components Charts Appendix A2: Ground Components Charts Appendix B: Navigation Subgroups Components and Specifications Appendix C: Vision Subgroup s Components and Specifications 15

16 Appendix A1: Plane Components Chart 16

17 Appendix A2: Ground Components Chart 17

18 Appendix B: Navigation Components and Specifications Item Price Weight Power Consumption Piccolo LT Autopilot and Accessories 109 g ~ 1W Ground Station Kit Developer s Kit Avionics Integration Kit Various Antennas Piccolo Interface Cable UHF and GPS Ground Plane NiMh Acell Battery Packs Pitot-Static Kit Cable Set, CDI Deadman Tach $8,500 $1,000 $5,500 $206 $125 $10 $160 $560 $90 DGPS Receiver $282 minimal minimal RC Transmitter with Servos $ lbs 2.4 W Appendix C: Vision Components and Specifications Item Price Weight Power Consumption TASE Camera and Accessories PTZ Gimbal with Stand Developer s Kit $15,000 $4, g N/A 18