2011 AUVSI Student UAS Competition. Team Spycat

Transcription

1 2011 AUVSI Student UAS Competition Team Spycat Kansas State University Team Members - Nick Clattenberg, Shawn Georg, Nathan Feldkamp, Bryan Rogler, Brian Blankenau, Jacob Wagner, Nathan Reichenberger, Jacob Kongs, Brandon Lackey, Eric Johnson, Jeremy Taylor, Jon Albrecht, and Jon Thompson (pilot) Faculty Advisors - Dr. Thompson, Dr. Schinstock Submitted - May 23, 2011 Proof of Flight Statement: We certify that the aircraft described in this paper has flown and is fully operational. For the 2011 Association for Unmanned Vehicle Systems International (AUVSI) Student Unmanned Aerial System (SUAS) competition, Team Spycat developed a UAS system to satisfy military needs and competition specifications for an aerial reconnaissance system. Using a systems engineering approach, the team evaluated multiple air vehicle platforms and ground station concepts according to a prioritized list of key performance parameters (KPP). The chosen overall system concept consisted of the Sig Kadet Senior aircraft and 3 laptop computers for the ground station. Subcomponent design, test, and evaluation (DTE) teams were employed in a two-stage process to ensure added payloads and modifications to the aircraft met safety and performance thresholds. The 3 Microsoft 1080p web cameras provided data at 2.4 GHz through a Wi-Fi connection to the ground station while the Piccolo II autopilot system communicates to the air vehicle operator (AVO) laptop at 900 MHz. Full system tests were conducted prior to the competition to ensure that the UAS meets safety standards and is able to complete all mission requirements.

2 Table of Contents Introduction... 3 I. Analysis of Mission Requirements... 3 II. Systems Engineering Approach... 3 Platform Analysis... 4 Air Vehicle Strategies and Selection... 5 Information Management Analysis... 5 Ground Station Setup & Flight-line Communication Strategy... 6 III. Subsystem Design, Testing and Evaluation (DTE)... 6 Camera Payload DTE... 7 Trade Study... 8 Data Processing DTE Autonomy, Navigation and Control DTE Data Link and Ground Station DTE Air Vehicle Modifications DTE IV. Risk Management Plan Risk Identification Risk Mitigation Pre-flight Checklist References Kansas State University - Team Spycat Page 2

3 Introduction For company level operations, the U.S. Marines have requested the use of unmanned aerial systems (UAS) to support a patrol as they conduct point or route reconnaissance. The mission of the UAS is to conduct sweeps for actionable intelligence, analyze possible targets and send the information to the UAS operator or supervisor. The UAS must also be able to accept special instructions for takeoffs, landings, and search procedures while maintaining airspace integrity. To accommodate this request, Team Spycat has designed and built an unmanned aerial system to compete in the 2011 AUVSI Seafarer Chapter Student Unmanned Aerial System Competition. The method for designing and implementing the UAS included analyzing mission requirements, developing system solutions, selecting the best system design and testing to ensure the final product meets or exceeds the performance and safety standards. I. Analysis of Mission Requirements The mission profile for the competition consists of sending an air vehicle to fly through waypoints or search areas to properly locate and identify target characteristics for an unknown number of targets and transmit intelligence requirements to the ground station operators within 40 minutes. To accomplish this task, any type of aerial vehicle that is of a heavier-than-air configuration may be considered. The list of Key Performance Parameters (KPP) was the primary tool used in defining the scope of the systems engineering problem. The six KPP s contain the minimum threshold requirements and desired objectives for each rated category of flight performance. The KPP categories are: autonomy, imagery, target location, mission time, operational availability, and in-flight re-tasking. Without direct numerical weightings assigned to each KPP, the relative KPP value must be determined in another way to aid in the design selection. The method for prioritizing the KPP categories was subject to the need of the U.S. Marines for accurate, real-time intelligence. This was interpreted to emphasize imagery, target location, and then mission time as the driving factors of the design selection process. Autonomy was highly considered as an important factor to refine the air vehicle navigation and imagery processing. This KPP alone had the greatest potential as a force multiplier by reducing the number of required operators in the field. The operational availability KPP was next in importance as it pertains to the endurance of the UAS to accomplish the mission at hand. The flexibility to incorporate in-flight retasking was assigned lowest priority as it would require the least amount of work to satisfy. II. Systems Engineering Approach With experience in developing UAS applications for both fixed wing and rotary air vehicles, Team Spycat had to make some key decisions early in the design process regarding the air vehicle platform. While the air vehicle had to maintain standards developed by the Academy of Model Aeronautics (AMA), the Kansas State University - Team Spycat Page 3

4 ground stations had no official restrictions. General considerations were made to have ground stations simple and portable, but the air vehicle had priority in the first stage of the UAS design process. Platform Analysis To complete the mission, rotary vehicles would utilize aspects of their design differently than the fixedwing aircraft. Therefore, separate strategies had to be developed alongside the apparent physical limitations of each air vehicle to adequately compare the two systems. Rotary vehicles come in a variety of configurations including small helicopters, multi-rotors (tri or quad rotors), and internally ducted fans. The key strength that is common for all rotary vehicles is the ability to hover and take high quality images to identify the target characteristics consisting of shape, alphanumeric, alphanumeric color, background color, and orientation. The limitations for this class of air vehicles typically involve low cruise velocities for covering large search areas, vulnerability to enemy fire while hovering, and low range/endurance maximums. Due to the increased level of complexity, internally ducted fans were not included in the air vehicle comparison. With significant differences in performance capabilities, it was deemed necessary to compare both the helicopter and multi-rotor air vehicles. Helicopter Ducted fan Multi-rotor Figure 1: Conceptual Rotary Air Vehicles Fixed-wing air vehicles have different configurations that vary the number of wings and location with respect to the center of gravity. These factors distinguish fixed-wing aircraft in their ability to satisfy the KPP s. Notably, Bi- or tri-wing configurations may increase the lift coefficient and decrease the overall cruise velocity. By operating at lower velocities, the air vehicle can enhance imagery and target location at a significant cost of simplicity and portability. Fixed-wings with center of gravity locations toward the rear usually require main wings toward the rear and canard surfaces at the nose of the aircraft. Unless carefully designed, the canards create problems in predicting the stall characteristics at high angles of attack for an autonomous landing. Therefore, the only selected fixed-wing configuration for comparison was a simple mono-wing with the center of gravity towards the front of the vehicle. Kansas State University - Team Spycat Page 4

5 Mono-wing Bi-wing Canard aircraft Figure 2: Conceptual Fixed-wing Air Vehicles Air Vehicle Strategies and Selection The most reasonable options to compare were the helicopter, multi-rotor and mono-wing air vehicles. Using the prioritized KPP s developed earlier, the team evaluated the expected mission performance through a weighted decision matrix. Air Vehicle Imagery (x2) Target Location (x2) Mission Time (x2) Autonomy Operational Availability In-flight Retasking Helicopter + S - S - S -1 Multi-rotor S S S + S S 1 Mono-wing - S S 2 Table 1: Air Vehicle Weighted Decision Matrix While the multi-rotor s relatively high cruise velocity made it an attractive rotor vehicle option, the simple mono-wing had the best capability of completing the mission within 20 minutes while not using any time outs. With offset scores in the most important KPP categories, operational availability was the deciding factor in choosing the fixed-wing over the multi-rotor. Team Spycat chose the familiar platform of the Sig Kadet for its aerodynamic configuration, sound structural design, and fuselage. Because the type of aircraft chosen was designed specifically to carry a variety of sensors, few modifications were necessary to carry the payloads our team needed for the mission. Information Management Analysis To accommodate the air vehicle, information for autopilot navigation and imagery needed separate communication frequencies from the air to the ground station. Once in the hands of the ground operators, imagery information was expected to be quickly accessed and analyzed. Simultaneous to the imagery processing, operators sought to designate search areas or waypoints for the air vehicle to locate and identify targets. The main goal was to ensure flight safety throughout the mission by reducing the workload per operator and creating redundancies where applicable in the flight programs and flight-line operations. The second goal was given to maximize the KPP s through increased speed, accuracy, and efficiency of imagery processing. The final goal was to maximize autonomy in the system for all aspects of the flight mission profile and imagery processing functions. Totals Kansas State University - Team Spycat Page 5

6 Ground Station Setup & Flight-line Communication Strategy The variables to be determined at this point in the design process were the required number of operators, ground station laptops, and software programs that could properly manage all aspects of the mission to include recovery and emergency procedures 1. The simplest solution was to divide the ground station into three separate laptop computers. One laptop was designated to operate the autopilot navigation and the other two were assigned to the operation of target location and imagery processing software programs. During the mission, the minimum number of operators must be three to continuously operate the three laptop computers. We chose to assign additional members as the RC pilot and the assistant to the target identification operator to hold the antennae for the imagery receiver. Lastly, the sixth and final member was chosen as the team supervisor and act as the ground crew s spokesman to judges and safety officials. Images Receives: Sensor data Transmits: GNC commands Actionable Intelligence Imagery Processing Identified targets Target Identification Autopilot Navigation Supervisor Imagery Analyst Target ID Operator Air Vehicle Operator (AVO) Antennae Assistant Figure 3: Ground Station Setup and Communication Concept RC Pilot III. Subsystem Design, Testing and Evaluation (DTE) Having identified the platform and communication structure for accomplishing the mission, we addressed details for the component designs in the order of our prioritized KPP list. Without prior years experience in this competition, determining simple methods to meet the threshold requirements was the greatest challenge. When testing the requirements of one subsystem and finding the need to make adjustments, the designs of the other multiple subsystems would have to accommodate these changes. Therefore, the process for designing the components or programs at this level was iterative and inseparable from testing and evaluation. 1 Emergency and mission recovery procedures are described in Section IV. Kansas State University - Team Spycat Page 6

7 We chose to implement a two-stage method for most DTE teams. The first stage identifies the KPP threshold parameters and the designs necessary to accomplish them. The second stage focuses on refining the previous system decisions in order to ensure maximum system safety, efficiency, and KPP score. The synthesis of this process yielded our final design decisions. System Concept Test Making it work Design How it should work No Evaluate Are KPP thresholds met? Is it feasible? Yes Design How to maximize score No Evaluate Are safety and KPP goals met? Test Making it work Final Design Yes Figure 4: Stage 1 of DTE Process Figure 5: Stage 2 of DTE Process From the system concept, Team Spycat was divided into teams that applied this two-stage DTE process. Camera Payload DTE The first team developed the camera subsystem. This component had to be tested for individual quality and interoperability with the data link, processing programs, and the air vehicle. Therefore, the camera team had to communicate requirements and changes in the design to the other teams as quickly as possible. Stage 1 To accomplish threshold requirements, the team tested a Toshiba security camera. It was capable of real time video feed transmitting to the ground station via ethernet connection. The security camera was small enough to fit within the fuselage, lightweight, and able to be gimbaled for capturing images off to the side. This qualified the camera system for potential use during the competition. However due to poor image quality it was deemed unusable for the target image data processing. For better image quality, the team explored the possibility of using a Canon Rebel T1i series camera. Taking 2-4 pictures per second with 15 megapixel resolution, images would have been excellent for the autonomous imagery analysis software. However, the air vehicle placed restrictions on the size and maneuverability of the camera. Gimbal designs to move the camera or rotate a set of reflective mirrors were both made in attempt to capture images off to the side at 60 degrees from underneath the aircraft. Ultimately, none of the designs for this camera were able to overcome fuselage space restrictions and the ability to capture images to the side of the aircraft. Kansas State University - Team Spycat Page 7

8 Stage 2 The team then searched for another camera with real time video feed transmission and found the 1080p Microsoft webcam. With greater image resolution, this camera had only two issues that needed to be addressed before being competition ready: autofocus resolution and camera configuration. During the testing phase, the autofocus would distort images of an example target. The data processing team had difficulty filtering lines or detecting targets. Unfortunately, attempts to adjust or electronically control the autofocus were not compatible with the webcam. To resolve this issue, the camera was taken apart to permanently disable the autofocus. Further increases in the image quality were achieved by replacing the original lens with one that had a smaller field of view. While this decreased the area covered by each picture, the sharpness of the image significantly improved and allowed for the possibility of autonomous target identification. The other important modification was finding a way to capture the image off to the side of the flight path. The 1080p webcam had a couple options to overcome the 60 degrees of rotation. In contrast to the Canon Rebel, gimbal designs would work for this webcam simply because of its small size within the fuselage. The other option was to mount multiple fixed webcams in different directions below the aircraft, thus eliminating the need for a gimbal. To properly compare these two options, a more refined analysis of the decision was conducted in terms of the camera system s effect on air vehicle endurance, simplicity, and reliability to accomplish the mission. Trade Study The outcome of this trade study was primarily used to determine which camera configuration would best suit Team Spycat s UAS. The secondary purpose of this trade study was used in general to measure the flight score s sensitivity to the changes in weight and power management of one subsystem. Implications of this trade study were used to develop a consistent approach to evaluating the relative cost and benefit of adding components to the air vehicle. The camera system affects the endurance of the air vehicle in two different ways, yet both can be measured in terms of weight. The first way is a direct measurement of the weight, comparing the difference between the two camera systems. Second, the weight of additional batteries can be measured if the amount of estimated power increases to operate one camera system versus the other. We made a reasonable assumption for this analysis to be accurate. The battery power to operate the gimbal or cameras would be separate from the battery power to run the aircraft motor; energy gained or lost would only be used for camera operation. So, relating back to the aircraft s power requirements, the effect of increased camera power consumption was measured in weight of the additional batteries as opposed to energy. However, would this added weight have a negative impact in terms of measurable scoring parameters? The answer rests with the definition of the cruise velocity as it relates to the air vehicle s flight endurance. For every aircraft, this endurance is maximized at a certain velocity where the induced drag and parasite drag are equivalent. For steady flight, this can be called a maximum endurance cruise Kansas State University - Team Spycat Page 8

9 velocity (as opposed to max range). The equation relating this velocity to the weight and lift coefficient is given below [1]. 2W V cruise = ρ SC L_cruise The design goal for reconnaissance aircraft is to stay aloft as long as possible scanning for targets at this cruise velocity. If endurance is the only factor in the decision making process, increased weight would always have a negative impact. For the competition, if the estimated flight is assumed to be greater than 20 minutes, the weight could be slightly increased to help reduce the mission time. If less than 20 minutes, less weight would be better to maximize endurance while scanning for targets. So, the goals of the trade study were to benchmark current vehicle specifications against an assumed mission profile distance and calculate the mission time. Following through with the comparison methodology, calculations for optimum cruise velocity depend upon the gross weight, cruise lift coefficient, wing area and atmospheric density. The wing area is fixed for the Sig Kadet at ft^2 and atmospheric density at slug/ft^3. The current weight is 11 lb. It is assumed that the Sig Kadet uses a Clark-Y flat bottomed airfoil. Experimental studies of the Sig Kadet airfoil have shown that lift curve versus angle of attack varies slightly depending upon the type of Clark-Y airfoil. In a wind tunnel, the type of Clark Y used by the Sig Kadet resembles the lift curve slope of a Clark-Y 11.7% smoothed or Clark-Y2. [2] Calculated below are the alternative airfoils using Xfoil at a Reynold s number of 350,000, which is close to the cruise velocity at operating conditions. Figure 6: Sig Kadet Airfoil Lift vs. AOA Kansas State University - Team Spycat Page 9

10 With a finite wing lift coefficient of at zero angle of attack, the resulting cruise velocity came to be 48.4 fps. Assuming that the total distance flown to find all the targets is set at 20 miles, the estimated mission time would for this cruise velocity would be 36 minutes and 22 seconds. Recalculating the cruise velocity for a 1 lb increase in the weight, the cruise velocity comes to 50.5 fps. If the autopilot chooses to fly at this cruise velocity, the mission could be completed in 34 minutes and 51 seconds. The positive aspect of gaining weight and choosing to fly at the optimum velocity means that the vehicle is saving approximately 1 and a half minutes for a one pound increase in weight! However, non-intuitive benefits of weight can be both positive only if the endurance of the 5 cell LiPo battery can sustain flight for minutes with poor wind conditions. For the Sig Kadet at zero angle of attack, the experimental drag coefficient is [1] The original cruise velocity at 48.4 fps produces 1.37 lbf of drag. A velocity of 50.5 fps produces a drag force of 1.49 lbf. For the Hacker A40-10S motor power requirements, the two drag forces would require the motor to produce 14.8 ft*lbf/s and 16.1 ft*lbf/s at their respective cruise velocities. Converting into SI units, power required for the first velocity was W and for the second velocity, W. The amount of available flight time was dependent upon the battery discharge data but conservative estimates allow for 42 minutes and 35 minutes of flight time, respectively. This will mandate the need for the monitoring of the motor battery current and voltage discharge data to make informed decisions during the competition. If weight can be reduced, it is necessary depending upon the distance to all the targets and number of passes to gather the required intelligence. An additional modification for the 3 camera system was to capture images off to the side of the flight path at a minimum of 60 degrees in any direction below the vehicle. In order to do this, one camera was pointed straight down and the other two were mounted at 45 degrees from the bottom of the aircraft. With modified lenses, each webcam now had a 40 degree field of view designed to fly at 200 ft to cover a span of 800 ft lateral distance as the aircraft flies overhead. Because of the Beagle board limitations, one camera would be enabled at a time. So, depending upon the position of the target with respect to the aircraft s flight path, the appropriate camera would be enabled by the target identification operator. z y Figure 7: 3-Webcam configuration fields of view from aircraft at 200ft altitude (y-z axes in ft) Kansas State University - Team Spycat Page 10

11 To make it work for spacial requirements within the air vehicle, a bracket was designed that would fit in a location previously cleared for the Rebel T1i. The bracket and housing was designed to be easily placed on the aircraft and give the cameras the required views specified in the previous figure. Figure 8: Camera Housing Design Three webcams were selected over the gimbaled camera mostly because they kept the system simple reducing the number of moving parts and reliable; depending only upon electrical communication rather than electrical and mechanical functions. Finishing the camera system DTE process, many lessons were learned regarding the nature of the competition and how we were to approach future system decisions. Data Processing DTE Target identification sequences, imagery analysis, and target location data processing was accomplished by the data processing team using the camera pictures and GPS output. They set the standards for image resolution and camera timing. The output of the data processing algorithms created the actionable intelligence for viewing by the judges. Stage 1 For the image processing, this DTE team used the MATLAB software with the Image Processing Toolbox preset functions. MATLAB easily allowed us to transform and run functions, such as edge detections, dilations, and centroids of targets, on jpeg images received back from the Beagle Board. The first attempts at establishing autonomous image recognition began with using MATLAB programs to filter straight lines and areas of constant pigmentation. This style of early recognition would identify the shape of the target and some alphanumeric colors. When testing the target recognition software, the photos needed to be relatively large within the viewing window. It was also determined that a human operator would need to confirm the target suggestions from the imagery software. This was due to limited accuracy at assessing potential targets. Stage 2 When images were received, the program was designed to run an edge segmentation function on the image; use a mask that "filters" out the green grassy backdrop from the target, and then filter out smaller objects based off of pixel size. With the few remaining highlighted areas we can look at the eccentricity in order to determine which marked objects are targets. MATLAB has an intrinsic [stats] code-set which determined many characteristics about the outlined target. This included the centroid (in pixel x and pixel y) combined with our telemetry data to determine the GPS coordinates of the target. Kansas State University - Team Spycat Page 11

12 The telemetry data was accessed directly from the server port of the computer connected to the Piccolo, and was retrieved by a C# program that was used as a time search function. For target location, algorithms were developed to find the center of the target and to filter the images. Retrieving GPS coordinates and attitude information from the operating camera, algorithms utilized Euler angles and trigonometry to calculate the target locations. From this point, a MATLAB conversion program allowed the output to be viewed in the proper format as (ddd.mm.ssss). Figure 9: Example target location, actionable intelligence The target types supported by this data analysis methodology are typical shapes such as squares or circles. Alphanumeric shapes that are supported by this imagery analysis have continuous internal color schemes such as the letters A, I, L, or H. Unsupported alphanumeric targets include B, A, Q, or P that have separate internal regions where color schemes of the target can be confused with the alphanumeric color scheme. Limitations to the current target identification algorithms can detect these alphanumerics as targets but cannot recognize them as ASCII characters. This will be determined by the ground operator. The expected levels of autonomy relating to the imagery analysis for the Team Spycat UAS include automatic target identification, target location, shape recognition and alphanumeric color of supported target types. Additional actionable intelligence will be determined by ground operators on imagery analysis laptops. Autonomy, Navigation and Control DTE Before analyzing the list of performance parameters in detail, the primary goal of this DTE team was to create an unmanned system with the highest level of aircraft autonomy (as opposed to methods of autonomy defined by the imagery DTE team) for takeoffs and landings. While this KPP is extremely important, the process to achieve all threshold parameters in stage 1 of the DTE process was initially overlooked. Being the first year in the competition, initial ambition to maximize this KPP for cash rewards eventually yielded to the stage 1 demand to meet threshold requirements. Lessons learned from this experience helped to develop the two-stage subcomponent DTE process for components that Kansas State University - Team Spycat Page 12

13 will drive in the system design process. Remaining DTE teams did not follow the stage 1 and 2 methodology as their purpose was to respond to the teams driving the overall design. Stage 1 In order to achieve autonomous takeoffs and landings, Team Spycat began developing and testing open source autopilot code from scratch. Notable achievements included the development of the longitudinal and later transfer functions for the complete set of rigid body equations. The purpose was to view the response characteristics of different controllers specifically for the Sig Kadet s weight, inertia, and control surfaces. Linearizing about an estimated 60 fps, the transfer functions for the complete set of rigid body equations were calculated using MATLAB and a trusted reference [3]. In order to control the pitch for an autonomous landing, the change in the Euler angle θ (or pitch with respect to a fixed reference frame) due to a change in elevator deflection δe is given below. 3 2 θ () s 2.806s 3.194s 1.085s () = δe s s s s s For an autopilot to obtain control through crosswinds on landing, the change in the velocity perpendicular to the flight path v due to aileron and/or rudder inputs δ ar, was expressed in the transfer function below. vs () = δ () s s s s ar, s s s Despite much progress in terms of control, timelines began to add pressure on the team to establish a suitable user-interface which could operate with the newly developed autopilot code. The Piccolo II autopilot system had a tested programming structure that would prove sufficient to accomplish threshold requirements for autonomous waypoint navigation, in-flight re-tasking, stability algorithms to control the aircraft and maintain flight pattern altitude and direction. Stage 2 In order to use the Piccolo II autopilot software, an aerodynamic model of the Sig Kadet was created using MIT Professor Mark Drela s Athena Vortex Lattice (AVL) software. Using input geometry and center of gravity information, stability coefficients and derivatives were used to define the air vehicle s dynamic model. Control gains were then determined automatically for the aircraft using the Piccolo software. The autopilot successfully controlled the Sig Kadet in full system tests. From the current simulations, we tested the airplane in cruising conditions. For autonomous takeoffs and landings, the plan is to hook up the laptops to the piccolo system (via CAN connection) and test takeoff and landing procedures. Procedures for testing the takeoff include placing the plane on the ground and giving it the command for max power and deflecting the elevator for a steady climb. For descent, a target location will be given from the GPS and a specific glide slope down to the point will be calculated. Testing will determine the parameters for the best takeoff and landing approach. Kansas State University - Team Spycat Page 13

14 To execute this safely, an RC pilot will be on standby to monitor the situation and take over if problems occur. Figure 10: Software-in-loop (SIL) autopilot simulation Data Link and Ground Station DTE The data link between the air vehicle and the ground station adapted to overcome issues pertaining to the size of the image files being produced by the camera system. In addition, the autopilot and RC control units were operated using the same frequency dependent upon the mission status. In situations with autopilot malfunctions or miscommunications, the air vehicle would be switched from autonomous to RC control provided one safety measure in the event of an autopilot malfunction. Data Link Initially, for the Canon Rebel camera, attempts were made to record data for post flight data processing. Then, once the camera selection changed, all efforts were focused upon making the Wi-Fi system bullets transmit Figure 11: Autopilot and Image Transmission Hardware Kansas State University - Team Spycat Page 14

15 information accurately from the air vehicle to the ground station from the 3 Microsoft 1080p web cameras. The Image Transmission System's data link is composed of two Ubiquity bullets operating on the 2.4 GHz spectrum. The bullets communicate wirelessly with each other over a Cat5 cable providing the communication between the ground station computers and the airborne Beagle Board. The bullets were tested and found to have a range over 4 miles with a decent strength of connection. We can transfer the images at a rate of approximately 1 image per second under ideal conditions. The false detection rate was set at 0.5 Hz to be less than the detection rate as specified in section C paragraph Factors that affect the speed of the image transmission are the CPU load of the Beagle Board, the read/write times on the SD card in the Beagle Board and bullet to bullet connection strength. Testing has shown that the best recovery method for a lost bullet connection is a quick power cycle restarting the Wi-Fi communication. RC Control Figure 12: Communication frequencies and organization Ground Station Setup The Ground station laptops are independent of the Autopilot system. The image transmission software is a Command Line Interface (CLI) and a software program, FileZilla, handles the File Transfer Protocol (FTP) for the transmitted images. The image processing laptop also receives the piccolo telemetry data and stores it so it can be used to find the GPS coordinates. This provided the structure to accomplish the minimum level tasks for the first stage. The Image Transmission System is comprised of 2 laptop computers connected to external monitors. One computer is dedicated to controlling the airborne Beagle Board and downloading the images from the images taken by the airborne camera system. The second computer analyses the images and calculates the target's location. The second computer also maintains a database which stores the Piccolo s telemetry data so it can easily be accessed to determine a target's location. The images will be stored in such a way that they will be accessible over the local network to the necessary computers. Kansas State University - Team Spycat Page 15

16 Air Vehicle Modifications DTE To accommodate the camera selections and autopilot payloads, air vehicle modifications were a necessity. However, for safety and performance considerations, the team identified structural limitations that could threaten the success of the UAS. Landing Gear For flexibility in the camera selection process, the original landing gear needed modifications from the original aluminum skids. In addition, improvements concerning weight reduction and improved structural performance for hard landings were deemed necessary after adding high value payloads within the aircraft. Initial attempts at improving the quality of the landing gear included carbon fiber landing gear layed up at +/- 45 degree angles to support the skids for added torsional strength. After trying to repair prototypes after collision testing, the team determined that on-site repairs for carbon fiber materials were going to pose difficulties at the competition. This led the team to consider a more versatile composite structure that could sustain the required collisions of a hard landing while protecting the air vehicle and internal components. The final landing gear tested was composed of aluminum struts and PVC piping nicknamed Shark Bite with a wooden dowel rod inserts. Flat pieces of aluminum were bent at 45 degrees and connected to the fuselage with metal bolts. The 45 degree angle provided enough clearance for the three cameras and did not obstruct the field of view at 60 degrees from underneath the aircraft. Aluminum was selected because of its pliable strength, relatively light weight, and ability to be repaired at the competition if an accident occurs. Figure 13: Shark Bite PVC, Aluminum strut, and fuselage connections To ensure the aluminum struts did not detach from the shark bite plastic tubing a dowel rod insert was used to ensure solid fitting. Basically, a short section of a dowel rod was inserted into the tubing and then a Dremel tool was used to notch a groove for the aluminum strut to slide in. Both the dowel rod inserts and aluminum struts were secured with 3M Epoxy to ensure a solid fitting. The front of the skids attached to the front of the fuselage and metal bolts were used to connect it to the fuselage. The plastic tubing was shaved to ensure a flush fitting to the fuselage surface. Kansas State University - Team Spycat Page 16

17 After the required modifications, the landing gear provided a 3.5 inch ground clearance; more than enough for the selected cameras. Figure 14: Skids showing camera ground clearance [left] and Shark Bite PVC with dowel rod insert [right]. Fuselage Safety considerations were paramount in the fuselage modifications. The primary goal was to secure all payloads including the motor battery, auxiliary battery, Piccolo II autopilot, and 3 Microsoft 1080p Webcams. The choice of materials needed to sustain the impact of hard landings and protect the internal components in case of a plane crash. High density foam was used to fabricate the motor and auxiliary battery brackets that were the main supports within the fuselage. They were secured using hot glue to the side of the composite sandwich panel fuselage. In addition, a carbon fiber box was made to house and secure the Piccolo II autopilot system. The main safety consideration when moving around components in the aircraft was the weight of each component and its effect on the static stability of the aircraft. While a full analysis was not necessary, the placement of these components needed to maintain a static margin of 10-15% by placing the center of gravity at the quarter-chord. Figure 15: Fuselage and Empennage Photos Kansas State University - Team Spycat Page 17

18 IV. Risk Management Plan To ensure safety during flight operations, risks to the mission, personnel, and equipment were listed. Once the levels of risk were determined, safety procedures and pre-flight checklists were created to mitigate or eliminate the threats. Risk Identification Potential risks were listed and assessed by level and type. The degree of the risk was determined by the severity of the accident and likelihood that it could occur during the mission. The severity of the risk, organized from most severe to least, was the potential for loss of life, personal injury, mission completion of threshold parameters, and equipment recovery. Accordingly, the risks were assessed in the following matrix. Identified Risks Loss of Communication Air Vehicle Damage Wind Gusts Loss of Motor Battery Power Loss of Payload Battery Power Electrical Failure Laptop malfunction Risk Level Low If less than 30 seconds Insecure parts No crosswind(xwind) Timeout needed Slow image transmitting Electrical short Imagery data loss Risk Level Medium Within 30s and 3 mins Holes,dents, mishandling X wind < 8/11 kts Cannot get all targets No image transmission Component overheating Autopilot malfunction Risk Level High Greater than 3 min Motor or wing failed Dust devils on landing Crash landing No vehicle control Fire RC Control required Table 2: Risk Identification Matrix Risk Type Personnel Risk Type Mission Risk Type Material Crash in crowd Intel not received Air vehicle in crash or specific component Mission Air vehicle time/co. in crash Completion, Emergency timeouts landing Target ID, Air vehicle mi. time/co. in crash Heat/burn Mission Fried parts, injury completion crash N/A Intel not Ground received station Risk Mitigation For each of the risks identified, the mitigation procedures were implemented in the shop, during the pre-flight checklist or within the in-flight recovery or emergency procedures. Shop mitigation procedures included bright colored battery covers, security clips on all wiring connections and the securing all components to the vehicle with locking fasteners. In addition, the aircraft is brightly colored for easy visibility. The following mitigation factors did not completely eliminate the risk, but provided satisfactory protection for the personnel, mission and materials involved in the competition. Loss of Communication The most likely risk that the team did encounter during flight testing was the loss of communication by either the autopilot or Wi-Fi bullets. Loss of the autopilot communication provided the greatest threat to the mission, personnel and equipment and was mitigated according to the level of the risk. If the autopilot communication was lost Kansas State University - Team Spycat Page 18

19 for less than 30 seconds, the flight resumed with the assumption that temporary disruption would pass by quickly without loss of control of the aircraft. At 30 seconds, a pre-programmed flight path is automatically given to the aircraft to avoid crowd locations and circle within a close distance to the home station and resume communication. At 3 minutes of continued loss of communication, the command will be given to terminate the flight into a death spiral far from crowds or SUAS personnel. The risk of losing communication with the Wi-Fi bullets is less severe with personnel and material, but most severe concerning the mission. If communication is lost less than 30 seconds, the first recovery procedure was to restart the Wi-Fi communication for a quick power cycle. Similar to the loss of autopilot communication, after 30 seconds of lost communication, the aircraft will initiate a return-tobase flight pattern to regain communication with the ground station. If after 3 minutes, the aircraft will land and a timeout will have to be used to fix communication problems to perform the mission requirements within the allotted time window. Air Vehicle Damage or Electrical Failure The danger of having parts of the aircraft become damaged or failing is possible to occur before and during the flight. To mitigate this risk before the flight, a pre-flight checklist following the risk mitigation section provides detailed checks for structural integrity, electrical connections, and mechanical connections within the air vehicle. During the flight, the probability is reduced but recovery measures are dependent upon the severity of the risk. If a payload component becomes loose or fails and does not affect the outcome of the mission, adjustments will be made to complete the mission and fix the aircraft afterward. If the damage does prohibit the mission, severity will be determined on-site as to the best course of action. Most likely this will entail landing for repairs and re-launch if time allows. Worst case scenario for irrecoverable damage is to initiate the flight termination sequence at a minimum of a 500 foot radial distance from competition personnel or equipment. Wind Gusts Risk due to wind gusts is most significant during takeoffs and landings. If there are crosswinds and gusts less than the rules specified maximums, rudder corrections are necessary for the autopilot to maintain a controlled landing on the specified flight path. If greater, the aircraft will not be expected to fly. Also, if dust devils or other environmental hazards are spotted nearby, the team may request another landing zone and/or inform the safety officials prior to landing. Loss of Motor Battery Power To monitor the amount of power available and rate of power consumption during the flight, Hall Effect current sensors were installed to provide feedback to the ground station operators. Despite the designs allotting for sufficient time to maneuver and capture target images, if battery power is low the plane will take a return-to-home flight path and land within 5 minutes of estimated available power remaining. For an emergency, the plane will become a controlled glider and land in a safe area in the grass below. Loss of Payload Battery Power Similar mitigation procedures will be employed for the payload battery using the Hall Effect current sensor. However, with control actuators being powered by the payload battery, it is mandatory that the plane be able to land within an extended 10 minute window to avoid a Kansas State University - Team Spycat Page 19

20 complete loss of control of the aircraft. The safety factor for the payload battery is doubled for this reason. Laptop Malfunction The greatest risk with a laptop malfunction is with the threat to mission accomplishment and maximizing the KPP score. If the autopilot laptop malfunctions, control is resumed by communication with a separate box apart from the laptop. If the laptop cannot communicate and changes still are needed in the flight path, the RC pilot will take control to resume the mission. If target identification laptop malfunctions, time will be allowed to address the issue or restart as the aircraft performs a holding pattern near the designated target. The mission should resume based upon the severity of the problem with the laptop. Pre-flight Checklist With proper safety measures taken prior to each flight, the risk to the mission, personnel and equipment is reduced to its lowest levels. The checklist for Team Spycat s competition aircraft identifies the communication requirements for each ground station operator and standard air vehicle checks for a safe flight. Table 3: Pre-Flight Checklist Inspected Item Fuselage components Electrical connections Structural tests Air vehicle/com. check Imagery/Wi-Fi Bullets Autopilot Functions Pre-launch, launch Actions Check foam, bolt, and Velcro attachments to each payload within vehicle Cover and secure any open wire connections with shrink-wrap or el. tape Inspect wing, tail, motor, propeller, & landing gear for structural integrity Perform range check and ensure control surfaces are operating correctly Test the cameras and make sure images are captured, delivered, received Inspect GPS signal, control sensors for accuracy, and Piccolo II response Static motor test, ensure operator readiness, then perform hand launch References 1. Jeppesen Sanderson Inc. (1981). Introduction to Aircraft Flight Test Engineering. Pg Parikh, Kartik K. (2009). CAE Tools for Modeling Inertia and Aerodynamic Properties of an R/C Airplane. Master s Thesis, University of Texas at Arlington. Pgs. 54, bitstream/handle/10106/1775/kiranparikh_ uta_2502m_ pdf?sequence=1. 3. Nelson, Robert C. (1989). Flight Stability and Automatic Control. 2nd ed. McGraw-Hill Companies, Inc. Pgs Printed in the USA. Kansas State University - Team Spycat Page 20