The Doomerang University of California, Los Angeles

The Doomerang University of California, Los Angeles 8 th Annual Student UAS Competition Association for Unmanned Vehicle Systems International

Table of Contents Team Structure... 2 Abstract... 3 Introduction... 4 Background and Overview... 4 Team... 4 Mission Specifications... 5 Restrictions... 5 Time Restrictions... 5 Payload Restriction... 5 Airframe... 6 Hardware Modification... 6 Autopilot and Peripherals... 8 IR Sensors... 9 Autopilot Summary... 10 Telemetry and GSC... 11 Uplink Commands... 14 Imaging... 15 Safety and Failsafe... 16 Power... 16 Flight Conditions... 16 Losing GPS Lock... 16 RF Interference... 16 Vibrations... 16 Pitot Tube... 16 SD Card... 17 Flight Test... 17 Concluding Statements... 18 Acknowledgements... 19 1

Team Structure Project Manager: Aadit Patel Systems Lead: Alexander Lim Structures/Airframe: Philip Lai Autopilot Leads: Angelica Camacho & Will Schoellkopf GCS and Telemetry: Jeff Nakahara & Sennan Sulaiman Imaging Lead: Eriberto Morales Missions Lead: Vincent Hu 2

Abstract The 2009-2010 UCLA AUAV Team has designed and implemented an autonomous unmanned aerial vehicle (AUAV) for the 2010 Unmanned Air Systems competition. The airframe of the airplane is an RC airplane, the TwinStar II, which was modified to allow for a larger payload capacity. The TwinStar II was then fitted with the Attopilot autopilot, which included an external GPS unit and on-board PID controller in order to maneuver the airplane fully autonomously. A pair of XBee Pro Series 2 modems operating at 2.4 GHz allows for real-time telemetry downlink from the AUAV to the ground station. Imaging of ground targets is taken through a commercial 8 megapixel digital camera, which is mounted on the underside of the left wing of the airplane. Real-time video downlink was not implemented at the time of writing, but may be incorporated by the competition date. A total of three test sessions have been completed (each session consisting of 5 flight tests). These sessions have allowed the team to evaluate the performance of autonomous flight and the quality of video recording from the camera. Overall, the UCLA AUAV is wellequipped and tested for the 2010 UAS competition. 3

Introduction Background and Overview The UCLA AUAV team was formed in 2007 to harbor interests in creating autonomous air systems at UCLA. The UCLA UAV team is more focused in systems implementation, between airframe, autopilot, GPS, ground station, data links, and imaging. Thus, most of the development cycle is focused on design research while the production cycle is mostly focused on sub-system integration and flight testing. Team The 2009-2010 academic year marks the UCLA AUAV Team s 3 rd year in development. The team consists of nine undergraduate students, all of whom are Aerospace Engineers. However, many students have technical expertise outside of aerospace engineering, such as in computer science and imaging. Their collective knowledge and dedication to the project has allowed the UCLA AUAV team to produce UCLA s first fully autonomous UAV. UCLA UAV Team at Apollo XI Test Field 4

Mission Specifications The mission for the AUVSI student competition is to perform ISR (Intelligence, Surveillance, and Reconnaissance) given GPS waypoint coordinates. Two missions are specified to be performed. The first mission is to fly on a route designated by the GPS waypoints. Two targets will be identified during this mission. The targets are colored basic geometry shaped plywood boards with a colored alphanumeric symbol on them. One target will be directly under the specified flight route. The other target will be placed up to 250 feet off the specified flight path. Payload Restriction The maximum takeoff weight must be less than 55 pounds. The second mission is an area search. Targets will be placed in the entire flight area, and the unmanned vehicle must search the area and find as many of the targets as possible in an allocated amount of time. Restrictions For the first mission, there are altitude restrictions in place. The first target must be identified while the unmanned aircraft is at an altitude of 500 feet +/- 50 feet. The second target must be identified at an altitude of 200 feet +/- 50 feet. The aircraft also must stay within +/- 100 feet from the designated flight path. For the area search, the aircraft is allowed to fly between 100 and 750 feet altitude but must remain within the mission area. There is a 200 feet margin between the mission area and the no-fly zone. Time Restrictions 40 minutes are allotted for preparation before the mission begins. The total mission time cannot exceed 60 minutes. 5

Airframe Since our main objective is to build an autonomous UAV, the team decided to purchase a commercially available RC plane kit. After some research, the Multiplex Twin Star II was selected for its excellent flight characteristic, robust structure, payload capacity and room for modification. Table 1: Specification of Multiplex Twin Star II US SI Wing Length 56 in. 142.24 cm Wing Area 667 sq. in. 4 303 sq cm Wing Loading 9.08 oz./ft2 43.94 kg /m2 Fuselage Length Empty Weight: Center of gravity (from the leading edge) 40.5 in. 102.87 cm 32 oz. 0.9 kg 3.35 in. 85mm Table 2: Weight of Major Components Components Airframe (Empty) Motor LiPo battery Weight 32 oz. (900g) 2.3 oz. (65g) each 11 oz. (311g) AutoPilot & GPS 0.85 oz. (24 g) module Digital Camera 5.3 oz (150 g ) Gimbal & Mount Approx. 30 oz (850 g) MAX Landing skids or gears Approx. 5 oz (140 g) MAX TOTAL WEIGHT Approx. 5 lb (2.3 kg) MAX Hardware Modification To perform the missions prescribed in the competition, our aircraft will be required to carry extra payload including a camera system (power, camera, and autopilot), autopilot module, and a power source. In view of the additional payload, we have decided that an upgrade to the original power-plant was needed. Trade studies showed that the added weight due to the increased power requirement has minimal setback on the overall payload increased. Table 3: Hardware Modification Chart COMPONENTORIGINAL MODIFICATION Motor Electronic Speed Controller (ESC) Battery Servos 2 X Speed 400 2 X Himax HC2812 Brushless OutRunner Pico Control 600 EMK Apogee 2S2P 2170-mAh Hitec RCD HS-81 Thunderbird Brushless ESC FlightPower Lithium Polymer 4350 mah Unchanged 6

Our airframe is designed to be hand launched, since the competition allows landings to be carried out in the grass strip adjacent to the runway. The main skid is designed as protective device for the fuselage and provides a smooth sliding upon landing. In the event of a rough landing at high speed, the skid is designed to detach in order to absorb the energy on impact to protect the airframe s main structure. While the wing tip skids are installed to provide extra clearance for the propellers, in addition, the engines will be shut off upon landing to avoid damage. Figure 2: Twin Star II mounting thermopile sensors Figure 1: Twin Star II airframe is wide enough to safely harbor the autopilot 7

Autopilot and Peripherals One of the primary mission objectives for the competition is autonomy, so an autopilot is needed to control the airframe for the mission. The requirements for an autopilot are: Robust autonomous control Autonomous waypoint navigation Autonomous takeoff and landing capabilities Ease of integration with airframe Cost effectiveness Table 4: Autopilot Weight Autopilot Size Without GPS Attopilot 1.5 x 2 x 0.5 ½ ounce control Attopilot records in-flight data either by sending telemetry back to the ground control station or by recording flight parameters, such as airspeed, GPS location, and altitude on an onboard SD card. The SD card can then be removed after flight and flight data can be analyzed using flight playback software. With GPS and all sensors 2 ounces Figure 3: The Attopilot Board The UCLA team decided to use Attopilot, which is a powerful, widely available autopilot system. It is a small, lightweight, fast, fully featured autopilot capable of meeting all the required objectives and parameters of this competition. It provides capabilities such as autonomous takeoff and landing, in flight programmable way points, and sensory control. Attopilot is a small autopilot system capable of flying the Twin Star II. Because of the adaptive algorithms, 50Hz attitude control, and automatic gain scheduling based on airspeed, AttoPilot is a versatile yet inexpensive autopilot solution. Figure 4: The Attopilot board with servos The UAV team utilizes a custom ground station software made for the Attopilot board. Furthermore, a KML-to-TXT converter compensates for the fact that Attopilot does not have a native way to input waypoints. Instead, Google Earth is utilized to add point-and-click waypoints, exporting them as a KML file that the supplied utility can convert into a form that AttoPilot can read. Google Earth waypoints are either plotted in either 3-D or 2-D. 8

2010 UAS Competition UCLA Figure 5: 3-D waypoint overlay Google Earth on All of the data processed by Attopilot can be sent back to the ground station for close observation of the telemetry data. The telemetry data can be read as raw readings or through graphical representations, to ensure that all of the flight instrumentation, power systems are within flight specifications. In the absence of a set-up utility, the main way to adjust Attopilot for an airframe is by editing a text file that is stored on the on-board SD card. Figure 6: 2-D waypoint overlay on Google Earth The ground station relays provides telemetry information about the real-time flight parameters via downlink through XBee Pro Series 2 Modems operating at 2.4 GHz. Figure 7: Attopilot Ground Control Station Figure 8: Screen shot of SET File that is stored on the on-board SD card. IR Sensors The IR sensors estimate the orientation of our UAV relative to the warm earth and cold sky. The system is immune to vibration and disorienting launches, wind gusts, or stalls that would otherwise confuse inertial-based autopilots. The concept of operation with regards to these sensors is that at a zero pitch or bank angle, the difference in heat between the two sensors should be zero. As a result of this difference, the relationships between angles are calculated during flight. Each of the 3 pairs of sensors measures one axis. If there is poor weather or terrain, 9

the x/y sensors may not be as accurate, so the third sensor provides a ground calibration to provide for accurate angle calculations as the aircraft travels over terrain with different IR radiation. Autopilot Summary Low weight: 36 grams (1.25 ounces) is the total added to our airplane including GPS and antenna and sensors. Dimensions: 25x20x10mm. Integrated data logging to micro SD card Capacity for 6 servos, including trigger, pan, and tilt servos Table 5: Maximum Limits of Attopilot Figure 9: IR Sensor calculates IR differentials Courtesy:http://paparazzi.enac.fr/wiki/Image:IR_ex ample2.jpg Max Distance from 300 km Home Position Turn Rate 60 deg/s Pitch and Roll Rate Airspeed Speed Maximum Limits 300 deg/s 125 mph via pitot tube Above 125 mph GPS is used 10

Telemetry and GSC The Attopilot autopilot relies solely on Google Earth to obtain, maintain, and modify waypoint data. Google Earth waypoint data is written in the web servo pulse rates. Analysis of these recorded data allowed these values to be adjusted to fit our requirements. A flight performance and navigation log file is saved onto the Attopilot s microchip for every flight performed. Used in conjunction with a prepared waypoint navigation testing programming sequence, this enabled more complicated manoeuvrings and navigation controls to be calibrated, most notably climb rates, aileron banking rates, pitch and roll angles and rates. The autonomous waypoint navigation code contains altitude, latitude, and longitude information obtained from Figure 10: Waypoints created directly on Google Earth in KML format. (Apollo 11 Model Aircraft Field) programming key mark-up language (KML). Attopilot software contains a waypoint editor that converts KML into text files, producing a list comprised of string commands, which is then saved onto the onboard Attopilot security detail microchip for autonomous flight. Prior to flight, autonomous information from the Attopilot microchip is read from a C# -to-kml-converted flight plan file and a tuning setup file - a text file containing raw values for flight performance and airframe configurations such as home waypoint altitude, waypoint activation distance, aileron and elevator control, and input data for auxiliary integrated systems. The Attopilot tuning setup file is vital to obtaining optimal autonomous flight performance. Our UAV was test flown on numerous occasions for the sole purpose of capitalizing flight performance and autopilot behavior. The Attopilot software comes with a test environment condition where, when enabled, records values pertaining to thrust, servo gains, and Figure 11: Projected image of UAV trajectory following waypoint sequence after a completed test flight Google Earth converted data. These waypoints in the code can be programmed to specify altitude changes and certain trajectory commands upon approaching the activation distance for a waypoint. This activation distance is a measurement in meters defined in the setup code file for the minimum distance the UAV must be to initiate the behavior programmed in the waypoint navigation code. Achieving an accurate trajectory to ascertain waypoint activation was a significant stipulation to meet while calibrating navigation controls. This allows subsequent waypoints requiring the UAV to turn at sharper angles to be done in one 11

pass, without having to make adjustments over a longer period with multiple passes. A limitation to the Attopilot software in regards to waypoint navigation is that the autopilot will follow the waypoint sequence as coded. One way to alleviate this limitation is to utilize real-time uplink commands via radiomodems to instruct the autopilot to perform differently than its programmed waypoint navigation. Specified navigation behavior upon activation of a waypoint is determined by the waypoint code trigger system, coupled with navigation values in the setup code. The flight plan for Attopilot (WP.txt on the micro SD card) has a special data field of triggers. The triggers allow scripting of events during the flight plan besides just simply navigating from point to point. Besides controlling events, the triggers adjust behaviors and responses to certain conditions. In these ways, the triggers make WP.txt flight plans very powerful by allowing modification of behavior while minimizing changes to the set code. UCLA UAV Team Test Flight 1: WP Latitude Longitude Alt Triggers $00000, 34.17530, -118.48250, 0250, 00000000*2E Figure 12: Waypoint data converted from KML-totext format $18, Servo 1 Upper Limit = 1695 Aileron $19, Servo 2 Upper Limit = 1774 Elevator $20, Servo 3 Upper Limit = 1819 Throttle $21, Servo 4 Upper Limit = 1972 Rudder $25, Servo 1 Lower Limit = 1322 Aileron $26, Servo 2 Lower Limit = 1210 Elevator $27, Servo 3 Lower Limit = 1062 Throttle $28, Servo 4 Lower Limit = 1058 Rudder Figure 13: Code for final adjusted values of servo pulse rates, in microseconds. $32, Path Angle Max = 30 (degrees) $33, Path Merge Line Distance = 30 (in metres) $34, Roll Angle Limit = 35 Max allowed roll (in degrees) $35, Pitch Angle Limit = 20 Max allowed pitch (in degrees) $36, Steer Proportional Band Width = 110 width to spread proportional control of roll response (in degrees) $37, Alt Proportional Band Width = 80 altitude window to spread proportional pitch response (in metres) $38, Steer_D = 3 Gain factor dampening term of heading control $39, Alt_D = 5 Gain factor dampening term of altitude control $40, MaxRate = 10000 Max allowed pitch and roll range o f change in degrees/second Figure 14: Code for final adjusted values for maneuvering controls 12

Telemetry data broadcasted by Attopilot can be read from the ground control interface, such as altitude, heading, velocity, and climb rate. In-flight string commands are enabled with the integration of a two-way telemetry network, allowing waypoint and flight path data and plane trajectory to be modified during flight. Furthermore, entirely new waypoints can be created and added, allowing rapid change in the flight plan. To achieve wireless data link, Zigbee XBee Pro 2.4 GHz RF modules are utilized and integrated with the Attopilot v1.8 control power control board (PCB). These commercial modems were chosen for their adequate line-of-sight range of one mile, compatibility with the Attopilot hardware, low power requirements, and reasonable cost within our budget. With a data uplink rate of 250 Kbps and a dual modem set-up network one installed directly onto the plane, powered by the Attopilot PCB, and one incorporated in the ground station via a universal serial bus two-way command strings and realtime flight telemetry data are able to be packetized. As per the Attopilot specifications, the modulation rate of the XBee telemetry network is programmed to 38400 bauds. Figure 15: Plotted telemetry data obtained from Attopilot, including airspeed, altitude, head error, pitch, and roll 13

Table 6: Command Uplink Table R L G Z W M Command Description Resume navigation from loiter Loiter at current location Fly to specified coordinates and loiter Come home and loiter Waypoint number reset pointer Upload new waypoint GPGGA GPS GGA NMEA sentence for convoy follow P D H A T B C Proportional gain adjust Differential gain adjust Heading proportional band adjust Altitude proportional band adjust Trigger fire Lockout R/C Tx and put mode to Auto Unlock Tx lockout Uplink Commands When the L loiter command is issued, the current location is used as centre of the loiter circle. The radius, direction, and altitude default to those specified in the set code used for return to launch home loiter. The airspeed used is that of current target. The observed flight behavior upon issuance of L is that UAV transitions from straight flight to a banking turn and spline merge to the circular path. The G loiter command has more flexibility. Latitude and longitude are specified as well as altitude, radius, and direction of loiter. The specified coordinates can be far away from the current location. The G loiter can be thought of as a go to and loiter. Upon resume, the UAV will proceed on a straight line from center of G loiter to the current waypoint target. Upon issuance of G if the UAV is far from the new target location it will proceed on a straight line path for center of the loiter circle. Once within waypoint activation (per set code) distance of the circle edge it will begin to spline merge onto the circle. At any time the mission can be paused and UAV forced home by using the Z come-home command. If any waypoint in the list has the reverse path control flag, then Attopilot knows that safe return home is by backtracking along the waypoint path rather than line of sight return to launch. Once home the UAV enters loiter. W allows the current waypoint number to be reset elsewhere in the list. For example if the UAV is currently at waypoint 6 you may skip ahead to waypoint 9 or go back to an earlier waypoint such as 3. Issuing this command causes direct line-of-sight path between current location and the commanded new waypoint number. 14

M allows new waypoints to be loaded to AttoPilot. Old waypoints may be over written or new waypoints added to the end of the list. Convoy follow command occurs when the GPGGA NMEA sentence is sent to AttoPilot via uplink from the ground station. Any time AttoPilot receives this sentence the current mission is paused and the coordinates of the GPGGA are used as a moving loiter target. Once the UAV is within waypoint activation distance of the coordinates + loiter radius, it enters loiter (UAV has caught up with convoy). As long as the UAV is more than waypoint activation distance from loiter circle edge, it purses the coordinates specified in GPGGA as a straight line path. Altitude is set as the current barometric altitude target above ground plus changes that happen to GPS altitude. For example, if the convoy drives over a hill, the altitude target will shift upwards as the ground convoy ascends, and shift downwards as the convoy descends on the far side of hill. Imaging Images of the search area will be taken by a commercial digital camera attached to the underside of the left wing. A counterweight of the same weight is placed on the opposite wing in order to balance the airplane. Before launching the airplane, the digital camera will be set to record at a specified zoom level. The video s timestamp will be synced with the time stamp of the ground station. When the airplane has safely landed, the SD card from the digital camera will be removed and analyzed on a dedicated computer. When a target is spotted, the time stamp will be matched with the location, utilizing information from the recorded flight on the GCS. The team will then be able to give target specifications and location after the flight is complete. The R resume command must be issued to resume flight plan. Typically, one would take manual control of the UAV via R/C Tx and land the UAV after convoy following is complete, as the UAV is most likely in a location that is not part of or near the original flight plan. The P and D gains for 50Hz attitude and 5Hz throttle control are changeable in flight mode. Gains for navigation are also settable during flight. These include proportional band widths for heading and altitude control. The option to force a manual trigger (i.e. a camera control trigger) at any time and in any flight mode is by issuing the T command. 15

Safety and Failsafe Power First and foremost, one of the most important requirements for an unmanned aerial vehicle (UAV) is that there is a failsafe feature. The failsafe feature should cover a ground station decision to override the autopilot and a dangerous low power from the battery on the UAV. The Attopilot autopilot hardware that is used offers the ability to set a limit of mah of the power coming in to the autopilot and also a forced override to manual remote control flight. Once the mah goes below a certain limit the UAV will go into a ready to land (RTL) feature and land safely. If for any reason the crew on the ground station determines that the UAV is behaving as intended, then a remote control override can be forced and the UAV can land manually landed or controlled. If for some reason the failsafe for low mah or low voltage does not activate the RTL and the ESC is cut off, Attopilot will keep the plane safe with nose-high altitude and level wings for a soft crash landing that will most likely result in minimal or no damage to the plane. Flight Conditions There are certain environment conditions in which the UAV should not be flown in that the fail safe will not be able to compensate for the situation. Flight in light rain or drizzle is prohibited because the horizon sensor will be blind due to the rain and the UAV will crash immediately unless manual override is implemented. In addition, the UAV with Attopilot should never be flown in clouds, fogs, dust storms, or areas where dust is prominent. Every pre-flight the windows of the horizon sensors should be double-checked to make sure that it is clean or a crash will result. Other environmental situation that is crucial to avoid using the UAV is when wind speeds are strong, because the navigation will be confused if the UAV is flying backwards with respect to the ground due to strong winds. Essentially, do not fly if the wind speed is 50% of the cruise speed of the UAV. Losing GPS Lock In cases where GPS lock is lost, Attopilot will lock the current altitude and do a slow banking left turn. This will stop the plane from veering off course too quickly and the slow turn will keep it in the general location. RF Interference If there is radio frequency interference when there is analog to digital converter present, the UAV will behave normally and respond accordingly on pre-flight ground tests. However, once in the air the UAV will do a violent nose dive in which case the only failsafe method to recover the UAV is to the take over manually and land the UAV. To avoid such problems Attopilot is installed in a metal box or some type of radio frequency shield. Vibrations In fail safe respect, there should not be worry about vibrations affecting Attopilot because Attopilot does not use an inertialmeasurement-unit (IMU), so Attopilot does not need dampers or other methods to protect it from vibrations. Pitot Tube If the pitot tube was blocked, the only failsafe mechanism would be to take manual control. This stuck Pitot tube will cause the PID control to loop and cause max throttle. The UAV will speed up and possibly oscillate due to the control gains being meant for lower speeds. The only 16

way to remedy this situation would be to take over manually. SD Card There is a fail safe for in flight loss of SD card. Attopilot reads the set file and the waypoint file and loads them into the onboard ram, so even if there is a in-flight loss of the SD card the UAV will still function normally. Flight Test Our test flights are carried out with the help of an experienced RC pilot in Apollo 11 field located in Van Nuys. Test flight #1: Manual Controlled Flight We performed this manually-controlled test flight to validate major airframe hardware modifications including the brushless motors and the undercarriage skid we attached on the bottom of the fuselage and at the wing tips. The objectives of the test is, first, to ensure that the added skids will not compromise the aerodynamic stability and maneuverability of the aircraft and secondly, to evaluate the effectiveness of the skids are in protecting the aircraft and it s propellers on landings as designed. In addition, this test also served as a proof of flight; therefore, a video of the flight is recorded. The flight began with a hand-launch with no power and was allowed to guild for some distance to check for major problems since we can safely land the aircraft at this point. Afterwards, power was gradually added and pilot checked for asymmetric power effect on overall control. After confirming all control surfaces are operating correctly, our pilot flew our plane for 2 laps to access the flight characteristic before coming in for a soft landing on the grass. Our pilot concluded that flight characteristic remained unchanged despite the addition of the skids. And our new motors are operating well, without major asymmetric power effect. Upon landing, our skids provided good sliding and held together (although they are designed to detach in case of rough landing) and the wing tip skids effectively protected the propeller from striking the ground. The test flight was considered to be successful. It accomplished our test and validation objectives. We planned to embed other equipments such as the camera on the undercarriage skids. Test flight #2: Manual Controlled & Autonomous Flight: The objective of this test flight is to assess the performance of our autopilot system as well as the fail-safe mechanism. Before the test flight, the aircraft was taken to a designated location at the airfield to set the home position according to the autopilot pre-flight procedure. The aircraft was then hand launched and taken to a safe altitude of about 100m AGL under manual control of our RC pilot. After checking all flight characteristics of the aircraft, controls were handed over to the autopilot. The flight plan was set for the aircraft to loiter around the home position. The flight profile of the aircraft was assessed to identify any abnormal control input by the autopilot. At the end of the test flight, our RC pilot took manual control and landed the aircraft Results of the test flight provide data for correction to the flight control settings of 17

the autopilot including maximum flight surfaces deflections and control gain to prevent under- or over- control. More flight tests are scheduled before competition in order to completely fine tune the autopilot. Concluding Statements The final steps before competition will include performing more evaluative flight tests at the Apollo XI field in Van Nuys, CA to ensure in-flight stability. The team will also further test uplink commands and find the maximum range for data link communication in order to quantify limits of the fail-safe technology. In addition, the team hopes to establish a real-time video feed from the AUAV to the ground station so ground targets can be analyzed in realtime. Currently, the UCLA AUAV is able to navigate autonomously, respond to uplink commands, as well as take in-flight video of ground targets. Thus, the UCLA AUAV is well-equipped for competing at the 2010 UAS Competition in June. 18

Acknowledgements The UCLA AUAV Team would like to thank their faculty advisor Damian Toohey for his assistance in this project. In addition the team would like to thank their RC pilot Tony di Leo for helping with flight testing at the Apollo XI Field in Van Nuys, CA. 19