Lightweight Fixed Wing UAV

Transcription

1 Lightweight Fixed Wing UAV Cindy Xiao, Rijesh Augustine, Andrew Jowsey, Sebastian Werner, Brian Hinrichsen, Duncan G. Elliott University of Alberta Abstract The University of Alberta Aerial Robotics (UAARG) is a student vehicle project in the Faculty of Engineering. Funded from both Shell Canada and the University of Alberta, UAARG designs and builds custom unmanned aerial systems (UAS). UAARG s past redesign of a lightweight, versatile, fixed-wing unmanned aerial vehicle (UAV) has proven to unique and a viable system. This autonomous UAV is capable of real-time image capture and transmission via integrated onboard electronics. UAARG has has made improvements to this year s system via extensive software modifications to the interoperability and imaging subsystem. Advances in image analysis have lead to identification and reduction in georeferencing errors. This report summarizes the design, build, and testing that has been devoted for a successful year at AUVSI Seafarer Chapter s 14th Annual Student UAS (SUAS) Competition. Table of Contents Introduction System Overview Airframe Selection Airframe Characteristics Airframe Customizations Avionics Communications Autopilot Imaging System Performance Autopilot Testing Imaging Test Results Achievable Competition Objectives Safety Considerations and Approach Safety Design Features Firmware Failsafes Conclusion Acknowledgements References University of Alberta - UAARG 1

2 Introduction In the past year, UAARG has made critical refinements to our system to better excel at the AUVSI Seafarer Chapter s Student UAS (SUAS) Competition. These refinements have streamlined and stabilized UAARG s proven process based on a lightweight, cost-effective, and portable UAV, featuring a wingspan of 1.8m and weighing 2.9kg. The electronics of the aircraft reside on an open board in a compact hull, allowing for easy troubleshooting and debugging. Several reliability and accuracy improvements, such as an external magnetometer, have also been made to the avionics system. More upgrades have been made in the imaging software to more efficiently process images, such as adding CSV and KML export features to expedite result generation and analysis of past flights. With these efficient improvements UAARG can expect to complete the primary autonomous flight and search area tasks to be completed faster, allowing for more time at the flight line to be used for the secondary tasks, including the Emergent Target, Off-Axis Target, Actionable Intelligence, Interoperability, and Sense and Detect tasks. Figure 1: System Block Diagram University of Alberta - UAARG 2

3 System Overview Airframe Selection Figure 2: HobbyKing EPP FPV Airframe UAARG will be operating a UAV with a modified flying-wing airframe called the EPP FPV from the online hobbyist retailer Hobbyking. A flying-wing vehicle has been traditionally used instead of multi-rotors because of the ranges and search areas required in reconnaissance miss ions. Furthermore, onboard electronics can be installed in many different locations to prevent issues such as interference and overheating. Modifying an existing airframe has the advantage of reducing build times and system repeatability. In 2011, UAAR G attempted to design and fabricate a custom UAV airframe but cost, flying weight, and lead time became a burden in early development phases. As a result, UAARG has continued the objective to proving the feasibility of UAV operation via low cost off-the-shelf and lightweight components. For the past 2 years, UAARG s primary UAV airframe has been the EPP FPV from HobbyKing. The Senior Telemaster Plus was used fo r over 2 years before then. The Skywalker X8 is a flying wing airframe that has been considered as the EPP FPV s possible successor but presents other design challenges. As seen in Table 1, a comparison of important factors between UAV airframes is presented. All three airframes listed above have be flown multiple times with their respective UAV payload design. The decision to switch and continue operating with the EPP FPV was based solely on testing, analysis, and maintainability. The Senior Telemaster Plus, which performed in previous AUVSI competitions, was sufficient in meeting the requirements of UAV operations but has a high cost of $380 per frame. This alone is a great challenge during autopilot testing as any crash is always unrepairable with its wooden composition. The EPP FPV was chosen as the successor because it is inexpensive and made of a durable expanded polypropylene (EPP) foam. EPP foam is able to absorb high impacts with little damage or deformation. Thus, maintenance costs and downtime are low after failed flight tests. Furthermore, EPP allows the overall frame to remain lightweight and easy to retrofit for new avionic components. One challenges with the EPP FPV, seen in Table 1, is that the payload space is relatively compact and the plane s payload weight in nearly maxed in its current configuration. The Skywalker X8 was evaluated as an alternative but its cost of $163 per frame, launching system, and challenges with center of gravity and flight characteristics without a tail boom require fundamental shift in UAARG s well established practices. As a result, the EPP FPV is continuing as UAARG s UAV Airframe. University of Alberta - UAARG 3

4 Table 1: UAV Airframe Choice Comparison Table Airframe: Senior Telemaster Plus EPP FPV (Current) Skywalker X8 Relative Cargo Space Medium Small Large Type Fixed Wing Fixed Wing Fixed Wing / Flying Wing Launch System Wheels and Runway Hand Launch Catapult Launch Landing System Wheels and Runway Belly Land Belly Land Composition Wood Foam, Carbon Fiber Foam, Carbon Fiber Max Cruise Speed 100km/h 70km/h 100km/h Max Range 1.5km 1.5km 1.5km Max Endurance 60min 40min 60min Cost US Crash Durability Fatal Salvageable Salvageable Relative Modifiability Low High Medium Weight (payload included) 11kg 3kg 5kg Airframe Characteristics An electronic motor configuration was chosen for the purpose of autopilot integration and reduced airframe retrofitting. Dimensions and components of the UAV can be found in Table 2. Operations will be carried out during daylight hours using Visual Flight Rules (VFR). The weather criteria to determine whether the flight-testing may proceed are listed below. Table 2: Summary of UAV Airframe Characteristics Propulsion System Dimensions Flight Specifications Motor RPM 750kv Weight 2.89 kg Max Wind Velocity Gusts up to 20km/h Flight Time 40 Minutes Wing Span 180 cm Precipitation None Batteries 2 x 5000 mah 4 Cell LiPo Length 131 cm Minimum Visibility 3 miles ESC 45A Turnigy Height 245 cm Ceiling for flight below 500ft AGL 1000ft AGL Max Cruising Speed 70 km/h Propellor 10x7 cm Ceiling for flight below 800ft AGL 1500ft AGL Airframe Customizations Multiple modifications were made to improve the structural integrity of the EPP FPV s stock airframe and to allow the incorporation of various electronic components. Improvements to the airframe includes high failure point reinforcement, fast assembly design, and flight performance repeatability. Cut-outs were made in the wings, fuselage and tail to install sensory and communication components that are essential for UAV operations. An extensive University of Alberta - UAARG 4

5 design review was conducted amongst club members and faculty advisors to ensure that structural integrity was not lost in critical structures. Improvements such as a new motor mount, tail boom reinforcement, and wing covers were added to eliminate failure points in the stock airframe. On hard landings, the original basswood motor mount would fracture and detach due to thin wood and low quality adhesives. Furthermore, the long and weight-balancing tail boom snapped due to its inherent high stress zone near the base of the fuselage. As seen in Figure 3, the motor mount was fortified by using a thicker basswood and adding thin wooden extrusions that puncture into the fuselage. A balsa wood rod is now placed within the tail boom and a thicker carbon-fiber tube envelopes the fracture point. These failure points were quickly determined and resolved during early payload flight tests. Figure 3: Failure-Point Corrections Figure 4: Airframe Improvements Flight performance was optimized in subsequent flight tests. Centre of gravity was no longer ideal after installing the autopilot and imaging electronics within the on-board cargo space. In response, center of gravity was analyzed using Table 3, and corrected by attaching a servo mount along the plane s tail-boom (Figure 4, left). The mount is constructed out of a carbon fibre sheet and is secured with epoxy. This also allows for more weight to be placed in the cargo space and less stress on servo s control arms. Stress on all servos were reduced by bending control rods to a specific angle. The vertical stabilizer was also improved with an extension that increases surface behind the propeller (Figure 3, left). Table 3: UAARG 2015 Weights and Balances University of Alberta - UAARG 5

6 Modifications to the airframe were also made to simplify operations and decrease deployment time. A detachable tail boom was constructed to improve the transportation of the aircraft without increasing setup time substantially (Figure 4, centre). To prevent changes in the airframe structure between flights, wing struts (Figure 4, centre) and reinforcements for the servo control rods were installed; this greatly reduced the need for re-trimming the aircraft during test flights. The wings also contain several major sensory components. The airspeed pitot tube is placed so that it meets the air 5cm from the leading surface of the left wing. This ensures that the static port is not at a high pressure system; the port is connected to the actual sensor via rubber tubes. A Canon Powershot A2400IS camera is embedded within the right wing a secured using thin sheets of basswood and t-nuts. Once again, its mount on only extrudes 1-2mm from the wing, thus maintaining its aerodynamics. Avionics Figure 5. Layout of all aircraft avionics systems The avionics of the UAV are designed in a highly modular fashion. The majority of the avionic hardware is laid out on a cut out perforated board. Components are primarily secured to this avionics board using M3 screws, spacers and bolts. This allows components to be swapped on and off with ease without disrupting other hardware. The laid out design also allows for quick and easy troubleshooting, as all components are visible and easily accessible. The avionics board is designed to slide easily into the front of the plane, as seen at the top of Figure 6. There are two primary contact points at the front and back of the board that are used to ensure alignment of the avionics board with the fuselage; the board is shaped to the fuselage at these points. Pressure from the batteries and fuselage wall provide lateral stability to the avionics board. Components on the avionics board include power switches, autopilot computer, onboard imaging computer, telemetry radio, logic level converter and switching power regulators. A printed circuit board (PCB) was designed University of Alberta - UAARG 6

7 for additional modularity. The PCB is used for power distribution for avionics hardware and telemetry data communication. The PCB houses two Murata 5V 2A switching regulators, the Digi Xbee Pro telemetry radio, and a logic level converter. The logic level converter pulls the telemetry data from the autopilot computer from a 3.3V logic level to a 1.8V logic level that the imaging computer is rated for. The avionics board also houses 4 switches used to fully power on the plane. Two of the switches are used to power on/off the autopilot board and the onboard imaging computer. The remaining two are used to regulate power flow to the servos and the motor. All connections are at the back of the avionics board. This includes USB connections, autopilot module connections, servo connections, and power connections. Figure 6: Avionics board layout Peripheral hardware that connects to the avionics board include two DSM2 Orange RX satellite receivers, an airspeed sensor, a current sensor, a GPS sensor, 2 Wi-Fi modules, a camera, and an ESC. The autopilot computer is the Lisa/MX V2.1 (Lisa), an STM32 microcontroller based board that is compatible with the Paparazzi autopilot system. The Lisa is capable of outputting up to 8 PWM signals to control up to 8 servos. Radio Control (R/C) data is received via satellite receivers that directly interface with the Lisa; up to two satellite receivers may be used. This allows for improved R/C reception at longer ranges and varying attitudes. The Lisa has multiple built in analog-to-digital converters (ADC). This allows for analysis and processing of data from various sensors and incorporating the information into autonomous navigation and geo-referencing. Last years model used a current sensor to compensate for changing magnetic fields within the fuselage. Using the Lisa I2C line an external magnetometer was placed into the wing far from aircraft generated magnetic fields. Communications Aircraft control is split between the Manual Aircraft Control link (referred to hereafter as R/C link) and the Autopilot Telemetry/Control link. The safety pilot handles a 2.4GHz Spektrum Transmitter. This links to two onboard satellite receivers, which communicates to the Lisa. The information transmitted is capable of directly controlling the aircraft's flight by manipulating its control surfaces. It is also capable of engaging and disengaging the autopilot. The autopilot telemetry/control link uses the 900MHz spectrum. This spectrum is part of an ISM band, with many available options for easy to use serial-enabled radio modules with reliable range and high transfer rates. For this system, two identical Digi Xbee Pro's work together for bidirectional communications between the plane and GCS. This link is used to send real-time telemetry from the plane to the GCS. This provides constant status and location of the UAV. The GCS can also send data and commands to the aircraft via this link. This includes adjusting different autopilot gains in flight, and modifying the aircraft's autonomous flight plan. University of Alberta - UAARG 7

8 The Imaging Data/Control link provides command transmission capabilities to the Odroid and acts as a data downlink for image and metadata transmission. Due to the long-range requirements of the link, the desire for link reliability, and the small data size requirements of the images being transmitted, this link is currently using the a Wi-fi standard. Table 4: Characteristics of UAS Communications Links Link Name Frequency Protocol Directionality Autopilot Telemetry/Control Manual Aircraft Control (R/C) 900 MHz Digi proprietary Bidirectional between aircraft and GCS 2.4 GHz Spektrum proprietary Unidirectional, aircraft is receive-only Imaging Data/Control 2.4 GHz a Bidirectional between aircraft and GCs GPS 1575 MHz (GPS L1) GPS L1 C/A Unidirectional, aircraft is receive-only Autopilot All tasks put forward by the competition require or heavily benefit from an autopilot and telemetry subsystem. The system needs to be able to achieve controlled flight and have a continuous telemetry downlink. The autopilot system should be able to solve these needs and allow for easy integration of custom software and hardware for task specific goals. Building an autopilot system from scratch is not the easiest route, especially when their are a few mature open source systems with all the required criteria. Paparazzi is an open source autopilot system that has been chosen for the UAV. Paparazzi has a relatively small but dedicated community behind it. The code is laid out cleanly to allow for easy integration of custom hardware and software. The Paparazzi GCS is straightforward, and mapping of obstacles for the sense detect and avoid task was quickly implement. The Paparazzi system communicates within its subsystems via the Ivy bus, a text based software communications bus that allows for external programs, written in Python, C or Ocaml, to easily interact with the different subsystems. Control of the aircraft and autonomous navigation is achieved using the Lisa MX autopilot board. This board is based off of an STM32F4 microcontroller and has a 10 degree of freedom inertial measurement unit (IMU). This IMU coupled with a GPS and a quaternion based attitude heading reference system in software makes up the inertial navigation system (INS). The INS information includes the attitude, altitude, location, speed and orientation of the aircraft. This same information is used for georeferencing. The INS is used by the autopilot for stabilized flight and autonomous navigation of the UAV. Each mission has different flight routines that are sequenced together. The autopilot mission may be programmed to consecutively run through a sequence of flight routines or wait for user inputs. These mission are preprogrammed and loaded onto the the onboard microcontroller. Failsafe exception conditions are hard coded, so a telemetry link is not required for the activation of a failsafe. The missions can be changed in realtime using the telemetry link. This allows completion of the emergent target task. The two way telemetry and command communications system updates the ground control station with the full status of the UAV, and the autopilot board with information about changes to the flight routine. University of Alberta - UAARG 8

9 Figure 7: Block diagram of autopilot system components Imaging The imaging subsystem is responsible for capturing images and image data from the UAV, relaying it to the ground, and processing that data to obtain target information. There are two major components to the system: - The onboard imaging computer. Responsible for image capture, data tagging, and relaying data to the ground imaging station. - The ground station computer and Wi-fi router. Responsible for initiating image capture, receiving the image data, and marking target locations. Figure 8. Block diagram of the imaging system University of Alberta - UAARG 9

10 The onboard imaging computer is an Odroid U3. This is a single-board computer with an ARM architecture, running a server version of Ubuntu We chose a Linux-based single-board computer for this task due to the availability of image processing libraries and the need to interface with and control many different types of hardware. The Odroid was chosen due to its high performance and low price relative to other single-board computer options on the market at the time (the original Raspberry Pi, the Pandaboard, the Beaglebone). We have one ground-facing camera onboard the aircraft, a Canon Powershot A2400IS. This is a consumer point-and-shoot camera that we have modified with the Canon Hack Development Kit (CHDK), a free and open source firmware upgrade for Canon cameras. CHDK allows us to set shutter speed, ISO and aperture. Additionally CHDK provides custom scripting, remote shooting and download from the camera. A consumer point-and-shoot is a cheaper solution than a specialized scientific imaging camera, but at the same time give us better quality and exposure control than a lower cost solution like a typical webcam would. In order to control the 3 processes of capturing, tagging, and transmitting images taken from this camera, the Odroid runs a custom multithreaded program, called Waldo. The Odroid receives telemetry data through its UART port from the autopilot. Waldo decodes this data and extracts the information about the aircraft s position and orientation, which are required for georectification of the image. Waldo then interpolates this data when the image capture is triggered and tags the resulting image with it. Images are then transmitted to the ground over the Wi-fi link, in parallel with the capturing and tagging processes. Waldo is initiated remotely from the ground imaging station; a user is able to monitor the status of image processing and transmission from the ground station. However, Waldo also runs regardless of whether a Wi-fi connection exists between the aircraft and the ground station; it saves images into a local buffer when link is lost and attempts to re-transmit images upon regaining a Wi-fi link, ensuring its reliability. The Imaging Ground station simply consists of a laptop computer and a Wi-Fi router. The Wi-Fi network is configured as an access point to both the ground station laptop and the Odroid and is the medium in which they exchange data. Waldo (on the Odroid) is configured to deposit the images along with telemetry information in a folder in the FTP server on the Wi-Fi network. Pigeon continuously monitors this folder and imports images as they available. Figure 9: Typical Pigeon UI when Analyzing Images. University of Alberta - UAARG 10

11 The Ground Imaging Station, which receives and processes these captured images, consists of a laptop computer and Wi-fi router. The Wi-fi router is configured as an access point to connect both the Ground Imaging Computer and the Odroid on a common wireless network. Image georeferencing is done here, via another custom program called Pigeon; this is written in Python for a GNU/UNIX environment, but can be extended to be cross-platform as well. The system is designed so that a human operator performs target recognition; Pigeon allows the operator to easily browse through images and mark the position of targets within the image. Pigeon is then able to take those coordinates in addition to the received position and orientation data, and georectify the position of the marker. Target properties, such as shape, orientation, and color, are also entered by the operator upon marking the image. Pigeon also has various export functionality; it is able to export target location via CSV and KML, to assist in data post-processing and analysis, as well as to expedite providing result to the judges in this competition. System Performance Autopilot Testing Figure 10 and 11 show desired roll/pitch as compared to actual aircraft roll/pitch during an autopilot-stabilized flight. These two loops are very low level and are the basic stabilized control loops; that is, they are essential for fully autonomous navigation. Figure 10: Desired Roll vs. Actual Roll during a stabilized flight University of Alberta - UAARG 11

12 Figure 11: Desired Pitch vs. Actual Pitch during a stabilized flight Figure 10 shows the aircraft responding well to roll commands. The pitch response still requires some improvement, but is suitable for basic autonomous navigation. The main deviations can be seen at about 290s and 370s where the actual pitch drops significantly from the desired pitch, likely due to a lack of airspeed and the aircraft approaching a stall state as a result. Further work and testing will be done to fine tune this loop, especially its reliability in varying wind conditions. Even with the poorly tuned pitch loop, however, the aircraft is still capable of reliable autonomous navigation. University of Alberta - UAARG 12

13 Figure 12: Autonomous Search Path of.8 square km This large search area was covered by our system in 9 minutes, with minimal deviation from the planned search path. The UAV operated at a maximum distance of just over 1 nm (2.1 km) away from the ground control station. Imaging Test Results In order to evaluate the imaging system s geolocation accuracy, the aircraft was flown at a test R/C flying field. The field has been previously surveyed to determine the locations of most visible features on the field, such as fence posts, cinderblocks, and sprinkler boxes. During a test flight, it is possible to use these as Ground Control Points (GCPs) - i.e. known locations that images taken of the field can be referenced against. In the following test results, large, aerially visible targets were placed over 4 sprinkler box GCPs. For each target, markers were placed in Pigeon for the first 5 appearances of that target - i.e. the first 5 passes of the aircraft over that target. The result from Pigeon s geolocation calculation was then exported and compared to the known location of the GCP; this was done for each individual marker associated with a GCP. An average position was then generated University of Alberta - UAARG 13

14 using Pigeon from the markers, and an average position error was calculated by comparing this position with the GCP s known location. Figure 13. Visible ground target (in white), ground control point location (marked as an X), and marked location of that target from a previous pass (flag icon) The results below show that most of the time, the imaging system is able to achieve within 75ft of localization accuracy. As seen in the 3rd table, the imaging system was not able to localize ground feature SPRINKLER015 to within 75ft, but it was able to localize it to within 150ft, which is the threshold requirement for localization. Table 5: Comparison of marked ground locations in software to reference ground control point locations GCP Name GCP Position (Lat) GCP Position (Lon) SPRINKLER Marker Name Position (Lat) Marker Position (Lon) Distance Error from GCP (ft) Bearing Error from GCP ( ) R P D E O Number of Markers Average Position (Lat) Average Position (Lon) Average Distance Error (ft) Average Bearing Error ( ) GCP Name GCP Position (Lat) GCP Position (Lon) SPRINKLER Marker Name Position (Lat) Marker Position Distance Error from Bearing Error from University of Alberta - UAARG 14

15 (Lon) GCP (ft) GCP( ) Q G U S F Number of Markers Average Position (Lat) Average Position (Lon) Average Distance Error (ft) Average Bearing Error ( ) GCP Name GCP Position (Lat) GCP Position (Lon) SPRINKLER Marker Name Position (Lat) Marker Position (Lon) Distance Error from GCP (ft) Bearing Error from GCP ( ) T H X V AD Number of Markers Average Position (Lat) Average Position (Lon) Average Distance Error (ft) Average Bearing Error ( ) GCP Name GCP Position (Lat) GCP Position (Lon) SPRINKLER Marker Name Position (Lat) Marker Position (Lon) Distance Error from GCP (ft) Bearing Error from GCP ( ) J I W A H Number of Markers Average Position (Lat) Average Position (Lon) Average Distance Error (ft) Average Bearing Error ( ) University of Alberta - UAARG 15

16 Achievable Competition Objectives 1. Autonomous Flight As UAARG has demonstrated numerous times, this UAS is capable of autonomous waypoint-guided flight. The hand-launch takeoff and landing will be performed with an R/C pilot, as these maneuvers have a much smaller margin of error. Once at competition the final flight plan will be developed which will include waypoints for autonomous navigation of primary and secondary tasks, search area boundaries, and no-fly zones. Each of these flight characteristics will be tested using Paparazzi s integrated simulator, to train the autopilot operator and test the software before flight. Once flight has commenced the operator and the judges will be continuously be updated of the plane's position, attitude, airspeed, and next autonomous actions. Images will be continuously taken and stored by the camera, where the Wi-Fi link is maintained the images will reach the ground station for processing. 2. Search Area Task Proceeding the autonomous flight task the flight plan will take the plane over the search area. As image capture is initiated upon takeoff and continues for the entire flight. A spreadsheet with an example calculation is shown below with our Canon Powershot A2400IS. This calculation is performed with an example cruising altitude of 70m and speed of 15 m/s, and shows that we are able to cover a first sweep of a 1km x 0.7km search area in under 10 minutes with ~188 pictures. This is with a required delay of approximately 3 seconds between images, well within the capture rate achievable by our camera s shutter speed settings and Waldo. Table 6. Example search area calculations University of Alberta - UAARG 16

17 3. Actionable Intelligence If the Wi-fi link is lost, Waldo will continue to capture images; it will attempt to automatically reestablish connection and continue to transmit image data once the Wi-fi link is reestablished. This method allows analysis to begin before landing. As a result, target characteristics are discovered well in advance to mission deadlines. Targets of interest can be further examined once discovered. 4. Off-axis Target In order to achieve this task, a high-altitude roll will be made perpendicular to the target by setting the autopilot to make a tight turn near the edge of the flight boundary. An example of this roll can be seen in Figure 14 where marginal resolution is present with the target at a 55 angle from the plane. To view a target 500 ft away at this angle means we need to make this roll at an altitude of 352 ft, well within the aircraft s capabilities. Figure 14: Pigeon with a High-Roll Image 5. Emergent Target Task When the last known location is provided to the Autopilot operator they will set the point into Paparazzi. The aircraft will be set to perform one pass directly over the target and will circle in concentric circles outwards until the Emergent Target is found. Depending on the remaining time, once the emergent target s location has been found at the Ground Imaging Station, its waypoint shall be programmed into the GCS; a second flight pass will then be performed at a lower altitude. 7. Interoperability and SDA Task The Paparazzi autopilot system s ground software components - the GCS, the telemetry data link, and the control uplink communicate with each other via a common message bus (the Ivy bus). All telemetry information received from the aircraft is broadcast onto the Ivy bus. Additional modules may be written to receive messages from this software bus and to broadcast messages onto it, allowing points in the GCS to be updated automatically. We plan to at a minimum read the UAV position from the Paparazzi Ivy Bus and transport this data to the competition server, as well as download object locations for the SDA task and display them in our GCS. Mission Commander, a Python program, will be used to act as the interoperability client. Using the Ivy software communications bus used by the Paparazzi GCS, the program will send messages to the GCS with University of Alberta - UAARG 17

18 obstacle information. These obstacles will then be displayed on the GCS and the user will then have the ability to reroute the aircraft to avoid obstacles. Mission Commander will eavesdrop on telemetry information and forward the relevant information to the interoperability server. Mission commander is capable of telemetry upload rates up to 71 Hz and averages at 66 Hz. Such high upload rates do consume significant cpu processing power and will not be utilized during the actual competition. Obstacle avoidance will be done by manually rerouting around stationary obstacles. Moving obstacles will be avoided by flying under them if lateral avoidance is challenging. Figure 15: Obstacle Information Displayed on Autopilot GCS Safety Considerations and Approach Safety is important to both mission success and UAARG members. UAARG has implemented the following three types of safety considerations to ensure no incidents occur during competition: These techniques have been continuously tested and will be presented to competition organizers during the Flight Readiness Review (FRR). Safety Design Features Manual Power Toggling Power toggle switches are conveniently installed in the nose of the airframe to provide easy manual power-down capabilities and prevent accidental shorting during troubleshooting on the avionics board. University of Alberta - UAARG 18

19 Figure 16: Perfboard Power-Toggle Switches (from L to R: Motor, Servos, Lisa, and Odroid) Kill Switch A kill switch is installed in addition to the power switches to prevent propellor incidents during assembly, testing, and troubleshooting. This kill switches works by grounding the motor signal going to the electronic speed controller. It will only be turned off at the moment of launch and re-engaged immediately after landing. Risk of Power Loss Due to weight and space constrictions no redundant power supply was implemented. The UAV will lose control if failures occur in the BEC or the 5V regulator that powers the autopilot control board. Although not redundant, the individual power supplies are rated well in excess of requirements.. Firmware Failsafes All failsafes are programmed and flashed into the autopilot board prior to flight. These failsafes are derived from both competition requirements in Section 9 and club protocols.the two primary failsafe states that the UAV may enter are Standby and Termination. In Standby state, the UAV will return to a predesignated home waypoint within close proximity to the takeoff area. This state's primary purpose is to postpone the mission until nominal communications return. The Termination state is defined by 0% throttle, full elevator up, full right rudder, and full roll on ailerons. During operations both the pilot and the ground control operator have direct ability to terminate the flight at any time. Furthermore, the pilot and GCS operator have direct radio contact at all times. All members present during the flight can request to end the flight, but must go through the flight commander before the flight can be terminated. To test the safety procedures, all possible scenarios will be simulated on the ground. This can be accomplished by physically disrupting various communication links and observing the reaction. We do this by setting up a test bench and powering off the communications modules responsible for each link, or by walking with the bench out of the wireless link s known range. Once the simulated failure occurs, the UAS should respond by going into the appropriate failsafe mode as mentioned above. The flight commander will constantly monitor the UAV during flight. If there is any danger to either the UAV and its components, or the personnel on the ground, the flight will end through manual R/C landing or termination through one of the many methods described. Certain conditions for delaying a flight may include: sudden change in weather or wind, UAV performance problems, critical component failure, falling debris, or smoke is seen coming from the UAV. Table 7 shows failure scenarios and the UAV and ground crew response. University of Alberta - UAARG 19

20 Table 7: Failure scenarios and response actions Failure: Recovery: Solution: Lost Telemetry High In the event of losing telemetry communications, the UAV must automatically return home after 30 seconds. Otherwise, the pilot must use R/C override to maintain manual control or land safely. If the UAV does return home but communications cannot not re-established with 3 minutes, the pilot must use R/C override to maintain a manual control or land safely. R/C uplink failure Lost Telemetry and R/C Uplink Communication s High In the event when R/C uplink is lost for greater than 10 seconds, the flight must be terminated. In the event of losing telemetry communications, the UAV must automatically return home after 30 seconds. Otherwise, the flight must be terminated. If the UAV does return home but communications cannot not re-established with 3 minutes, the flight must be terminated. Lost GPS Medium In the event that GPS location is lost, the UAV control will return to R/C uplink. If the pilot cannot safely operate the UAV, its flight will be immediately terminated. Outside mission boundary Conclusion Medium If the UAV leaves the mission boundaries, the UAV will be directed back within boundaries. In the event of loss of control and data link, flight will be automatically terminated By continuing to build on the successes and failures experienced with this system, UAARG has been able to address the main concerns with the system, thereby improving the systems accuracy and functionality. These improvements include a CSV export tool in Pigeon to quicken image processing, moving the magnetometer from the autopilot board to the wing to remove the magnetic interference of other nearby components, and integrating Mission Commander to Paparazzi to sense and detect the objects we are to avoid. With these improvements we expect to have more time at the flight line and better report our findings. Acknowledgements We would like to thank Peter Cary, the Edmonton Radio Control Society, the University of Alberta, Shell Canada and all members of the University of Alberta Aerial Robotics Group as for their support and guidance. References [1] AUVSI Seafarer Chapter. AUVSI Seafarer Chapter 14 th Annual Student UAS Competition Internet: _Rev_1.0_FINAL_( ).pdf University of Alberta - UAARG 20