<%BANNER%>

Improving Operational Efficiency of Small Unmanned Aircraft Systems for Remote Sensing Applications

Permanent Link: http://ufdc.ufl.edu/UFE0043857/00001

Material Information

Title: Improving Operational Efficiency of Small Unmanned Aircraft Systems for Remote Sensing Applications
Physical Description: 1 online resource (114 p.)
Language: english
Creator: Evers, Brandon S
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: aerial -- aircraft -- autonomous -- flight -- photogrammetry -- planning -- remote -- sensing -- systems -- uas -- uav -- unmanned
Mechanical and Aerospace Engineering -- Dissertations, Academic -- UF
Genre: Aerospace Engineering thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Aerial photogrammetry is traditionally a specialized application for small manned aircraft with large operating expenses. Emerging small unmanned aerial systems (sUAS) have the capability to augment existing manned missions with low-level, cost effective unmanned missions. One of the main barriers to the use of sUAS is their limited endurance with current propulsive methods. To optimize data post-processing, care should be exercised to ensure that images are taken orthogonal to the direction of the flight path. To minimize crosswind-induced crab angles, flight paths should always be flown directly into or with the wind, thus simplifying the wind interaction to a purely additive or subtractive term for ground coverage. Thus, neglecting turns and altitude changes, the optimization problem is simplified to total ground coverage with vehicle speed and wind speed as the only variables. With this flight regime determined prior to vehicle construction, optimal range can be increased by the inclusion of trailing edge (camber/reflex) control, maximizing the Lift to Drag ratio (L/D) to minimize power consumption over a range of airspeeds. The results presented here couple aerodynamic analysis, ground coverage efficiency, and power consumption data to provide a method for calculating optimal flight speeds in any given wind condition for maximum coverage. Due to the complex interaction between the aerodynamic and propulsive efficiencies of the sUAS, interesting and non-intuitive results emerge. In an approach analogous to the speed-to-fly theory used in cross-country soaring, a dynamic flight management algorithm was developed on a modified Procerus® KestrelT autopilot. This algorithm allows the sUAS to monitor the winds aloft and modify the flight plan transects and commanded airspeeds accordingly. Furthermore, this algorithm sets the camber of the airfoil at its optimum position for each portion of the flight plan. This combination of orientation, commanded airspeed, and camber position maximizes the total area mapped per flight. Initial tests of the flight management algorithm yielded significant improvements in ground coverage efficiency in terms of both energy consumption and required flight time. By reducing energy and flight time requirements by up to 14% and 43%, respectively, these flight management strategies significantly improve the operational feasibility of sUAS for remote sensing applications.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Brandon S Evers.
Thesis: Thesis (M.S.)--University of Florida, 2011.
Local: Adviser: Ifju, Peter.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0043857:00001

Permanent Link: http://ufdc.ufl.edu/UFE0043857/00001

Material Information

Title: Improving Operational Efficiency of Small Unmanned Aircraft Systems for Remote Sensing Applications
Physical Description: 1 online resource (114 p.)
Language: english
Creator: Evers, Brandon S
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

Subjects / Keywords: aerial -- aircraft -- autonomous -- flight -- photogrammetry -- planning -- remote -- sensing -- systems -- uas -- uav -- unmanned
Mechanical and Aerospace Engineering -- Dissertations, Academic -- UF
Genre: Aerospace Engineering thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Aerial photogrammetry is traditionally a specialized application for small manned aircraft with large operating expenses. Emerging small unmanned aerial systems (sUAS) have the capability to augment existing manned missions with low-level, cost effective unmanned missions. One of the main barriers to the use of sUAS is their limited endurance with current propulsive methods. To optimize data post-processing, care should be exercised to ensure that images are taken orthogonal to the direction of the flight path. To minimize crosswind-induced crab angles, flight paths should always be flown directly into or with the wind, thus simplifying the wind interaction to a purely additive or subtractive term for ground coverage. Thus, neglecting turns and altitude changes, the optimization problem is simplified to total ground coverage with vehicle speed and wind speed as the only variables. With this flight regime determined prior to vehicle construction, optimal range can be increased by the inclusion of trailing edge (camber/reflex) control, maximizing the Lift to Drag ratio (L/D) to minimize power consumption over a range of airspeeds. The results presented here couple aerodynamic analysis, ground coverage efficiency, and power consumption data to provide a method for calculating optimal flight speeds in any given wind condition for maximum coverage. Due to the complex interaction between the aerodynamic and propulsive efficiencies of the sUAS, interesting and non-intuitive results emerge. In an approach analogous to the speed-to-fly theory used in cross-country soaring, a dynamic flight management algorithm was developed on a modified Procerus® KestrelT autopilot. This algorithm allows the sUAS to monitor the winds aloft and modify the flight plan transects and commanded airspeeds accordingly. Furthermore, this algorithm sets the camber of the airfoil at its optimum position for each portion of the flight plan. This combination of orientation, commanded airspeed, and camber position maximizes the total area mapped per flight. Initial tests of the flight management algorithm yielded significant improvements in ground coverage efficiency in terms of both energy consumption and required flight time. By reducing energy and flight time requirements by up to 14% and 43%, respectively, these flight management strategies significantly improve the operational feasibility of sUAS for remote sensing applications.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Brandon S Evers.
Thesis: Thesis (M.S.)--University of Florida, 2011.
Local: Adviser: Ifju, Peter.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0043857:00001


This item has the following downloads:


Full Text

PAGE 1

1 IMPROVING OPERATIONAL EFFICIENCY OF SMALL UNMANNED AIRCRAFT SYSTEMS FOR REMOTE SENSING APPLICATIO NS By BRANDON SCOTT EVERS A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIV ERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2011

PAGE 2

2 2011 Brandon Scott Evers

PAGE 3

3 To Mom Dad Megan and Heather

PAGE 4

4 ACKNOWLEDGMENTS Thank you, Mom and Dad, for giving m e the early tools to succeed and the patience and encouragement to pursue an advanced degree without the pressure to get ure together brings. I owe a huge debt of gratitude to the interdisciplinary team of researc hers contributing to the UF UASRG, whose passion and support have fostered a truly unique and successful program. Dr. Ifju, thank you for providing me with the ama zing opportunity to work in such an exciting field and for your academic guidance throughout my college career. Dr Percival thank you for your years of work keeping this project moving and focused on quality science despite a constantly changing funding environment. Matt, your infinite patience and ability to juggle conflicting demands from all directions has kept us all working smoothly and I wish you luck as you begin your doctoral studies. Drs. Dewitt and Smith, I truly feel lucky to have worked with s uch a broad range of researchers from varying disciplines; the collaboration among sciences has created something great. Finally, I thank my colleagues for their innumerable hours spent working on this project. John, thanks for all your work pushing the pr oject towards operational success and for the confidence to press on with another generation of development. Thomas Rambo, thank you for your efforts in the field and for your huge contribution s to the autopilot integration code. Thomas Reed, leasure to crank out airframes with you and your work as on call machinist has been invaluable. I look forward to working with you all for some time to come.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 LIST OF ABBREVIATIONS ................................ ................................ ........................... 12 ABSTRACT ................................ ................................ ................................ ................... 16 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 18 Motivation ................................ ................................ ................................ ............... 18 The UF Nova Project ................................ ................................ .............................. 19 Directly Georeferenced Aerial Photogrammetry ................................ ..................... 20 2 NOVA PR OJECT PLATFORM DEVELOPMENT ................................ ................... 24 Legacy Systems ................................ ................................ ................................ ..... 24 Lessons Learned ................................ ................................ ................................ .... 28 Platform Requirements and Design Goals ................................ .............................. 31 3 NOVA 2.1 AIRFRAME ................................ ................................ ............................ 35 Wing Design ................................ ................................ ................................ ........... 35 Tail Design and Stability Considerations ................................ ................................ 42 Fuselage Design and Center of Gravity Considerations ................................ ......... 46 Control Redundancy and RDS Linkages ................................ ................................ 48 Avionics ................................ ................................ ................................ .................. 50 Power System and Cooling ................................ ................................ ..................... 52 Fabrication and Assembly ................................ ................................ ....................... 56 Performance Evaluation ................................ ................................ .......................... 59 Platform Overvie w ................................ ................................ ................................ .. 60 Airframe ................................ ................................ ................................ ............ 60 Electronics ................................ ................................ ................................ ........ 60 Deployment ................................ ................................ ................................ ...... 60 4 SYSTEM DEPLOYMENT AND EVALUATION ................................ ....................... 62 ISR Mission Profile ................................ ................................ ................................ 62 Photogrammetry Mission Profile ................................ ................................ ............. 63 Flight Planning Implications ................................ ................................ .................... 64

PAGE 6

6 Flight Planning Optimization ................................ ................................ ................... 70 Operational Feasibility of sUAS for Remote Sensing Applications ......................... 73 5 SPEED TO FLY APPLICATION ................................ ................................ ............. 75 Theory ................................ ................................ ................................ ..................... 75 Trailing Edge Control ................................ ................................ .............................. 79 Initial Power System Model ................................ ................................ ..................... 87 Experimental Power Consumption Testing ................................ ............................. 91 STF Application ................................ ................................ ................................ ...... 96 Autopilot Integratio n of Variable Camber for STF ................................ ................. 103 6 SUMMARY AND CONCLUSIONS ................................ ................................ ........ 107 Summary ................................ ................................ ................................ .............. 107 Conclusions ................................ ................................ ................................ .......... 108 Recommendations ................................ ................................ ................................ 108 LIST OF REFERENCES ................................ ................................ ............................. 111 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 114

PAGE 7

7 LIST OF TABLES Table p age 3 1 Nova 2.1 wing design parameters ................................ ................................ ...... 39 3 2 Nova 2.1 tail design parameters ................................ ................................ ......... 44 3 3 Comparison of motor specifications for Neu 1509 1.5Y and Scorpion 4025 12 53 3 4 Comparison of as built Nova 2 and Nova 2.1 platforms ................................ ..... 60 5 1 MH32 wing conf iguration parameters ................................ ................................ 85

PAGE 8

8 LIST OF FIGURES Figure p age 1 1 Typical forward and lateral overlap values on adjacent flight lines for ae rial photogrammetry missions ................................ ................................ .................. 21 1 2 Nova 2.1 flight planning guide for determining flight line spacing and airspeed for desired overlap ................................ ................................ ............................. 22 1 3 Illustration of loss of stereoscopic coverage between (a) uncrabbed adjacent images and (b) overlapping images with crosswind induced crab angle. ........... 23 2 1 Failed launch attempt of MLB Foldbat at Pine Island, FL. ................................ .. 25 2 2 Assembled Tadpole WSUAV. ................................ ................................ ............. 26 2 3 Top and side view of completed Nova 1 airframe. ................................ .............. 27 2 4 Assembled Nova 2 sUAS. ................................ ................................ .................. 28 2 5 Hand launch and water recovery of Nova 2.1 sUAS from airboat on Lake Okeechobee, FL demonstrating operational constraints imposed by the unique operating environment. ................................ ................................ ........... 34 3 1 Drag polar of SD7032 airfoil at cruise and dash average Reynolds numbers of 350k and 700k, respectively. ................................ ................................ .......... 38 3 2 Nova 2.1 wing planform wi th XFLR5 mesh distribution displayed. ..................... 39 3 3 Comparison of lift to drag ratios for Nova 2 and Nova 2.1 wing designs. ........... 40 3 4 Wing bending moment diagram illustrating 85% reduction in bending moment resulting from outboard joiner location. ................................ ............................... 41 3 5 Nova 2.1 wing polyhedral configurati on in front view. ................................ ......... 42 3 6 Overview of Nova 2.1 tail design. ................................ ................................ ....... 45 3 7 XLFR5 streamline analysis illustrati ng tail placement away from downwash and fuselage junction vortices. ................................ ................................ ........... 45 3 8 Preliminary component placement for CG calculations was completed prior to final fuselage design. ................................ ................................ .......................... 46 3 9 Nova 2.1 complete airframe configuration. ................................ ......................... 47 3 10 Schematic of aileron RDS linkage showing tool and electrical free connection. ................................ ................................ ................................ ......... 49

PAGE 9

9 3 11 Tail RDS linkages. ................................ ................................ .............................. 50 3 12 Custom machined autopilot enclosure al lows for complete waterproofing of critical components while maintaining adequate heat management. .................. 51 3 13 Fuselage top populated with electronics prior to joining. ................................ .... 52 3 14 Efficiency plots for Neu 1509 1.5Y (inrunner) and Scorpion 4025 12 (outrunner) motor options for 18.5 V input. ................................ ......................... 53 3 15 Propeller efficiencies for various P/D and J ratios ................................ ............. 54 3 16 Nova 2.1 propulsion system, with Aeronaut CAM 15x13 folding propeller, externally vented Neu 1509 1.5Y, and Phoenix Ice 100 external heat sink. ....... 56 3 17 Local unidirectional carbon reinforcements around wing mounts and hatch locations. ................................ ................................ ................................ ............ 57 3 18 Schematic of fuselage halves built with molded ...................... 58 3 19 Completed fuselage pr ior to avionics install in preparation for joining. ............... 58 3 20 Wing construction picture showing selective local reinforcement with unidirectional and bidirectional carbon fabrics. ................................ ................... 59 3 21 Disassembled Nova 2.1 illustrating small deployment volume. .......................... 61 3 22 Assembled Nova 2.1 aircraft ................................ ................................ .............. 61 4 1 Typical sUAS flight profile for ISR missions. ................................ ....................... 62 4 2 Dipole flight path over areal target on Lake Okeechobee, FL with each point representing a single image exposure. ................................ ............................... 66 4 3 Linear flight path over levee target with takeoff and landing spirals visible. ........ 67 4 4 Flight line patterns for remote sensing applications. ................................ ........... 69 4 5 Flight plan design procedure and relevant parameters. ................................ ...... 70 4 6 Parallel track flight plan for mapping road degradation in Picayune Strand State Forest, FL. ................................ ................................ ................................ 71 4 7 Flight profile for a typical sUAS based aerial mapping mission. ......................... 72 4 8 Optimized flight profile with more aggressive takeoff/climb routing and a power off descent to landing ................................ ................................ .............. 73

PAGE 10

10 5 1 XFLR5 generated power curve (assuming 100% propulsion efficiency) for Nova 2.1 illustrating graphical computation of optimal STF in still air. By flying at V mp ( mp = mini mum power), power consumption is held at a minimum and maximum duration is achieved. However, flying at V mr ( mr = maximum range), minimizes the Distance Specific Energy Consumption (given by the quotient P/V ) and maximizes the distance traveled per unit o f input energy. ..... 77 5 2 XFLR5 generated Nova 2.1 power curve illustrating application of STF intercept shifts to the right indi cating a loss in effective groundspeed and the new optimal airspeed, V mr is found to be higher than the still air value, V* mr ................................ ........ 78 5 3 Visual comparison of SD7032 and MH32 airfoil s. ................................ ............... 81 5 4 Comparison of drag polars (Re=350k) for original SD7032 and MH32 airfoils. .. 82 5 5 Trailing e dge configurations for various flight mode settings. ............................. 83 5 6 MH32 wing in clean, takeoff, and brake configurations. Camber change is especially noticeable at the inboard flap edges ................................ ................. 84 5 7 Computed drag curves for original SD7032 wing and MH32 wing in clean, float and dash modes demonstrating improved performance in high speed cruise. ................................ ................................ ................................ ................. 86 5 8 XFLR5 generated power curves for original SD7032 wing and composite curve for selected MH32 flight modes. The full selection of flight modes has been omitted for clarity. ................................ ................................ ...................... 87 5 9 Power system block diagram. ................................ ................................ ............. 88 5 10 Static thrust test results for Neu 1509/1.5Y (6.7:1 GB), Phoenix Ice 100, Aeronaut CAM 15x13 folding prop. ................................ ................................ ..... 89 5 11 Comparison of computational power curve model with experimental values for original SD7032 wing. ................................ ................................ ................... 90 5 12 Power testing flight plan consisting of repeated 1.6 km long parallel transects. ................................ ................................ ................................ ............ 92 5 13 Airspeed intervals for MH32 clean configuration flight ................................ ....... 93 5 14 Experimentally derived power consumption results compared for original Nova 2.1 wing and modified MH32 configurations. ................................ ............. 94 5 15 Airspeed intervals for SD7032 configuration flight showing much noisier data and a climb fault occurring during a 24 m/s transect. ................................ ......... 95

PAGE 11

11 5 16 Varia tion of advance ratio, J across airspeed range for MH32 clean configuration. ................................ ................................ ................................ ...... 96 5 17 Experimentally obtained MH32 power curve with still air DSEC shown. ............. 97 5 18 Schematic of flight performance for 2 km transects in still air conditions detailing time of flight and power consumption. ................................ .................. 98 5 19 Schematic of flight performance for 2 km transects in 10 m/s steady wind conditions, illustrating efficiency losses without STF application. ....................... 99 5 20 Improved time and ground coverage efficiencies for 2 km transects in 10 m/s steady wind with STF application. ................................ ................................ .... 100 5 21 Demonstration of performance gains using variable camber enhanced STF on MH 32 composite power curve. ................................ ................................ ... 101 5 22 Schematic of flight performance with variable camber enhanced STF application. ................................ ................................ ................................ ....... 102 5 23 Positive feedback in closed loop flap control. ................................ ................... 105

PAGE 12

12 LIST OF ABBREVIATION S A Ampere Ah Ampere hour AR Aspect Ratio b Wingspan cc Cubic Centimeter CCD Charge Coupled Device CG Center of Gravity CIR Color Infrared C d Section Drag Coefficient C l Section Lift Coefficient C l,max Section Maximum Lift Coefficient C L,max Wing Maximum Lift Coefficient cm Centimeter COTS Consumer Off The Shelf D Propeller Diameter DEM Digital Elevation Model dm Decimeter D SLR Digital Single Lens Reflex DSE C Distance Specific Energy Consumption EDA Equivalent Dihedral Angle ESC Electronic Speed Control FAA Federal Aviation Administration g Gram gsm Grams per Square Meter

PAGE 13

13 GCP Ground Control Point GCS Ground Control Statio n GPS Global Positioning System HALE High Altitude, Long Endurance Hz Hertz IMU Inertial Measurement Unit INS Inertial Navigation System ISR Intelligence, Surveillance, and Reconnaissance ITAR International Traffic in Arms Regulations J Propeller Advance Ratio kg Kilogram km Kilometer kV Rotations per Minute per Volt L/D Lift to Drag Ratio m Meters MAC Mean Aerodynamic Chord mAh Milliampere hour MAV Micro Air Vehicle mi Mile MP Megapixel m/s Meters per Second n Revolutions per Minute N Newton N m Newton m eter P Power

PAGE 14

14 P/D Propeller Pitch to Diameter Ratio PID Proportional Integral Derivative P min Minimum Input Power RDS Rotary Drive System Re Reynolds Number RF Radio Frequency R min Minimum Turn Radius RMS Root Mean Squared RPM Revolutions per Minute S Wing Area sUAS Small Unmanned Aerial System STF Speed to Fly UAS Unmanned Aerial System UASRG Unmanned Aerial Systems Research Group UAV Unmanned Aerial Vehicle UF University of Florida USACE US Army Corps of Engineers V Volt V Airspeed VLOS Visual Line of Sight V mp Minimum Power Airspeed V mr Maximum Range Airspeed W Watt Wh Watt hour WS Wingspan

PAGE 15

15 W/S Wing Loading Air Density t Tail Dihedral Angle

PAGE 16

16 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science IMPROVING OPERATIONAL EFFICIENCY OF SMALL UNMANNED AIRCRAFT SYSTEMS FOR REMOTE SENSING APPLICATI ONS By Brandon Scott Evers December 2011 Chair: Peter Ifju Major: Aerospace Engineering Aerial photogrammetry is trad itionally a specialized application for small manned aircraft with large operating expenses. Emerging small unmanned aerial systems ( sUAS ) have the capability to augment existing manned missions with low level, cost effective unmanned missions. One of the main barriers to the use of sUAS is their limited endurance with current propulsive methods. To optimize data post processing, care should be exercised to ensure that images are taken orthogonal to the direction of the flight path. To minimize crosswind in duced crab angles, flight paths should always be flown directly into or with the wind, thus simplifying the wind interaction to a purely additive or subtractive term for ground coverage. Thus, n eglecting turns and altitude changes, the optimization problem is simplified to total ground coverage with vehicle speed and wind speed as the only variables. With this flight regime determined prior to vehicle construction, optimal range can be increased by the inclusion of trailing edge (camber/reflex) control, max imizing the Lift to Drag ratio ( L/D) to minimize power consumption over a range of airspeeds. The results presented here couple aerodynamic analysis, ground coverage efficiency, and power consumption data to provide a method for calculating optimal flight speeds in any given wind condition for

PAGE 17

17 maximum coverage. Due to the complex interaction between the aerodynamic and propulsive efficiencies of the sUAS, interesting and non intuitive results emerge. In an approach analogous to the speed to fly theory used in cross country soaring, a dynamic flight management algorithm was developed on a modified Procerus autopilot. This algorithm allows the sUAS to monitor the winds aloft and modify the flight plan transects and commanded airspeeds accordingly. Fu rthermore, this algorithm sets the camber of the airfoil at its optimum position for each portion of the flight plan. This combination of orientation, commanded airspeed, and camber position maximizes the total area mapped per flight. Initial tests of the flight management algorithm yielded significant improvements in ground coverage efficiency in terms of both energy consumption and required flight time. By reducing energy and flight time requirements by up to 14% and 43%, respectively, these flight manag ement strategies significantly improve the ope rational feasibility of sUAS for remote sensing applications.

PAGE 18

18 CHAPTER 1 INTRODUCTION Motivation High precision mapping and engineering products produced using aerial photogrammetry techniques have typically been serv iced by small manned aircraft deploying large frame cameras coupled with expensive inertial and geolocation equipment (Wolf and Dewitt 2000) However, recent advances in the quality of small optical sensors an d inertial navigation equipment have led to the ad option of small unmanned aerial vehicles (UAVs ) as a platform for remote sensing applications. By directly georeferencing the imagery as it is captured, analysts can later construct high accuracy digital el evation models (DEMs) and mosaicked orthophotos acceptable for a wide range of environmental, engineering, and management applications (J. H. Perry 2009) Because this imagery can be processed without the requirement of presur veyed ground control points (GCPs) data collection can proceed much faster than traditional methods and in areas previously unreachable for the survey of GCPs An additional advantage is found in the low operating altitudes of sUAS for remote sensing, whi ch enable the collection of very high resolution imagery (sub 1 cm pixel size) which traditional platforms cannot readily achieve. Finally, the acquisition, mobilization, and deployment costs associated with sUAS are much lower than those of traditional pl atforms, which places remote sensing data in the hands of more engineers and researchers and permits more frequent acquisitions, further increasing their scientific value.

PAGE 19

19 The UF Nova Project For more than a decade, interdisciplinary researchers at the Un iversity of Florida (UF) Unmanned Aircraft Systems Research Group (UASRG) have advanced an interdisciplinary project to investigate and refine techniques to utilize sUAS for remote sensing applications (Jones 2003) While early efforts were focused on real time video based payloads for wildlife monitoring, a programmatic shift towards production of high accuracy, high resolution orthomosaics developed, motivated by the increased number of applications for this data type (Wilkinson 2007) While real time video streams are valuable for immediate awareness during situations like wildfire monitoring and natural disaster response, the vast majority of natural resource applications require high accuracy hig h resolution imagery over large geographic areas. Although hours of full motion video can yield the same information about a given target area, the resulting deluge of data, its storage and subsequent processing put it at an operational disadvantage to the at a glance situational awareness afforded by still orthomosaics (Magnuson 2010) Additionally, for projects involving traditional remote sensing postprocessing techniques like automated vegetation classification, current meth ods are optimized for utilization of still imagery (Jensen 2004) The high resolution imagery products able to be produced by an sUAS based data collection system yield themselves to a variety of applications, from vegetation i maging for crop assessment, (MacArthur, Schueller and Crane 2005) environmental rehabilitation (Nyquist 1996) habitat evaluation (Breckenridge, Harbour, et al. 2006) and rangeland monitoring (Breckenridge 2007) In addition to research focused deployments o f the technology, p otential operational applications for invasive species mitigation, infrastructure inspection, and environmental

PAGE 20

20 impact monitoring attracted additional support from the US Army Corps of Engineers (USACE), Jacksonville District (Bowman, Watts, et al. 2008) As a result, an evolving focus of the project became the advancement of the system technology for data collection and processing to the point where operators external to the UASRG development team could reli ably produce scientifically us able data on a repeatable basis (Burgess, Ifju, et al. 2009) This operational requirement, relatively unique for a research organization, has led to a number of developmental advantages that would no t otherwise have been encountered (Vorse 2011) Directly Georeferenced Aerial Photogrammetr y Using directly georeferenced aerial imagery, geomaticists can produce highly accurate orthomosaics and DEMs by using an aerotriangula tion process. At its core, direct georeferencing refers to the tagging of ea ch image with a set of exterior orientation parameters (EOPs), estimated camera position and orientation data given by an INS/GPS (J. H. Perry 2009) T hrough multiple observations of common tie points between adjacent images, aerotriangulation is then used to improve the fidelity of those initial estimations. These improved EOPs are then used to create a geometric model of the terrain needed for accurate mosaicking of the overlapping images. Because the EOPs are used to compute the scene geometry and vice versa a simultaneous bundle adjustment is performed, which combines the two models and performs a least squares minimization on the estimation error (Wolf and Dewitt 2000) This simultaneous bundle adjustment is one of the key advantages of the photogrammetric technique, as it enables the accuracy of the completed DEMs and orthomosaics to be improved over the initial accurac y of the raw INS/GPS sensor data

PAGE 21

21 The overall time required to complete this lengthy processing technique can be minimized by improving the accuracy of the initial parameter estimates, which requires sufficient coverage of the target area by multiple obse rvations. Full stereoscopic coverage, needed for generation of 3D scene geometry, requires that every point in the target area be imaged at least twice. To guarantee sufficient coverage, minimum image overlaps between adjacent exposures must be maintained both along flight lines and across adjacent lines. As shown in Figure 1 1, forward overlap (endlap) and lateral overlap (sidelap) refer respectively to the percentage overlaps between images along the same fl ight line or on adjacent lines while forward and side airbase refer to the respective spacing between adjacent images. Figure 1 1. Typical forward and lateral overlap values on adjacent flight lines for aerial photogrammetry missions While camera parameters like the interval between exposures, lens focal length, and sensor geometry contribute to determining image overlap, flight parameters like spacing between adjacent flight lines, altitude, and airspeed also have definite effects.

PAGE 22

22 For mos t flights with the Nova project payload, altitude is primarily dictated by desired output resolution (flying lower for higher resolution), leaving airspeed and flight line spacing to be determined to maintain desired endlap and sidelap percenta ges. For a given set of camera parameters, a flight planning guide can be developed to determine appropriate values for these parameters with overlap requirements based on desired output products. A sample flight planning guide for the current Nova 2.1 pay load is shown in Figure 1 2 (J. H. Perry 2009) Figure 1 2. Nova 2.1 flight planning guide for determining flight line spacing and airspeed for desired overlap [Adapted from P erry, J.H. 2009 A Synthesized Directly Georeferenced Remote Sensing Technique for Small Unmanned Aerial Vehicles Master's Thesis (Page 82 Figure 4 11) University of Florida, Gainesville, FL.] A major factor in flight planning for sUAS is the direction and intensity of winds airspeed must be accounted for to ensure that endlap req uirements are met. Because the Nova 2.1 payload does not have an azimuthal mount to continuou sly correct for

PAGE 23

23 crosswind induced sideslip, significant losses in stereoscopic coverage can occur when the aircraft heading is not in line with the ground track, as shown in Figure 1 3. As a result, it is most efficient for flight lines to be oriented in l ine with prevailing winds to minimize potential coverage losses due to this sideslip, or crab. Figure 1 3 Illustration of loss of stereoscopic coverage between (a) uncrabbed adjacent images and (b) overlapping images with crosswind induced crab angle.

PAGE 24

24 CHAPTER 2 NOVA PROJECT PLATFORM DEVELOPMENT Legacy Systems Since 2001 the UF UASRG has fielded 5 generations of sUAS, with each successive iteration building upon lessons learned from deployme nts of previous generations. A programmatic feedback loop was c reated wherein increasingly reliable and capable airframes enabled the development of more sophisticated payloads that delivered increasingly valuable data products. The emergent demand for increased geographic coverage with these improved data deliverable s then drove the push for further improving airframe deployability and endurance completing the cycle and beginning a new generational loop The initial sensor platform was the FoldBat, a commercial offering from the MLB Company. With a 1.5 m wi ngspan and a flying weight of 4.3 kg, the FoldBat deployed a trio of 330 line CMOS cameras, transmitting live video to a ground station for tracking and record ing. The low resolution provided by these chip imagers soon motivated a payload modification to a single 525 line progressive scan camera, at the expense of increased payload weight. Propulsion was provided by a nitromethane fueled 5 .2cc glow engine, which yielded 45 minutes of endurance. Unfortunately, t he high wing loading of the FoldBat led to poor takeoff and landing success rates especially during field operations from challenging launch sites, as seen in Figure 2 1. Additionally, the lightweight fragility and lack of waterproofing made it unsuitable for operation in wetlands and marin e areas targeted fo r study by the research group (Jones 2003)

PAGE 25

25 Figure 2 1 Failed launch attempt of MLB Foldbat at Pine Island, FL. To address the operational deficiencies of the FoldBat system, members of the UF Micro Air Ve hicles (MAV) Lab developed the Tadpole, an electrically powered sUAS designed for wildlife surveillance (Lee 2004) The resulting 2 m wingspan, 3.4 kg Tadpole shown in Figure 2 2, carried two CCD cameras (nadir and side look) with an onboard digital video recorder While much more successful during field deployments adequate g eolocation for the captured video data motivated development of a subsequen t platform (Wilkinson 2007)

PAGE 26

26 Figure 2 2 Assembled Tadpole WSUAV. The resulting Polaris/Nova 1 system was developed expressly to provide accurately georeferenced imagery and the overall system architecture began to reflect th at developing mission (Bowman 2008) Built on a modified Tadpole fuselage, the 2.44 m wingspan Nova 1 shown in Figure 2 3, boasted an onbo ard payload control computer that combined images from a stripped down 12.1 MP Canon point and shoot still camera with GPS position and orientation metadata from the autopilot IMU. Although limited by the low precision and update rates of the metadata, t hi s direct georeferencing enabled limited photogrammetric post processing of the imagery to permit geolocation of visible targets with a 67.6 m RMS error (Bowman 2008) The success of the Nova 1 system in producing directly geo referenced imagery from an sUAS led to a development effort intended to push the technology to an operational phase where deployment crews from the USACE and similar organ izations could repeatably and reliably collect georeferenced imagery for a variety of projects.

PAGE 27

27 Figure 2 3 Top and side view of completed Nova 1 airframe. To accomplish this goal, a dramatically improved photogrammetric payload was developed by the ge omatics team of the UF UASRG. The new payload, designed to use commercial off the shelf (COTS) components centered around an Olympus E 420 digital single lens reflex (D SLR) camera, which offered a much larger sensor and improved optics over previous ima gers. To deliver improved navigation and sensor orientation data, an Xsens MTi G GPS/INS was selected, offering major improvements in both accuracy and update rate over the relatively coarse sensors used by the autopilot. This is key for improving photogra mmetric processing because at a flying height of 200 m, a 3 degree error in the estimated orientation of the camera results in a 10.5 m error in the initial positional estimate This ultimately results in increased search spaces and dramat ically longer processing times (J. H. Perry 2009) Additionally, the MTi G is not subject to cumbersome International Traffic in Arms Regulations (ITAR), meaning that both the physical deployment of the device and its collecte d data are not ITAR restricted and more readily publishable. The Nova 2 platform was a significantly larger aircraft than its predecessors, maintaining the 2.44 m wingspan of the Nova 1, but enlarging the planform area and fuselage to present a much large r aircraft better suited to the increased s pace

PAGE 28

28 requirements of the upgraded payload. As indicated by the USACE logo and paint scheme on the Nova 2 airframe in Figure 2 4, the Nova 2 was the first sUAS to successfully demonstrate the operational deployment of a UF UASRG developed system by external operators In moving the deployment of the platform from a research and development stage into this new operational phase, a unique set of associated design challenges were presented (B urgess, Percival, et al. 2009) Figure 2 4 Assembled Nova 2 sUAS. Lessons Learned Although relatively successful as an initial test/deployment platform for the current generation mapping payload, there were a number of significant design and operat ional issues with the Nova 2 system. To realize system deployment by non expert operators external to the UF development team, the platform reliability needed to be enhanced and the overall assembly and deployment

PAGE 29

29 s implified. Additionally, research flights focused on operational procedure optimization required a stable platform whose reliability was never in question. The key flaw in the Nova 2 design was excessive platform weight. With a wingspan of 2.44 m and a pl anform area of 67.4 dm 2 the Nova 2 tipped the scales at 5.1 kg without payload and 6.1 kg with a typical loadout supporting a single E 420 D SLR Neglecting payload t his was a lready in excess of the 4. 5 kg design weight around which the wing area was opt imized and is mostly attributable to inefficient laminate design and manufacturing techniques. For expediency, the entire vehicle was constructed of a laminate of carbon fibe r fabric with an epoxy matrix. The fuselage was hollow molded with three layers of 190 gsm bidirectional carbon fabric while the foam core wing was reinforced with a single vacuum bagged layer of 190 gsm bidirectional carbon fabric. While strong, this layup schedule resulted in a heavier than necessary airframe that was also brittle dur ing landing recoveries on unimproved fields. Additionally, because carbon fiber is a conductor, significant RF interference and blanking was encountered during flight testing due to shadowing of the RF link between the ground stati on and the airborne vehicle. Inadequate early collaboration between airframe and payload specialists also meant that the airframe was delivered without consideration of payload location and its As a result, the optimal operational configuration resulted in the Nova 2 being severely nose heavy, which, when combined with the offset thrust line and high wing loading, contributed to a number of failed hand launch attempts.

PAGE 30

30 As a result of th o verweight operational condition, the flight perfo rmance of the platform suffered. This takeoff and la nding performance, owing to the high stall speed of 17 m/s This is a relatively high value for platforms in t his size class, especially for hand launched, belly landed aircraft and led to numerous operational incidents during these flight pha ses An additional consequence of the increased flying weight was the Nova 2 minimum cruise airspeed of 22 m/s, which wa s found to be too high to achieve the minimum specified airbase at lower altitudes, making all downwind legs practically worthless for effective data collection. For operational deployments, such a large effective loss is critical as it halves potential gr ound coverage per flight, minimizing the endurance, with typical flight durations of 35 mi turnaround time between missions, this severely limited the amount of ground coverage a crew could achieve per day. Finally, from a practical deployment perspective, a number of critical design flaws hampered the Nov First, water intrusion proved to be a major issue during wetlands missions contributing to at least one terminal in flight failure. Numerous ingress points at mechanical junctions like the fusela ge/boom joint and wing and tail mating surfaces as well as cooling inlets around the motor were inadequately sealed, leading to significant accumulation of water in the fuselage and subsequent elevator servo failure during operations on Lake Okeechobee, FL Additionally, insufficient cooling of electronics was found to be an issue, with the hot

PAGE 31

31 operating environment and waterproof enclosure requirements compounding the situation. As a result, poor reliability was encountered, with an emergency flight recove ry required on one mission due to an in flight ESC fire. Lastly field assembly and preflight procedures for the vehicle were cumbersome and time consuming, a situation frequently exacerbated by the frequent requirement to remove and refit the wing or fron t hatch to reset a preflight system fault. Removing the wing required the replacement of 14 rubber bands used to mount the wing, while front hatch access demanded replacing a taped seal needed to waterproof the fuselage, neither task a welcome one on the d eck of an airboat. Platform Requirements and Design Goals As a result of lessons learned during initial operational deployments of the Nova 2 system, a refined platform design was deemed necessary, with a number of key performance criteria guiding the desi gn. First among which, and a key lesson in overall sUAS work, is to begin vehicle design by evaluating what data product is to be produced with the system. Almost all s UAS research programs begin with an available platform which program managers then try t o adapt to their specific needs, while often focused development (Owen 2011) Instead, program managers should determine who will be us ing the resulting data products and consult wi th the end users to determine ultimate delivery data specifications. From there, appropriate payload design and specifications should be These operational specifications can include the need for a gyro stabilized gimbaled camera, a hovering requirement for persistent observation, high altitude operations, etc. Once a suitable payload is determined, vehicle design should then proceed to maximize payload

PAGE 32

32 effectiveness and the overall mission feasib ility through range and duration maximization Moreover to improve flexibility in deployment logistics, the platform should be designed to enable operation with a minimum of runway launcher and recovery areas and equipment. With this process in mind de sign of the Nova 2.1 began with determining what payload configuration would maximize mission effectiveness for the natural resource and infrastructure monitoring missions that the program sought to execute. To deliver the high accuracy, high resolution or thomosaics identified as key deliverables, the that the camera is affixed to the fuselage and therefore has the same heading and orientation as the aircraft. To ens ure sufficient overlap for full photogrammetric coverage, this requires that the payload be maintained in a near constant nadir v iew while over the target area. In addition to specific payload deployment needs, r emo te sensing applications similarly call for sp ecialized vehicle requirements. For example, typical Intelligence, Surveillance, and Reconnaissance ( ISR ) missions require maximum duration on target and normally remain at a fixed airspeed except during brief dash periods as required by the mission. As a result, vehicle optimization centers on maximizing duration, typically at a low cruise airspeed to minimize power consumption and enable the vehicle to orbit compact target areas. However, remote sensing applications focus on maximizing total ground covera ge per flight to minimize operational downtime between flights, which heavily influences the amount of area covered by a single flight crew per day. Also, the desired airspeed is completely dependent on the imagery product specifications, and

PAGE 33

33 may range fro m a very low speed required to guarantee overlap for low altitude high resolution studies to a high cruise speed intended to minimize redundant data for higher altitude medium resolution flights. Drawing on the lessons learned from deployment of the Nova 2 a minimum cruise speed of 15 m/s was determined to adequately fulfill the minimum forward overlap betwe en exposures at a 150m altitude. M any remote sensing applications for environmental monitoring purposes involve operating in rugged terrain where suitable launch and recovery sites may be unavailable As a result, the Nova 2.1 platform specifications were developed cooperatively with the USACE and the geomatics specialists at UF to provide a small, hand launched sUA S able to be launched and recovered from an airboat operating in the swamps and waterways of S outh Florida as shown in Figure 2 5 These capabilities, unique to the Nova series of aircraft, improve operational availability without the additional complexit y of custom launch and recovery equipment or location requirements Inherent in the hand launch requirement is a sufficiently low stall speed to minimize failed launch events, which are common to many marginally hand launchable systems. Additionally, relia ble repeated operation following water recoveries was targeted as a key goal, which emphasized waterproofing of the fuselage, as well as for safe operation and recov ery of the vehicle. Finally, a push for operational simplicity lead to a requirement for simple, tool free field assembly and readily accessible components for possible troubleshooting in the field.

PAGE 34

34 Figure 2 5 Hand launch and water recovery of Nova 2.1 sUAS from airboa t on Lake Okeechobee, FL demonstrating o perational constraints imposed by t he unique operating environment.

PAGE 35

35 CHAPTER 3 NOVA 2.1 AIRFRAME Wing Design In order to obtain the desired flight characteristics established by the desig n goals, a key focus in the development of the Nova 2.1 wing design was the reduction of wing loading ( W/ S ). Although optimal airfoil selection and planform design to these optimizations if the base design is not within an acceptable window. For critical phases of flight like takeoff and landing, wing loading is a primary factor in determining minimum sta ll speed, according to Equation 3 1 (3 1 ) Additionally, wing loading determines how hard the w ing is working during cruise, with lightly loaded wings having more efficient cruise performance at low speeds than designs with higher wing loadings (Anderson 2007) Turn performance is also affected, which is a key factor affecting mission planning for remote sensing applications. Given by Equation 3 2 the minimum turn radius ( R min ) an aircraft can perform at a given ban k angle, is directly proportional to its wing loading; that is, lightly loaded wings can perform tighter turns. (3 2 )

PAGE 36

36 t in aircraft development. Because additional features and equipment inevitably find their way into the airframe, conservatively low design wing loadings help to reduce subsequent losses in takeoff and turn performance as weight is added until the final co nfiguration is reached. Because the Nova 2.1 would deploy an existing photogrammetric payload already developed for the Nova 2, the maximum payload weight could be readily calculated. To provide mission flexibility, the additional weight of a second E 420 imager was accounted for in the event that simultaneous capture of color infrared (CIR) imagery or deployment of an alternate auxiliary sensor was dictated by mission requirements. With the maximum payload weight established, an initial vehicle weight esti mate was determined by selecting approximate flight component weights and estimating total laminate weights for the wings, fuse lage and tail Solidworks was used to determine surface area in order to compute laminate weight using estimated layup schedules This resulted in an estimated maximum vehicle weight of 5.9 kg. Because the USACE did not want to increase pilot workload, flaps were omitted, minimizing operational complexity at the cost of overall flight performance. Unfortunately, without flaps to i C L,max the wing design must be focused on maintaining hand launchability, at the cost of a reduction in absolute cruise efficiency. In order to maintain an acceptably low stall speed to permit easy hand launching, the wing design must carry higher than normal planform area than if it were designed around a cruise operating point. This excessive

PAGE 37

37 area cause s higher parasitic drag, which reduces flight efficiency, especially at higher cruise airspeeds. The initial wing design, including planform and airfoil optimization, was performed using XFLR5 software, which utilizes a hybrid 3D panel/vortex lattice method (Deperrois 2011) Minimization of required wing area for takeoff motiv ated the selection of an airfoil with a relatively high section C l,max To achieve this high section C l max while retaining a relatively thin profile for higher speed cruise at lower C l values a Selig Donovan SD7032 airfoil (M. S. Selig 1989) was selected. Designed to provide high performance in a low Reynolds number (Re 350 k at cruise) regime the SD7032 has a low thickness (charact eristic of low Re airfoils) of 9.96 % and a camber of 3.66%. This relatively high amount of camber results in a C l,max of 1.45 and a minimum drag C l of 0.45 with the section drag polar presen ted in Figure 3 1. This minimum drag C l is in which its optimal performance is reached (Abbott and Von Doenhoff 1959) This value is normally used by designers to select airfoils based on their cruise lift requirements or inversely, to determine optimal operating speeds for a given wing design.

PAGE 38

38 Figure 3 1. Drag po lar of SD7032 airfoil at cruise and dash average Reynolds numbers of 350k and 700k, re spectively Because of the hand launch requirement, a low minimum stall speed of 12 m/s at maximum weight ( 10 m/s at typical loadout) was selected to maximize launch success C l, max the r esulting design wing area was calculated as 81.5 dm 2 Although a high aspect ratio wing is desirable to enhance overall efficiency by reducing induced drag especially at cruise, the Nova 2.1 was c onstrained to a 2.74 m span for storage and tra nsport purposes resulting in an aspect ratio of 9.2. With the overall geometry set, t he planform design was then revised to shape the lift distribution toward a near elliptical arrangement. Because the wing would be constructed using hot wired foam cores, which only permit straight tapered trapezoidal panels, a six panel planform was selected as a compromise to allow sufficient control over lift distribution while minimizing fabrication complexity. At the outer stations, the

PAGE 39

39 SD7032 airfoil was progressivel y thinned to 8.49% thickness and decambered to 2.98% to provide aerodynamic washout while one degree of geometric washout was also incorporated. Thinning the airfoil towards the tips helps to prevent premature separation due to the lower Reynolds numbers t hat result from narrower chord widths at these locations. These modification to move towards a more elliptical lift distribution while preventing the tips from being too small, which would negatively affect tipstall ch aracteristics as well as create section efficiency losses due to low Re ynolds number effects. The overall w ing design is pr esented in Figure 3 2 with key design parameters reviewed in Table 3 1. Figure 3 2 Nova 2.1 wing planform with XFLR5 mesh distrib ution displayed. Table 3 1. Nova 2.1 w ing d esign p arameters Wingspan b 2.74 m Planform Area S 81.6 dm 2 Design Max Wing Loading, W/S 7.23 kg/m 2 Mean Aerodynamic Chord 30.6 cm Aspect Ratio, AR 9.2 Taper Ratio 2.33 Root Tip Sweep 0.53 The overall wing performance was evaluated using a fixed lift polar analysis within XFLR5. The lift to drag ratios ( L/D ) for both the Nova 2 and Nova 2.1 are presented in Figure 3 3. For simplicity in comparing wing designs, the fuselage and tails were not included in this analysis, but were later simulated for more accurate performance

PAGE 40

40 assessments. As a result, the computed L /D ratios are higher than actual ly encountered in flight, but are still valuable for relative comparison. From inspection, the overall benefit fr higher efficiency at lower speeds. Additionally, a reduction in stall speed can be seen in the new design, as well as a downward shift in the maximum efficiency cruise airspeed. Also evident however, is the inevitable tradeoff in high speed cruise performance, m/s. Figure 3 3 Comparison of lift to drag ratio s for Nova 2 and Nova 2.1 wing designs A three piece wing arrangement was selected to enhance portability while As evidenced in the wing be nding moment diagram in Figure 3 4 at cruise, moving the wing joiner position from the cent er station ( as in a 2 piece wing) to the first panel break located 76.2 cm from the root (3 piece wing configuration ) results in an 85% reduction in wing bending

PAGE 41

41 moment from 13.7 N m to 2.1 N m. This allows for dramatically lighter reinforcement around th e wing joiner areas, resulting in simplified construction and structural weight reduction. This also aids in waterproofing during field assembly, as there is no longer a wing joint centrally located over the fuselage to be repeatably and re liab ly waterproo fed with each setup. Figure 3 4 Wing bending moment diagram illustrating 85% reduction in bending moment resulting from outboard joiner location. To provide adequate static stability about the roll axis, polyhedral was incorporated into the wing design with a flat center section and the outer wing panels raised to 5 as seen in Figure 3 5 This results in an Equivalent Dihedral Angle (EDA) of 2.86 While modest, this is sufficient to provide adequate proverse roll response to yaw inputs, which allows for effective control redundancy in the event of aileron failure. An unplanned evaluation of this redundancy was encountered during a high speed power consumption flight, during which both aileron linkages failed, leaving the undriven

PAGE 42

42 ailerons to flutter u ncontrollably. Despite the complete loss of direct roll control, adequate orientation and directional control was mai ntained using only rudder input, and a safe centerline landing was achieved. Figure 3 5 Nova 2.1 wing polyhedral configuration in front view Tail Design and Stability Considerations One of the design lessons imparted by the field deployment of the previous Nova 2 platform was the structural vulnerability presented by a T tail configuration during rough landings. The combined vertical imp act and horizontal deceleration loads typical of off field landings were exacerbated by the structural weight of the horizontal stab being located far from the junction between t he vertical fin and the tail boom. As a result, frequent fractures occurred at this junction, as well as at the stab attachment points on top of the fin. An alternate V tail configuration was selected due to a number of practical, structural, and aerodynamic reasons. Operationally, the V tail configuration raises the tail surfaces u p and helps minimize submersion and fouling during landings in either water or tall brush. Additionally, because the tail attachment point is directly located on the fuselage, the mounting is structurally tougher than a T tail arrangement. As the two V tai l surfaces combine the functions of three traditional surfaces in a traditional configuration (left stab, fin, right stab), the V tail is lighter and has less wetted area, resu lting in parasitic drag reduction. Finally, because both ruddervators are actuat ed for

PAGE 43

43 either pitch or yaw inputs, there is built in control redundancy to allow for safe recovery of the vehicle in the event of a single actuator failure. The tail sizing was chosen to provide adequate stability and damping in both the pitch and yaw axes Conventional approaches to V tail sizing have assumed a geometric projection onto the horizontal and vertical planes to compute normal tail volumes as with traditional configurations. While simple, this method fails to account for interference and endpla te effects and typically results in undersized tails with correspondingly poor stability behavior of the aircraft. Instead, the equivalent horizontal and vertical tail areas, S h and S v are computed using Equation 3 3 (Drela 2000) (3 3 ) With a V tail configuration, an additional degree of design flexibility is found in the adjustment of the tail dihedral angle t adjusting area distributions between ve rtical and horizontal planes as needed to achieve the desired horizontal and vertical tail volumes, V h and V v which are calculated using Equation 3 4 (3 4 ) As a result of this sizing, a n overall V tail area, S Vtail of 14.5 dm 2 was chosen, with a tail dihedral angle, t of 35 (110 included angle). These selections result in generous tail volumes, summarized in Table 3 2.

PAGE 44

44 Table 3 2. Nova 2.1 tail design parameters Tail Span 76.2 cm Tot al Area, S Vtail 14.5 dm 2 Tail Dihedral Angle, t 35 (measured from horizontal) Horizontal Tail Volume, V h 0.38 Vertical Tail Volume, V v 0.029 Developed specifically for low Reynolds number operations, the HT series of tail airfoils was selected to pr ovide excellent performance and low weight. At the root, a 7.5% thick HT14 foil was selected to provide sufficient thickness for structural purposes. This is progressively blended across the span to the 5.0% thick HT12 used at the tip for minimum drag and structural weight reduction. Both foils are unique in their forward positioning of the maximum thickness (HT14 at 18.9%, HT12 at 18.6%), which allows for zero deadband and fully laminar flow, enabling improved control linearity and overall drag reduction. Additionally, the forward maximum thickness position allows for large ruddervator surfaces (40% chord versus 20 25% typical) without structurally weakening the fixed portion of the stabs. These large surfaces enable improved low speed handling, which was a n issue with the Nova 2 design. The overall Nova 2.1 tail design is presented in Figure 3 6

PAGE 45

45 Figure 3 6 Overview of Nova 2.1 tail design. Finally, to provide minimum drag and optimum control efficiency during all phases of flight, the V tail height rela tive to the wing was adjusted to move the tails out of the Figure 3 7. Because the wash moves downwards during high lift phases like takeoff and landing, this adjustment primarily improves cruise efficie ncy and control effectiveness. Figure 3 7 XLFR5 streamline analysis illustrating ta il placement away from downwash and fuselage junction v ortices

PAGE 46

46 F use lage Design and Center of Gravity Considerations Mirroring the rest data collection requirements. To provide maximum mission flexibility, a readily accessible payload bay 15 cm 15 cm 28 cm was chosen to allow deployment of the current payload as well as pr oviding space for optional simultaneous deployment of a second imaging sensor for CIR or other data collection. Initial rough sizing and component placement were performed in Solidworks, with an early configuration model shown in Figure 3 8. To ensure that the CG could be maintained for a variety of mission loadouts, an extended battery tray was included to allow the flight pack to shift fore and aft to compensate for varying payload weights. Figure 3 8. Preliminary component placement for CG calculations was completed prior to final fuselage design. Like previous Nova variants, raised skids were molded on either side of the payload bay to protect the delicate optics from direct impact during operations from unimproved areas. Additionally, care was taken during the final shaping phase to design

PAGE 47

47 in a suitable hand hold area located at the CG to allow the launcher maximum control over the vehicle during flight operations in high winds. This molded in relief located below the wing can be seen in the airframe configuration overview presented in Figure 3 9 Because the Nova 2.1 was also to be used as a payload development platform, maximum accessibility was required for rapid reconfiguration. As a r esult, two large access hatches were incorporated into the top o f the vehicle, each compression sealed with thumbscrews during final assembly. Figure 3 9 Nova 2.1 complete airframe configuration. To provide adequate static stability along the pitch axis, a static margin of 5% MAC was selected, with the battery plac ement adjusted for each payload configuration This static margin is lower than typically found in similar designs in order to minimize the trim drag associated with the nose heavy condition required by such a configuration. As a result, the airframe s sti ck f ree recovery from pitch disturbances will be weakly stable implying that although the disturbance will not increase, actual pilot correction may be required to return the aircraft to level flight. Because the Nova 2.1 was designed to be

PAGE 48

48 operated compl etely autonomously in all phases of flight, a higher degree of pitch stability is unnecessary. Control Redundancy and RDS Linkages For roll control, spoilerons were incorporated into the wing at the tip panels, with the option to be raised together as spoi lers to provide improved glideslope control for short field landings. The separate actuators in a spoileron configuration also offer control redundancy in the event of an individual servo failure. Combined with the independently driven ruddervators on the V tail, this arrangement offers control redundancy on all axes, with two independently wired servos per axis. Because each actuator is wired with its own independent power and signal line, there is further redundancy in the event of wiring failure. To elim inate the electrical connection across the wing panels normally required for aileron actuation with traditional pushrod controls, a rotary drive system (RDS) linkage was employed. This mechanical servo linkage allows for rapid field assembly, with the angl ed servo actuating rod sliding into a premolded aileron pocket as the wing panels are joined. This slop free linkage, shown in Figure 3 10, allows for complete waterproofing of the servos as they are completely contained within the wing center panel and al l electrical lines run through the sealed under wing hatch.

PAGE 49

49 Figure 3 10 Schematic of aileron RDS linkage showing tool and electrical free connection The V tail linkages, shown in Figure 3 11, also employ an RDS installation, although in a straight lin e configuration which is fixed in place during construction. While more demanding during wing construction and servo installation than traditional pushrod style linkages, RDS linkages offer improved durability and drag reduction in addition to the previous ly mentioned waterproofing improvements. As found in work performed on comparably sized competition RC gliders, the drag from 4 external pushrod linkages was found to be approximately equal to the contribution from the entire horizontal tail (Hepperle 2003 ), so moving these internally results in definite aerodynamic improvements (Selig, et al. 1996). The improved durability is especially noted because the prototype Nova 2.1 employed standard pushrod linkages, which were frequently damaged during landings in tall grass/low brush. After incorporating RDS linkages, this control surface vulnerability dropped dramatically, with the only recorded linkage damage among three operationally deployed vehicles occurring during a direct impact with a tree upon landing.

PAGE 50

50 Figure 3 1 1 Tail RDS linkages Avionics Based upon lessons learned during previous deployments of the Nova 2 and other amphibious UAS protection of flight critical systems from water ingress was highlighted as a key design area (Pisanich and Morris 2002) These systems include the autopilot hardware and its associated GPS and communications modem, the Electronic Speed Control (ESC) that manages motor operation and regulates power to the actuators, and the actual serv o actuators and their associated wiring. To fully protect the autopilot while maintaining component temperatures within acceptable limits, a finned aluminum enclosure shown in Figure 3 12, was constructed with IP 67 environmentally rated connecto rs used to interface with all related hardware.

PAGE 51

51 Figure 3 1 2 Custom machined autopilot enclosure allows for complete waterproofing of critical components while maintaining adequate heat management. To further protect flight critical systems, all wiring was environmentally sealed with jacketed cable to eliminate potential shorting hazards. Finally, to minimize water exposure of any flight critical component, these items were mounted to the inside top surface of the fuselage, which also aided in cable mana gement efforts. To maintain flexibility as a remote sensing payload development platform, waterproofing for the payload was accomplished with minimal enclosure requirements beyond maintaining the environmental integrity of the fuselage and its hatches. The internal hardware arrangement and hatch locations can be seen in the populated fuselage top seen in Figure 3 13 prior to joining with the bottom fuselage half.

PAGE 52

52 Figure 3 1 3 Fuselage top populated with electronics prior to joining. Power Sys tem and C ooli ng Selection of the power system was motivated primarily by the need to maximize flight duration while minimizing system weight. To maintain hand launch capabil ity, a power loading of 220 W/ kg is required, resulting in a power requirement of 1300W duri ng takeoff and initial climbout By using a brushless inrunner motor coupled to a sealed planetary gearbox potential weight savings and efficiency advantage s can be gained over comparable outrunner motors with the acceptable downsides of increased cost an d periodic gearbox maintenance. The two options are compared below in Table 3 3. Using a three parameter model, overall motor efficiency can be estimated, with the selected Neu 1509 1.5Y and a comparable Scorpion 4025 12 outrunner (used on the Nova 2) pres ented in Figure 3 14 At maximum power, the Neu 1509 1.5Y burns 68.9 W less power as waste heat, conserving energy and minimizing cooling requirements

PAGE 53

53 Add itionally, the Neu motor is 77 g lighter than the Scorpion, presenting a 22% weight savings. Table 3 3. Comparison of motor specifications for Neu 1509 1.5Y and Scorpion 4025 12 Neu 1509 1.5Y Scorpion 4025 12 Winding Resistance, R m 0.007 0.022 No load Current, I o 2.00 A 1.80 A Kv (RPM/V) 358 (geared) 440 Efficiency at 1300 W 94.6% 89.3% Waste P ower at 1300 W 70.2 W 139.1 W Weight 270 g 347 g Figure 3 14. Efficiency plots for Neu 1509 1.5Y (inrunner) and Scorpion 4025 12 (outrunner) motor options for 18.5 V input The selection of an appropriate propeller is an inexact science, and must ba la nce conflicting requirements for static thrust, cruise efficiency, and motor loading. For static thrust optimization, large diameter, low pitch blades are most effective but limi t cruise efficiency. For cruise phase propeller efficiency calculations, the n ondimensional propeller advance ratio, J is used, and is given by Equation 3 5 :

PAGE 54

54 (3 5 ) Physically, the advance ratio describes the ratio between the distance the propeller moves through the air during a single revolution, and the diameter of t he propeller. For stable cruise flight, peak propeller efficiency occurs when the advance ratio is slightly less (~0.1) than P/D) (Durand 1927) This behavior is shown in Figure 3 15, wi th maximum overall efficiency increasing for propellers with higher P/D ratios (coarser pitch). Figure 3 1 5 Propeller efficiencies for various P/D and J ratios [Adapted from Durand, W. F. 1927. Tests on Thirteen Navy Type Model Propellers. NACA Technic al Report 237 (Page 165, Figure 28) ] Although highly pitched propellers have definite efficiency advantages during cruise, they are frequently stalled when run at full power in a static condition, delivering s ignificantly less thrust. To ensure sufficient static thrust for hand launches while maintaining adequate cruise performance, an Aeronaut CAM 15x13 (385 mm diameter, 330 mm pitch P/ D = 0.857 ) f olding prop eller was selected. This propeller adequately

PAGE 55

55 loads the selected motor to the desired input power levels and offers an operational benefit of folding back during landing to minimize potential blade damage. To power this motor/propeller combination, a Phoenix Ice 100 ESC was selected, offering onboard data logging and an onboard switching regulator to efficiently power 1 Ah rechargeable lithium polymer pack from MaxAmps, was carried over from t he Nova 2 system, as it offered superior energy de nsity and satisfied design duration requirements. Due to the unique waterproofing requirements of the system, cooling of the main heat sources in the propulsion system, the motor and the ESC, required slight departures from typical mounting procedures. Abl e to be safely submerged, the motor was mounted in front of a waterproof firewall, and was exposed to external cooling air through vents in the nose of the fuselage. Additional cooling was provided by the t the ESC from water while providing ample cooling, an external heat sink was flush mounted to the fuselage exterior, and the ESC was bonded to this with thermally conductive epoxy. The overall configuration of the power system is shown in Figure 3 16. Thi s cooling strategy proved extremely effective, with all propulsion components operating just above ambient temperatures, even during flight tests in 35 C ambient conditions.

PAGE 56

56 Figure 3 16 Nova 2.1 propulsion system, with Aeronaut CAM 15x13 folding prope ller, externally vented Neu 1509 1.5Y, and Phoenix Ice 100 external heat sink Fabrication and Assembly To minimize weight and provide maximum RF transparency, the fuselage of the Nova 2.1 was constructed using a hybrid layup of fiberglass and aramid fabri cs with epoxy resin. After prepainting the fuselage female molds a cosmetic layer of 70 gsm bidirectional E glass was first laid up followed by two layers of 170 gsm bidirectional aramid fabric. A 4 harness weave was used to help the aramid fabric confor m to the fuselage curves without excessive relief cuts. Local unidirectional carbon fiber reinforcements shown in Figure 3 17, were placed around the wing mount and hatch locations to provide adequate strength without excessive weight or RF blocking. This layup was then vacuum bagged to remove excess resin and post cured for 8 hours at 50 C for increased stiffness. By using aramid fabric for the fuselage, a light and tough structure could be created, with excellent damage tolerance. Unlike carbon, which s hatters upon impact, aramid composites tend to locally wrinkle without tearing, keeping

PAGE 57

57 damaged sections together. This enables rapid field repairs, increasing operational availability. Figure 3 17. Local unidirectional carbon reinforcements around wing mounts and hatch locations. molded into the fuselage top, depicted in Figure 3 18 (Drela 2003) This joggle off sets the overlapping fabric at the part seam to allow consistent overlap for strength and to minimize the weight of any required adhesive for secondary bonding of the fuselage halves

PAGE 58

58 Figure 3 18. Schematic of fuselage halves built with molded joggle After curing, the completed fuselage halves we re trimmed and demolded in preparation for avionics installation into the fuselage top. Because the parts were painted in the mold, the epoxy resin bonds chemically and mechanically to the paint, leaving a tough exterior surface with no further finish work required. Figure 3 1 9 Completed fuselage prior to avionics install in preparation for joining.

PAGE 59

59 The wings and tails were constructed using CNC hotwired foam cores with precut slots for joiner tubes, pitot lines, and actuator wiring. These cores were joined into center and tip panels, with reinforcements added for RDS pockets, wing bolts, and other mounting hardware then finish sanded in preparation for bagging. Strips of adhesive backed bias cut aramid fabric were then applied for leading edge reinforcement and to serve as skin hinges. The actual wing skins, 54 gsm E glass with 160 gsm unidirectional carbon reinforcements, were then la id up onto prepainted mylar carriers, as shown in Figure 3 20 These carriers were then placed over the cores and vacuum bagged to ensure adequate bonding. Like the fuselage, the wings and tails were postcured for 8 hours at 50 C. Figure 3 20 Wing const ruction picture showing selective local reinforcement with unidirectional and bidirectional carbon fabrics. Performance Evaluation Overall, the Nova 2.1 demonstrated significant performance gains over its predecessor. In its production form, the Nova 2.1 d eployed an identical payload to the

PAGE 60

6 0 Nova 2, but with a 16% lower flying weight, despite having added 21% more wing area. The resulting reduction in wing loading, summarized in Table 3 4 led to dramatic improvements in flying characteristics, especially n oted in the relative ease of hand launching due to the lower stall speed. Additionally, overall flight stability and rapid wings level correction following gust perturbations was noted, improving data collection and allowing for l ess conservative sidelap v alues and higher ground coverage efficiency. Table 3 4 Comparison of as built Nova 2 and Nova 2.1 platforms Nova 2 Nova 2.1 Wingspan, b 2.44 m 2.74 m Wing Area, S 67.4 dm 2 81.6 dm 2 Typical Weight, W typ 6.1 kg 5.1 kg Wing Loading, W/S 9.05 kg/m 2 6.25 kg/m 2 Cubic Wing Loading, W/(S 3/2 ) 11.0 kg/m 3 6.92 kg/m 3 Relative to its predecessor, the Nova 2.1 also proved to be a much cleaner aircraft, with a significantly smaller frontal area as a result of the tractor configuration. Although this has obvious b enefits for cruise efficiency, it also meant that landing approaches were very long and shallow. From a final breakout altitude of 40 m, dead stick autoland approaches often took 450 m to complete. Although manual landings could be somewhat shortened, this presented a potential problem for operations from many forested areas which motivated work to integrate flaps onto later aircraft.

PAGE 61

61 Figure 3 21 Disassembled Nova 2.1 illustrating small deployment volume Figure 3 22 Assembled Nova 2.1 aircraft

PAGE 62

62 CHAPTER 4 SYSTEM DEPLOYMENT AN D EVALUATION ISR Mission Profile For typical ISR deployments long the mainstay of sUAS applications, the operatio nal objective is straightforward: provide persistent visual intelligence over selected targets of interest. The level of mission planning necessary to achieve this objective is minimal and simply requires designa tion of GPS waypoints for each reconnaissanc e target while selecting adequate takeoff and recovery areas. As shown in the sample flight profile in Figure 4 1, mission execution typically begins with a climb to a cruise altitude (typically constrained by the resolution of the video sensors), followed by a cruise to the selected target area. To provide maximum duration for deployment of the selected sensor package, all flight is normally conducted at the minimum power airspeed, V mp Upon reaching the designated target waypoint, the sUAS will enter a co ntinuous loiter about the target, providing constant visual intelligence until an alternate target is selected or available endurance is exhausted. At that point, the sUAS then returns to the selected recovery area at V mp before descending and executing it s final recovery flight path (traditional, deep stall, parachute, etc.). Figure 4 1. Typical sUAS flight profile for ISR missions

PAGE 63

63 Because the overall flight path is relatively simple, there is little room for flight planning optimization beyond logical ordering of selected targets, either by flight distance or target priority. Additionally, because overall flight optimization focuses on maximization becomes a key objective. As a result, unless mission circumstances d ictate otherwise (e.g. brief dashes between priority targets or convoy tracking), all cruise and loiter flight phases are conducted at V mp to conserve power. This fixed operating condition also simplifies vehicle design, as all aerodynamic and propulsion optimization can be centered on maximizing performance in a narrow airspeed window. Photogrammetry Mission Profile Conversely, for aerial photogrammetry missions, flight paths are often highly complicated in order to satisfy coverage requirements. In comparison to the handful of target locations needed for a typical sUAS ISR mission, average photogrammetry flights for the Nova 2.1 system require over 60 waypoints to establish accurate flight lines that satisfy data collection specificatio ns and maintain designated image overlap. Depending on desired output resolution and the resultant flight altitude, optimal cruise speed may vary; it may be beneficial to fly as slow as possible to guarantee sufficient endlap for low altitude, high resolution missions or to fly quickly in an effort to minimize redundant data collection on high altitude, low resolution flights. This broad operational speed range, coupled with the operational requirement for a low stall speed to maintai n hand launchability, presents a significant design challenge for vehicle and propulsion optimization.

PAGE 64

64 Flight Planning Implications For sUAS based remote sensing applications, f light planning is an essential part of overall flight optimization. Although on site and on demand planning is important, the most efficient deployments are achieved with as much advanced preparation as possible. Flight plans must take into account environmental variables like wind speed and direction, and mission parameters like the orientation of the target area and desired output resolution. Finally, to improve operational efficiency, efforts should be made mission overhead, which includes take off and landing flight phases as well as any turns made at the ends of flight lines. A systematic approach to flight planning has been developed by the UF UASRG to ensure a semi optimum data collection for any situation, concluding in a flight plan that ca n be further optimized by speed to fly theory. Regulatory restrictions are a key factor in mission management and are the first real planning consideration following a n overall feasibility analysis. With some limitations around restricted areas, sUAS oper ation is possible in most classes of airspace o n a case by case basis. However, as a general rule, flight should be limited to Class G airspace. Because t he typical 200 m flight altitude for the Nova payload exceeds the 400 ft (122 m) AGL limit for sUAS im posed by the F ederal A viation A dministration (FAA) a licensed manned pilot is required to serve as Pilot in Command of the system Although somewhat cumbersome, t his requirement has the benefit of allowing the sUAS to bypass the restriction of flight clos er than 5 mi les (8 km) from any airport significantly expanding available deployment areas Additional airspace access is gained by fitting an optional altitude encoding transponder, w hich allows the Nova 2.1 system access inside t he 30 n autical m ile (56 km) veil of many larger airports around

PAGE 65

65 the country. With these two capabilities the Nova 2.1 sUAS has achieved unprecedented access to the majority of Class G airspace throughout the country with the UF UASRG spearheading regulatory acceptance of sUAS r esearch deployments The next step in the mission optimization process is dependent on final image quality parameters. The output resolution ( typically a specified ground sample distance) desired for final data deliverables is first used to select an appr opriate flight altitude for the vehicle. Dependent on the desire for 2D or 3D photogrammetric processing, the required side lap percentage is then calculated, either directly or using a chart like in Figure 1 2 to determine t he side airbase and adjacent fl ight line spacing. Once the se flight parameters are set, the actual generation of flight lines is delegated to the G round C ontrol S tation (GCS) and the flight planning software. Although the software automatically generates the precise waypoints, t he opera tor retains overall control of the process both for regulatory compliance and operational efficiency At this stage, t he preliminary flight plan consists of all necessary flight li nes extending to the FAA imposed 1 mile ( 1.6 km ) VLOS limit, and the flight lines can be translated and rotated as a whole onto any target area. The next factor to be considered in the mission planning process is target area geo metry Flight paths for optimal data collection are highly dependent on target specific geometry and ma y result in significantly different planning strategies, as shown in Figures 4 2 and 4 3 for broad area and linear targets. While areal targets frequently permit rotation of the flight lines to minimize crab angle, target areas with excessively high aspect ratios (linear features like pipelines and levees) will force the flight lines to be parallel to the longer transect regardless of wind direction. This is to minimize the

PAGE 66

66 number of turns during which data collection is not ongoing. Additional geometric co nstraints may arise from nearby obstacles or regulatory restrictions that prohibit flight over regions directly adjacent to the target area, necessitating flight parallel to the prohibited area to prevent over flight during turnaround maneuvers. Figure 4 2. Dipole flight path over areal target on Lake Okeechobee, FL with each point representing a single image exposure.

PAGE 67

67 Figure 4 3. Linear flight path over levee target with takeoff and landing spirals visible. For all other areas without over flight or aspect ratio restrictions, wind direction is the predominant geometric consideration and flight lines can be oriented in any direction. The most efficient way to process aerial mapping data for mosaicking is with images that are orthogonal to each other m aximizing image overlap and minimizing redundant data storage The it is most efficient for the flight lines to be oriented with the wind so the sensor orientation aligns with the ground track ins This ensures that the full width of each orthogonal image is available for overlapping coverage At this point, the flight plan consists of a set of parallel flight lines that will guarantee sufficient photogramm etric coverage in an improved layout with geographic and geometric constraints. The final piece of flight planning and optimization is to set turns and properly order flight lines. At normal cruise speeds and with a conservative

PAGE 68

68 maximum bank angle of 35 t he Nova 2.1 has a minimum turning radius of about 75 meters This simple performance measure is by far the most influential in determining optimal flight line sequencing. In an ideal flight plan all end of line turn s finish with the UAV established in a wings level position aligned with the next flight line, immediately prior to entering the mapping target area This ensures the area boundaries and eliminates excessive collection of non target data. To accomplish this, a va riety of flight patterns are available, with the most common dipole and parallel track options shown in Figure 4 4. If the flight line spacing combined with downwind drift is less than the specified minimum turn radius then the dipole flight pattern takes priority allowing the UAV to skip one or more adjacent flight line s sacrificing the additional flight distance during the turnaround to prevent the UAV from overshooting its desired path If the combined spacing is greater than the turn radius of the ai rcraft, th e n the parallel track flight path a common pattern for aerial searches, takes priority (Santamaria, et al. 2008) The most efficient option due to its short turnarounds, the parallel track flight pattern is the prefe rred arrangement, and is used for most flights unless very low altitudes (and thus, small flight line spacing) are required. In actual implementation, the flight planning software will take the forecasted wind direction and magnitude into account and proje ct the additive/deleterious effects on the turn radius. The flight management algorithm then takes the required array of flight lines from the preliminary flight plan and orders them as called for by the specified flight pattern.

PAGE 69

69 Figure 4 4. Flight lin e patterns for remote sensing applications. On site, if the flight team determines that the wind has diverged significantly from the forecasted direction or magnitude, then a strategy is put in place for in flight assignment of the flight plan. Once launch ed, the aircraft will enter a circular orbit about a preselected takeoff point. After making a complete orbit, the autopilot is then able to accurately estimate wind direction and magnitude. At this point, updated parameters can be automatically fed into t he flight planning software and a new flight plan properly oriented into the wind can be uploaded instantaneously. Finally, because the autopilot continuously updates its estimations of the winds aloft, any significant changes in conditions can be easily dealt with by uploading a new automatically generated flight plan, saving significant time and effort over manual adjustment by the ground station operator. Regardless of an y in flight wind direction change, the final step to flight optimization occurs on ce airborne. The speed to fly theory explained in later sections sets the desired airspeed for each upwind and downwind fligh t line for maximum

PAGE 70

70 efficiency. The overall flow for flight plan development is summarized below in Figure 4 5, with relevant parame ters indicated for each design stage. Figure 4 5 Flight plan design procedure and relevant parameters Flight Planning Optimization In addition to ensuring sufficient photogrammetric coverage through appropriate arrangement and ordering of flight lines, significant operational efficiency gains can be made by minimizing mission overhead. Loosely defined as any flight time not spent collecting data over the target area, mission overhead includes takeoff, climb, descent, and landing phases, transit between the target area and the launch/recovery site, and any orienting turns required to align the UAV with its target flight lines. Without active effort to minimize overhead, it can easily exceed 50% of overall flight time. Even after months of system deploymen t, non optimal path s like that shown in Figure 4 6 were used, highlighting the gains still to be made by minimizing mission overhead. Imagery Parameters Ground Sample Distance Altitude % Side Overlap Flight Line Spacing Geometric Parameters Target Geometry Bounding Box of Flight Restricted Airspace Direction/Location of Turn Points Environmental Parameters Wind Direction Orientation of Flight Lines Wind Direction and Magnitude Sequencing of Flight Lines Speed To Fly Theory Commanded Airspeed

PAGE 71

71 Figure 4 6 Parallel track flight plan for mapping road degradation in Picayune Strand State Forest, FL. To fully optimi ze total mission overhead, power consumption during all flight phases must be addressed. A normal flight profile for autonomous aerial mapping flights, shown in Figure 4 7, has significant room for improvements to reduce overall time and power consumption. For takeoff and initial climb out phases, best climb speeds should be used instead of more typical full throttle maneuvers in order to minimize the total energy expended getting to cruise altitude. Additionally, while the initial takeoff heading is determ ined first by local site factors (power lines, trees, roads) and then by wind direction, the traditional spiraling climb to altitude should be supplanted by a direct ascending route to the first flight line, minimizing the time and power needed to transiti on to the data collection portion of the flight.

PAGE 72

72 Figure 4 7 Flight profile for a typical sUAS based aerial mapping mission As discussed previously, routing of turns between subsequent flight lines was initially done by hand, and excessive margin was le ft at the ends of the area boundaries to guarantee adequate alignment of the payload prior to entering the target area, as shown in Figure 4 6. By more aggressively routing necessary turnaround maneuvers, the automated flight planning software will minimiz e mission overhead while maintaining necessary overlap parameters. During mapping projects over large continuous areas where multiple flights are needed, it is not possible to readily calculate the exact number of complete transects available per flight du e to various environmental un certainties. As a result, it becomes important to accurately estimate remaining battery capacity throughout the flight in order to maximize available coverage while preventing over discharge of expensive system batteries. If su fficient capacity is available to completely cover the given target area, however, the mission should be flown as fast as possible (while maintaining endlap minimums) in order to maximize the number of possible flights per crew, per day. One final note for overall mission efficiency regards the final descent and landing phases of flight. Instead of returning to the recovery area at the data collection altitude, burning precious energy that could have added additional transects, the initial descent

PAGE 73

73 should be gin as the UAV completes its final flight line, allowing a power off glide to the landing area. While not practical in areas where minimum altitudes must be maintained to clear obstacles or terrain, this final optimization provides an additional target for improving total flight efficiency. A more direct optimized mission profile is shown in Figure 4 8, demonstrating the gains possible through the elimination of mission overhead. Figure 4 8 Optimized flight profile with more aggressive takeoff/climb ro uting and a power off descent to land ing Operational Feasibility of sUAS for Remote Sensing Applications While 1 hour flights covering 500 m 500 m target areas are acceptable for preliminary research applications, true operational feasibility of batter y powered sUAS for remote sensing applications will require, at a minimum, full coverage of 2 km 2 km targets. This will maximize areal coverage within current FAA mandated 1 mile (1.6 km) VLOS restrictions, and provide operators with the ability to map 1000 acres (40 0 ha) of Although this endurance level is certainly achievable for sUAS with current battery technologies, a design challenge is brought on by a ma average weight and the desire to maintain hand launch capability. Because the scope of typical system deployments often exceed single flight coverage, with multi day missions

PAGE 74

74 common, minimization of the required time for each fl ight is critical to increase flight crew effectiveness. In addition to actual flying time required for each mission, significant ground intervals may be required for system mobilization to new target areas, excluding a normal 10 minute airframe turnaround. As a result, both mission planning and airframe optimization should focus on maximizing coverage efficiency with respect to both time and power requirements. Ultimately, this leads to ward a need to widen the available speed range of the sUAS to provide a high optimal cruise speed for rapid coverage while maintaining operational flexibility with a sufficiently low stall speed to permit hand launching.

PAGE 75

75 CHAPTER 5 SPEED TO FLY APPLICA TION Theory Because hand launched sUAS tend to have lower cruise airspeed s ( 15 20 m/s), wind conditions play a large role in determining their ultimate ground coverage. Additional endurance constraints arising from current electric propulsion systems mean that unlike traditional wet fueled platforms, operators cannot simply fue l up more and fly longer in significant wind conditions. Taking into account heavier optical payloads associated with aerial mapping applications, little margin remains for additional battery weight. As a result, a novel flight planning strategy is needed to aid in path optimization for remote sensing missions. For parallel track flight plans, with flight paths always being flown directly into or with the wind, the wind interaction is simplified to a purely additive or subtractive term for ground coverage. Neglecting turns, altitude changes, and small disturbances, this further simplifies the optimization problem to total ground coverage with vehicle speed and wind speed as the only variables. With the ultimate goal being to reduce overall energy expenditu re per linear kilometer flown in order to yield the highest number of mapping transects (defined as one pair of upwind and downwind legs across the target area), an analogous approach was found in another branch of aviation familiar with limited power sour ces: cross country soaring Without a direct motive source, glider performance is extremely sensitive to energy management. For cross country gliding flights, maximizing distance traveled with the potential energy (altitude) gained by climbing in thermal u pdrafts is a primary concern and this optimization is dramatically affected by prevailing winds. As a result, glider pilots use a speed polar to determine the optimum speed to fly ( STF ) in

PAGE 76

76 headwind or tailwind conditions (Federa l Aviation Administration 2007) In a similar fashion, the power curve of a powered aircraft can also be used to calculate optimum range airspe eds in variable wind conditions (Phillips 2004) To perform this analysis, a samp le Nova 2.1 power curve, presented in Figure 5 1, was generated using XFLR5 generated drag estimates and assuming 100% propulsion efficiency for simplicity. Immediate visual analysis of this sample curve presents two important data points. The first is the minimum power consumption airspeed, V mp ( mp = minimum power), which is the commanded airspeed to be maintained in order to maximize level flight duration. This airspeed remains constant independent of wind conditions and requires a minimum input power le vel, P min to be maintained. This is the optimum cruise airspeed for zero wind conditions and is typically maintained during most sUAS missions, especially for ISR applications. The second, less obvious value is the point at which a tangent line drawn from the origin intersects the power curve. This line represents the minimum energy consumption per unit distance, which is given by the slope of the tangent line (= P/V [J/m or Wh/km ]) For convenience, this value will be termed the Distance Specific Energy C onsumption (DSE C ). The abscissa at the point of intersection corresponds to the airspeed value termed V mr ( mr = maximum range), which yields the optimal range. To maintain this airspeed, an optimal input power value of P mr is required. In still air conditi ons, where the airspeed and groundspeed are equal, such analysis holds true. In headwind or tailwind conditions, however, the calculations proceed slightly differently.

PAGE 77

77 Figure 5 1 XFLR5 generated power curve (assuming 100% propulsion efficiency) for N ova 2.1 illustrating graphical computation of optimal STF in still air. By flying at V mp ( mp = minimum power), power consumption is held at a minimum and maximum duration is achieved. However, flying at V mr ( mr = maximum range), minimizes the Distance Spec ific Energy Consumption (given by the quotient P/V ) and maximizes the distance traveled per unit of input energy. In pure headwind/tailwind conditions, the effective groundspeed is found through a simple additive relationship with the airspeed and wind sp eed, with headwinds reducing effective groundspeed. Graphically, this results in a positive shift of the tangent line along the x axis, corresponding to the magnitude of t he headwind, as seen in Figure 5 2 To compute V mr corresponding to a 10 m/s headwind (typically the highest wind speed to permit safe operation) the x intercept is shifted 10 m/s right and a new tangent has had a negative effect on efficiency, increa sing the DSEC according to Equation 5 1: (5 1) P min P mr V mp V mr

PAGE 78

78 Additionally, the new V mr value is found to be higher than the original still air figure, V* mr indicating a need to accelerate into a headwind for increased ground coverage efficiency. Figure 5 2. XFLR5 generated No va 2.1 power curve illustrating application of STF intercept shifts to the right indicating a loss in effective groundspeed and the new optimal airspeed, V mr is found to be higher than the still air val ue, V* mr Although increasing the commanded airspeed on upwind passes does increase the immediate energy consumption rate, the overall energy expenditure for the flight line is lower if flown at V mr A simple thought experiment of a worst case scenario bea rs this out. Assume that a Nova 2.1 is flying at 16.5 m/s into a 15.5 m/s headwind. At a resultant groundspeed of 1 m/s it will take 100 s to cover 100 m, burning 4500 J. If that same Nova 2.1 were to fly at the upper limit of 30.5 m/s into the same 15.5 m/s headwind, the groundspeed would become 15 m/s. The sUAS now covers the same 100 m flight line in 6.6 s, burning only 957 J. Although the wattage required to maintain the faster airspeed is around four times higher, the overall energy consumption for th e

PAGE 79

79 flight line is lower. In addition to the obvious power consumption benefits, flying at higher speeds into the headwind also improves the time efficiency of data collection, as the time lost flying slower upwind without STF is not totally recovered during subsequent downwind legs. This effect will be discussed further detail in a later section. Tailwinds will have the opposite effect, increasing effective groundspeed and necessitating a negative shift along the x axis corresponding to the tailwind speed. A s a result, V mr will decrease from its still air value and the tangent line will become flatter, indicating an efficiency gain with lower DSEC As a rule, V mr will always increase into headwinds and decrease into tailwinds. Although ever increasing headwin ds will necessitate higher and higher V mr values until increasing tailwinds will not. Due to the parabolic shape of the power curve, V mr values in increasingly strong tailwind conditions will decrease unt il they reach the lower bound indicated by V mp the minimum power airspeed. Further increases in tailwind speed will not necessitate further deceleration, but will still result in ground coverage improvements because of the increased groundspeed occurring at the minimum power level. In the unlikely event that the tailwind enhanced groundspeed exceeds maximum allowable airbase limits, it will become necessary to fly slower than V mp in order to maintain adequate image overlap. Crosswinds are compensated for b y computing the effective geometric headwind/tailwind component and proceeding accordingly on the power curve (Arrow 1947) Trailing Edge Control Because the high wind (10 m/s) V mr for the Nova 2.1 fell near the maximum stable airspeed in the current flight envelope, an effort was needed to expand the high speed

PAGE 80

80 cruise capability of the sUAS to increase effective ground coverage in high wind conditions. Additionally, for lower resolution, high altitude mapping flights, it is mor e operationally efficient to fly as quickly as possible since forward image overlap is sufficient even at high speeds. This serves to maximize crew efficiency as it allows faster area coverage with the goal of enabling more flights per day on large survey projects (J. H. Perry 2011) With a fixed trailing edge, the options for improving high speed cruise for hand launched sUAS are limited. Because the wing design must use a higher cambered airfoil to minimize launch incidents while maintaining a reasonable planform area, the minimum drag section C l is inevitably higher than desired for high speed cruise. As a result, the majority of the flight is spent in a suboptimal flight regime, paying a penalty required to satisfy the requ irements of the first few seconds of flight. Although flaps were omitted on the original spec Nova 2.1 to save weight and operational complexity for unskilled operators, they were selected as a simple avenue to optimize the wing design for the wide airspe ed range required for remote sensing applications. To provide a more direct comparison between wing designs, the wing planform was held constant and a simple airfoil substitution was made replacing the original SD7032 with an MH32 foil, originally designe d for pylon racers (Hepperle 1991) The two airfoils are visually compared in Figure 5 and camber immediately evident. Interestingly, man y of the same design criteria for racing airfoils are common to remote sensing sUAS, namely a similar Re regime, low drag when flying fast, and strong performance when worked at high angles of attack.

PAGE 81

81 Figure 5 3. Visual comparison of SD7032 and MH32 airf oils. An immediate inspection of the cruise (Re = 350k) drag polars in Figure 5 4 reveals that although the SD7032 benefits from a higher C l,max in a clean configuration, the MH32 performs much better in a high speed cruise regime ( C l vs. 9.96 %) and camber (2.40% vs. 3.66%), which reduces its minimum drag C l to ~0.2 from ~0.5 for the SD7032. Also, the MH32 is designed to be highly responsive to camber change, a trait n ot shared by the SD7032. This provides a significant benefit to STF optimization, because it provides an opportunity for fine wing adjustment for each commanded airspeed, rather than requiring large trailing edge deflections to established coarsely spaced flight modes.

PAGE 82

82 Figure 5 4 Comparison of drag polars (Re=350k) for original SD7032 and MH32 airfoils. By utilizing full trailing edge control (adjustable camber/reflex across inboard flaps and outboard spoilerons ), the airfoil can then be dynamically m odified in flight to suit mission specific airspeed requirements. For low speed missions, the trailing edges are deflected down to provide increased camber and improved duration. As higher airspeeds are required, the trailing edge can be raised into a refl exe d position, dropping excess camber and reducing drag for improved headwind penetration. Finally, these flaps are also deployed in a more typical manner to improve takeoff and landing performance and regain the high C l,max offered by the original SD7032 configuration. A number of deflection geometries and their associated flight modes are presented in Figure 5 5.

PAGE 83

83 Figure 5 5 Trailing edge configurations for various flight mode settings. With independently adjustable inboard and outboard camber setting s there is sufficient flexibility to finely tune the spanwise lift distribution to optimize the wing for each commanded airspeed or flight condition. This optimization was performed in XFLR5, with preliminary flight conditions focused on providing reduced power consumption in float, crawl, and dash modes. Independent inboard and outboard flaps also improve landing performance over simple inboard flaps or even full span flaperons by enabling a butterfly landing configuration. This drops the inboard flaps to + 60 while raising the outboard spoilerons 15, generating a strong vortex at the panel junction in addition to the drag generated by the inboard flap deflection. The reduced effective angle of attack at the tip panels forces the center portion of the wi ng to stall first, promoting a straight ahead stall that is much easier to recover from than a low altitude stall spin event. To recover the low speed, high lift performance provided by the previous SD7032 configuration, a takeoff flap setting was used, in C L,max and preventing an increase in stall speed. Although the small camber and reflex deflections associated with the float and dash flight modes are not immediately visible, the larger flap changes for the takeoff and brake mo des are shown in Figure 5 6, illustrating the flexibility provided by a four surface wing.

PAGE 84

84 Figure 5 6. MH32 wing in clean, takeoff, and brake configurations. Camber change is especially noticeable at the inboard flap edges. XFLR5 was used to generate per formance polars for a range of potential wing configurations, applying a fixed lift (Type 2) analysis using the Nova 2.1 flying weight to determine cruise characteristics. Of the 30 flap configurations tested, three initial flight modes were selected, in a ddition to takeoff and brake configurations These updated flap configurations summarized in Table 5 1, broader speed range, improving slightly on the original power requirements in slow cruise while providing signi ficant reductions in high speed power consumption, especially in the reflexed dash mode. Although minor percentage gains could

PAGE 85

85 theoretically be made by further subsetting the low speed flight regime with intermediate levels of camber, this was deemed unnec essary as the very minor gains only occurred across narrow speed ranges ( 0.5 m/s) Additionally, while significant gains stand to be made at higher airspeeds ( > 30 m/s) with increasing levels of reflex, the current power loading of the Nova 2.1 precludes f light at these speeds, primarily due to limitations in propeller pitch speed and Table 5 1. MH32 wing configuration parameters with flap deflections given in degrees, with positive values indicating downward deflection Fla p Mode Airspeed Range Inboard Deflection Outboard Deflection Brake +60 15 Takeoff +10 +8 Float 9 1 3 m/s +3 +2 Clean MH32 1 3 21m/s 0 0 Dash 21 30 m/s 1 1 The computed drag polars for the original SD7032 configuration and selected MH32 flight mo des are presented in Figure 5 7. Immediately evident is the significant drag reduction of the clean (no camber or reflex) MH32 configuration over the original Nova 2.1 wing at higher airspeeds (>16 m/s), with further improvements available with the selecti on of the reflexed Dash configuration. Conversely, relatively uniform performance between the SD7032 and MH32 configurations is noted throughout the middle speed range (12 16 m/s), with the SD7032 having superior performance in the low speed reg ime due to its higher base camber level.

PAGE 86

86 Figure 5 7. Computed drag curves for original SD7032 wing and MH32 wing in clean, float and dash modes demonstrating improved performance in high speed cruise. These independent flight modes were then combined in to a composite power curve, with the global minimum power consumption taken from each of the independent modes. The comparison between the original SD7032 configuration and the updated MH32 composite configuration presented in Figure 5 8 illustrates the dr amatic effect dynamic camber adjustment can have across the flight envelope. The small upward kinks in the composite power curve between configurations indicate areas where intermediate adjustments can be made to further tailor the wing camber to better ma tch the flight regime.

PAGE 87

87 Figure 5 8. XFLR5 generated power curves for original SD7032 wing and composite curve for selected MH32 flight modes. The full selection of flight modes has been omitted for clarity. Even without a complicated array of camber conf igurations, significant gains to flight efficiency can still be made by using only a few simple flight modes, and are readily integrated into most autopilot systems with minimal difficulty and mechanical complexity. These modifications not only improve glo bal flight performance across the airspeed envelope, but prove especially beneficial for remote sensing applications where STF flight management strategies demand significant airspeed variations. Initial Power System Model While preliminary power models g enerated using computed drag values and 100% propulsion efficiency assumptions are sufficient for comparative analysis between wing designs, greater fidelity is required to optimally integrate STF strategies Towards this end, accurate power curves must be generated that incorporate appropriate efficiency models for all propulsion related systems. Initially, efforts were made to determine the se

PAGE 88

88 power models computationally. U sing a model described by the block diagram in Figure 5 9 estimated thrust require ments were combined with individual efficiency curves for the propeller, motor, and ESC to develop an overall relationship between cruise airspeed and required input power. Figure 5 9 Power system block diagram. Because lit tle to no efficiency data are provided by manufacturers for any of the propulsion system components, a static thrust stand was constructed to provide a single model for the combined performance of the Nova 2.1 propeller, motor, and ESC The results of th e static thrust test are shown in Figure 5 10 with a maximum static thrust value of 31.1N yielding a takeoff thrust to weight ratio of 0.62. With the peak thrust significantly lower than predicted by the propeller RPM at that throttle setting, this indica tes that the highly pitched 15x13 propeller is stalled, and limit s corr elation with in flight behavior.

PAGE 89

89 Figure 5 10 Static thrust test results for Neu 1509/1.5Y (6.7:1 GB), Phoenix Ice 100, Aeronaut CAM 15x13 f olding prop. Full estimation o f the propeller efficiency at various airspeeds is a complicated affair requiring collection of thrust and power measurements to determine appropriate performance coefficients, C T and C P Used in Equations 5 2 through 5 4 these parameters also re quire kno wledge of advance ratio, J which must be collected in flight. As a result, thrust values found for static tests are insufficient for accurate efficiency calculations. Propeller thrust coefficient, C T (5 2 ) Propeller power coeff icient, C P (5 3 )

PAGE 90

90 Propeller efficiency, (5 4 ) The shortcomings of the power estimation were not limited to an inadequate propeller model. Combining the thrust stand results with the computed drag curves p reviously described in Figure 5 7, an estimated power model for the Nova 2.1 was generated, and a test flight was conducted for validation purposes. During the flight, autopilot stabilized flight lines were flown at fixed airspeeds, and the level flight po wer consumption was measured. experimental data, shown in Figure 5 11, demonstrate the significant disagreement between simple calculations and real world results. Figure 5 11. Comparison of com putational power curve model with experimental values for original SD7032 wing.

PAGE 91

91 Although computational models like those used in XFLR5 nearly always under predict drag, it was found that the primary contributor to the experimental disagreement with the mo del power curve was the frequent and irregular power spikes encountered as the autopilot modulated the throttle to maintain each commanded airspeed. While this can be somewhat minimized through optimal tuning of throttle slew rates, it comes with the trade off of poorer airspeed and altitude tracking. The irregularity of these thermal up/downdrafts, prevented the determination of an accurate correction function. As a result, all further efforts turned to generation of experimentally obtained power curves that better match real world performance. Experimental Power Consumption Testing To gather in flight power consumption information, an Eagle Tree Systems eLogger V3 was utilize d to record system voltage and current draw Connected in line just after the system battery, this configuration allowed the additional monitoring of payload and autopilot baseline power requirements that, although minor, have measurable im pacts on overall system duration. These values were recorded at a 2 Hz rate and synchronized with logged RPM data from the ESC and telemetry recorded by the Kestrel autopilot after each flight. Experimental power data for the Nova 2.1 system was collected by flying numerous linear transects at constant, autopilot controlled airspeeds while energy consumption rates were measured. In keeping with FAA mandated VLOS restrictions, the transects were limited in length to 1.6 km, as shown in the ground station sc reenshot in Figure 5 12.

PAGE 92

92 Figure 5 12. Power testing flight plan consisting of repeated 1.6 km long parallel transects. Care was taken to minimize the effects of thermal updrafts and, following a linear regression on the initial data set, a three sigma re jection criteria was utilized to reject possible outliers. These flights were also performed with the payload running, which contributes a constant power draw of ~6.5 W, giving real world numbers that are much more valuable for practical flight planning pu rposes. Including reflights due to thermal interference, a total of 6 data collection flights, each roughly 50 minutes in duration, were flown to collect power data for the SD7032, MH32 Clean, MH32 Float, and MH32 Dash configurations. Because both test win gs fit interchangeably with the Nova 2.1 fuselage, the remainder of the system configuration could be held constant between flights to provide a direct comparison. A sample power profile for an MH32 Clean test flight is presented below in Figure 5 13. The commanded airspeeds for this flight are also plotted to indicate the relative timing of each airspeed

PAGE 93

93 window, and the stable stairstep behavior of the power indicates acceptable overall tracking. The spikes visible at the ends of each airspeed interval cor respond to required accelerations, while the brief peaks at the middle of each window are correlated to the 180 turnaround at the end of the transect. Figure 5 13. Airspeed intervals for MH32 clean configuration flight. To process the collected power d ata, e ach commanded airspeed was considered achieved when the autopilot telemetry indicated that the aircraft had reached the new commanded airspeed and held steady within 0.5 m/s while not varying altitude by more than 2 m for at least 2 seconds to reject over and undershoots as the airspeed stabilized. When this was not achievable, typically due to phugoid instability at the high and low ends of the airspeed range the data for that airspeed was omitted and that cordingly. This limit is mainly a result of the Kestrel

PAGE 94

94 The experimentally derived power consumption data, presented in Figure 5 14, compares the original wing with selected configurations of the updated flapped variant. As expected, the updated base MH32 wing showed improved high speed performance over the original design, even expanding the available ai rspeed envelope from 24 m/s to 30 m/s before pitch instability or climb faults became a factor. Additionally, reduced high and low speed power consumption was achieved in Dash and Float flight modes. Decent agreement between the predicted crossover points on the power curves was observed, as the Dash configuration began surpassing the base MH32 wing at 22 m/s (21 m/s was predicted). Similarly, the performance transition towards the Float configuration occurred at 17 m/s, slightly faster than the predicted 1 3 m/s. Figure 5 14 Experimentally derived power consumption results compared for original Nova 2.1 wing and modified MH32 configurations. Surprisingly, even the low speed performance of the SD7032 showed much poorer performance than the updated wing ev en after reflights to confirm the unexpected result These higher power consumption rates are associated with poor airspeed

PAGE 95

95 tracking observed as significant throttle variations that occur even during straight and level transects. This unsteady throttle in put, with variations of up to 40%, was much higher than observed during flights of the updated wing, where level flight throttle inputs were frequently held 5%. The resulting power data, shown in Figure 5 15, would initially hint at poor tuning. This is s urprising, considering that the throttle tracking is even noticeably poor in th e middle of the airspeed range, which is the regime aroun d which the Kestrel autopilot was initially tuned. Figure 5 15 Airspeed intervals for SD7032 configuration flight s howing much noisier data and a climb fault occurring during a 24 m/s transect. By combining airspeed data with logged RPM values throughout the flight, propeller selection for cruise optimization can be evaluated by computing the average operating advance ratio, J encountered during normal airspeeds for remote sensing P/D ratio should approximately match J (~0.1 less according to Figure 3 15) in order to provide the optimum propulsive efficiency. Wit h variation in J across the airspeed range shown for a clean MH32

PAGE 96

96 configuration in Figure 5 16, the average advance ratio for all MH32 flight modes was found to be 0.77. With the nearest propeller choices being 15x13 ( P/D = 0.857) and 15x10 ( P/D = 0.660), the selected Aeronaut 15x13 CAM propeller is the best available option for cruise optimization. Finally, although the 15x10 would offer a considerable static thrust advantage for hand launch improvements and may be better suited for slower ISR flights, the higher speeds available for STF integration confirm the 15x13 as a superior choice for sUAS based remote sensing applications. Figure 5 16 Variation of advance ratio, J across airspeed range for MH32 clean configuration. STF Application To verify the effectiveness of STF application to flight management, ex perimental power curves for the clean MH32 and camber adjusted variant s were used, allowin g simultaneous evaluation of predicted improvements to STF due to integration of camber change. To better il lustrate the deleterious effects steady winds have on aerial mapping flights, simulated transects over a 2 km long target area will be evaluated for both time

PAGE 97

97 of flight and energy requirements. First, the experimental power curve for a fixed camber MH32 w ing presented in Figure 5 17, is considered. Figure 5 17 E xperimentally obtained MH32 power curve with still air DSEC shown. From inspection, V mp in this configuration falls at 12 m/s, with a corresponding 122.4 W average power draw. In zero wind cond itions, V* mr for this particular wing falls at 16 m/s, yielding a DSEC of 2.49 Wh/km. A sample pair of 2 km long transects (upwind and downwind legs) flown in these conditions will require 4 minutes, 10 seconds and 9.96 Wh of energy to complete, as shown in Figure 5 18.

PAGE 98

98 Figure 5 18. Schematic of flight performance for 2 km transects in still air conditions detailing time of flight and power consumption. If, for the same configuration, a steady 10 m/s wind were encountered and the commanded airspeed went unchanged in response, significant time and energy penalties would ensue as summarized in Figure 5 19 During the upwind leg, effective groundspeed will be reduced to 6 m/s, dramatically increasing the time of flight and raising the DSEC to 6.63 Wh/km, a 266% increase over the still air figure. Although the downwind leg will proceed much faster, with an effective groundspeed of 26 m/s, the overall time of flight for the pair of transects is increased over the still air figure by 65%. Similarly, the reduced DSEC of the downwind leg fails to make up for the losses encountered during the upwind transect, resulting in a total power consumption of 16.32 Wh, an increase of 64%. Additionally, the high groundspeed encountered on the downwind leg may exceed minimum airbase requirements, resulting in still larger efficiency losses as only upwind legs become usable for data collection.

PAGE 99

99 Figure 5 19 Schematic of flight performance for 2 km transects i n 10 m/s steady wind conditions, illustrating efficiency losses wit hout STF application. To minimize these losses, application of STF strategies begin by reexamining the aircraft power curve. With the example 10 m/s headwind, the new value for V mr rises to 20 m/s, with a headwind reduced DSEC of 5.69 Wh/km, presenting a 14% improvement over the previous non STF upwind case. For the downwind leg, STF application requires that the Nova 2.1 now fly at V mp (12 m/s). The resulting flight plan with relative gains over the non STF case is presented in Figure 5 2 0 By applying ST F strategies, a total time of flight reduction of 29% and a total energy savings of 11% were achieved relative to the unmodified flight plan, demonstrating significant operational advantages to the technique.

PAGE 100

100 Figure 5 2 0 Improved time and ground cover age efficiencies for 2 km transects in 10 m/s steady wind with STF application. Through variable camber integration into STF application, minor additional gains in both time and energy consumption will also be possible. For this analysis, the composite MH 32 power curve in Figure 5 2 1 is used to compute appropriate flight speeds for both the headwind (red line) and tailwind (black line) conditions. At both airspeeds, the lower position of the composite power curve demonstrates an efficiency gain over the fi xed camber MH32 variant.

PAGE 101

101 Figure 5 2 1 Demonstration of performance gains using variable camber enhanced STF on MH 32 composite power curve. During the upwind leg of the transect, the wing is set in its Dash position, enabling the Nova 2.1 to penetrate up wind at a V mr of 24 m/s. Although only slightly more efficient than the fixed camber case with respect to energy consumption, this increased speed reduces the time of flight by nearly a minute, and contributes to the entirety of the 20% flight time reducti on for this example. During the downwind phase, the flight speed remains the same, pegged at a V mp of 12 m/s. However, because the wing can be set into the more efficient Float configuration for this airspeed, a small power savings can also be obtained. As summarized in Figure 5 2 2 although camber adjustment offers only slight reductions in overall energy consumption (3%) over the fixed configuration STF example, the increased airspeed range results in significant time savings, so much so that the total ti me is less than that of the initial zero wind case.

PAGE 102

102 Figure 5 2 2 Schematic of flight performance with variable camber enhanced STF application. Although STF with variable camber has the greatest effect with high wind speeds, it can still offer modest (6 .5%) reductions in DSEC even in zero wind conditions. In addition to the additional ground coverage offered by STF integration, a somewhat unexpected benefit is found in the dramatically reduced time of flight for mapping operations. For large projects, th is is hugely important, as it reduces total deployment time, offering significant logistical and financial incentives With variable camber and application of STF now justified for both upwind and downwind components, the questio n arises of scalability. Are STF strategies practical for commercial or civil aviation? Probably not, since the cruise speeds for both are typically much higher than the prevailing wind speed for any given flight. While the theory still holds true, the efficiency gains of using STF a re marginal for a vehicle whose V mp is often an order of magnitude higher than the prevailing wind. Additionally, unlike

PAGE 103

103 civil and commercial flights whose routes are predefined based on destination optimal aerial mapping flight plans orient all flight li nes in the same direction of the wind, allowing for maximum implementation of STF theory. Furthermore, the small distance scale of an sUAS flight causes the local wind field to appear semi constant, whereas the effective wind field for a commercial flight will be much more varied, lending itself to alternative flight planning strategies to take benefit of regionally favorable wind conditions (Rubio and Kragelund 2003) An additional consideration for the suitability of STF strat egies is the degree to which power consumption increases with speed. The sharper the power consumption curve, the larger the effect will be on the slope of the energy per unit distance line. The 1/x efficiency curve for electric power systems is quite diff erent from wet fueled engines, penalty for off peak operation. As a result, STF is most effective when V mp is within an order of magnitude of the wind speed. This is most pr onounced in gliders, H igh Altitude Long Endurance (H ALE ) UAS, and sUAS applications. Surprisingly, even migratory birds display this behavior of slowing down to V mp in tailwinds and accelerating to V mr in headwinds (Alerstam 1990 ) Autopilot Integration of Variable Camber for STF Integration of STF theory into the current flight path navigation algorithm is relatively straightforward. The same avionics equipment mentioned above is used, with the exception of custom firmware for the to give more precise control of both inboard and outboard control surfaces. With this firmware, complete independent control of each surface is granted to the autopilot, essentially creating an infinite number of discrete configurations (over 210 100 with integer resolution deflection settings).

PAGE 104

104 In an effort to prevent this flight management strategy from creating a controls issue, f lap integration wa s accomplished with an open loop controller for surface deflection This meant that flap defle ction was not continuously varied in response to normal airspeed fluctuations, but remained in specified configurations based on commanded airspeeds. For additional safety latching was implemented to prevent inadvertent flap re configurations that are dang erous to flight. In this approach, a discrete controller is used with cruise, acceleration, and deceleration modes. During acceleration mode, when the desired airspeed is more than 4 m/s greater than the current airspeed, the flaps will incrementally retra ct as airspeed increases. This prevents the flaps from retracting prematurely and causing a stall. Conversely, in deceleration mode, when the commanded airspeed is more than 4 m/s less than the current speed, the flaps will incrementally extend as airspeed decreases. This also prevents a premature stall from commanded airspeeds lower than the flap s deployed s tall speed. Once the vehicle is within 4 m/s of the desired airspeed, the cruise latch is triggered, automatically transitioning the aircraft out of ac celeration or deceleration mode. During cruise mode, once the flaps are in their airspeed appropriate configuration, they will not move until a different commanded airspeed is given. Using this method, each discrete airspeed or airspeed range can be paired with an optimized flap setting to provide improved cruise efficiency to the limit of the wing design. This implementation strategy avoids the undesirable positive feedback loop associated with closed loop flap control. Without this strategy, wind gusts or other perturbations could cause undamped oscillations as shown in Figure 5 2 3 If a N ova 2.1 were cruising at a desired airspeed of 15 m/s and a perturbation occurred to cause the

PAGE 105

105 airspeed to dip below 15 m/s, closed loop control would cause flaps to ex tend, further slowing the aircraft Shortly thereafter, the thro ttle would increase effort in response to the increasing error between actual and commanded airspeeds When this additional power increases the airspeed, flaps will retract in response to a ne w optimal condition. The sudden decrease in drag will then cause the airspeed to increase even further before the throttle decreases, causing overshoot. Th is underdamped positive feedback loop currently rules out closed loop flap control integration at thi s time. Figure 5 2 3 Positive feedback in closed loop flap control Full mission camber optimization must also account for the takeoff and landing phases of flight s During the first phase of takeoff, any STF logic is disabled and a constant takeoff fl ap setting is desired. This is to prevent premature flap retraction in cases of wind gusts and ground effect below the flap s deployed stall speed. Similarly cruise optimized STF logic should also be overridden during landing During the initial descent ph ase of the landing procedure, takeoff flaps are again deployed, maximizing

PAGE 106

106 available lift to lower the stall speed while also increasing drag to increase the descent angle Following the transition to a final landing approach full brake is deployed exten ding both inboard flaps and outboard spoilerons to decelerate and descend as quickly as possible, maximizing landing performance to enable short field operations. During initial tests, full brake deployment resulted in a decrease in required final approach distance of over 50%, with more improvements available with improved tuning and more aggressive final routing.

PAGE 107

107 CHAPTER 6 SUMMARY AND CONCLUSI ONS Summary The purpose of this thesis was to determine an appropriate sUAS design process to optimize overall sy stem effectiveness for emerging remote sensing applications. Towards this end, a logical hierarchy of system design constraints was developed from desired output data specifications, payload limitations, and the realities of operational deployments. The in itial development and operational assessment of the Nova 2.1 sUAS was examined, with special focus on its suitability for aerial mapping applications. To further improve the feasibility of deploying sUAS for remote sensing applications, various strategies to enhance the efficiency of flight path management were explored. Due to the unique power constraints of electric powered sUAS, significant importance was placed on overall energy management through the minimization of unnecessary mission overhead. In an effort to improve ground coverage efficiency throughout the main data collection portions of aerial mapping flights, both in terms of flight t ime and overall mappable area, speed to f ly theory was incorporated into the Nova 2.1 flight management package. By optimizing flight speeds based on current wind conditions, significant improvements could be made in overall ground coverage efficiency. To further improve the Nova 2.1 system for its target applications, trailing edge control was integrated into the ST F algorithm in addition to the general operation of the system, reducing cruise drag and improving takeoff and landing performance.

PAGE 108

108 Conclusions Building upon five generations of institutional experience developing sUAS for remote sensing applications, the Nova 2.1 was the first aircraft in its class capable of providing high accuracy, high resolution geospatial data produ cts on an operational basis. Due to the implementation of a design process focused on eliminating operational deficiencies, the Nova 2.1 p roved highly successful during the last two years of field deployments by crews within the UF UASRG as well as external operators within the USACE. While initial integration of overhead minimization strategies was limited to on the fly adjustments by the ground station operator, updated flight management software has been developed to automate much of this process, resulting in a 20% reduction in energy consumption during the beginning and terminal phases of mapping flights. Following p ower consumption test flights for the original wing as well as an updated flapped variant, STF application was found to offer up to 14% increased ground coverage and a 43% reduction in required flight time for mapping flights in high wind conditions. Addit ionally, camber adjustment was shown to significantly increase available airspeed range, widening the operational envelope for a variety of sUAS applications. Finally, preliminary integration of a landing flap configuration reduce d the s landing length requirement by 55 % reducing deployment requirements and improving operational availability in challenging terrain. Recommendations A natural product of continued operation of a complex system like a remote sensing sUAS is the constant cyc le of improvement that occurs as new applications give rise to novel problems. Currently, remote sensing missions using the Nova 2.1

PAGE 109

109 sUAS require extensive operator experience not for the safe navigation of the aircraft, but rather to ensure that usable d ata are being generated with each flight. Because of the steep learning curve associated with effective flight planning for aerial photogrammetry applications, a number of initial missions were completed successfully, yet produced no usable data due to i nsufficient stereoscopic coverage or similar technical issues that preclud ed postprocessing of collected imagery To improve the feasibility of widespread use of sUAS as remote sensing tools, flight planning should be automated as much as possible until on ly the most basic user inputs are required. In an ideal situation an operator would only need to input their desired data product specifications and provide an outline of the target area in order to gen erate an optimized flight plan, with all relevant pay load control also being managed internally to the system. To ensure that the maximum performance capabilities of a given sUAS are utilized to provide the highest operational capacity for remote sensing applications, integration of STF strategies should be designed in from the start. On the airframe side, this means taking full advantage of the additional flexibility afforded by fully variable camber to keep overall wing area small, both for high speed cruise and practical packing considerations. Regarding a utopilot development, work should continue towards providing hands off STF optimization of flight speeds and wing configurations throughout the flight without external operator input, as existing implementations currently require As with flight planning, efforts should be focused on moving all major mission management to the software backend wherever practical. Finally, to improve operational flexibility, the internal autopilot control algorithms should be modified to allow gain scheduling, which would pe rmit a wider operational

PAGE 110

110 airspeed envelope and allow more accurate control with large flap deflections. While the latter certainly limits overall takeoff and landing efficiency, the actual routing of takeoff and landing flight paths can also limit absolute performance. To improve overall takeoff and landing performance, these flight paths should be made much more aggressive to reduce operating area requirements, alleviating what is currently a primary operational limitation of the Nova 2.1 system.

PAGE 111

111 LIST OF REFERENCES Abbott, I., and A. Von Doenhoff. Theory of Wing Sections: Including a Summary of Airfoil Data. Courier Dover Publications, 1959. Alerstam, T. Bird Migration. Cambridge University Press, 1990. Anderson, J. D. Fundamentals of Aerodynamics. McGraw Hill, 2007. Arrow, K. "On the Use of Winds in Flight Planning." Journal of Meteorology 1947. Bowman, W. S. "Design and Validation of Autonomous Rapid Mapping System Using Small UAV." Master's Thesis. University of Florida, 2008. B owman, W. S., A. C. Watts, J. H. Perry, M. Morton, and P. G. Ifju. "UAS as monitoring tools for assessing the Nation's levees and montoring invasive plant populations." Workshop held at the US Army Corps of Engineers Research and Development Center. Vicksb urg, MS: Presentation slides, March 19, 2008. Breckenridge, R. "Improving Rangeland Monitoring and Assessment: Integrating Remote Sensing, GIS, and Unmanned Aerial Vehicle Systems." Dissertation. University of Idaho, 2007. Breckenridge, R., J. Harbour, R Lee, M. Dakins, and S. Bunting. "Using Unmanned Aerial Vehicles to Assess Vegetative Cover and Identify Biotic Resources in Sagebrush Steppe Ecosystems: Preliminary Evaluation." Eleventh Forest Service Remote Sensing Applications Conference. 2006. Burge ss, M. A., et al. "A Small Unmanned Aircraft System for Ecological Research." AUVSI's Unmanned Systems North America. Poster, 2009. Burgess, M. A., P. G. Ifju, H. F. Percival, J. H. Perry, and S. E. Smith. "Development of a survey grade sUAS for natural r esource assessment and monitoring." Florida Cooperative Fish and Wildife Research Unit Annual Coordinating Committee Meeting. Gainesville, FL: Slide presentation, May 23, 2009. Deperrois, A. XFLR5 v6.03 [Software]. 2011. http://xflr5.sourceforge.net/xflr5 .htm. Drela, M. Mark Drela Tail Airfoils. July 2001. http://www.charlesriverrc.org/articles/drelaairfoils/markdrela_tailairfoils.htm (accessed October 2011). Quick V Tail Sizing. March 2000. http://www.charlesriverrc.org/articles/design/markdrela_vtai lsizing.htm (accessed October 2011).

PAGE 112

112 SuperGee I Pod Molding. October 2003. http://www.charlesriverrc.org/articles/supergee/pod_molding.pdf (accessed October 2011). Durand, W. F. "Tests on Thirteen Navy Type Model Propellers." NACA Technical Report 237 1927. Federal Aviation Administration. Glider Flying Handbook. Skyhorse Publishing, 2007. Hepperle, M. Drag of Linkages. August 2003. http://www.mh aerotools.de/airfoils/linkage.htm (accessed October 2011). Hepperle, M. "Neue Profile fr Elektro Pylonm odelle." FMT Kolleg 10 1991. Jensen, J. Introductory Digital Image Processing. Prentice Hall, 2004. Jones, G. P. "The Feasibility of Using Small Unmanned Aerial Vehicles for Wildlife Research." Master's Thesis. University of Florida, 2003. Lee, K. H. Development of Unmanned Aerial Vehicle (UAV) for Wildlife Surveillance." Master's Thesis. University of Florida, 2004. MacArthur, D., J. K. Schueller, and C. D. Crane. "Remotely Piloted Mini Helicopter Imaging of Citrus." American Society of Agricultural and Biological Engineers 2005. Magnuson, S. "Military Swimming in Sensors and Drowning in Data." National Defense January 2010. Nyquist, J. "Applications of Low Cost Radio Controlled Airplanes to Envionmental Restoration at Oak Ridge National Laboratory." 1996. Owen, P. "When the Ravens Met th e Sandhill Cranes." Unmanned Systems June 2011. Perry, J. H. "A Synthesized Directly Georeferenced Remote Sensing Technique for Small Unmanned Aerial Vehicles." Master's Thesis. University of Florida, 2009. Perry, J. H., Evers, B. S., Rambo, T. J., Reed T. M., Burgess, M. "Natural Resource Management with a Small Unmanned Aircraft." North America. 2011. Phillips, W. F. Mechanics of Flight. John Wiley & Sons, 2004. Pisanich, G., and S. Morris. "Fielding an Amphibious UAV: Devel opment, Results, and Lessons Learned." Digital Avionics Systems Conference. 2002.

PAGE 113

113 Rubio, J., and S. Kragelund. "The Trans Pacific Crossing: Long Range Adaptive Path Planning for UAVs through Variable Wind Fields." Digital Avionics Systems Conference. 2003 Santamaria, E., P. Royo, C. Barrado, E. Pastor, J Lopez, and X. Prats. "Mission Aware Flight Planning for Unmanned Aerial Systems." AIAA Guidance, Navigation and Control Conference 2008. Selig, M. S., C. Lyon, P. Giguere, C. Ninham, and J. Guglielmo. Summary of Low Speed Airfoil Data, Volume 2. SoarTech, 1996. Selig, M. S., Donovan, J., and Fraser, D. Airfoils at Low Speeds. Soartech, 1989. Vorse, B. "Unmanned Aerial Vehicle Taking Fight on Invasive Species to New Heights." JaxStrong February 2011. Wilkinson, B. E. "The Design of Georeferencing Techniques for Unmanned Autonomous Aerial Vehicle Video for Use with Wildlife Inventory Surveys: A Case Study of the National Bison Range, Montana." Master's Thesis. University of Florida, 2007. Wolf, P. R., and B. A. Dewitt. Elements of Photogrammetry with Applications in GIS. McGraw Hill, 2000.

PAGE 114

114 BIOGRAPHICAL SKETCH Brandon Scott Evers was born and grew up in Lake Wales, FL. From an early age, his passion for flight and all things mechanical led him to wards a career in aviation. Upon graduation from the International Baccalaureate School in Bartow, FL, Brandon attended the University of Florida on a National Merit Scholarship. Pursuing a degree in ir Vehicles Lab, where his practical engineering education began. Contributing to projects ranging from an air dropped bomb damage assessment MAV to an entry in the US European MAV Competition, he developed valuable experience in rapid design and composite manufacturing techniques while working under Dr. Peter Ifju. Brandon obtained his B.S. in Aerospace Engineering in 2009 and immediately began his graduate studies at the University of Florida. He continued to work on UAV related projects, designing the No va 2.1 sUAS in collaboration with fellow UF UASRG researchers. In addition to his roles in system design and fabrication, Brandon also served as the remote pilot for field deployments of the Nova 2.1 and was responsible for training and certification of ex ternal system operators from the USACE. With FAA sUAS restrictions providing additional incentive to pursue a lifelong goal, Brandon obtained his Private Pilot Glider rating in 2011, introducing him to the world of silent flight, which provided some of the inspiration for his thesis work. During his graduate studies, Brandon also rediscovered a love for the outdoors, and enjoys spending more time than is practical mountain biking, backpacking, and kayaking with his incredibly willing fiance throughout Nort h Florida.