A HYBRID VEHICLE FOR AERIAL AND TERRESTRIAL LOCOMOTION
by
Richard J. Bachmann
Submitted in partial fulfillment of the requirements
for the degree of Doctor of Philosophy
Dissertation Advisor: Dr. Roger D. Quinn
Department of Mechanical Engineering
Case Western Reserve University
May, 2008 CASE WESTERN RESERVE UNIVERSITY
SCHOOL OF GRADUATE STUDIES
We hereby approve the thesis/dissertation of
______
candidate for the ______degree *.
(signed)______(chair of the committee)
______
______
______
______
______
(date) ______
*We also certify that written approval has been obtained for any proprietary material contained therein. Table of Contents Section Page Chapter 1 Introduction...... 1 1.1 Motivation...... 2 1.1.1 Chemical or Biological Weapon Source Localization...... 2 1.1.2 Ad hoc Sensor Network Deployment ...... 2 1.1.3 Search and Rescue...... 3 1.1.4 Hazardous Environment Exploration...... 3 1.1.5 Hostage Situations...... 4 1.1.6 Medium field military situational awareness...... 4 1.2 Biological Inspiration...... 5 1.3 Organization of this document...... 6 Chapter 2 Background...... 7 2.1 Micro Terrestrial Vehicles ...... 7 2.2 Miniature Air Vehicles ...... 11 2.2.1 Flapping Wings...... 12 2.2.2 Rigid Wings ...... 13 2.2.3 Flexible Fixed Wings...... 14 2.3 Hybrid Amphibious Vehicles ...... 15 2.4 Hybrid Aerial/Terrestrial Vehicles ...... 16 Chapter 3 The First Prototype ...... 18 3.1 Locomotion Subsystem Selection...... 18 3.2 Multi-Modal Mobility Tradeoffs ...... 20 3.2.1 Single Source Propulsion...... 22 3.2.2 Wheel-legs ...... 23 3.2.3 Steering ...... 24 3.2.4 Wings ...... 25 3.3 Design Process...... 26 3.3.1 Terrestrial Drive System...... 26 3.3.2 Sensor system...... 27 3.3.3 Fuselage ...... 27 3.3.4 Wing...... 28 3.4 Fabrication Process...... 29 3.4.1 Wing and Tail ...... 29 3.4.2 Fuselage ...... 31 3.4.3 Wheel-legs ...... 33 3.5 Results...... 34 3.6 Critical Analysis of Prototype Hybrid Vehicle...... 40 3.6.1 Primary Goals...... 41 3.6.1.1 Autonomous Operation...... 41 3.6.1.2 Improved Durability...... 42 3.6.1.3 Repeatable Fabrication and Assembly Procedures...... 43 3.6.2 Secondary Goals...... 43 3.6.2.1 Improved Deployment ...... 43 3.6.2.2 Enhanced Flight Performance...... 44
i Table of Contents Section Page 3.6.2.3 Improved Wing Folding...... 44 3.6.2.4 Ground Take-off...... 44 3.6.2.5 Tail Hook...... 45 Chapter 4 Airframe Mechanical Design...... 46 4.1 Fuselage Mold Fabrication ...... 47 4.2 Component Mounting...... 50 4.2.1 Motor Nacelle...... 51 4.2.2 Tail Boom...... 53 4.3 Tail ...... 55 4.3.1 Assembly...... 55 4.3.2 Alignment ...... 56 4.3.3 New Tail Design ...... 57 4.4 Component Placement ...... 58 4.5 Results...... 59 Chapter 5 Terrestrial Running Gear Design...... 61 5.1 Larger direct drive system ...... 62 5.2 Indirect Drive System ...... 63 5.3 Slip-clutch Hub...... 65 5.4 Angled Slip-clutch Hub ...... 66 5.5 Parallel Shaft O-ring Slip-clutch...... 67 5.5.1 Test Apparatus for Torque Transmission Measurement...... 67 5.5.2 Test Procedure for Torque Limiter Evaluation...... 68 5.6 Concentric Shaft O-ring Slip-Clutch ...... 71 5.7 Wheel-legs ...... 74 5.8 Results...... 75 Chapter 6 Airframe Aerodynamic Design...... 76 6.1 Wing and Aerodynamic Property Tests...... 76 6.1.1 Purpose...... 76 6.1.2 Test Apparatus and Set-up...... 77 6.1.3 Data Collection and Analysis...... 80 6.1.4 Wing Testing Procedure ...... 81 6.1.4.1 Independent Wing Test Results ...... 82 6.1.4.2 Wind Tunnel Testing of Aerial Platform...... 83 6.2 Power Plant...... 85 6.2.1 Purpose...... 85 6.2.2 Test Apparatus for Thrust Measurement ...... 86 6.2.3 Test Procedure...... 89 6.2.4 Data Analysis...... 90 6.2.4.1 Component Selection...... 92 6.2.4.2 Data Verification...... 93 6.2.4.3 Drive Voltage Analysis...... 95 6.2.4.4 Speed Controller Analysis ...... 96 6.2.4.5 Motor Analysis...... 97
ii Table of Contents Section Page 6.2.4.6 Headwind Comparison ...... 97 6.2.4.7 Propeller Analysis...... 98 Chapter 7 Auxiliary Subsystem Design...... 102 7.1 Tail Hook...... 102 7.2 Wing Folding...... 106 7.2.1 Confinement Subsystem...... 107 7.2.2 Foldable Wing...... 110 7.2.3 Actuation Subsystem ...... 112 7.3 Ground Take-off...... 120 7.4 Rapid Deployment ...... 124 7.4.1 Procerus Kestrel™ Autopilot...... 124 7.4.2 Palmtop Autopilot Interface...... 124 7.4.3 Interface Control Hierarchy ...... 125 7.4.4 Interface Appearance...... 127 7.4.5 User Relative Ad Hoc Path Appointment...... 128 Chapter 8 Conclusions and Discussion ...... 130 Appendix A Power Plant Thrust Data ...... 136 A.1 Thrust Data for Mighty Micro 010 14-Turn Motor ...... 136 A.2 Thrust Data for Mighty Micro 010 10-Turn Motor ...... 138 A.2.1 7.4V Data...... 138 A.2.2 11.1V Data...... 139
iii List of Figures Figure Page Figure 1. Relative sizes of a Mini-Whegs robot and a Blaberus giganteus cockroach...... 11 Figure 2. The flexible airfoil allows for passive adaptive washout...... 15 Figure 3. The first layer of carbon fiber...... 30 Figure 4. The fabric is sandwiched between layers of carbon fiber ...... 31 Figure 5. The fully wrapped fuselage tool, including indent for the propeller motor ...... 32 Figure 6. How to create the flanges to which the wing will be glued...... 33 Figure 7. Wheel-legs and two lay-up tools ...... 34 Figure 8. First prototype MALV...... 35 Figure 9. Surveillance image from on-board video camera...... 37 Figure 10. Wing folding sequence of the non-actuated model ...... 38 Figure 11. Folding wing...... 39 Figure 12. Wing-folding mechanism implemented on prototype...... 39 Figure 13. Wing-folding prototype vehicle ...... 40 Figure 14. Rendering of MMALV fuselage mold ...... 49 Figure 15. Close-up of the motor nacelle and mounting face...... 52 Figure 16. Servo placement is both durable and repeatable within the tail boom...... 54 Figure 17. Control rod accessibility on new design...... 55 Figure 18. Double-slot assembly of the first prototype tail ...... 56 Figure 19. Deflection of vertical stabilizer halves (exaggerated)...... 56 Figure 20. New Tail Design...... 57 Figure 21. Component mounting-hole fixture ...... 59 Figure 22. Photograph of the new MALV aerial platform ...... 60 Figure 23. Even large, metal-geared servos are damaged in landings...... 62 Figure 24. Initial Terrestrial Drive System in Model A Fuselage ...... 64 Figure 25. Slip-clutch hub design ...... 66 Figure 26. Angled slip-clutch hub design...... 67 Figure 27. Test Rig for Measuring Torque Limiter Transmission Capacity ...... 68 Figure 28. Schematic Representation of Pre-deployment Assembly Process ...... 70 Figure 29. Assembled view of the terrestrial drive system power train ...... 72 Figure 30. Exploded view of the terrestrial drive system power train...... 72 Figure 31. Exploded view of terrestrial drive system housing ...... 73 Figure 32. The current MALV design ...... 75 Figure 33. Rendering of sting with wing mounting adapter ...... 78 Figure 34. Modified sting set-up...... 80 Figure 35. Fuselage B and Fuselage C...... 84 Figure 36. Assembly for full aerial platform testing...... 84 Figure 37. Rendering of the thrust measurement apparatus ...... 87 Figure 38. The thrust measurement apparatus mounted in the wind tunnel ...... 88 Figure 39. Force Schematic of Thrust Balance...... 90 Figure 40. Virtual Displacements of the Thrust Balance...... 91 Figure 41. Temporal Consistency Verification Results...... 94 Figure 42. Thermal Consistency Verification Results...... 95 Figure 43. Speed Controller Comparison Data...... 96
iv List of Figures Figure Page Figure 44. Motor Thrust and Efficiency Comparison...... 97 Figure 45. Power Comparison for One Configuration in Different Headwinds...... 98 Figure 46. Propeller Comparisons on Mighty Micro 010 14-turn Motor at 25 MPH ...... 99 Figure 47. Propeller Comparisons on Mighty Micro 010 14-turn Motor at 30 MPH .....100 Figure 48. Propeller Comparisons on Mighty Micro 010 14-turn Motor at 35 MPH .....100 Figure 49. Propeller Comparisons on Mighty Micro 010 14-turn Motor at 40 MPH .....101 Figure 50. Isometric view of the tail-hook design ...... 103 Figure 51. Tail-hook in deployed position...... 104 Figure 52. Tail-hook in midst of retraction process...... 105 Figure 53. Tail-hook in fully retracted position...... 106 Figure 54. Original folding wing, with confinement structure highlighted...... 108 Figure 55. Schematic of an integrated wing confinement system ...... 109 Figure 56. Folding wing with reinforced top enclosure...... 110 Figure 57. Four baton designs tested by Mr. Brian Taylor...... 111 Figure 58. Biased baton concept for a foldable wing ...... 112 Figure 59. Mock-up of the First Wing Folding Mechanism Design...... 113 Figure 60. The Retractable Landing Gear Mechanism...... 114 Figure 61. Six Bar Linkage in situ...... 115 Figure 62. Custom Output Bracket for Retractable Gear Implementation ...... 116 Figure 63. The final wing folding linkage ...... 117 Figure 64. The wing folding mechanism integrated within the fuselage...... 117 Figure 65. Schematic of the Wing-folding Four-bar Linkage ...... 119 Figure 66. Schematic of the Vehicle Re-orientation Mechanism...... 123 Figure 67. Static Take-off Capable MAV Re-orienting Itself for Take-off ...... 123 Figure 68. Communications link schematic...... 125 Figure 69. Three Menu Views of the MAVControl Interface ...... 127 Figure 70. Bearing/Distance/Altitude Entry Screens...... 128 Figure 71. The final MALV vehicle ...... 131
v List of Tables Table Page Table 1. Electromechanical Components of the First Prototype MALV...... 28 Table 2. Performance Characteristics of First Prototype MALV ...... 36 Table 3. Mechanical Properties of First Prototype MALV...... 36 Table 4. Aerodynamic Properties of First Prototype MALV ...... 37 Table 5. Hypothetical Vehicle Comparison...... 77 Table 6. Parameters of the Wings Tested ...... 82 Table 7. Summary of Data for Independent Wing Tests ...... 83 Table 8. Summary of Data for Vehicle Tests ...... 85 Table 9. Design Variables of Wing Folding Four-bar Linkage...... 120 Table 10. Static Thrust Testing for Ground Take-off Vehicle...... 122 Table 11. Break-out of MAVControl Command Hierarchy...... 126 Table 12. Mechanical Properties of 16-inch Wingspan MALV...... 131 Table 13. Aerodynamic Properties of 16-inch Wingspan MALV...... 132 Table 14. Performance Characteristics of 16-inch Wingspan MALV...... 133
vi
A Hybrid Vehicle for Aerial and Terrestrial Locomotion
by
Richard J. Bachmann
A durable hybrid vehicle has been developed capable of both aerial and terrestrial locomotion. The motivation for the work was a wide range of sensor deployment scenarios that would benefit from a vehicle capable of 1) flying long distances to a target area and 2) walking around the target to perform near-field inspection. A technology survey was performed to identify the candidate terrestrial and aerial locomotion technologies for integration. The Mini-Whegs robot, developed at Case Western Reserve
University, was selected as the terrestrial running gear, and the flexible-wing micro air vehicle (MAV), developed at University of Florida, was selected as the aerial platform. A rigorous trade-off analysis led to a remote control prototype that had a fully functional airframe augmented with two R/C servos, modified for continuous rotation, driving independent music wire wheel-legs at the front of the vehicle. This vehicle achieved most of the original performance requirements. It could fly, land, crawl, and regain flight by crawling off the edge of a rooftop. A critical performance evaluation illuminated improvements to the design and fabrication necessary to create a viable hybrid vehicle for field deployment. The vehicle design and fabrication processes were overhauled to improve the durability and reproducibility of the final design. A custom-built terrestrial locomotion subsystem, with compliance in the drive train, was crucial to improved durability. CNC fabrication of the fuselage mold and a one-piece tail design were central to repeatability. A commercially available autopilot was implemented for autonomous operation. Vehicle mass increased from 118 to 365 grams. The wingspan was
vii
subsequently increased to 16", but wing loading increased from 37 to 64 N/m2. A corresponding decrease in controllability was observed. Winglets were found to increase lift, but decrease stability by mitigating wing flexibility. The final vehicle was able to fly, land, and crawl repeatedly. Over 8 flights (and landings) have been performed by the vehicle, and the vehicle has yet to show any signs of damage. The vehicle cruises at 14 m/s and crawls at 0.33 m/s (0.8 body lengths per second). For comparison, a typical
Mini-Whegs runs at 5 body lengths per second.
viii
Chapter 1 Introduction
The purpose of the work described in this dissertation was to develop a miniature
mobile sensor platform capable of both aerial and terrestrial locomotion. Inherent in the
desire for the vehicle to locomote both in the air and on the ground is the need for it to transition from flight to crawling without sustaining damage, which proved to be one of the greatest challenges in the project. The original motivation for this research also resulted in improved performance characteristics for hand-launchability and the ability to regain flight from a high perch. Finally, in the interest of developing a reproducible craft, considerable effort was put into the fabrication process. In short, the goal of this research was to develop the smallest reproducible vehicle that could fly, land, crawl, and regain flight in special circumstances while carrying sufficient payload to provide sensor feedback.
Recent technological advances, especially those in computing, sensing, and power storage, have facilitated the development of miniature robotic vehicles. However, until recently, the agility of these vehicles has been limited by reliance on a single mode of locomotion. Furthermore, until this work, those vehicles that did demonstrate multiple modes of locomotion operated solely in the amphibious realm, capable of swimming and walking/crawling.
1
1.1 Motivation
A range of applications would benefit from a small sensor platform capable of both aerial and terrestrial locomotion. Generally speaking, these applications fall into one
of two groups: 1) close approach to a target in an area that is difficult or dangerous to
access, and 2) ad hoc sensor networks, as characterized by rapid, intermittent, or short
term sensor deployment. A few of these applications are presented here.
1.1.1 Chemical or Biological Weapon Source Localization
In the case of an outdoor release of a chemical or biological weapon, rapid
localization of the source is critical to saving lives, and the search process would be
expedited by the deployment of multiple search platforms. Unfortunately, monetary
considerations limit the number of helicopters that can be brought rapidly into service by
a single emergency organization (fire, police, national guard). Furthermore, having
manned aircraft fly through the airborne plume could place those personnel in danger.
However, a small fleet of unmanned aerial vehicles (UAVs) could generally locate the
source by blanketing the target area with sensor platforms. Imbuing the UAVs with
terrestrial locomotion capabilities would allow the vehicles to land and precisely localize
and inspect the source. At that point, the minimum number of emergency personnel could
respond to the threat.
1.1.2 Ad hoc Sensor Network Deployment
There are several instances where a rapidly deployed, reconfigurable sensor
network would be useful. Police executing a manhunt or secret service securing a
motorcade route would be desirous of quickly blanketing a small geographical area with
2
sensors, and being able to relocate the sensors on the fly. The temporal and
reconfigurable aspect of these networks would preclude manual sensor placement or even
deployment of stationary sensors from an aerial vehicle. The hybrid vehicle presented here would allow an ad hoc sensor network to be in place within an hour of the deployment order, and the network could also be reconfigured at will.
1.1.3 Search and Rescue
Similar to the chem/bio scenario described above, mountain/forest search and
rescue operations are limited by the cost of deploying the search vehicles (primarily
helicopters) and personnel. Again, a small fleet of UAVs would significantly decrease the
time necessary to thoroughly explore a given area. Furthermore, these vehicles could
even be deployed in the case of inclement weather that would otherwise preclude search
operations. Finally, if the vehicle were capable of terrestrial locomotion, it could land in
order to 1) explore any occluded areas (caves, under forest canopy) and 2) approach the
stranded adventurer to deliver aid or provide two-way communications.
1.1.4 Hazardous Environment Exploration
Other situations arise that involve sensor deployment over hundreds of yards rather than miles, such as train and industrial accidents involving hazardous chemicals. In most cases, a quarantine area (of perhaps a ¼ mile radius) is established around the spill
site and all personnel entering the quarantine area must wear contamination suits. While
emergency personnel will eventually be required to enter the area for repair and
remediation, useful information could be acquired by an unmanned sensor platform
deployed into the area. Whereas a fully equipped terrestrial robot may provide more
operational utility (with manipulators and greater sensor payload), environmental factors
3
may preclude a terrestrial robot from reaching the site. However, a platform capable of
flying to the site and then landing and crawling around the site would be impervious to terrain irregularities and could reach the site faster and with less operator interaction.
1.1.5 Hostage Situations
During hostage situations, law enforcement personnel could gain a definite
tactical advantage by being able to “look” into the building. Currently, no good option
exists for gaining this view. Snipers on adjacent buildings can be blocked by curtains,
personnel approaching the building could be put in serious danger, and robots
approaching the building could be easily observed. However, a small hybrid vehicle
could be deployed from a safe distance, and would be quiet enough to approach the
building unnoticed. The vehicle could possibly enter the building through an open
window or land on the roof and rappel down the side of the building for a better view of
the interior.
1.1.6 Medium field military situational awareness
While any of the civilian applications enumerated above would be sufficient
reason to develop a multi-locomotory mode capable vehicle, this research was, as with
much engineering research, originally motivated by military applications. The research
presented here was funded by the U.S. Air Force under two separate Small Business
Innovation Research (SBIR) contracts awarded to BioRobots, LLC, a local small
business concern. The SBIR program is a two phase program – Phase I is a short research effort intended as a proof-of-concept, and Phase II is for the development of a functional prototype. The author served as the Principal Investigator on both the Phase I and the
Phase II efforts. As such, while he did not personally perform all of the research
4
presented here, he did perform much of it, and he conceived of, directly oversaw, and
verified that research that he did not perform.
The original application envisioned by the Air Force was as a reconnaissance and
surveillance tool for foot-borne soldiers. The program solicitation called for a 12"
maximum dimension (for both wingspan and overall length), one pound maximum weight vehicle that could execute the following mission sequence:
an operator hand launches the vehicle
the vehicle flies to the target area up to one mile away
the vehicle lands on the roof of a building
the vehicle reconfigures for terrestrial operation and moves to the edge of the roof
the vehicle collects video surveillance, and either stores or transmits the data
the vehicle reconfigures for aerial operation and walks off the edge of the roof
the vehicle returns home or continues to a new target location
The working title of this project was Morphing Micro Air-Land Vehicle, or
MMALV. Since wing-folding (morphing) has yet to be integrated into the design, the
acronym MALV is used throughout this document to refer to the hybrid vehicle
developed as a result of this dissertation research.
1.2 Biological Inspiration
Considerable attention has been paid recently to the implementation of biological
locomotion principles in mobile robots. Animals in a range of sizes rely on the two
locomotion modes that are central to this research, and could, therefore, act as models. Of
course, those animals have access to unparalleled actuators, structural components, and
5
energy conversion systems. Testing the efficacy of biological principles was not a
primary concern of this research. However, biological locomotion principles were
considered during the design process, and the constituent technologies (as presented in
Chapter 2) did benefit from varying degrees of biological emulation. Furthermore, some
principles of biological locomotion were fundamental to the success of the final design,
and these will be discussed as they arise in the presentation of the methods and results.
1.3 Organization of this document
Chapter 2 presents an overview of related and pertinent research, specifically
aerial, terrestrial, and multi-mode capable robots. Chapter 3 discusses the proof-of
concept vehicle. It begins with the original trade-off analysis, proceeds through the
design process, and finishes with a critical analysis of the vehicle’s performance and
fabrication process. Chapter 4 presents the final mechanical design of the airframe,
including fabrication processes, Chapter 5 considers the terrestrial drive system of the final design, and Chapter 6 discusses the improvements to the aerodynamic design of the
airframe, including the wing and powerplant. Chapter 7 presents multiple auxiliary
systems developed for completing the operational utility of the vehicle, including a tail-
hook, ground take-off capabilities, and wing folding. Chapter 8 discusses the overall
results of the program, and potential future developments.
6
Chapter 2 Background
The research presented here resulted in the first-of-its-kind hand-launched hybrid
vehicle capable of both aerial and terrestrial locomotion. That said, the research did not attempt to “reinvent” the leg or the wing, but focused on the integration and fabrication
process. Thus, the background presented here focuses primarily on existing mono-modal
vehicles capable of either aerial or terrestrial locomotion, as building blocks for the
integration process. Hybrid vehicles capable of swimming and crawling are discussed briefly for insight into the design trade-off process inherent to the development of multi- mode capable vehicles. Finally, parallel research efforts are presented that focused on hybrid aerial/terrestrial vehicles.
2.1 Micro Terrestrial Vehicles
The objective of a being able to hand-launch the hybrid vehicle significantly
limits its size and weight, and consequently the size and weight of the terrestrial running
gear that can be integrated into the design. However, two major factors remain significant
challenges to the deployment and field utility of terrestrial micro robots. First, the relative
size of real-world obstacles (e.g. stairs, gravel, terrain fluctuations, etc.) makes movement
a daunting task for small robots. For example, RHex [1], at approximately 50 centimeters
7
and in length, is the shortest existing ground robot that can climb standard stairs without jumping, flying, or using gripping mechanisms on its feet. The second obstacle to field deployment is that power-source miniaturization has not kept pace with other critical technologies, such as actuation, sensing, and computation. This restricts most micro robots to be used as laboratory research platforms. The ultimate result of these challenges is that the literature holds few examples of terrestrial robots weighing less than 200
grams.
K-Team has developed three generations of the Khepera [2], a three-wheeled
robot that uses differential activation of the two front wheels for steering. With an array
of sensors and user-programmable control onboard, these micro-robots are popular with
group behavior researchers. The latest version is able to locomote on rough floors,
carpets, and even over doorsteps, which is a great improvement over previous
generations, which were restricted to table-tops operation, due to the original 1.4
centimeter diameter wheels.
Millibots [3] are tracked vehicles with a body size of approximately six
centimeters cubed. Inability to implement a modern track suspension at this scale led
researchers at Carnegie-Mellon University to create a multi-segmented “worm-like” version. However, the necessary addition of motors to “lift” adjacent modules resulted in a final module mass of 266 grams, and no performance specifications have been published for the modular design.
Researchers at the Aichi Institute of Technology developed a 40 gram, 28 millimeter diameter tri-pedal robot [4]. Piezoelectric wires that connect adjacent legs are differentially actuated to produce sequential motion in individual legs. However, the
8
kinematic arrangement of the vehicle produces no vertical motion of the foot, which would be needed to overcome obstacles. Furthermore, the 0.001 body-length per second top speed limits the utility of the robot.
Birch et al. [5] developed a 7.5 centimeter long, 94 gram hexapod inspired by the cricket. Coupling allowed the vehicles 16 joints to be driven by 8 actuators, and the robot achieved 0.4 body lengths per second with off-board power. This robot required the development of miniature braided pneumatic actuators, and a custom miniature compressor. While a half-legged, half-wheeled version was able to carry its power supply on-board, the hexapod version has not yet done so.
Sprawlita [6] is a 16 centimeter long hexapod developed at Stanford University.
Shape Deposition Manufacturing was used to embed sensors and actuators within structural elements, and to produce structural elements with heterogeneous properties.
Powered by a combination of servomotors and pneumatic cylinders, Sprawlita was able to walk at 3 body lengths per second, and overcome 3.5 centimeter high obstacles.
However, an operating air pressure of 6 bars makes it unlikely that the robot will become autonomous in its current form. More recently, the researchers responsible for Sprawlita developed iSprawl [7], which uses single motor propulsion and benefits from abstracted biological principles. While it can attain 15 body lengths per second, its obstacle climbing ability is severely limited by small excursions of its feet.
Locomotion studies on cockroaches have elucidated several critical behaviors that endow the insect with its remarkable mobility [8]. During a standard swing phase, the cockroach raises its front legs high in front of its body, allowing it to take smaller obstacles in stride. The animal uses a tripod gait for unperturbed walking, where adjacent
9
legs are 180° out of phase, but when climbing larger obstacles, the animal moves adjacent
legs into phase to increase stability.
These cockroach locomotion principles were captured in the mechanics of the
Whegs platform [9]. Whegs robots use six multi-spoke wheel-leg appendages, the front
pair of which is mounted at the extreme front of the vehicle. Adjacent wheel-legs are
typically 180º out of phase, thus producing the standard tripod gait. Torsional compliance
in the wheel-legs increases stability and traction by allowing adjacent legs to move into
phase when an obstacle is encountered [10]. Through these devices, all of the aforementioned cockroach locomotion principles are accomplished using only a single drive motor. Torsional and radial compliance in the legs serve to absorb shock from unexpected obstacle contact. Therefore, in accomplishing abstracted biological behaviors, fundamental biological principles are implemented. Alexander [11] identifies three
fundamental roles of compliance (i.e. springs and dampers) in biological legged
locomotion – energy storage and release, impulse load reduction, and body stabilization.
Whegs’ mobility capability is reliant on the latter two of these three roles.
The wheel-leg concept also proved to be scalable. Mini-Whegs™ (Figure 1) is a 9
centimeter quadruped [12] robot that offers a combination of speed, mobility, durability,
power autonomy, and payload. Mini-Whegs predates iSprawl in using a single drive
motor. Mini-Whegs can attain a speed of 10 body lengths per second, and can easily run
over 3.5 centimeter tall obstacles – higher than the top of its body compartment. Mini-
Whegs™ have also carried over twice their body weight in payload. Due to its high
mobility and straight-forward, lightweight implementation, Mini-Whegs was selected as the basis for the terrestrial locomotion system on MALV.
10
1 1cm
Figure 1. Relative sizes of a Mini-Whegs robot and a Blaberus giganteus cockroach. (Photo courtsy of Andrew Horchler)
2.2 Miniature Air Vehicles
The exact parameters of the final hybrid vehicle will be determined by an in-depth
design process. However, consideration of the program objectives (and consultation with
unmanned aerial vehicle experts) provides a basis for the size of vehicles that should be
considered in the background survey. Discussions with experienced R/C pilots suggest
that a wingspan of 25 to 50 centimeters (10 to 20 inches) will be needed to allow the craft
to fly at airspeeds of 6.7 to 8.9 meters per second (15 to 20 miles per hour). A low airspeed is desirable for video feedback clarity and to minimize loads experienced by the
terrestrial running gear during landing. Assuming a chord length of 10 to 20 centimeters
(4 to 8 inches), the Reynolds Number (Re) associated with these design and performance estimates ranges from 4.3x104 to 1.1x105. This coincides with the range of Re (from 104 to 106) where the aerodynamic efficiency (ratio of coefficient of lift to coefficient of drag) of smooth, rigid airfoils suffers more than an order of magnitude decrease [13]. At
11
this range, the laminar flow that prevails is easily separated, creating large separation
bubbles, especially at higher angles of attack.
Other major obstacles exist for flight at this scale [14]. Earth’s atmosphere
naturally exhibits turbulence that can result in significant variations in airspeed, and therefore lift, from one wing to the other. The resulting erratic roll torques, in conjunction
with the small mass moment of inertia inherent to aircraft at this scale, adversely affect
stability and significantly increase the roll stabilization effort required by the pilot [14].
The majority of research to develop traditional (non-rotary) winged miniature air
vehicles can be categorized into three fundamental approaches.
2.2.1 Flapping Wings
Nature suggests one method of dealing with the aerodynamic phenomenon
described above, whereby efficiency drops by over an order of magnitude as Re
decreases from 106 to 104. Consider, for example, the behaviors of birds of various sizes.
Birds with large wingspan are able to develop the airspeed necessary to maintain Re >
106, and thus tend to soar for prolonged periods of time. Medium-sized birds must flap
vigorously for take-off, and periodically flap their wings during flight in order to
maintain sufficient airspeed to avoid allowing their Re to drop within the dangerous
range. The smallest birds and insect are incapable of developing the velocity sufficient to
maintain Re > 104. Therefore, they beat their wings continuously and rapidly, which
serves to increase their Re independent of their velocity, and also use more advanced
aerodynamic effects to develop lift. Several micro air vehicles [15]-[18] have been
developed that use flapping wings to produce both lift and thrust. Furthermore,
researchers have demonstrated flapping wing MAVs that can fly and even hover [19]
12
using the “clap and fling” mechanism as described by Ellington [20]. However, these
MAVs are susceptible to failure in even light winds and their payload capacity is very
small. This approach remains attractive for future work, in particular for low speed, low
wind applications such as inside buildings.
2.2.2 Rigid Wings
Until now, the most widely used technique for SUAV (small unmanned aerial
vehicle) design has been to configure the airframe in a manner similar to larger aircraft.
In this method, thrust is developed using a conventional propeller, and lift is generated
with a thick, rigid airfoil. At this small scale, researchers usually fall back on the flying
wing design, in an attempt to maximize the wing area. Even with optimized wing parameters, these vehicles require stability augmentation systems and/or superior pilot skill.
Among the most successful examples of rigid wing MAVs designed with this approach is Aerovironment’s “Black Widow” [21], an electric 15 centimeter flying wing.
Virtually every component on the aircraft is custom built, including a sophisticated gyro- assisted control system. The final vehicle attained a flight duration of 30 minutes, resulting in a range of 17 kilometers at airspeeds of 17 to 24 meters per second (38 to 53 miles per hour). Other successful examples of rigid wing designs include the “Trochoid”
[22] and the “Microstar MAV” [23]. All of these vehicles have gyro-assisted stabilization systems, without which they would be impossible to control. Furthermore, none of the companies that developed these vehicles list them on their websites, suggesting they have been eschewed in favor of larger vehicles with more utility.
13
2.2.3 Flexible Fixed Wings
The third approach to small UAV design implements abstracted biological inspiration in the form of under-cambered (thin) flexible wings that are able to move and deform passively in response to applied loads. Researchers have shown that these bat-like wings enjoyed more favorable aerodynamic performance in fluctuating low Re environments [24]-[29]. This flexible wing concept has been successfully applied by Ifju and others to miniature air vehicles over the past eight years [14],[30]-[33]. A chord-wise compliant wing with a rigid leading edge facilitates a behavior known as passive adaptive washout, whereby perturbations are significantly mitigated. For example, an airborne vehicle may encounter a turbulent headwind, such that the airspeed over only the right wing is suddenly increased. The compliant wing structure responds to the instantaneous lift generated by the gust to deform in a manner similar to Figure 2 (courtesy of Peter
Ifju). This results in a reduction in the apparent angle of attack, and a subsequent decrease in lifting efficiency, as compared to the non-deforming wing. This reduced lift efficiency partially counteracts the increased airflow over the wing, resulting in nearly equivalent lifting force generated between the two wings. Similarly, as the airflow over the wing stabilizes, the wing returns to its original shape. This behavior significantly reduces the control load experienced by the pilot [14].
14
Figure 2. The flexible airfoil allows for passive adaptive washout. (Photo courtesy of Peter Ifju [14])
2.3 Hybrid Amphibious Vehicles
Prior to the research presented here, some of the most successful hybrid robots were those that operate in the aquatic and terrestrial domains. One of the first of these was a water-tight version of RHex [34]. The primary compromise in the development of the hybrid vehicle was in leg design. The optimal swimming leg design (flat, fin-like) and the optimal crawling leg design (rim-like) are diametrically opposed.
Boxybot [35] was developed at the Swiss Federal Institute of Technology,
Lausanne mostly as a platform for testing central pattern generator efficacy in hybrid swimming/crawling robots. Caudal fin like structures generate a range of swimming behaviors and can produce an inverted pendulum terrestrial gait during which the body undergoes cycloidal motion. Swimming behavior was further augmented at the expense of crawling by the implementation of a vertical tail.
The salamander employs the same body undulations during both swimming and crawling. The developers of Boxybot took advantage of this behavior in the design of
15
their next robot [36]. The 83 centimeter long vehicle was capable of terrestrial velocities up to 0.11 body lengths per second and swimming velocities up to 0.14 body lengths per second using a single primitive central pattern generator.
2.4 Hybrid Aerial/Terrestrial Vehicles
Few have researched small, multi-mode robots capable of aerial and terrestrial locomotion. Among these, the Entomopter [37] was perhaps the most ambitious.
Reciprocating Chemical Muscle [38] was developed to drive Entomopter’s 15 centimeter flapping wings at 35 Hertz. Stereolithography and Fused Deposition Modeling were used to create intricate wing structures that mimic the structural characteristics of insect wings, with an added feature. The stiffening “ribs” within the wing were hollow and used to deliver gas to specific points in the wing for purposes of attitude control and increased lift development. The gas was a byproduct of the normal operation of the Reciprocating
Chemical Muscle. However, it is a work in progress – only one manuscript has been published on Entomopter since 2003, and video and actual performance characteristics have yet to be published.
The only other program with the goal of developing a hybrid vehicle capable of powered flight and powered walking was funded under the same SBIR program that funded this dissertation. Pursued by Orbital Research, there are no refereed publications describing either the design process or the performance characteristics of the resulting vehicle. A company brochure shows that a multiple fixed wing design would be used for flight, and monopod hopping would be used for terrestrial locomotion.
Since the completion of the work described herein, one other “hybrid” vehicle has been developed. Developed at Swiss Federal Institute of Technology, Lausanne, the Self
16
Deploying Microglider [39] is a 10 gram vehicle that can jump and glide moderate distances. The 7 gram jumping platform can propel the vehicle over 1.4 meters into the air. Near the peak of the ballistic path, a 3 gram folding wing, fabricated of Mylar and carbon fiber, deploys. The vehicle then glides slowly to the ground. The development of a 10g vehicle capable of these behaviors is a testament to the abilities of the development team. However, the vehicle’s flight phase is purely passive, and the vehicle would provide little utility in the applications enumerated in Chapter 1.
17
Chapter 3 The First Prototype
3.1 Locomotion Subsystem Selection
The goal of this research was to develop a sensor platform that could fly a long distance to a target area, land within the area, and then locomote on the ground a short distance to provide close visual feedback of an objective. Considering the time that has already been dedicated to the development of flying and crawling vehicles, the decision was made to not reinvent wings or legs, but rather focus on integration of existing aerial and terrestrial locomotion technologies. The pinnacle of existing locomotion technologies has been presented in Chapter 2, and these were critically evaluated for applicability to this research.
Based on their dominance of International Air Vehicle Competitions, the flexible wing technology, as put forth by the University of Florida, was selected as the basis for flight locomotion of MALV. The flexible wing provides a superior combination of controllability, payload, speed, and efficiency in the critical size range, as compared to rigid and flapping wings.
Three possibilities were considered for terrestrial locomotion: wheels, legs, and wheel-legs. Each of these running gear have advantages and disadvantages for the given task. In biology, legs are able to reach on top of obstacles and produce the desired normal
18
forces that are superior to frictional forces for climbing. However, as the literature shows,
legs have not been implemented in robotics with the same success, especially at this
small scale. Due to the need for bearing surfaces and material to handle the inherent force
concentrations, legs tend to be either heavy or fragile. Small robots that have successfully
implemented legs on a power autonomous vehicle suffer from limited rough terrain
mobility due to small joint excursions.
Wheels are particularly attractive because they offer the option of allowing the
terrestrial running gear to spin freely, so that propeller thrust can be used as the driving force during terrestrial locomotion, similar to larger, manned aircraft. Early stages of the trade-off analysis (next section) suggested that this method could save considerable mass on the integrated vehicle, allowing for a smaller vehicle of increased sensor payload.
Four freely spinning wheels were attached to a flexible wing MAV, and experiments demonstrated that this vehicle could successfully take-off and land on smooth, hard surfaces. Furthermore, terrestrial locomotion could also be accomplished over this same terrain. Only limited steering could be accomplished by the rudder on the ground, because the four wheels were not steerable. Had this running gear been deemed appropriate, the rear wheels would have been be replaced with a caster, thus increasing the steerability of the vehicle on the ground. While locomotion and control proved excellent on smooth, hard terrain, the vehicle performed poorly on irregular terrain. The propeller had a strong tendency to collide with obstacles and terrain features and the vehicle had a tendency to pitch forward on irregular terrain and even more so when an obstacle was encountered. To allow the propeller to spin without impacting the ground, the line of action of the propeller thrust was necessarily several centimeters above the
19
point of ground contact. Friction on irregular terrain and normal forces created by
obstacles occurred at or near the point of ground contact. This force couple resulted in a
pitching moment that caused “nosing over” behavior. The greatest danger of both nosing
over and the general tendency to hit the terrain with the propeller was the opportunity for
damaging the propeller. If the propeller were to lose a blade, the developed thrust would
be greatly compromised and the imbalance would destroy the motor.
To overcome the disadvantages of passively mounted wheels, actuated wheels
were briefly considered. Quinn et al. [10] have demonstrated that the advantage that
wheels enjoy on smooth flat surfaces is far outweighed by the advantages that wheel-legs
enjoy on rough terrain. While some wheeled vehicles (such as the NASA Mars Rover)
have attained mobility similar to Whegs robots, these vehicles all rely on a fully actuated
suspension. Such a suspension has yet to be implemented on a vehicle the scale of
MALV, and the weight and complexity of such a suspension precludes it from use here.
Each of the mission scenarios described above includes a high level of uncertainty as to
the terrain that will be encountered by the vehicle during terrestrial locomotion. Even
small surface irregularities can be relatively large when encountered by a small robot.
Therefore, the Whegs locomotion technology was selected as the basis for the terrestrial
running gear on MALV.
3.2 Multi-Modal Mobility Tradeoffs
Having settled upon the flexible wing and wheel-leg technologies as the bases for the aerial and terrestrial locomotion subsystems, respectively, it was necessary to perform a trade-off analysis to develop a functional vehicle. By increasing the wingspan or cruising airspeed, it may have been possible to mount a complete Mini-Whegs robot (200
20
grams) onto the bottom of a MAV. However, the simple process of incorporating the
Mini-Whegs chassis into the design of the MAV fuselage would noticeably decrease the
overall mass, and thereby increase the performance of the integrated vehicle. While this
obvious example would improve both aerial and terrestrial mobility of the hybrid vehicle,
other interactions would not be mutually advantageous. For example, wheel-legs will
increase drag and reduce controllability during aerial locomotion. Indeed, no commercial
aircraft have unnecessary protuberance from their fuselages. Furthermore, the mass of the
wheel-legs reduces the payload capacity and could adversely affect stability of the
vehicle. Similarly, on the ground, wings, propellers, and control surfaces and their servos
limit payload and could impede mobility in confined spaces. While it is obvious that wheel-legs, wings, propellers, and control surfaces are all absolutely necessary for the
intended hybrid vehicle, and cannot be compromised, other decisions were not so clear.
For example, the terrestrial mobility of a four wheel-legged robot is considerably better than a similar vehicle with only two wheel-legs. However, those additional wheel-legs contribute to even more drag, instability, and payload reduction during aerial locomotion.
Therefore, a tradeoff analysis was performed to identify the most important parameters for terrestrial locomotion and aerial locomotion. In the case of non-critical parameters, advantages and disadvantages of each were identified and weighed against one another and against the overall utility of the vehicle. In the case of a conflict where mechanisms
required for one mode of locomotion were severely deleterious to the other, options were identified for resolving this conflict. In the tradeoff analysis flight was assumed to be the limiting condition because of energy demands and the larger payload of crawling vehicles.
21
3.2.1 Single Source Propulsion
The first trade-off considered was the potential of using the propeller to develop thrust for ground locomotion. While this was being investigated from a terrestrial locomotion standpoint, it was also considered here on the basis of saving considerable mass. During flight, the prop/motor combination will be required to produce sufficient thrust in 20-30 mph airspeed to overcome drag on the vehicle. To do this, the propeller will need to spin rapidly, with the speed being dependent on the pitch of the propeller.
Motors draw less current at higher rpm, so the motor/prop will be tuned to produce the necessary thrust at the desired airspeed with the motor spinning near its no-load speed
(the speed at which the motor spins when no torque is applied to the motor), thus consuming as little current as possible. However, under these conditions, producing thrust at very low airspeed (as during terrestrial locomotion) will cause the motor to spin at a much lesser rpm than during flight. This will result in a high current draw, higher than would be required for separate terrestrial drive motors tuned to produce the necessary torque near their own no-load speed. Due to the inefficiency of existing power storage systems, current consumption is a critical consideration. In combination with ground mobility difficulties enumerated above, this power inefficiency definitively eliminates using propeller thrust for terrestrial locomotion.
Alternatively, a single motor could be used to both drive the propeller in flight and drive the wheel-legs for walking in a manner akin to insects using large muscles to drive their body-coxa leg joints and flap their wings [40][41]. While this method would again save the mass of the terrestrial drive motors, a necessary clutch and transmission would significantly increase the complexity of the system, and offset some (or all) of the
22
mass savings. The clutch would be needed to at least disengage the propeller during
crawling, and potentially disengage the wheel-legs during flight. A transmission would
be required because the wheel-legs rotate at a maximum of 100 rpm, while the propeller
spins at nearly 20,000 rpm. The complexity was considered to outweigh the potential
mass savings, rendering this method undesirable for the early MALV prototype.
3.2.2 Wheel-legs
Wheel-legs are necessary for ground mobility. However, their implementation
was reconsidered to improve the overall performance of the vehicle. Past Mini-Whegs
robots used four wheel-legs driven by a single propulsion motor [12]. While this
arrangement confers an excellent combination of speed and mobility on the vehicle, a
four wheel-leg arrangement was contradictory to both the geometry of the existing MAV
design and the aerodynamic performance of the hybrid vehicle. The front wheel-legs are
more important because they reach in front of the vehicle and on top of obstacles in the
vehicle’s path to lift and pull the vehicle forward. Mini-Whegs™ are designed this way to
model the front legs of cockroach, which lift high and in front of the animal to overcome
obstacles [8]. To reduce the mass of the vehicle and bring the terrestrial drive system into
compatibility with the geometry of the existing MAV fuselage, testing of the ground
mobility of a mock vehicle was executed with two wheel-legs instead of four. The wheel-
legs were placed in the front and to the side of the propeller, and the rear of the fuselage
was permitted to drag on the ground. This mock vehicle was able to move forward over
obstacles similar in height to a Mini-Whegs robot with comparably sized wheel-legs.
Furthermore, it was observed that the fuselage benefited obstacle climbing by acting like a tail that prevented the robot from flipping onto its back, which happens when a purely
23
terrestrial vehicle attempts to surmount obstacles tall relative to its length. The drawback
to this design is that the mock vehicle’s mobility in reverse on rugged terrain was poor
because the fuselage impacted irregularities and impeded motion. However, the weight
savings justified the two-wheel leg design.
3.2.3 Steering
Mini-Whegs robots use a rack and pinion or four-bar mechanism to steer the front
legs. Another common steering technique for small and large robots is differential
activation, in which the running gear on one side of the robot is driven at a different speed (or in a different direction) than the running gear on the opposite side. A trade-off analysis was performed between these two steering techniques. Each design requires two
motors – two standard DC servomotors on the differential steering method, and one
standard DC servomotor for propulsion and one R/C servo for steering on the rack and
pinion method. While the R/C servo weighs less than the second DC servomotor, the
mass savings is not large. Any mass savings is further mitigated by the universal (UV)
joints that are required to transmit the motor torque to the wheel-legs through the steering
angle on the rack and pinion method. While the mass was roughly equal between the two
steering techniques, complexity, durability, and aerodynamics strongly favored the
differential drive method.
The aforementioned universal joints significantly increase the complexity of the
rack and pinion system. Commercial off-the-shelf (COTS) UV joints are not available at
a size appropriate to the proposed vehicle, and custom-made ball and cup joints are used
on all existing Mini-Whegs robots.
24
Birds develop nearly zero airspeed immediately prior to perching. While MALV
could reduce its airspeed prior to landing using a technique called flaring, it will not be able to approach zero airspeed. Therefore, the terrestrial running gear are likely to experience large impulse loads during landing, and the UV joints would be a prime candidate for damage.
To imbue MALV with the best possible obstacle mobility, the wheel-legs must reach in front of the propeller [8]. This means that the minimum track (lateral distance between the wheel-legs) is the diameter of the propeller. The rack and pinion method requires that the wheel-legs be mounted even further apart, so that they do not interfere with the propeller when they are turned. The result of this geometry is that either 1) the
UV joints and the steering knuckles that support the UV joints will be mounted outside the fuselage, or 2) the fuselage width will be increased to include the UV joints and steering knuckles. Either of these conditions will adversely affect the aerodynamics of the hybrid vehicle.
3.2.4 Wings
The word critical perhaps underestimates the importance of wings to flight phase of the hybrid vehicle being developed. However, on the ground, the wing is an impediment to locomotion, especially in narrow spaces. Birds and insects fold their wings when they are on the ground for a variety of reason, one of which is to eliminate impediments to motion. A wing folding mechanism was therefore identified as a crucial subsystem for the overall utility of MMALV.
25
3.3 Design Process
The first step of the design process was to estimate the masses of the terrestrial
locomotion subsystem and the sensor subsystem. These would be the payload of the
aerial platform. Drawing on the experience of MAV designers at University of Florida
(UF), the mass of the aerial locomotion subsystem would be estimated, and a wing
designed that could develop the necessary lift within a maximum wingspan of 30.5
centimeters (12 inches). Naturally, the wing parameters would depend on the desired
cruise speed of the vehicle, which was limited by the aerial power plant (motor and propeller). Once again, MAV experience suggested a reasonable cruise speed of 7-9 meters per second (15-20 miles per hour).
3.3.1 Terrestrial Drive System
Having determined that two independently driven wheel-legs would serve as the
terrestrial locomotion subsystem for the hybrid vehicle, specification of the motor and
drive train were still required. DC servomotors of an appropriate size for this application
all have a consistent shortcoming; these motors are intended to interface with axles that
are supported by a pair of bearings – their output shafts are short (< 1 centimeter) and are
not threaded on the end. There is no acceptable method to mount the running gear
directly to the motor output shaft. R/C servos, on the other hand, are designed specifically
to support small mechanical elements without the need for external bearings. These self-
contained electromechanical devices have a small motor, a transmission, a potentiometer,
and the electronics necessary to perform angular position control of the splined output
shaft. A threaded hole in the end of the output shaft allows for positive attachment of the mating servo horns, which are available in circular, two-arm, and four-arm varieties.
26
While the electronics normally limit the rotational motion of the output shaft to ± 60°, the
servo can be modified for continuous rotation by disconnecting the potentiometer from
the electronics and replacing it with a set of appropriate size resistors. A technology
survey was performed to identify the optimal R/C servo available within the appropriate
size range. The MPI MX-50HP was selected for this application. The independently driven wheel-leg system required the use of a 5-channel R/C receiver, rather than the 4- channel receiver used on previous MAVs. The total design mass associated with the terrestrial drive system was therefore 25 grams
3.3.2 Sensor system
A survey of COTS electronics identified a micro video camera available at 3 grams, an 80 milliwatt transmitter available at 1 gram, and a 5 Volt regulator at 1 gram.
The base sensor system was available for 5 grams.
3.3.3 Fuselage
The critical design criterion of the fuselage is to facilitate placement of the center
of gravity (CG) at the desired location with respect to the wing. Previously, this was
accomplished by placing the relatively massive battery at the extreme fore of the craft.
Because the wheel-leg drive motors must necessarily occupy that space on the hybrid vehicle, the wing was moved aft. An approximate mass balance was performed using the components listed in Table 1, to determine the necessary length of the nose on the hybrid
vehicle.
27
Table 1. Electromechanical Components of the First Prototype MALV
Component Description Mfg P/N Propeller motor Feigao 1208425L Terrestrial drive motor Maxx Products MPI MX-50HP FM receiver FMA Direct M5 (Futaba shift) Propeller motor controller Castle Creations Phoenix-10 Steering servos AeroMicro Cirrus CS-4.4 Batteries Thunder Power TP480-2SJ Video camera Supercircuits PC208 Video transmitter RF-Links SDX-22 Voltage regulator Dimension Engineering DE-SW050
3.3.4 Wing
Prior experience suggested that the aerial platform would weigh approximately 80
grams. Adding the masses for the terrestrial locomotion (25 grams) and sensor (5 grams)
subsystems, along with a 10 gram factor of safety, the predicted vehicle weight totaled
120 grams. Beginning with an estimated total vehicle weight of 120 grams and the established maximum dimension of 30.5 centimeters, a wing was designed that would provide the necessary flight characteristics for the vehicle. This was done using wing design software previously developed at University of Florida. As the wing parameters
(including wingspan, root chord, sweep, ellipse ratio, and curvature) are entered, the graphical user interface displays in real-time the top, side, and front projections of the wing. The program also outputs key geometric features (planform area, aspect ratio and aerodynamic center) and 2D aerodynamic estimates (coefficients of lift, moment, and
drag, and the lift-to-drag ratio). The low aspect ratio and the inherent flexibility of the
wing naturally compromise the accuracy of the 2-D aerodynamic estimates produced by
28
the software. However, this software had previously been used by UF to produce
successful MAVs. The wingspan was fixed by the original SBIR solicitation, and a 15
centimeter root chord was selected based upon the vehicle overall length restriction (30
centimeters), the estimated length of the nose (as determined in the previous section), and
the predicted size of the vertical stabilizer (as estimated from prior experience). The
fabrication process does not support dihedral in the wing. The remaining search space is
the airfoil profile. A brief search was performed over this space to identify an airfoil that
would produce the desired lift, minimize pitching moment, and minimize the maximum
camber of the wing.
3.4 Fabrication Process
The first prototype was made using the fabrication process developed by UF over
many years of MAV development [14]. It is presented here only with sufficient detail to
understand the alterations that were made to the process, as discussed in Chapter 4.
3.4.1 Wing and Tail
The wing design software outputs a script file containing the desired shape of the airfoil. This is imported into a CAD/CAM package that outputs tool paths to machine the desired shape into a piece of epoxy modeling foam. The tail section is laid-up on a flat piece of the foam. Other than the shapes of the substrates, the lay-up process is identical for the wing and the stabilizer sections. The lay-up materials are mounted to the substrate in the following order:
Teflon film – to prevent the resins and adhesives from bonding to the foam
Wing/tail schematic – as a roadmap for placement of the carbon fiber materials
29
Teflon film – to prevent the schematic from bonding to the carbon fiber
Carbon fiber – one layer of unidirectional and one layer of woven carbon fiber in
all appropriate areas (Figure 3)
Fabric – pulled taut, ensuring no wrinkles appear in the fabric
Carbon fiber – the remaining layers of carbon fiber, for strength, and to sandwich
the fabric between layers of carbon fiber (Figure 4)
Figure 3. The first layer of carbon fiber
30
Figure 4. The fabric is sandwiched between layers of carbon fiber
3.4.2 Fuselage
Fuselage fabrication is more complicated. Top and side views of the desired fuselage shape are printed to scale and then adhered to the rectangular solid piece of
epoxy modeling board in the appropriate orientations. Lengthwise lines on the fuselage
projection prints are used to align the print-out with the blank, and the lateral lines are
used to align the two projection views with one another. After the projections are
properly mounted to the blank, a jigsaw is used to cut out the top profile of the fuselage.
This results in the original block being cut into three pieces (the fuselage and a scrap
piece on either side), which are then taped together with masking tape. This is done
because the side profile is currently glued to one of the pieces of scrap, while the other
piece is necessary to orient the blank orthogonally in the jigsaw. The side profile is cut out using the jigsaw, and the corners are rounded using a table-top belt/disc sander. The
31
mold is then mounted in a milling machine and a groove is milled into the nose of the
mold to create an indentation in the fuselage to which the motor will be mounted.
The resulting fuselage tool is wrapped tightly in Teflon film, and then wrapped in
two layers of woven carbon fiber. Excess carbon fiber is cut away, as shown in Figure 5.
Flanges on the fuselage at the interface with the wing are necessary to provide surface
area onto which to bond the wing. A thin copper plate or piece of heavy paperboard
serves as a support against which to form the flange. The support must be held snug to
the fuselage to ensure that the flange shape matches the wing airfoil shape. First, the
support is mounted to the mold with a piece of two-sided tape, and then tape around the fuselage reinforces this, as shown in Figure 6.
Figure 5. The fully wrapped fuselage tool, including indent for the propeller motor
32
Figure 6. How to create the flanges to which the wing will be glued
3.4.3 Wheel-legs
As stated previously, R/C servos have horns that mate to the output shaft spline and can be screwed to the shaft. The modularity of these horns allowed multiple wheel- leg styles to be tested. Figure 7 shows three wheel-leg styles, along with the tools used to
fabricate the carbon fiber wheel-legs. The conic shape in the center of each tool allows
the wheel-legs to reach around and in front of the propeller, while minimizing fuselage
width and keeping the servos mounted closely to the fuselage. A trimmed servo horn is
placed at the top of each splaying cone, allowing the horn to be molded directly into the
wheel-leg. This both reinforces the servo horn and avoids separation of the spokes from
the horn. A later version of the wing tool includes channels that ensure the consistency of
the carbon fiber spokes and facilitate the formation of a “foot” at the end of each spoke.
Also shown in the lower-right of the figure is a third wheel-leg design, comprising four
0.025 inch diameter music wire spokes bonded to a round servo horn.
33
Figure 7. Wheel-legs and two lay-up tools
3.5 Results
The design and fabrication process outlined above resulted in the MALV shown
in Figure 8. As a radio-controlled (R/C) vehicle, this prototype performed most of the
mission scenario elements enumerated above. After being hand-launched, it could fly, land, and crawl with a payload of two miniature cameras, one video transmitter, and a switch to control which camera image was transmitted. The vehicle could also walk off the edge of a parking garage and regain flight after dropping approximately 7 meters.
34
Figure 8. First prototype MALV
The first prototype flew at a cruising airspeed of 11 meters per second (24 miles
per hour), which correlates to a Reynolds number of approximately 1 x 105. This is well
within the range where the efficiency is compromised. Tests demonstrated a maximum
flight time of approximately 15 minutes. At its cruising speed, the vehicle then has a
maximum range of 4.9 kilometers (3 miles). The terrestrial drive system could develop a
maximum speed of 0.33 meters per second (1 body length per second), and can surmount
obstacles of maximum height 44 centimeters (1.75 inches). Power consumption was naturally much lower on the ground, and studies estimated a maximum terrestrial-only mission time of 100 minutes. These results are summarized in Table 2.
35
Table 2. Performance Characteristics of First Prototype MALV
Parameter Value Cruising air speed 11 m/s Reynolds number ~1 x 105 Maximum flight time 15 min Range (round trip) 4.9 km Maximum terrestrial speed 0.33 m/s Maximum crawling time 100 min Range (round trip) 0.99 km Maximum obstacle height 4.4 cm
The total mass of the first prototype was 118 grams, including the sensor system.
The original design goal of a 12 inch maximum dimension was met, but both the
wingspan and the overall length were 12 inches. A fuselage width of 5.1 centimeters was
just sufficient to house the two terrestrial drive motors back-to-back. A leg length of 4.2
centimeters was the maximum that could be attained without interfering with the wings.
A wheel track of 12.4 centimeters allows the wheel-legs to reach around and forward of
the propeller. These parameters are summarized in Table 3.
Table 3. Mechanical Properties of First Prototype MALV
Parameter Value Total mass 118 g Overall length 30.5 cm Wingspan 30.5 cm Fuselage width 5.1 cm Leg length 4.2 cm Track (distance between wheel-legs) 12.4 cm
36
Table 4 summarized the parameters of the aerodynamic components of the
vehicle. Figure 9 shows an image capture of the video feedback from an airborne MALV
prototype. Roads, cars, and individuals are clearly discernible. This confirms that a vehicle of this size can fly stably enough to provide useful surveillance.
Table 4. Aerodynamic Properties of First Prototype MALV
Parameter Value Wingspan (b) 30.5 cm Aspect Ratio (AR) 2.55 Wing loading 31.7 N/m2 Wing area (S) 364.4 cm2 Location of CG from wing leading edge - 2.6 cm Horizontal stabilizer area 52.3 cm2 Elevator area 24.4 cm2 Vertical stabilizer area 52.9 cm2 Rudder area 5.0 cm2
Figure 9. Surveillance image from on-board video camera
37
In addition to the hybrid vehicle with a fixed wing, a prototype wing-folding
vehicle was fabricated. Figure 10 shows a non-actuated mock-up vehicle in various
stages of wing retraction. High friction levels led to a new fabrication process. Figure 11
shows the second wind design. The bearing surface is created by laying-up multiple layers of woven carbon fiber with a layer of Teflon film in between. Figure 12 shows an
implementation of this folding wing on a prototype MALV. One modified servo winds up
two cables that rotate two pulleys to cause the wing folding. In this model, the wings
must be manually deployed, as there is no return mechanism. Inaccuracies inherent to the
carbon fiber fabrication process cause the two wings to deploy inconsistent from one
another, as can be seen in Figure 13.
Figure 10. Wing folding sequence of the non-actuated model
38
Figure 11. Folding wing
Hubs
Retraction cables
Wing retraction servo and pulley
Figure 12. Wing-folding mechanism implemented on prototype
39
Figure 13. Wing-folding prototype vehicle
3.6 Critical Analysis of Prototype Hybrid Vehicle
A rigorous design review process was performed to identify potential improvements to the prototype design. The vehicle design was evaluated at each step in the fabrication and assembly process to identify any negative interactions between the design and the assembly/fabrication process. The vehicle design and fabrication processes were evaluated as they relate to the trimming procedure and functionality within the scenarios describe in Chapter 1. Finally, the fabrication process itself was assessed for its impact on the evaluation criteria. The following evaluation criteria were considered:
functionality of the vehicle
reproducibility of the vehicle
durability of the vehicle
ease of vehicle fabrication and assembly
The critical design evaluation yielded multiple areas for improvement over the initial prototype; these desired improvements were divided into primary and secondary goals. Primary goals are those that were deemed absolutely necessary to the successful
40
field deployment of MMALV. Secondary goals are those that would improve the
operational utility of the vehicle.
3.6.1 Primary Goals
3.6.1.1 Autonomous Operation
While the flexible wing theoretically improves vehicle stability over a rigid-wing
configuration, piloting the prototype still required the full attention of the operator. The
small size of the craft made it practically undetectable beyond 200 meters. To overcome
this, the operator wore a set of goggles that provided him with the view from the
“cockpit” of the vehicle. These goggles significantly restricted the operator’s local field
of view. Even at shorter ranges, discerning the vehicle’s orientation was difficult, and the
lag required to make this determination could easily result in a crash. It was necessary for
the operator to remain intently focused on the vehicle. The applications enumerated in
Chapter 1 require that an autopilot either assist an operator or that the vehicle be
completely autonomous. Therefore, the functionality review identified autopilot implementation as a primary goal for this work.
A survey was conducted of commercial-off-the-shelf (COTS) autopilots, and the
Procerus Kestrel was selected based on mass and volume. Although the Kestrel is the lightest full-function autopilot available, the total mass increase associated with this model is 62 grams. While the first prototype was able to perform many of the necessary tasks, the sensor system nearly consumed its entire payload capacity. Early tests suggested that 16 inch wingspan wings would be able to produce sufficient lift for the autonomous vehicle. This aspect was addressed first because the increased size and mass
41
would need to be considered in the execution of the remaining performance
improvements identified by the critical performance review.
3.6.1.2 Improved Durability
Lack of durability was the most significant shortcoming of the first prototype,
primarily because it represented multiple hurdles to the field deployment of MMALV.
Operationally, low durability would result in short vehicle life-span. Due to the
characteristics of the vehicle, even the most controlled landing strongly resembles a
crash, from a dynamic loading standpoint. For example, video shows that the vehicle
decelerates horizontally from 11 meters per second to zero meters per second in less than
0.25 seconds. This corresponds to an acceleration of ~4.5 times the acceleration due to
gravity and an average braking force of 5 Newtons. This example is for landing on
asphalt only. Landings on gravel and grass resulted in much shorter deceleration times,
and therefore higher average decelerations. Furthermore, the vertical deceleration is
nearly instantaneous. Due to the nature of impact mechanics, the maximum force
experienced is higher than the average impulse load.
This impulse loading is why testing of the first prototype demonstrated that the
vehicle could survive only a limited number of deployments. The review panel agreed
that, for MMALV to gain acceptance among warfighters and first responders in the field, the durability of the vehicle would need to be improved. In addition to adversely affecting the acceptance of MMALV, low durability made it difficult to field each vehicle on an individual basis. The “hand-crafted” fabrication and assembly process associated with MMALV requires that each vehicle undergo several trim flights. During this process, R/C controller settings and vehicle control surfaces are adjusted to produce the
42
desired flight characteristics, namely straight level flight when the elevator/rudder control
input is zero. Variations between distinct vehicles caused very different initial flight
characteristics between those vehicles, often resulting in crashes on the first trim flight.
Due to low durability, these crashes often resulted in component separation.
3.6.1.3 Repeatable Fabrication and Assembly Procedures
MMALV’s size lends itself to hand fabrication, and this production process will naturally lead to differences between nominally identical vehicles. However, it is critical to maintain the highest possible level of fidelity between consecutive vehicles, for purposes of the trimming process and subsequent aerodynamic performance. The critical design review identified several facets of the first prototype design that represented particular hurdles to maintaining this fidelity during the fabrication and assembly process.
motor and component mounting
tail assembly
fuselage mold design and fabrication
3.6.2 Secondary Goals
3.6.2.1 Improved Deployment
The importance of being able to deploy MMALV rapidly in military situations is
clear. However, several of the scenarios above require deployment of multiple vehicles,
which would also benefit from being able to deploy the sensor platform very quickly.
Furthermore, these scenarios require that the pre-flight process be not only short, but straight-forward as well, providing as few opportunities for error as possible. Lastly, the autopilot initialization procedure must be performed on a palm-top computer, as the first responder is unlikely to have available in the field a laptop or notebook PC.
43
3.6.2.2 Enhanced Flight Performance
While the first prototype produced sufficient lift to carry the cameras and
transmitter as discussed above, those components nearly consumed the entire payload
capacity of the vehicle. The additional mass associated with an autopilot and added
durability would easily overload the 30 centimeter wingspan wings. However, any
increase to the wingspan of the vehicle would adversely affect MMALV’s mobility
during terrestrial operations. Therefore, optimization of the wing lift would be important to minimize the wingspan and, subsequently, maximize the mobility.
3.6.2.3 Improved Wing Folding
The initial program description called for a vehicle capable of reconfiguring for terrestrial locomotion. While the first prototype demonstrated terrestrial locomotion
without the need for reconfiguration, the mission utility review reinforced wing folding
as a goal for the program, due to its multiple benefits:
increased storage and portability
wing fabric protection
increased ground stealth
increased mobility in confined spaces
3.6.2.4 Ground Take-off
While the initial Air Force mission scenario called for MMALV to regain flight
by walking off the edge of a rooftop, imbuing MMALV with the ability to take-off from
the ground would significantly increase its mission applicability. In several of the
missions enumerated in Chapter 1, the vehicle would have to be retrieved by emergency
personnel, or sacrificed to the elements. Ground take-off would allow the vehicle to
44
“retrieve itself.” Furthermore, other military missions for which ground take-off would
enhance MMALV’s utility are under-vehicle IED inspection and long-term surveillance.
The functionality review identified the former as a near-term application for MMALV, as
it would allow for rapid deployment of a sensor platform from a safe stand-off distance.
However, to retrieve MMALV may not be feasible, and allowing the technology to fall into enemy hands is highly undesirable. Long-term surveillance would involve MMALV landing on the ground, concealing itself, and observing a route or installation for multiple days or even weeks. This mission was suggested during the review at AFRL/MNAV.
Again, retrieval of the vehicle would likely be impossible. In both cases, ground take-off would allow MMALV to return to its launch point or a safe location for retrieval.
3.6.2.5 Tail Hook
To execute the initial mission scenario of rooftop surveillance, the review team agreed that a tail hook would be required. Nearly all flat-roofed buildings have a short parapet wall around the roof. For MMALV to crawl to the edge of the roof and perform surveillance would require mounting this parapet wall. The preferred option for executing this was identified as a tail hook. Furthermore, the tail hook would allow a general
“perching” behavior, which was identified as a useful element in several other mission scenarios. One such mission was for maritime interdiction. Allowing MMALV to hang from one of the high cables of the interdicted ship would provide the boarding party with a crow’s nest view of the activities on the deck of the ship being boarded.
45
Chapter 4 Airframe Mechanical Design
The critical design review process identified lack of durability and lack of repeatability as primary hurdles to the actual deployment of MMALV. The work presented in this chapter was intended to precisely identify and eliminate the causes of these shortcomings, thus conferring the maximum level of durability and repeatability on subsequent editions of the hybrid vehicle. Several design and fabrication aspects were identified as the specific sources of poor durability and inconsistency.
Fuselage Mold Fabrication – Hand fabrication of the fuselage was identified as
both a direct and indirect source of inconsistency and as an indirect source of poor
durability. The new fuselage mold is fabricated on a computer numerical
controlled (CNC) milling machine. Section 4.1 discusses how this directly and
indirectly improves durability and consistency.
Component Mounting – A primary source or poor durability was the method by
which components were attached to the fuselage, particularly the propeller motor
and the control surface servos. Section 4.2 presents changes to the design of the
fuselage (facilitated by CNC fabrication) that led to improvements in component
mounting.
46
Tail Design – As will be discussed in Section 4.3, the design of the tail, vis-à-vis
hand fabrication, led to inherent difficulties in repeatable assembly, which was a
major problem because the tail has a strong influence on stability. While other
unrepeatable facets of the assembly process could be made within adequate
tolerances with only a reasonable level of care, the difficulty of getting the tail
assembly adequately accurate was such that it further aggravated the non-
repeatability of the component.
Tail Alignment – As their very purpose is to control the yaw and pitch of the
vehicle, the importance of mounting the control surfaces (and tail) in alignment
with the fuselage is readily apparent. Section 4.3 demonstrates how tail alignment
is significantly improved in the new design.
Component Placement – The aerodynamic performance of a vehicle this size is
very sensitive to the placement of the center of gravity (CG), and the location of
the CG is sensitive to the location of each of the components within the vehicle.
Section 4.4 presents the process by which the accuracy of component placement
is enhanced in the new design.
4.1 Fuselage Mold Fabrication
Hand fabrication of the fuselage mold was identified as a direct hindrance to vehicle repeatability. If the fuselage mold were to become damaged, it would be impossible to create an exact replica. So, even if repeatability were attainable given the current fuselage mold, this repeatability could not be guaranteed in the long term.
Furthermore, the fabrication process, as described in Chapter 3 does not guarantee
47
symmetry of the fuselage, which is important to aerodynamic performance. Finally, as
will be discussed in the following sections, hand fabrication of the mold was identified as
a contributing factor to difficulties in component mounting, component placement, and
tail alignment.
CNC fabrication is an obvious solution to the problem of a hand-fabricated
fuselage mold. However, a few hurdles needed to be overcome in the design of the mold.
One such hurdle was the inclusion of wing mounting flanges on the top of the fuselage. If
the fuselage sides were flat and parallel from the tip to the tail, machining a contour into the mold to control the shape of the flanges would be simple enough. However, parallel sides from tip to tail are contradictory to the inclusion of a tail boom (Section 4.2.2).
Furthermore, it is advantageous to also have flanges at the top of the tail boom, for the
securing of the tail. This additional geometry is further in opposition to parallel sides
from tip to tail. Finally, the process of removing the fuselage from the mold involves
prying the fuselage away from the mold sufficiently to introduce a solvent, such as
denatured alcohol, into the space between the fuselage mold and the Teflon film, to
dissolve the spray adhesive used to adhere the film to the mold. Flanges on the mold (to
produce flanges on the fuselage) would limit the ability of the fabricator to introduce the
solvent and to obtain a positive grip on the fuselage for the purposes of pulling the
fuselage off of the mold.
Figure 14 depicts a rendering of the final fuselage mold. The mold comprises a
three piece system in which the center section controls the shape of the fuselage, and one
“fairing” on either side controls the shape of the wing mounting flanges. Three blind
holes at each interface house dowel pins to provide accurate alignment of the three
48
components, while two through holes allow threaded rod to be used to securely assemble
the mold. Each piece is wrapped in Teflon film, and then the parts are assembled. During
the lay-up process, carbon fiber at the top of the fuselage sides is folded over against the
bottom surface of the outer mold pieces. After the fuselage has been baked, the fairings
are removed from the main mold. This allows for easy access to the sides of the fuselage, for the introduction of solvent. Finally, removing the fairings provides direct access to the flat surfaces of the wing and tail mounting flanges, against which sufficient force can be applied to remove the fuselage from the mold.
Right fairing - attached to center mold section
Left fairing – aligned with but separated from Center fuselage mold center mold section
Figure 14. Rendering of MMALV fuselage mold
Two general differences which are readily apparent between the first prototype
and the final design are 1) the flowing contours of first prototype from tip to tail vs. the
flat sides of the final design and 2) the implementation of concave surfaces on the final
design, which are absent on the first prototype. Both of these differences are the result of
fabricating the improved fuselage mold on a CNC milling machine. The previous
fabrication technique involved hand machining the fuselage mold from a single piece of
49
epoxy modeling board. The shaping process was carried out on a band saw and a belt sander. While this did allow for the production of smooth contours over the fuselage, there were several drawbacks. First, the symmetry of the fuselage is highly dependent on the skill of the fabricator. Fuselage asymmetry could adversely affect aerodynamic properties. Secondly, producing the mold with repeatable accuracy was nearly impossible. This is important to batch production of the vehicle, and for replacement of the fuselage mold if it becomes damaged. However, the third and most critical drawback of the old process was the inability to include concave surfaces on the fuselage, which significantly limited the shapes and structures that could be built into the fuselage. The significance of this will be discussed in more detail below. The general shape of the fuselage is a result of the CNC milling process. Prior to machining, the entire mold had to be designed in a 3-D modeling environment, such as Solidworks™. While most 3-D modeling packages have the ability to create smooth contours on a complex volume through the use of blended surfaces, it is very difficult to maintain strict control of the blended surfaces. High fidelity structures and surfaces on the fuselage are critical for positive component placement. Therefore, the fuselage mold was designed using standard geometric shapes, which facilitate the design and fabrication of mating parts that will contour exactly to the fuselage.
4.2 Component Mounting
In addition to the general advantages garnered from implementation of CNC fabrication of the fuselage mold, several specific design and fabrication process enhancements were realized over the first prototype to address the shortcomings
50
enumerated previously. One of those collateral improvements was the security with
which components are mounted to the fuselage.
4.2.1 Motor Nacelle
On the first prototype, the motor mounting process presented a significant hurdle to both durability and repeatability. The motor was mounted into either a groove or slot on the top of the nose using cyanoacrylate (CA) glue. This was a significant source of compromised durability; crashes frequently resulted in the motor separating from the fuselage. Furthermore, the slot in the top of the nose was cut by hand using a rotary tool.
The accuracy with which this groove aligned with the fuselage longitudinal axis was highly dependent on the skill of the fabricator.
The final design includes a conspicuous motor nacelle (Figure 15). The front face of the fuselage provides a strong, fixed surface on which to mount the propeller motor.
This provides significant durability to the mounting system. In fact, a 350 gram prototype of the final design crashed nose first into the ground at full speed, and the motor remained securely attached to the nacelle. A portion of the nacelle is cut away to assist in motor mounting, and to provide cooling airflow over the motor. The speed controller is wired directly to the motor, and resides in the back portion of the motor nacelle.
51
Front mounting face
Motor nacelle
Figure 15. Close-up of the motor nacelle and mounting face
The nacelle design also allows for the precise, repeatable adjustment of the thrust line of the motor. This is useful to counteract the natural roll torque produced by the propeller. A vehicle the size of MMALV is very susceptible to the roll torque created by the propeller spinning through the air. Whereas large aircraft utilize ailerons mounted near the wingtips to counteract motor roll torque, a smaller craft such as MMALV uses instantaneous yaw to produce the desired roll. Yaw motion of the vehicle causes the airspeed over one wing to be higher than that over the opposing wing. The resulting difference in lift between the wings creates the controlling roll torque. During typical maneuvers, this initial yaw motion is produced using the rudder. However, using the rudder to counteract the continuous roll torque generated by the propeller would require
52
that the steady-state position of the rudder be at some slight deflection. In addition to the
added drag that constant rudder deflection would create, this technique will result in
dramatically different characteristics between left and right turns. The motor nacelle
allows the motor thrust line to be angled (about the yaw axis) with respect to the fuselage
axis, resulting in a continuous yaw torque that can be tuned to offset the roll torque of the
propeller during normal cruise flight.
4.2.2 Tail Boom
On the first prototype, the control servos were attached to the inside rear of the fuselage with a combination of CA glue and thread. Two holes were drilled in each side of the fuselage. Kevlar thread was woven in one hole, around the servo body, and out the other hole on the same side. This process was repeated several times. Once the servo body was pulled snug against the fuselage wall, a couple drops of CA glue were applied to the contact surface. This mounting process damaged both the durability and repeatability of the vehicle. Regardless of the amount of tension applied to the Kevlar thread, it was impossible to eliminate ALL motion of the servo body using only this mounting technique – thus the CA glue. However, a high impulse landing (crash) could cause the CA glue to fail in sheer, where it is weakest. The result would be control servos that were not firmly attached to the fuselage, meaning that subsequent control commands could result in an inconsistent combination of control surface motion and control servo motion. This would make the craft more difficult, or even impossible, to control.
The CNC fabrication process also allowed for the implementation of an integrated tail boom on the fuselage. As can be seen in Figure 16, a servo mounting bracket can be inserted into the tail boom (as shown by two arrows) and screwed down. Screwing the
53
servos to the bracket provide a positive attachment between the fuselage and the servos.
The rudder control servo mounts into the open slot on the mounting bracket as shown by
the single arrow. In addition to providing for durable attachment of the servo, the tail
boom also provides for consistent, repeatable placement of the servos. Furthermore, the
tail boom facilitates accurate, repeatable alignment of the tail, which will be discussed in
Section 4.3.
Rudder control servo
Elevator control servo Servo mounting bracket Rear portion of fuselage
Tail boom
Figure 16. Servo placement is both durable and repeatable within the tail boom
In addition to improving vehicle durability as discussed above, the tail boom streamlined both the assembly process and the trimming procedure. In the first prototype, the control servos were mounted inside the aft of the fuselage and the control rods passed through the fuselage wall. A control rod is a piece of stiff wire that transmits the motion
54
of the servo output horn to the control horn mounted on the control surface. Due to geometric constraints, attaching the control rods to the control horns and servo horns was quite difficult on the first prototype. Furthermore, adjustment of the control rod length during the trimming process was difficult. On the new design, the control rods are easily accessible, as shown in Figure 17, making them both easy to install and easy to adjust during trimming.
Rudder control rods
Figure 17. Control rod accessibility on new design
4.3 Tail
4.3.1 Assembly
The first prototype tail was a two part assembly constructed using a double-slot configuration, as shown in Figure 18. With the exception of motor placement, this was the single greatest obstacle to repeatability. While it would be possible to control the
55
width of the slot in each piece, controlling the thickness of the mating “tab” area was not possible. The result was that the upper and lower halves of the vertical stabilizer typically deflected in opposite directions, as shown (with exaggeration) in Figure 19.
Figure 18. Double-slot assembly of the first prototype tail
Figure 19. Deflection of vertical stabilizer halves (exaggerated)
4.3.2 Alignment
For alignment and mounting, the taper at the front of the horizontal stabilizer was intended to mate to the taper at the aft of the fuselage. A slot in the bottom of the fuselage
56
accommodated the vertical stabilizer. The assembled tail was aligned by hand and held
against the fuselage while CA glue was applied to the contact surface.
4.3.3 New Tail Design
Figure 20 shows the new tail design. While maintaining the horizontal and
vertical stabilizer areas, the new tail design eliminates the need to perform the double-slot
assembly process. Furthermore, the design facilitates accurate alignment of the tail during assembly. The rectangular opening in the middle of the horizontal stabilizer fits snuggly
around the control servos, which in turn fit snugly into the tail boom, as described above.
Also, by holding the control surfaces at a fixed distance from the control servos, the
lengths of the control rods are standardized from vehicle to vehicle.
Right Rudder
Elevator
Left Rudder
Servo Pocket
Figure 20. New Tail Design
57
4.4 Component Placement
The previous sections describe how the tail-boom positively dictates the position
of the control servos, which in turn dictate the position of the tail. It is critical that the remaining components of the craft be affixed in a similarly conclusive manner. While the motor nacelle provides a surface on which to firmly mount the motor, cutting away the top of the nacelle for motor placement and cooling purposes necessitates that the location of the mounting holes through the front surface be highly accurate. Furthermore, several other components (terrestrial drive system, battery, and autopilot) must be positioned within the fuselage so as to establish the center of gravity (CG) in the appropriate location.
Adoption of the CNC method for fuselage mold production provided a direct means by which to ensure accurate, repeatable component placement. As described above, the three piece mold system includes a series of five holes designed for alignment and assembly of the three mold pieces. Figure 21 depicts a rendering of a fixture by which the mounting holes for the various subsystems and components can be placed exactly where they are needed. An auxiliary mold center section is fabricated for this process. The auxiliary center section must be identical to the actual mold on which the fuselage is wrapped. After the fuselage is removed from the original mold, it is slipped onto the auxiliary mold. The bracket system, comprising three 1/16" thick aluminum plates, is attached to the auxiliary mold using the hole pattern discussed above. Attached to the 1/16" plates are multiple drill guides, which will ensure that the component mounting holes are located in the appropriate locations. Furthermore, the pocket cut through the side plate is used to mark the battery door cut-out. While hand machining of
58
the cut-out with a rotary tool will result is some inaccuracy, the battery door should
sufficiently cover those inaccuracies so as to not affect the aerodynamics of the vehicle.
Furthermore, a similar process could be used to mount the fuselage into a vise, so that the battery door cut-out can be machined using a CNC mill.
Battery door cut-out guide
Figure 21. Component mounting-hole fixture
4.5 Results
The changes described in this chapter successfully resolved hurdles to field
deployment created by lack of durability and lack of repeatability. CNC fabrication of the
fuselage mold ensures an unlimited supply of identical fuselage molds. This process also
confers the opportunity to implement small, precise changes to the design, if these are
deemed necessary in the future. In addition to these direct benefits, CNC fabrication
59
facilitates several other improvements to the design, fabrications, and assembly of the vehicle. A well defined motor nacelle and mounting face allows for accurate, durable attachment of the motor, including motor (mis)alignment to counteract the unwanted roll torque generated by the propeller. A tail boom provides accurate, durable attachment of the control servos, and simple inclusion and tuning of the elevator and rudder control rods. The tail subassembly process is eliminated by the single piece tail design, and the interactions between the tail boom and tail guarantee accurate alignment of the tail.
Finally, the CNC fabrication process directly facilitates the implementation of a fixture for the accurate placement of assembly holes to control the locations of the motor, terrestrial running gear, battery, and autopilot. Component placement is critical to control of the center of gravity of the vehicle. Figure 22 shows the resulting aerial platform.
Figure 22. Photograph of the new MALV aerial platform
60
Chapter 5 Terrestrial Running Gear Design
Terrestrial drive system fragility was the most persistent obstacle to field deployment of MMALV. While other sources of limited durability were apparent within the hybrid vehicle, those weaknesses typically came to light during less than perfect landing (crashes). For the terrestrial drive system, though, even the best landing resembles a crash, in as much as the large impulse loads that are experienced. This is due in part to the properties of the vehicle itself. Birds dramatically increase their angle of attack (called flaring) immediately prior to landing, which serves to both slow their speed
(through increased drag) and maintain a reasonable amount of lift. Birds, however, are able to flap their wings to both generate lift at zero airspeed and develop control torques at zero airspeed. While MMALV could use its elevator to emulate the flaring behavior, there are several reasons why this is not feasible. Driven by the desire to minimize the physical size of the vehicle, the wing loading is very high for MMALV. This causes control difficulties during low speed, high angle of attack (high-α) maneuvers.
Furthermore, even if the vehicle could perform such high-α maneuvers, it would need to be timed to occur when the vehicle was very close to the ground. However, small autopilots such as the one selected for MMALV use a pressure transducer to estimate the altitude of the vehicle. This technique suffers from sufficient inaccuracy that the manufacturer is investigating using an infrared rangefinder. This option, however, is not
61
yet available, and would increase the mass of the unit by 25 grams. Furthermore, it has
not been demonstrated that the autopilot can sufficiently control the MMALV during high alpha landings so as to not degenerate into a crash.
5.1 Larger direct drive system
Several iterations of wheel-leg drive systems were investigated before a
functional combination was identified. The first prototype implemented an R/C servo,
which had been modified to allow for continuous rotation, to drive each wheel-leg. For
convenience, this was the first method attempted on MALV 2. The wingspan was
increased partly to compensate for the increased mass of the terrestrial drive system. This
allowed for the selection of a larger R/C servo. The Blue Bird BMS-380 Max was chosen
because it has a metal geared transmission. However, while this did mitigate certain
failure modes of the terrestrial running gear, others were found, as shown in Figure 23.
Figure 23. Even large, metal-geared servos are damaged in landings
62
All R/C servo bodies are made of injection molded plastic, which is simply unable to withstand the impulse loads experienced at landing. Therefore, the first step in improving the durability of the system was to replace the R/C servo with a standard DC gearmotor. This requires the addition of a speed controller, and has other drawbacks.
Unfortunately, even the best COTS gearmotors are only designed to support a small radial load on the output shaft. For example, on a 10 mm Maxon gearmotor with a 64:1 transmission ratio (a reasonable candidate for the MMALV terrestrial drive system), the recommended maximum radial load on the output shaft is 1531 grams at 5 mm. This corresponds to a 5g deceleration for a 300 gram vehicle, which is insufficient to support the impulse load at landing. Furthermore, available gearmotor output shafts tend to be on the order of 1 centimeter long, and there is no way to positively fix an appendage onto the shaft. Considering these characteristics of gearmotors, it was apparent that an indirect drive system would be required.
5.2 Indirect Drive System
Figure 24 shows a photo of the initial wheel-leg drive system mounted in a scaled-up version (Fuselage A) of the prototype fuselage. Early testing confirmed that a timing belt system could successfully transmit motor output torque to the wheel-legs.
This system was investigated first for two reasons; it provides high durability, and it offers maximum flexibility on the placement of the gearmotor with respect to the wheel- leg axle. The initial Phase II fuselage, Model A, was insufficiently wide to place the motors end to end, and insufficiently tall to stack the motors. The belt drive system facilitates the nested motor arrangement depicted in Figure 24. The timing belt system is
63
also durable because multiple belt teeth are fully engaged with the grooves on the pulley
at all times, which functions to distribute the load across those teeth.
Left wheel- Right wheel- leg axle leg axle
Left wheel- Left wheel- leg drive leg drive motor pulley
Right wheel- Right wheel- leg drive leg drive motor pulley
Figure 24. Initial Terrestrial Drive System in Model A Fuselage
While radial loading of the gearmotor output shaft is avoided through this
implementation, tangential loading of the wheel-legs is still transmitted to the gearmotor.
Those gearmotors that would be candidates for the MMALV terrestrial drive system power plant all have very small gears. The teeth on these gears would likely be damaged by the high tangential loads experienced during landing. Therefore, the decision was made to implement a slip-clutch to limit the torque that is transmitted from the wheel-legs back to the gearmotor. A preferred torque limit of 120% of the drive gearmotor’s stall torque was identified, so that the gearmotor could transmit all available torque to the wheel-legs, while the wheel-legs could only transmit 20% more than the stall torque back
64
to the motor. Assuming a reasonable factor of safety in the gearmotor design, the teeth in
the transmission should be able to withstand 20% more than the stall torque.
5.3 Slip-clutch Hub
Initial implementations of the slip-clutch were carried out at the wheel-leg hubs.
In the first design (Figure 25) a soft polymer tube was placed between the wheel-leg axle
and the wheel-leg hub. While this arrangement was simple and showed potential during
bench tests, it was disqualified by early flight tests. Once the initial CNC fuselage (Model
B) was fabricated, sequential flight tests were performed. The initial flight test, during
which the craft demonstrated excellent flight characteristics, included only the aerial
locomotion platform. As ballast was added, to simulate the masses of the wheel-leg drive
system and the autopilot, the vehicle maintained desirable handling characteristics.
However, addition of the wheel-leg drive axles noticeably impaired the vehicle’s aerial
stability. Furthermore, inclusion of the wheel-leg hubs completely devastated the
prototype’s controllability. This was likely due to the aerodynamic properties of
cylinders. Each 0.5 inch diameter hub produced drag equivalent to a 1 inch thick airfoil 6
inches long, and this drag acted below the CG of the aircraft, making ascent difficult.
Furthermore, the shortness of the hub likely resulted in chaotic vortex shedding, completely disrupting the vehicle’s controllability. Furthermore, the breakaway torque of the system was impossible to control. Interference calculations showed that to obtain the desired breakaway toque required unattainable tolerances in the friction material.
65
Figure 25. Slip-clutch hub design
5.4 Angled Slip-clutch Hub
A second hub-clutch design (Figure 26) produced a more disk-like hub, rather than the cylindrical hub depicted above. The compressed spring creates a normal contract force between the outer stator (purple) and the wheel-leg hub (dark gray), and between the hub and the inner stator (burnt orange). The outer pin (green) transmits torque to the outer stator, and holds it onto the end of the shaft, while the inner pin (yellow) transmits torque to the inner stator. By varying the spring properties and the contact angle between the rotor and the inner stator, the normal force, and the resulting frictional force and breakaway torque of the clutch, could be tuned. However, this design suffered from low durability, as demonstrated during subsequent flight tests. Increasing the size of the components to impart the necessary durability would likely have resulted in similar flight characteristics as the previous method. The decision was made to move the clutch assembly internal to the fuselage.
66
Figure 26. Angled slip-clutch hub design
5.5 Parallel Shaft O-ring Slip-clutch
5.5.1 Test Apparatus for Torque Transmission Measurement
Figure 27 shows two renderings of a test rig for determining the torque transmission capacity of the various torque limiter arrangements. The torque limiter design being tested comprises two parallel, non-concentric spools that contact through a
soft polymer material. The first (or drive) spool (shown in red) is connected to the motor
output shaft (blue), and the second (or driven) spool (salmon) is connected to the wheel-
leg axle (yellow). Decreasing the gap between the drive and driven spools increases the
compression of the polymer material (dark gray), resulting in an increase in the contact
force, and subsequent increases in the maximum frictional force and torque transmitted
between the two spools.
67
Figure 27. Test Rig for Measuring Torque Limiter Transmission Capacity
5.5.2 Test Procedure for Torque Limiter Evaluation
The design of the test rig was also intended to mimic the deployment process for the vehicle. To maximize portability, it is imperative that the wheel-legs not be attached to the vehicle during storage and transport. However, any assembly required prior to deployment must be straightforward and convenient. One option would be to leave the
68
wheel-leg axle engaged with the drive system, and remove only the wheel-leg. However,
this would require a screw to attach the wheel-leg to the axle, or some other quick-release
mechanism. Screws are easy to lose, and would require the user to retrieve the appropriate tool. Quick-release mechanisms tend to be bulky, and the resulting hub size would adversely affect the aerial performance of the vehicle. If the wheel-legs are to
remain attached to the axles, then the axles must be disengaged from the drive
mechanism during storage and transportation. Once the wheel-leg axle is removed from the driven spool, the polymer material will push the two spools apart, as depicted in
Figure 28-1. During the pre-deployment assembly process, the polymer material must be recompressed. The following steps comprise the pre-deployment assembly process. The wheel-leg axle (yellow) is inserted through the outer bearing (light green) and into the driven spool (salmon) (Figure 28-2). A force (F) applied to the outer end of the wheel-leg
axle compresses the polymer (dark-gray) such that the wheel-leg axle aligns with the
inner bearing (green) (Figure 28-3). The axle is then inserted through the inner bearing
(Figure 28-4).
69
1) 2)
F
3) 4)
Figure 28. Schematic Representation of Pre-deployment Assembly Process
The test procedure was as follows. The motor shaft bearings (purple) were inserted into their respective pockets, and the two plates were assembled. The motor shaft was inserted through the bearings and the drive spool. The shaft clamp (cyan) was attached to prevent the motor shaft from rotating. A driven spool was selected, and the diameter of the driven spool recorded. The polymer material was placed on the driven spool. The driven spool was held between the plates while the pre-deployment assembly process, as enumerated above, was performed. Once the test rig was fully assembled, the lever arm (burnt orange) was attached to the wheel-leg axle, and progressively heavier weights were hung from the lever arm until the wheel-leg axle began to rotate, denoting slippage between the polymer and the drive spool. The torque at slippage was recorded.
70
Testing revealed the impracticality of this design. The first tests used an o-ring that produced very little interference with the two hubs. The torque delivered by this arrangement was well below the desired torque. As the thickness of the o-ring increased, the torque transmitted by the assembly increased. However, while the transmitted torque was still well below the desired value, the force required to assemble the unit (F in Figure
28-3) became unacceptably high, requiring a shaft extender to allow the testing personnel to apply the necessary force. A slight alteration to this slip-clutch resulted in the final design.
5.6 Concentric Shaft O-ring Slip-Clutch
Figure 29 shows an assembled view and Figure 30 shows an exploded view of the final design of the terrestrial drive system. Each wheel-leg is powered by a Solarbotics™
GM13a gearmotor (red), which produces 14.9 in-oz of torque at start-up, and 113 rpm under no load. A 26 tooth 48 diametral pitch gear (yellow) mounted to the motor output shaft adaptor (cyan) impels a 36 tooth gear (green) that is press fit onto the outer cylinder
(peach) of the friction clutch mechanism. The clamshell design of the friction clutch outer cylinder eases assembly of the unit. Three 5/16" I.D. x 1/2" O.D. quad-profile o-rings
(purple) transmit power from the outer cylinder to the inner cylinder (light blue). The normal pressure between the o-rings and the outer cylinder produces a break-away torque of approximately 166% of the motor stall torque (120% magnified by the 36:26 gear ratio). The O.D. of the inner cylinder is sized to produce optimal compression of the three o-rings. The two-piece inner cylinder has internal flats that positively engage the wheel- leg axle (light gray). The snap ring (black) at the end of the inner cylinder fits into the
71
groove on the wheel-leg axle, allowing for quick attachment and removal of the wheel- leg.
Figure 29. Assembled view of the terrestrial drive system power train
Figure 30. Exploded view of the terrestrial drive system power train
72
Figure 31 shows the three piece housing for the final terrestrial drive system. The
outer layer (salmon) houses two bearing, one for the output shaft of the motor output shaft adapter and one for the inner cylinder of the slip-clutch mechanism. The large hole in the middle section (blue) encloses the slip-clutch mechanism, and the small hole is a pass-through for the motor output shaft. Not visible from this side is a rectangular pocket
behind the motor axle pass-through hole. This pocket fits snugly around the outer end of
the transmission on the drive motor, and mates to the rectangular pocket on the inner
layer (green). Together, these two pockets secure the motor placement without requiring
screws. The inner layer also holds the inner bearing for the inner cylinder of the clutch
mechanism. Not shown in the figure is the snap ring housing, which mounts to the in-
board side of the inner layer to hold the snap ring in place as the wheel-leg axle is
inserted into the mechanism.
Figure 31. Exploded view of terrestrial drive system housing
73
5.7 Wheel-legs
Mini-Whegs robots implement machined acetal wheel-legs. The prevailing design is approximately a tear-drop shape, with a ¼ inch wide tread. While they provide excellent traction for the terrestrial robots, this design was deemed too heavy and insufficiently durable for implementation on MALV. The original prototype implemented straight-spoked music wire wheel-legs, as shown in the lower right of Figure 7. The spokes tended to snag on terrain features during landing, resulting in higher impulse loads than normal. Two designs were investigated during the vehicle redesign phase. Both were tear-drop shapes. The first design was fabricated from unidirectional carbon fiber (similar to the wheel-legs shown in Figure 8), while the second design is made of music wire.
Flight (landing) testing revealed that the stiffness of the carbon fiber design resulted in the impulse loads at landing being transferred almost completely to the drive train, including the axle. The result was that the axle bent on landing. Trial and error testing of the music wire design revealed a 0.062 inch wire diameter provided the desired combination of strength (to maintain shape during terrestrial locomotion) and compliance
(to reduce the impulse load transmitted to the drive train). However, the bare spokes were unable to provide sufficient traction on grass to provide consistent forward progress.
Acetal “feet” were machined and attached to the spokes using Kevlar thread (as shown in
Figure 32). These feet have survived two landings on asphalt, and improve locomotion speed on grass.
74
5.8 Results
The iterative design process described above resulted in a terrestrial drive system that successfully withstands impulse loads experienced at landing. In flight tests, the vehicle shown in Figure 32 has survived two crashes on grass, two normal landings on grass, and two landings on asphalt, and there is no sign of damage to the terrestrial drive system. The vehicle is capable of forward walking and walking turns. The wheel-legs can be inserted and removed from the drive system in a few seconds.
Figure 32. The current MALV design
75
Chapter 6 Airframe Aerodynamic Design
6.1 Wing and Aerodynamic Property Tests
6.1.1 Purpose
Wing selection for the first prototype MALV focused on the efficiency estimate calculated by the wing design software. Efficiency here is defined as the ratio of the coefficient of lift (CL) to the coefficient of drag (CD). The added mass of the autopilot and wheel-leg drive system improvements require a larger wing. The opportunity was taken to re-evaluate the wing design process. The nature of the hybrid vehicle renders total lift the dominant parameter, over efficiency. Consider, for example, two vehicles whose aerodynamic parameters are summarized in Table 5. The two vehicles are identical except that vehicle A includes a wing with CL = 1.0 and CL/CD = 5.0 at cruise speed, while vehicle B includes a wing with CL = 1.2 and CL/CD = 4.0 at the same cruise speed.
The apparent paradox that a less efficient wing can improve vehicle performance arises from the incongruity that the wing accounts for all of the vehicle lift, but only a portion of the vehicle drag. Therefore, a 20% increase in lift is only accompanied by an 18% increase in drag (in this particular example), even with a less efficient wing. However, the 20% increase in lift accounts for a 100% increase in battery capacity, which more
76
than compensates for the increase in lift. This extra lift could be used to increase the mission duration or to increase the payload of the vehicle.
Table 5. Hypothetical Vehicle Comparison
Vehicle A Vehicle B CL at cruise speed VC 1.0 CL at cruise speed VC 1.2 CD at cruise speed 0.2 CD at cruise speed 0.3 Efficiency (CL/CD) 5.0 Efficiency (CL/CD) 4.0 Lift at cruise speed 300 g Lift at cruise speed 360 g Form drag 100 g Form drag 100 g Wing (aerodynamic) drag 60 g Wing (aerodynamic) drag 90 g Total drag 160 g Total drag 190 g Power requirement 8 A Power draw 9.5 A Non-battery mass 240 g Non-battery mass 240 g Battery mass 60 g Battery mass 120 g Battery capacity 1800mAh Battery capacity 3600 mAh Mission duration 13 min Mission duration 22 min.
While the added mass is disadvantageous to the soldier and increases the wing loading, these drawbacks must be weighed against increased duration or increased payload. Rather than rely on a simulation, wind tunnel tests were performed on actual wings and on actual wings mounted to actual fuselages. The bulk of the work summarized in this section was performed, under the author’s guidance, by Mr. Timothy
Witushynsky as part of his Master of Science degree, after the author developed the test apparatus described next. The complete work can be found in [42].
6.1.2 Test Apparatus and Set-up
The mounting/measurement apparatus (referred to as the “sting”) for the wind tunnel test is depicted in Figure 33. The two dog-bone shaped cutouts create ideal locations for the placement of strain gauges. Four strain gauges (Figure 33 – depicted in red) on each dog-bone, arranged in a full bridge orientation, produce a voltage output that
77
is linearly proportional to lift (in the horizontal arm) and drag (in the vertical arm). The
necked section near the front of the horizontal arm produce surface strains that are linearly proportional to the moment applied to the sting. The shape of the dog-bones
theoretically decouple the lift and drag calculations, and result in translation of the beam
at the gauge mounting point, rather than bending. Beam bending at the vertical arm dog- bone would impart a pitch angle to the horizontal arm, thereby distorting the sensitivity of the lift and moment bridges.
Moment bridge
Vertical bridge
Strain gauge Horizontal bridge
Figure 33. Rendering of sting with wing mounting adapter
The outputs from the strain gauge full- and half-bridges were passed through a strain gauge amplifier to a data acquisition card. LABVIEW was used to sample the card and store the data to file. To average out inherent electronic noise and small vibrations in the apparatus, a 1000 Hz sampling frequency and twenty second collection interval were
78
used. The sting calibration procedure (presented in Appendix A) yielded the strain bridge sensitivity equations (Equation 1).
4 VLV 4.29 10 F 0.7771 45 VHH 4.39 10 F 1.02 10 F V 5.8424 (1) 3 VMtot 3.36 10 M 1.4929
VV, VH, and VM are the output voltages (in Volts) of the vertical, horizontal, and moment strain gauge bridges, respectively, FV and FH are the vertical and horizontal loads
(in mN) on the sting, respectively, and Mapp is the total moment (in mNm) measured about the physical midpoint between the moment bridge gauges.
The sting was mounted in the wind tunnel parallel to the direction of airflow. A wing was mounted to the sting at various angles of attack and tested as various airspeeds.
Any combination of high angle of attack and high airspeed led to unacceptable vibrations in the apparatus, with flexure occurring mainly at the moment bridge. A damper was mounted to the sting, as shown in Figure 34, to dampen out the vibrations.
79
Figure 34. Modified sting set-up
6.1.3 Data Collection and Analysis
Once airfoil vibrations were eliminated, it was necessary to verify that the airfoil maintained its intended shape. A narrow (5 centimeter wide) wing was fabricated from 10 layers of woven carbon fiber. This wing was mounted to a table and loaded at the leading edge with a mass 1.5 times the expected vehicle mass. Digital analysis of a lateral-view photograph confirmed that the airfoil maintained its original shape. All wings tested were fabricated with a 10 layer reinforced center section.
Having ensured that the airfoil maintains its shape, the remaining step was to verify the validity of the angle of attack measurements. High contrast markers were placed on the leading and trailing edges of each wing. A camera outside the tunnel was mounted exactly perpendicular to the airflow, at the trailing edge of the wing. The angle of attack was measured by importing the photograph into Photoshop™, drawing a line between the leading and trailing edges, and using the software to determine the angle of
80
the line. However, there was concern as to whether deflection of the wing would cause
parallax at the camera, thereby corrupting the angle measurement process. The entire
experiment was simulated in a 3-D modeling package that provides realistic perspective
renderings. These renderings were analyzed in the same manner as the photographs, and
the measured angle of attack agreed with the known actual angle of attack.
6.1.4 Wing Testing Procedure
Ten wings were fabricated for testing. While each of these ten wings had a 16 inch wingspan and 6 inch root chord, all other wing parameters were varied across the
test set. Two different airfoil shapes were used: 1) a reflexive airfoil shape used in the
Phase I prototype (referred to as the University of Florida, or UF airfoil) and 2) the curve corresponding to the mean camber line of the NACA 8300 airfoil. Three different planforms were used – rectangular, elliptical, and truncated elliptical. The rectangular planform was selected with the goal of maximizing lift, even if the coefficient of lift was not a maximum for that wing. The leading edge of the elliptical airfoil had an eight inch major axis and a 1.5 inch minor axis, while the trailing edge has an eight inch major axis and 4.5 inch minor axis. For the truncated elliptical wing, the major axes for the leading and trailing edges were changed to ten inches. Since the wing was 16 inches wide, this
resulted in a two inch chord length at the wingtip. The truncated elliptical planform was
employed so that winglets could be implemented on these wings. The winglets were one
inch tall by approximately two inches long. Two of the wings had a modified leading
edge. One concern for the thin wings was the knife-like leading edge. With this sharp
edge the possibility existed that a clear stagnation point did not form at the leading edge.
To investigate this, a one-eighth inch diameter carbon fiber tube was baked into the
81
leading edge, to provide a more classical rounded leading edge. Table 6 enumerates the various parameters of the ten wings selected for testing.
Table 6. Parameters of the Wings Tested
Wing Material/ Rounded Airfoil Planform Winglets Designation Flexibility LE Wing 0 NACA 8300 Rectangular Flexible No No Wing 1 UF Rectangular Flexible No No Wing 2 UF Rectangular Rigid CF No No Wing 3 NACA 8300 Rectangular Rigid CF No No Wing 4 NACA 8300 Elliptical Flexible Yes No Wing 5 UF Elliptical Flexible Yes No Wing 6 UF Tapered Flexible No Yes Wing 7 NACA 8300 Rectangular Flexible No Yes Wing 8 NACA 8300 Elliptical Flexible No No Wing 9 UF Elliptical Flexible No No
Each wing was tested at angles of attack ranging from approximately the zero lift angle of attack to the stall angle of attack, in one degree increments. Each wing at each angle of attack was tested under 0, 2.5, 5, 7.5, 10, and 12.5 m/s airspeeds.
6.1.4.1 Independent Wing Test Results
Table 7 summarizes values of interest for the various wings. Based upon these data, Wing 0 and Wing 4 were chosen for full vehicle testing. Wing 4 was selected because it demonstrated the largest CL and second highest efficiency. Wing 0 was selected based on a balance of overall lift, CL, and efficiency. While Wing 3 produced more lift and demonstrated a larger CL than Wing 0, the max efficiency of Wing 0 was appreciably higher than that of Wing 3. Furthermore, Wing 0 is a flexible wing, while
82
Wing 3 is solid carbon fiber, which would prohibit wing folding, an important capability
for MMALV field deployment.
Table 7. Summary of Data for Independent Wing Tests
Lift (mN) at Maximum α at Max Maximum α at Max Wing V = 12.5 m/s CL CL L/D L/D Flat Plate 6376 1.1 13.9° 13.3 3.8° Wing 0 8279 1.5 16.3° 11.5 5.5° Wing 1 7328 1.3 17.3° 10.9 7.9° Wing 2 7867 1.3 17.9° 9.9 7.8° Wing 3 9113 1.6 16.2° 10.3 4.5° Wing 4 7475 1.7 14.3° 11.9 7.5° Wing 5 6464 1.5 18.0° 11.4 8.8° Wing 6 6131 1.3 15.9° 10.9 8.5° Wing 7 7965 1.5 14.6° 10.2 4.7° Wing 8 6562 1.5 16.3° 12.5 4.8° Wing 9 5405 1.2 16.8° 9.5 5.2°
6.1.4.2 Wind Tunnel Testing of Aerial Platform
Based on results from the data analysis of the independent wing tests, two wings
were chosen for vehicle testing. Each wing was mounted to two different fuselages
(Figure 35) developed during the iterative design process. The primary differences
between the two fuselages are the width (wB = 2.0", wC = 2.8"), height (hB = 2.5", hC =
1.6"), and mounting angle of attack of the wing (αB = 8°, αC = 5°). A rigid tale section
was fabricated with eight layers of carbon fiber and mounted to the fuselage (Figure 36).
The vehicle was mounted onto the sting using the same angle of attack adapters as for the
independent wing tests. A similar image analysis was performed to determine the angle
83
of attack of the fuselage in each test; two points along the flat bottom of the tail boom
were manually selected to determine the angle of attack of the fuselage.
Fuselage B Fuselage C
Figure 35. Fuselage B and Fuselage C
Figure 36. Assembly for full aerial platform testing
Table 8 summarizes values of interest for the various vehicles. All angles of attack data refer to the angle of attack of the wing, which is equal to the angle of attack of the fuselage plus the mounting angle of attack of the wing with respect to the fuselage.
84
Also, the planform area used in calculating the aerodynamic coefficients was the full area
of the wing, even though part of each wing underside is occluded by the fuselage.
Table 8. Summary of Data for Vehicle Tests
Maximum α at Max Maximum α at Max Maximum α at Max Fuse Wing Lift (mN) Lift CL CL L/D L/D B W0 6916 14.3 1.5 17.8° 6.9 7.5° B W4 6680 13.9 1.2 17.0° 7.3 8.4° C W0 6896 13.8 1.9 16.6° 7.1 8.0° C W4 6445 13.5 1.6 15.4° 7.3 7.9°
The table shows that the four vehicles were fairly similar in their performance
characteristics, with the exception of the maximum CL for the Fuselage-B/Wing-4
combination. However, few conclusions can be reached from these data due to limited
test capability. By mounting the plane to the sting in the same manner as the wing, the moment arm of the lift force was significantly increased. This caused much larger angular deflections of the sting at the moment bridge location. Compounded by large mounting angles of attack of the wing with respect to the fuselage, very few data points were available for most of the aircraft at the higher airspeeds, which are closer to MMALV actual operating airspeed.
6.2 Power Plant
6.2.1 Purpose
The very nature of micro flying vehicles suggests that each component on board
be optimized in terms of mass. In the case of the power plant (motor/propeller/speed
controller combination), however, this single variable optimization is insufficient. As the
85
primary consumer of energy, it is important that the efficiency (ratio of power produced
to power consumed) of the power plant also be taken into account. This would generally
result in a multi-variable optimization involving the mass of the power plant, the thrust
produced by the power plant, CL and CD for the vehicle, and the energy density of the onboard energy storage system (in the case of MMALV, batteries). In some cases, a
heavier and less efficient power plant may be desirable, if the resulting increase in lift
(from greater airspeed due to more thrust) is sufficient to 1) offset the increased mass of
the power plant, 2) support additional power storage to overcome the decreased
efficiency, and 3) still provide additional payload. However, in the case of MMALV, the
landing factor is paramount. Increased airspeed implies that the impulse loads
encountered at landing will be greater, and the terrestrial running gear must be
proportionally heavier to sustain the increased loads. Furthermore, the landing distance
also increases proportionally to airspeed. Therefore, the decision was made to keep the
weight of the power plant as small as possible, and focus on maximizing the efficiency of
the power plant. The data collection presented in this section was performed by Mr.
Kenneth Moses, under the guidance of the author.
6.2.2 Test Apparatus for Thrust Measurement
Figure 37 shows a rendering of the mounting/measurement apparatus (referred to
as the thrust balance) for the power plant efficiency tests. The thrust balance comprises
two symmetric, parallel four bar linkages. The test motor is mounted in the slot of Link
A2, pointing upstream (to the right). An identical motor, equipped with an Advanced
Precision Composite 5.5x4.5 propeller, is mounted to Link B2 to counterbalance the mass
of the test motor/propeller. The design keeps Link 2 parallel to the airflow, and ensures
86
that Linkage A is geometrically identical to Linkage B. Therefore, the deflection angle of
Link A1 is identical to that of Link B1 (ψ), and can be measured by aligning a machined
slot along the axis of Link B1 with the 1° graduated protractor. The length of each Link 1
is 10 inches, placing the motor near the center of the 12-inch tall test section when Link 1
is deflected by 45°, as can be envisioned from Figure 38, which shows the thrust balance
mounted in the wind tunnel.
Link A2 Link 0 (Ground link)
Link A1 Slot
ψ
Link B1
Figure 37. Rendering of the thrust measurement apparatus
The entire thrust balance is designed to mitigate error introduced by the apparatus.
Links 1 and 3 are machined from 1/32 inch thick aluminum, to minimize the drag on these arms. Link 2 is designed to “hide” almost entirely behind the motor. This is critical to minimize occlusion of the prop-wash, which would impair thrust production of the power plant. Another potential source of error is the wires that provide power to the motor, which can be seen in Figure 38. As the thrust balance moves, the wires are
87
deformed, which results in a reaction force being applied to the apparatus. To mitigate
this error source, stranded copper wire with a highly flexible insulation was used. The
remaining test equipment are a BK Precision 1690 power supply, electronic speed
controller, a Futaba T9CAP R/C transmitter, and a matching GWS GW/R4PII R/C
receiver. Two speed controllers were investigated, the Phoenix-10 and the Phoenix-25,
both from Castle Creations. The output voltage of the power supply is tunable to 0.1 volt
increments, and the current draw is displayed to 0.1 amp precision. The power supply is
connected to the electronic speed controller (ESC), which provides 5 volt power to the
R/C receiver and performs pulse-width-modulation (PWM) control of the motor based
upon a command signal sent from the receiver to the speed controller. The receiver obtains an RF signal from the transmitter, which the test operator uses to command various motor speeds.
Figure 38. The thrust measurement apparatus mounted in the wind tunnel
88
6.2.3 Test Procedure
A candidate motor and propeller were selected and mounted into the Link A2. A
small bin was hung from Link B2. The motor model, propeller designation, and bin mass
were recorded. The R/C transmitter was turned on and the throttle placed in the off
position. The power supply was turned on and set to the proper voltage. Experiments
were run at 7.4 volts and 11.1 volts (mimicking a 2-cell and a 3-cell Lithium-polymer battery, respectively). Testing was done under static conditions and under various strength headwinds. If a headwind was desired, the wind tunnel was turned on and set to the appropriate wind speed. The drive voltage and the headwind velocity of the experiment were recorded. Once the wind tunnel had reached the commanded speed, the transmitter throttle was slowly increased until the prop motor began to rotate. The transmitter throttle has sixteen discrete positions. The throttle position, the angle ψ of the
Links 1 (Figure 37), and the current draw were recorded. A small amount of mass was then added to the bin hanging from Link B2, and the mass, ψ, and current draw were
recorded. This process was repeated until adding mass would cause ψ to surpass 60°. The
1° graduations on the protractor are only present for 30° ≤ ψ ≤ 60° (Figure 37). The
values of ψ were restricted to this range because 45° is the angle at which the error in
reading ψ from the protractor induces the least error in the calculated thrust. The throttle
was increased one click at time, and multiple data points were collected at each position.
At each new throttle position, mass was removed from the bin until ψ was at the lower
end of the acceptable range, and then slowly added back until ψ was near the higher end
of the acceptable range. The throttle position, mass, ψ, and current draw were recorded
for each of these data points.
89
6.2.4 Data Analysis
Figure 39 shows a partial free body diagram of the thrust balance. Because the
system includes five unknowns (F and the reaction forces XA, YA, XB, and YB) static
analysis would require solving the large system of equations created by summing the
forces and moments acting on each individual link.
F
WL2 Wmotor + mP1*g
YB YA ψ
-XB -XA
WL3 WL1
Wmotor + mP0*g WL4 + mA*g
Figure 39. Force Schematic of Thrust Balance
Rather, a virtual work analysis quickly provides the relationship between the applied offset mass (mA) and the thrust (F) of the propeller. Virtual work theory dictates
that
U= Fii r = m j *g* h j. (7) ij
90
In the case of the thrust balance, the thrust from the propeller is the only force that can do work on the system. Technically, the mass of each arm of the apparatus, both motors and propellers, and the offset mass should be summed on the right side of
Equation 7. However, close consideration of the apparatus shows that, with the exception of the offset mass and the two propeller masses, for each mass with a positive δh, there is an equal mass with an identical negative δh. Therefore, Equation 7 reduces to
Fr FFAP0P1m F x m m m g h, (8)
where mP0 and mP1 are the masses of the offset prop and the test prop, respectively. The virtual displacements δxF and δhm are caused by a virtual rotation (δψ) of the apparatus.
Figure 40 shows a close-up of the virtual displacement of the motor and the offset mass.
δrm δhm ψ ψ
δxF
δrF
Figure 40. Virtual Displacements of the Thrust Balance
91
As points on a rotating body, the virtual displacements δrm and δrF are both equal to L1* δψ, and, from Figure 40, it is clear that
xFF r *cos 90
r*sinF Lsin 1 . (9) hrsin90mF
rcosF
Lcos1
Substituting Equation 9 into Equation 8 and canceling common terms yields
Fsin mAP0P1 m m gcos mm m g . (10) F AP0P1 tan
Values for each of the variables on the right side of the equation were recorded during the several tests. All of the appropriate data were entered into a spreadsheet, and F was computed.
6.2.4.1 Component Selection
The Phase I prototype design utilized a Feigao 120 Brushless Motor and a GWS
EP3030 propeller. While this combination provided sufficient thrust for the 12 inch Phase
I prototype, which weighs approximately 150 grams, it was determined to be insufficient for the larger platform being developed as part of this work. An extensive internet search yielded the following information concerning available brushless DC motors. A wide
92
variety of motors are available whose mass is in the neighborhood of 15 grams. Since the
Feigao 120 weighs 17 grams and it is reasonable to assume that all of these motors would produce similar thrusts, this group of motors was eliminated from consideration. Another large group of motors are available whose mass is in the neighborhood of 50 grams.
While it is almost guaranteed that these motors would produce enough thrust, their excessive mass makes them undesirable. Weighing 32 grams, the Astro Flight Mighty
Micro 010 (MM010) motor is the only motor available in this mass range. The MM010 is available in 10-turn and 14-turn models, both of which were tested.
A variety of propellers were tested. The most common and accessible small motor propellers are made by Advanced Precision Composite (APC) and by Grand Wing
System U.S.A. Inc. (GWS). APC props are much more rigid, weigh more, and tend to have a higher pitch than the GWS props. Propellers with a range of diameter and pitch angle were selected for evaluation. The prop diameter was limited to six inches, to limit the length of the wheel-leg drive axles.
6.2.4.2 Data Verification
Data were collected on a daily basis over a several month period. Furthermore, over the span of a single test, the temperature of the motor and speed controller could change. To ensure that changing test conditions didn’t affect test results, two tests were repeated under disparate conditions. Figure 41 shows the results from the tests performed to verify the temporal consistency of the procedure. The two tests were performed using the Mighty Micro 010 10-turn motor, Castle Creations Phoenix-10 speed controller, and
APC 5.1 x 4.5 propeller, driven at 7.4 volts, in static air. The figure shows that the test results were reproducible over time.
93
Test Repeatability 1 APC 5.1x4.5, 10 Turn MM010 Phoenix 10, 7.4V, No Headwind 1600
1200
Test 1 800 6/22/2006 Test 2 6/23/2006 Thrust (mN) Thrust 400
0 02468 Current (Amps)
Figure 41. Temporal Consistency Verification Results
Figure 42 shows the results from the tests performed to investigate the dependency of the test results on the thermal conditions of the motor and speed controller. The two tests were performed using the Mighty Micro 010 10-turn motor,
Castle Creations Phoenix-10 speed controller, and GWS EP5043 propeller, driven at 7.4 volts, in static air. The figure shows that the test results were not affected by the thermal condition of the test equipment. Test 2 was performed with the speed controller at room temperature, and Test 1 was performed after the motor and speed controller had been used for multiple tests, and the temperatures of the components were elevated.
94
Test Repeatability 2 GWS EP5043, 10-Turn MM010, Phoenix 10, 7.4V, No Headwind 1800
1350
Test 1 6/23/2006 900 Test 2 (Cold) 6/28/2006 Thrust (mN) Thrust 450
0 02468 Current (Amps)
Figure 42. Thermal Consistency Verification Results
6.2.4.3 Drive Voltage Analysis
Originally, two different drive voltages were considered, 7.4V and 11.1V. These voltages emulate a 2-cell and a 3-cell lithium-polymer battery, respectively. The mass of the 2-cell battery is approximately two-thirds the mass of the 3-cell battery, so there is strong motivation to consider it as a power source. Early flight tests were being performed concurrent with the initial thrust tests. The early flight tests showed that, while the 2-cell lithium polymer battery was able to produce sufficient thrust for steady flight, the burst current was not available to produce the added thrust necessary for higher power maneuvers, such as tight banking and recovery from temporary disorientation. While the
7.4V data is presented in Appendix B, it is not discussed further in this section.
95
6.2.4.4 Speed Controller Analysis
Two speed controllers were initially considered. Figure 43 shows the results from two tests run with a Mighty Micro 10-Turn motor, APC 4.75x4.75 propeller, driven with
11.1 volts, in a 30 MPH headwind. The only difference between the test conditions was the speed controller used. These results are representative of similar data from tests run using different propellers in different headwinds. The data show that, while the Phoenix
10 ESC is rated to a 10 Amp continuous output, this level of performance was not attainable in tests. Later data demonstrate that the 14-Turn motor is more efficient than the 10-Turn motor, but complete test data show that the current draw of even the 14-Turn motor approached the current cutoff seen in Figure 43.
Speed Controller Comparison 10 Turn MM010, APC 4.75x4.75, 11.1V, 30 MPH headwind 1800
Phx 10 1350 ESC Phx 25 ESC 900 Trend (Phx 10) Thrust (mN) Thrust 450 Trend (Phx 25)
0 0 5 10 15 Current (Amps)
Figure 43. Speed Controller Comparison Data
96
6.2.4.5 Motor Analysis
Figure 44 compares the efficiency and thrust output from the two motors tested, when equipped with an APC 4.7x4.25 propeller in a 35 MPH headwind. While the 14-
Turn motor is more efficient within its thrust range, the 10-Turn motor can produce a greater overall thrust. Similar results were obtained for all propellers at each of the headwind test speeds. Full test data can be found in Appendix B.
Motor Comparison Phoenix 25, APC 4.7x4.25, 11.1V, 35 MPH headwind 1600
10 Turn 1200 MM010 14 Turn 800 MM010 Trendline (10 Turn) Thrust (mN) Thrust 400 Trendline (14 Turn)
0 051015 Current (Amps)
Figure 44. Motor Thrust and Efficiency Comparison
6.2.4.6 Headwind Comparison
Figure 45 shows the thrust produced by the APC 4.7x4.25 propeller on the 10-
Turn motor, driven at 11.1V in various headwinds. Not surprisingly, the thrust produced at a given current by this arrangement decreases with increasing headwind. Result for the other motor/propeller combinations showed similar trends.
97
Power Comparison 10 Turn MM010, APC 4.7x4.25 Phoenix 25, 11.1V 2000
1600
1200 25 MPH 30 MPH 800 35 MPH Thrust (mN) Thrust 40 MPH 400
0 02468101214 Current (amps)
Figure 45. Power Comparison for One Configuration in Different Headwinds
6.2.4.7 Propeller Analysis
Figure 46 through Figure 49 summarize the data for several propellers mounted on the MM010 14-Turn motor, driven at 11.1V, in differing headwinds. These propellers were chosen from the original field as being representative of the two lessons learned from this analysis. The first lesson is that, at lower airspeeds, propeller efficiency varies inversely with pitch; i.e. a shallower pitch results in higher efficiency. The second lesson is that, as airspeed increased, the more efficient propellers lose the ability to generate thrust. Both of these lessons are a result of the interaction between the motor speed- torque curve and the propeller speed-thrust-torque relationship. At any given (low) airspeed, a shallower pitched propeller will have to rotate at greater rpm than its steeper counterpart to produce the same thrust. However, due to the nature of DC motors, the faster spinning motor draws less current. This is apparent in Figure 46, where the APC
4.75x4.75 has the steepest pitch, and the GWS 5043 has the shallowest. To produce a
98
given thrust (say 1200 mN) the GWS must spin the fastest, which inherently draws the least current. So, while the output power (thrust x airspeed) is the same, the input power
(current x voltage) is lower. However, as the airspeed increases, the air molecules are already flowing past vehicle, and each propeller must therefore spin faster in order to
“catch-up with” the air and produce the same thrust as previously (at the lower airspeed).
However, motors can’t spin infinitely fast, and as the needed rpm for a given propeller approaches the no load speed of the motor, the motor simply can’t produce the necessary rpm, and the maximum thrust from the propeller at that airspeed falls off. This can be seen in Figure 48, where, in a 35 mph headwind, each propeller is unable to produce the same thrust that it could in a 25 mph headwind. The 40 mph headwind data (Figure 49) accentuates this phenomenon, where the maximum thrust from the GWS 5043 has fallen off dramatically. Therefore, the best propeller for a given vehicle is the shallowest propeller than can produce the necessary thrust at the vehicle’s cruising airspeed. The
APC 4.7x4.25 was selected as the propeller for the 16-inch wingspan MALV.
Propeller Comparison 14-Turn MM010, 11.1V, Phoenix 25, 25 mph 1800
1500
1200 APC 4.75x4.75 900 APC 4.7x4.25 GWS EP5043 Thrust (mN) Thrust 600
300
0 012345678 Current (Amps)
Figure 46. Propeller Comparisons on Mighty Micro 010 14-turn Motor at 25 MPH
99
Propeller Comparison 14-Turn MM010, 11.1V, Phoenix 25, 30 mph 1600
1200
APC 4.75x4.75 800 APC 4.7x4.25 GWS EP5043 Thrust (mN) Thrust
400
0 012345678 Current (Amps)
Figure 47. Propeller Comparisons on Mighty Micro 010 14-turn Motor at 30 MPH
Propeller Comparison 14-Turn MM010, 11.1V Phoenix 25, 35 mph 1200
900
APC 4.75x4.75 600 APC 4.7x4.25 GWS EP5043 Thrust (mN) Thrust
300
0 012345678 Current (Amps)
Figure 48. Propeller Comparisons on Mighty Micro 010 14-turn Motor at 35 MPH
100
Propeller Comparison 14-Turn MM010, 11.1V Phoenix 25, 40 mph 1000
750
APC 4.75x4.75 500 APC 4.7x4.25 GWS EP5043 Thrust (mN) Thrust
250
0 012345678 Current (Amps)
Figure 49. Propeller Comparisons on Mighty Micro 010 14-turn Motor at 40 MPH
101
Chapter 7 Auxiliary Subsystem Design
Several auxiliary systems were identified during the critical performance review that would be necessary to provide MMALV with the widest range of utility. While missions can be envisioned that do not strictly require any of the capabilities addressed in this chapter, each subsystem described here is necessary for at least one highly useful mission and would benefit multiple others.
7.1 Tail Hook
Careful consideration of the original mission scenario illuminated a major hurdle to the task of crawling to the edge of a rooftop – most flat roofs are bordered by a parapet, or knee-wall. A tail-hook was identified as the most likely means of allowing the vehicle to surmount the parapet. However, a passive grappling hook would not be acceptable, because MMALV would need to back-up beyond the edge of the wall so as to release the hook. The tail-hook must be able to be actively released. However, the arms of the hook must be able to support the load associated with arresting the momentum of the vehicle. Figure 50 shows a rendering of the tail-hook that was designed for MMALV.
The cable (labeled in the figure) enters the vehicle fuselage and wraps around a winch that is driven through a slip-clutch, similar to that used in the terrestrial drive system. By
102
limiting the possible tension in the cable, the slip-clutch limits the deceleration that will be experienced by the vehicle, as well as the loads experience by the hook assembly.
Cable
Body
Figure 50. Isometric view of the tail-hook design
Figures 51 through 53 demonstrate the operation of the tail-hook assembly. In the deployed orientation (Figure 51), the ball (gray) protrudes from the outside of housing into the groove on the inside surface of the slide pivot (green), prohibiting motion of the slide. As the winch collects the cable, the tail-hook body is restrained by the fuselage of the vehicle. Subsequent collection of the cable causes the internal spring to depress, until the plunger reaches the location shown in Figure 52, where it no longer constrains the ball to the outer position.
103
Figure 51. Tail-hook in deployed position
Continued collection of the cable results in the slide pivot pushing the ball into the inner position, allowing the slide pivot to compress the outer spring, until it reaches the position shown in Figure 53. At this point, the arms of the tail-hook are completely retracted.
104
Figure 52. Tail-hook in midst of retraction process
When the cable is released, the springs cause the process to happen in reverse.
The inner and outer springs both extend, until the groove on the slide pivot is aligned with the ball. As the inner spring is still compressed at this point, it drives the plunger back into the starting position, which pushes the ball into the outer position, engaged with the slide pivot.
105
Figure 53. Tail-hook in fully retracted position
7.2 Wing Folding
As identified in the original solicitation and reinforced by the critical performance review of the first prototype, wing-folding capability would enhance the utility and performance of MMALV by increasing portability, protecting the wings, and increasing
106
ground mobility in confined spaces. Further consideration of the wing-folding behavior illuminated three subsystems into which the folding mechanism can be divided. The most obvious subsystem is the actuation subsystem, which could be as simple as an R/C servo, but could also include a linkage or other hardware to convert actuator motion into motion of the wing. The second subsystem is the wing itself. The final subsystem exists at the wing/actuator interface, which will be called the wing confinement subsystem. Each subsystem has unique requirements.
The actuation subsystem must:
o Produce sufficient and symmetrical torque to tension the wing fabric
o Passively sustain in-flight forces from drag
o Provide a broad range of motion
The wing confinement subsystem must:
o Maintain the relative separation of the fixed section of the wing and the
moving section of the wing
o Prohibit drooping and upward bending of the wing due to the wing’s
weight or applied lift
The foldable wings must be able to:
o Ensure that the fabric does not hang below the fuselage in the folded state
7.2.1 Confinement Subsystem
The prototype wing-folding mechanism fabricated the confinement subsystem into the wing structure. The confinement system in the original wing-folding prototype molded the carbon fiber into a concave section in the wing mold. The resulting structure is circled in Figure 54. However, any irregularities in the bottom carbon fiber layer
107
created a matching irregularity in the upper layer, resulting in an interlocking mating surface, which resisted rotation. While this technique had drawbacks, it was still considered attractive from a mass savings standpoint. Therefore, the option was further investigated by Mr. Brian Taylor, under the guidance of the author.
Figure 54. Original folding wing, with confinement structure highlighted
Figure 55 shows a schematic of the concept as it was investigated by Mr. Taylor.
Rather than creating an irregularity in the wing mold, the new concept creates the confinement region against a flat section of the wing tool. The confinement system comprises three layers of carbon fiber, the static top and bottom enclosures, which serve to sandwich the wing leading edge between them. A hole through all three layers seats a dowel pin, about which the leading edge rotates. A second hole in the leading edge is where the actuation mechanism interfaces with the wing. The window cut into the top and bottom enclosures provides a clear path for the actuation mechanism to move through its entire range of motion.
108
Figure 55. Schematic of an integrated wing confinement system
This technique suffered from shortcomings similar to those experienced by the prototype. The technique relies on laying-up all three layers of carbon fiber simultaneously. However, the first layer (in this case the top enclosure) creates an irregularity in the substrate on which the middle layer (or wing leading edge) is laid-up.
The second layer also creates a feature in the substrate on which the remaining layer is laid up. To minimize the contours in the wing leading edge, this system was first laid-up using only two layers of woven carbon fiber for the top canopy. In this case, the top enclosure was deemed too compliant to support the lift generated by the wing. The potential for reinforcing the top enclosure was investigated. An aluminum plate was adhered to the top enclosure, as shown in Figure 56. However, any reinforcement structure that conferred sufficient stiffness on the top enclosure to support the lift force also created excess friction between the top enclosure and the wing leading edge. This was again a result of the contours in the wing leading edge. Even with only two layers of carbon fiber in the top enclosure, a detectable ridge resulted in the wing leading edge.
109
The vacuum bagging process compresses the layers together, resulting in close contact between the three layers. The un-reinforced enclosure flexed out of the way sufficiently for the wing to rotate freely. However, the reinforced wing restricted motion of the wing.
This wing confinement technique was deemed infeasible. Attempts to fabricate the reinforcement into the top enclosure resulted in more pronounced contours in the wing leading edge, and the subsequent rotation restrictions.
Figure 56. Folding wing with reinforced top enclosure
7.2.2 Foldable Wing
The original prototype tended to allow the wing fabric to sag during the folding process. This is unacceptable for field deployment in that it could allow the wing to suffer damage during terrestrial locomotion. Mr. Taylor investigated four separate baton configurations (Figure 57) to evaluate how each configuration affected the fold-ability of the wing and the consistency with which the fabric would fold up onto the top of the wing. On each wing, all batons originate from the axis of rotation to provide a fan-like folding process. Batons configuration A was easy to fold, but difficult to control.
Configuration B restricted wing folding due to the stiffness near the axis of rotation.
110
Baton configuration C and D were investigated as compromises between configurations
A and B. While configuration C was too stiff near the joint to allow suitable folding, configuration D folded acceptably and could be coaxed to repeatedly fold the fabric onto the top of the wing. The coaxing process involved folding the wing into the desired shape, and applying pressure to the wing fabric for eight hours. While this did create a crease in the fabric, this crease was indistinguishable when the wing was fully deployed
A B
C D
Figure 57. Four baton designs tested by Mr. Brian Taylor
Mr. Taylor’s results led the author to modify the wing baton structure. The bias conferred on the successful baton structure was not a fan-like geometry. Rather, it was a simple three layer fold. This same bias could be created forcing the outer half of the wing to “act” rigid during the folding process. Getting the fabric onto the top of the wing could be accomplished by integrating a biased member within the wing. Both of these could be
111
accomplished by fabricating a thin piece of music wire into the wing from Point A to
Point B, as shown in Figure 58. Curing the music wire into the leading edge will cause the front triangle of fabric to behave like a rigid wing during the folding process.
Furthermore, by only attaching the music wire to the wing at the leading edge and at
Point B, the music wire will try to return to a straight orientation. This means that the end of the wire at B will deflect upward once tension is removed from the wing. This structure should bias the fabric into a “doubled-over” orientation during folding, with one fold at the edge of the canopy and the other fold along line AB.
A
B
Figure 58. Biased baton concept for a foldable wing
7.2.3 Actuation Subsystem
A mock-up of the first wing-folding mechanism design is shown in Figure 59.
The motivation for the long transmission link was to keep the servos away from the wing rotation point. The space directly under the wing leading edge was to be occupied by the
112
battery and the autopilot, to maintain the CG in the appropriate location. When the wing is fully deployed the driven link is perpendicular to the transmission link, which acts through the center of rotation of the drive link, providing passive locking in this configuration. The major drawback to this design was the mass required for successful implementation. In its depicted form, the prototype suffered from a lack of stiffness of the transmission link. This flexibility could have been overcome by using a stiffer control arm, but the increase in mass was undesirable. Furthermore, to ensure proper functionality of the passive locking configuration, it is critical that each component be located accurately. Considering the distance between the wing rotation axis and the servo, the required bracket would be unacceptably massive.
Transmission link Drive link
Figure 59. Mock-up of the First Wing Folding Mechanism Design
Experience with the folding mechanism shown in Figure 59 led to the search for an off-the-shelf four bar linkage that had a locked configuration. One common such linkage is the retractable landing gear mechanism (Figure 60 – referred to as the “retract mechanism”) for model airplanes.
113
Output link
Input actuation Locking joint
Wing folded Wing deployed
Figure 60. The Retractable Landing Gear Mechanism
The retract mechanism was implemented in a six bar linkage configuration, as shown in Figure 61. A six bar linkage comprises a standard four bar linkage, where one of the links is driven indirectly by the drive link (Link 5) through the transmission link
(Link 4). In this implementation, the servo output horn is Link 5 and the control rod is
Link 4. This implementation enjoyed several benefits over the original implementation.
The self-contained nature of the retract mechanism offers flexibility in the placement of the actuation servo (circled in green). This flexibility negates the need for a stiff bracket and transmission link, leading to a much less massive implementation. Finally, using an off-the-shelf locking four bar mechanism is much easier than fabricating a custom linkage.
114
Figure 61. Six Bar Linkage in situ
The drawbacks of this implementation outweighed the advantages. The form factor of the retract mechanism required the implementation of a mounting bracket, as shown in Figure 61 The necessary bracket adds complexity, mass, and volume to the unit. The interaction between the retract mechanism and the wing was untenable. The original output arm of the retract mechanism is 0.060 inch diameter music wire.
Originally, this wire was bent upward 90° and inserted through a small hole in the wing leading edge. This arrangement realizes a bearing surface area of about 1.8x10-3 square inches, which would lead to deformation of the carbon fiber. Although an aluminum arm was implemented (Figure 63) that overcame this hazard, this adds to the mass and complexity of the unit.
115
Modified output arm
Figure 62. Custom Output Bracket for Retractable Gear Implementation
Figure 63 shows a rendering of the final wing folding mechanism, as designed by the author. Figure 64 shows how the unit fits within the fuselage. The folding mechanism is based upon a four bar linkage. The frame (peach) comprises the ground link (Link 0), which holds the servo (blue) and the bearings (purple) that support the wing rotation axle
(pink). The servo output horn is attached directly to the driving link (Link 1 – cyan). The transmission link (Link 2 – green) impels the driven link (Link 3 – red), which directly actuates the wing through the flanged shaft (pink). The 0.6 inch diameter circular hole pattern on the wing mounting flange will provide sufficient rigidity to the wing base to resist vertical motion of the wing under loading. Not shown in the figure is the top plate of the frame, which holds four bearings, one for each flanged shaft and one to support each servo output horn.
116
Figure 63. The final wing folding linkage
Figure 64. The wing folding mechanism integrated within the fuselage
117
A top view schematic of the wing folding linkage is shown in Figure 65. Several of the functional requirements listed above translate directly to linkage geometry attributes. First, the transmission angle (α) must be 90° when Link 1 and Link 2 are collinear (γ = 180°). This provides perfect passive locking, as all torque applied to Link 3
(such as drag during flight) will act through nodes 0 and 1, developing zero torque about these nodes. The total excursion (θ) of Link 3 should be as close to 90° as reasonably possible. This corresponds to folding the wing completely back along the fuselage. The mechanical advantage of the servo in relation to the wing, i.e. the mechanical advantage of Link 1 to Link3, should be maximized. The mechanical advantage amplifies the servo torque as it applies to tensioning the wing fabric. One way to provide the servo with a better mechanical advantage is to maximize the ratio of the length (L2) of Link 2 to the length (L1) of Link 1. However, maximizing this ratio is directly opposed to several of the objectives outlined in this paragraph. Fortunately, the passive locking geometry imparts a large mechanical advantage to Link 1 as γ approaches 180°, which is exactly where the most torque will be needed to tension the wing fabric. In addition to these attributes, other general geometric principles of four bar linkages apply. The necessary excursion (β) of Link 1must be within the operating range of a standard servo. The transmission angle (α) should be kept as close to 90° as possible to avoid binding of the mechanism during actuation.
118
α θ
α γ
β
Vehicle direction of travel
Figure 65. Schematic of the Wing-folding Four-bar Linkage
The actual link lengths and other geometric characteristics of the final folding mechanism design are enumerated in Table 9. As seen in the table, the last 1° of rotation of Link 1 produces only 0.11° of rotation in Link 3, corresponding to an average mechanical advantage of 8.8, which will help with tensioning the wing fabric. This feature will also assist in the trimming process, during which a servo position must be found that produces a transmission angle as close to 90° as possible. While the increments are small, servo trim positions are nonetheless quantized, and the mechanical advantage divides the angular displacement of Link 1 as it is transmitted to Link 3.
119
Table 9. Design Variables of Wing Folding Four-bar Linkage
Quantity Symbol Value Length of Link 0 L0 1.75 Length of Link 1 L1 1.33 Length of Link 2 L2 0.35 Length of Link 3 L3 0.49 Excursion of Link 3 θ > 90°
Maximum transmission angle αmax 117°
Minimum transmission angle αmin 57° Servo excursion β 38° Average mechanical advantage over last -- 1.7 5° of rotation Average mechanical advantage over last -- 8.8 1° or rotation.
7.3 Ground Take-off
Several progressive ground take-off tests were performed. First, a wheeled cart was fabricated that served as a sled for a MAV. The MAV tail was held firmly while the propeller was throttled up. Once full throttle was reached, the tail was released. The thrust from the propeller accelerated the MAV and sled to lift-off speed in approximately forty feet. For the next test, the wheels were mounted to the underside of the fuselage, and a single tail-wheel was mounted to the craft. This vehicle required a take-off distance of approximately 65 feet. These tests were both performed with aircraft built on the first prototype model. The purpose of these tests were to investigate the feasibility of rolling ground take-off and whether proximity to the ground would effect the aerodynamics of the vehicle in such a manner as to make take-off difficult or impossible. The tests discounted the notion of ground-effect difficulties.
120
The two tests described above were performed on a hardwood basketball court and on a concrete sidewalk, which are smooth surfaces compared to grassy fields, dirt paths, or even gravel roads. Take-off attempts with these same vehicles in short grass proved unsuccessful. Since it is unlikely that MMALV will always, or even occasionally, be able to locate a smooth surface for ground take-off, attention was turned to the possibility of static take-off. Static take-off refers to the condition where the vehicle separates from the ground when it has zero velocity. While the vehicle started from rest in each of the previous tests, it accelerated on the ground, until sufficient wing lift was produced to create separation from the ground. Initial tests were performed using a 16 inch wingspan aerial platform, with the Astro Flight Might Micro 010 14-turn motor.
These tests demonstrated that the selected power plant produced insufficient thrust for a controllable static take-off. The minimum requirement for static take-off is a power plant that can produce sufficient thrust to lift the vehicle straight upward, but static take-off requires more than just a thrust to weight ratio greater than one. As the vehicle begins to lift off from the ground, the airspeed is very near zero. With so little air flowing over the control surfaces, it is very difficult to stabilize the aircraft. The power plant must produce sufficient propeller wash so that when the prop wash reaches the tail, deflection of the control surfaces can produce enough force to stabilize the vehicle. Alternative motor/propeller combinations were investigated. Table 10 enumerates the results of static tests on two motors and several different propellers.
121
Table 10. Static Thrust Testing for Ground Take-off Vehicle
Current EMF Thrust Motor Propeller (amps) (volts) (grams) Mega 16/7/8 GWS EP 5043 23 11.1 330 Mega 16/7/8 APC 4.1 x 4.5 <25 11.1 >320 Mega 16/7/8 APC 4.5 x 4.5 >25 11.1 300 Mega 16/7/8 APC 5.1 x 4.5 16.5 11.5 390 Mega 16/7/8 APC 5.5 x 2.5 13.5 11.5 453 Mega 16/7/8 APC 7.0 x 4.0 21 11.5 495 Mega 16/15/3 APC 4.5 x 4.1 16.4 11.5 477 Mega 16/15/3 APC 5.1 x 4.5 22.2 11.5 481.5 Mega 16/15/3 APC 5.5 x 2.5 17 11.5 490 Mega 16/15/3 APC 7.0 x 4.0 25 11.5 497
Based upon this information, the Mega 16/15/3 motor with the APC 5.5 x 2.5 propeller was selected as the power plant. The low current draw allowed for retention of the Castle Creations Phoenix-25 speed controller. While the Mega 16/15/3 motor adds 45 grams to the vehicle mass, it more than compensates for this with additional thrust and the associated prop wash.
The performance of the new power plant was evaluated in skid take-off tests first.
Un-powered wheel-legs were placed on the vehicle, and it lifted-off from the ground in approximately 15 feet on a hard, smooth surface, but was unable to generate sufficient speed on grass. Multiple tests were performed in which the vehicle was manually placed in progressively more vertical orientations, and static take-off was attempted. The best behavior was observed at an inclination of 60° from the horizontal. At angles above this, the propeller wash over the airfoil created sufficient lift to cause the vehicle to pitch over on its dorsal side before it could lift off. A mechanism was designed to allow the vehicle to re-orient itself into the desired inclination autonomously. The apparatus consists of a
122
servo placed seven inches from the nose, with a seven inch arm attached to the servo.
Figure 66 shows a schematic of the mechanism. Figure 67 shows a static take-off capable
MAV at various stages of pre-take-off re-orientation.
Figure 66. Schematic of the Vehicle Re-orientation Mechanism
Figure 67. Static Take-off Capable MAV Re-orienting Itself for Take-off
123
7.4 Rapid Deployment
7.4.1 Procerus Kestrel™ Autopilot
The Kestrel™ autopilot is a full-featured autopilot system that allows the user to abstract low-level flight control to high-level concepts such as waypoint following and other behaviors, including loitering and landing. Procerus provides a control system called Virtual Cockpit, which runs on a PC (usually a laptop that can be taken into the field) and permits full configuration of the aircraft. Full specifications for the Kestrel™ autopilot can be found on the Procerus webpage [43].
The Kestrel™ autopilot consists of three physical components: the vehicle-borne autopilot module, a ground-based communications relay box (comm-box), and a PC that runs the Virtual Cockpit software. Communications between the vehicle and the comm- box can take place through a 900 MHz link, and the connection between the comm-box and the PC is achieved through an RS-232 serial cable.
7.4.2 Palmtop Autopilot Interface
The MAVControl palmtop interface platform replaces the PC with a palmtop computer. The serial connection is replaced with a Bluetooth radio connection (Figure
68). The added flexibility provided by both the palmtop computer and the Bluetooth connection allows the user to launch a vehicle unaided, holding the vehicle in one hand and selecting the launch options with the palmtop computer in the other. The Bluetooth connection also confers a level of flexibility to the deployment team – separate individuals could possess the palmtop computer and the comm-box, without a cable tying them together. The Bluetooth system in most consumer palmtop computers has a range of
10 meters.
124
Figure 68. Communications link schematic
7.4.3 Interface Control Hierarchy
While the flexibility of control offered by the Virtual Cockpit™ interface is useful in some situations, the critical performance review confirmed the need for a simplified control interface that can be used with a smaller control device such as a palmtop computer. The MAVControl palmtop interface platform has been developed to allow a palmtop computer to communicate with the Kestrel™ autopilot. The goal of the interface scheme is to present only the necessary data and controls to the user, so that an aircraft can be powered up and launched swiftly, with minimal interaction, possibly from a moving vehicle or during an emergency situation.
The flight control options have been placed in menus such that frequently-used functions will be easily accessible, while those functions which require more input on the part of the user are pushed back to submenus. By reducing the number of selections required to accomplish the most frequent tasks, a user familiar with the interface could instruct a craft to perform certain tasks without needing to look at the screen.
In particular, the take-off procedure has been greatly abbreviated by pushing back the bulk of the maintenance checks and calibrations to the on-base maintenance procedure, which can be done using Virtual Cockpit. Part of the remaining portion handled by MAVControl can also be performed on-base, with the remaining portion of the launch procedure consisting of only seven steps, including the required on-site
125
pressure sensor calibration and full-throttle battery test. The breakdown of the command hierarchy is enumerated in Table 11. Additionally, most of the commonly used in-flight functions – “follow me,” “land,” and “redirect” – are accessible as top-level options of the vehicle control menu, with other more complicated options appearing in a “change mode” submenu.
Table 11. Break-out of MAVControl Command Hierarchy
On-Base On-Base In-Field Take-off Preparation/Maintenance Preparation/Maintenance Procedure using using Virtual Cockpit using MAVControl MAVControl
Calibrate sensors Set up Bluetooth link Download flight plan Calibrate controls/trims Check ground station to vehicle Create flight plan battery Zero pressure sensors Download flight plan to Add entries for each Do full-throttle battery palmtop vehicle check Check vehicle battery Launch vehicle Test attitude sensors Test/adjust control surface trims
MAVControl also permits control and monitoring of multiple craft. The Kestrel™ autopilot is already configured to permit one ground station to control a large number of craft simultaneously. MAVControl permits the operator to access any of the craft launched through his ground station, monitor their progress, and issue commands.
Finally, MAVControl works with Virtual Cockpit by permitting pre-planned waypoint lists to be downloaded to the vehicle autopilot. Any number of waypoint flight plans can be created in advance and selected from within MAVControl as conditions require. These flight plans can be edited if necessary within MAVControl using the
126
bearing/distance/altitude waypoint method, providing additional flexibility to the operator in the field.
7.4.4 Interface Appearance
Most palmtop computers include a stylus for making selections on dense on- screen interfaces. The precision required for this process conflicts with the need for an easy, rapid deployment process. MAVControl has been designed such that the user can make selections on the screen by touching it with his finger. The on-screen buttons are large and easy to touch, even with a gloved hand. By disposing of the stylus, this also allows one soldier to launch a vehicle by holding the vehicle in one hand while controlling the launch procedure with the other hand.
Many situations requiring MMALV deployment will be in bright outdoor conditions. Some palmtop displays are notoriously difficult to read in such situations, so
MAVControl uses large text displays and color-coded buttons to increase readability. For nighttime conditions, MAVControl includes a night mode consisting of only red tones for reduced impact on night vision hardware. Figure 69 shows samples interface screens.
Figure 69. Three Menu Views of the MAVControl Interface
127
7.4.5 User Relative Ad Hoc Path Appointment
The Procerus Virtual Cockpit software provides facilities for full tuning, configuration, planning, and operation of the vehicle. However, most of this functionality is not necessary for the user in the field. The vehicle should be maintained such that tuning and configuration are already resolved before the vehicle is taken into the field, and an operation should generally include pre-planning for determining the function and course of an autonomous MMALV before the operation begins.
However, it is also likely that MMALV will need to be dispatched in spur-of-the- moment situations. Therefore, MAVControl includes the ability to specify user-relative waypoints. The user enters the bearing, distance, and altitude to a target (Figure 70).
MAVControl obtains the current GPS position from the vehicle, performs the necessary calculations, and uploads the mission data to the autopilot. Any user with a compass and a rangefinder will be able to use MAVControl to command MMALV to take off and circle a target so that aerial surveillance data can be obtained.
Figure 70. Bearing/Distance/Altitude Entry Screens
128
Discussions with active duty military personnel highlighted the importance of ease of use to the field deployment of MMALV. In emergency situations, the tool that gets used in not necessarily the tool that works best, but the tool that works first. To that end, the MAVControl user interface streamlines the deployment process for MMALV by relegating the maximum number of pre-flight tasks as on-base mission preparation tasks.
Thus, the number of steps required to be performed in the field in minimized.
129
Chapter 8 Conclusions and Discussion
The work presented in this dissertation conclusively demonstrates the potential of multi-modal air-land mobility in a small, lightweight vehicle. Motivated by military surveillance missions and other scenarios that would benefit from a hybrid sensor platform, the goal of the work was to develop the smallest vehicle that could successfully fly, land, and crawl. An initial prototype accomplished these goals in a 12 inch maximum dimension vehicle. However, the craft suffered from fragility in its terrestrial drive system, and could only be tele-operated via remote control. A critical performance review illuminated several areas of necessary improvement, including implementation of an autopilot and enhancement of the overall durability and manufacturability of the craft.
Implementation of the review critiques required improvements to both the vehicle design and the fabrication process. These improvements resulted in the 16-inch wingspan vehicle shown in Figure 71. The vehicle mass increased from 118 grams to 365 grams.
The fuselage width of 7.6 centimeters allows the battery to be loaded laterally into the fuselage. While the wheel-leg spokes could be longer, the 5.7 centimeter diameter is similar to wheel-legs on similarly sized Mini-Whegs. Table 12 summarizes the mechanical properties of the vehicle.
130
Figure 71. The final MALV vehicle
Table 12. Mechanical Properties of 16-inch Wingspan MALV
Parameter Value Total mass 365 g Overall length 40.6 cm Wingspan 40.6 cm Fuselage width 7.6 cm Leg length 5.7 cm Track (distance between wheel-legs) 14 cm
While the wingspan only increased by 33%, from 30 to 40 centimeters, reshaping the wing planform resulted in a 50% increase in wing area, to 560 cm2. The three-fold increase in mass accompanied by a 50% increase in wing area doubled the wing loading.
131
This characteristic was reflected in the controllability of the vehicles. Pilots noticed a decrease in stability from the first prototype to the 16-inch version, especially during landing. Attempts to flare the 16-inch wingspan craft prior to landing typically resulted in instability that led to a roll-over and nose-first landing. Table 13 summarizes the aerodynamic properties of the vehicle.
Table 13. Aerodynamic Properties of 16-inch Wingspan MALV
Parameter Value Wingspan (b) 40.5 cm Aspect Ratio (AR) 2.9 Wing loading 63.9 N/m2 Wing area (S) 560 cm2 Location of CG from wing leading edge - 3.2 cm Horizontal stabilizer area 155 cm2 Elevator area 31 cm2 Vertical stabilizer area 89 cm2 Rudder area 19 cm2
The increased mass of the 16-inch wingspan version required an increase in airspeed from 11 meters per second to 14 meters per second. An increase in the root chord maintained the Reynolds number near 1x10-5. A significantly larger battery maintained a flight duration near that for the original prototype. The increased mass had a detriment on the terrestrial performance of the vehicle. The maximum ground speed decreased, the maximum obstacle height decreased, and the vehicle was only able to perform “walking turns,” while the first prototype could yaw in place. The cause of this was the three-fold mass increase. Furthermore, the straight-spoked wheel-legs of the first
132
prototype were replaced by a tear-drop shaped wheel-leg, for purposes of durability.
Table 14 summarized the performance characteristics of the 16-inch wingspan MALV.
Table 14. Performance Characteristics of 16-inch Wingspan MALV
Parameter Value Cruising air speed 14 m/s Reynolds number ~1 x 105 Maximum flight time 12 min Range (round trip) 5.0 km Maximum terrestrial speed 0.33 m/s Maximum crawling time 60 min Range (round trip) 0.59 km Maximum obstacle height 3 cm
While an intense design process was required, the success of this research was predicated on the constituent technologies, both of which have emerged within the past decade. Flight vehicles are sensitive to weight, and the Mini-Whegs technology represents a critical compromise between simplicity and mobility. At the proposed scale of the hybrid vehicle, wheels and tracks could not confer a useful level of mobility on the vehicle. Aerial vehicles in this size range suffer from an order of magnitude decrease in their efficiency. This is exacerbated by the protrusion from the fuselage of the terrestrial running gear. The flexible wing MAV technology is crucial to dealing with this phenomenon at this size.
A sixteen inch wingspan craft was successfully designed and fabricated that can carry the components necessary for a field capable MMALV. The autopilot was successfully implemented and demonstrated on an aerial vehicle without a terrestrial
133
locomotion system. A robust terrestrial locomotion system has been designed.
Furthermore, several subsystems have been designed that will improve the fieldability of the vehicle. A wing folding mechanism will reduce MMALV’s portage size and terrestrial footprint. The MAVControl palmtop interface will allow the user to configure and deploy MMALV quickly and efficiently in the field. A retractable tail hook will allow MMALV to “perch” from a wire, and then release the wire to regain flight.
Continued efforts in this research area are strongly recommended. In particular, the following steps should be taken:
Autopilot optimization – The autopilot must be specialized to the MMALV
platform. First, weight and volume savings may be realized if the ability to
control ailerons and flaps were sacrificed. Furthermore, imbuing the autopilot
with a full-fledged terrestrial locomotion mode, complete with GPS navigation
and dead reckoning, is critical.
Wing optimization – While the flexible wing theoretically enhances stability,
wind tunnel tests should be performed to confirm this. The flexible wing is the
largest remaining source of compromised durability on the vehicle – if it can be
replaced, it should be.
Enhanced terrestrial locomotion – Recent work has been done on using a four
bar linkage to produce insect-like foot motion. Replacing the wheel-legs with
such a mechanism would allow MMALV to retract the legs up against the body
during flight and landing, which would provide two major benefits: 1) the wheel-
leg’s detrimental effect on aerodynamic performance could be eliminated, and 2)
damage to the terrestrial running gear during landing could be eliminated.
134
Vehicle capability integration – Several important vehicle capabilities were
demonstrated in this work. As many of these capabilities as possible should be
integrated onto the vehicle. While each subsystem should clearly be optimized for
weight, it may still be necessary, or as least advisable, to increase the vehicle’s
wingspan.
Vehicle signature characterization – Thus far, the vehicle’s wingspan has been
minimized to the extent possible. While a 16 inch wingspan is capable of
providing sufficient lift, an 18 (or even 20) inch wingspan could significantly
increase mission duration. Any detriment this increase in wingspan would have to
MMALV’s portability would be largely mitigated by the wing folding capability.
However, the signature of the vehicle during mission performance could be
significantly increased. Studies should be performed to characterize the visual and
acoustic signatures of the vehicle, for comparison to alternative vehicle
configurations, including stronger motors and larger wingspans.
135
Appendix A Power Plant Thrust Data
A.1 Thrust Data for Mighty Micro 010 14-Turn Motor
Propeller Comparison - Full Field 14-Turn MM010, 11.1V Phoenix 25, 30 MPH 1600 APC 4.75x4.75 APC 5.0x5.0 1200 APC 4.75x4.5 APC 4.5x4.1 APC 4.7x4.25 800 APC 5.1x4.5 APC 5.25x4.75 Thrust (mN) Thrust GWS EP5043 400 GWS EP4540 GWS EP4530
0 0246810 Current (Amps)
Figure A-1. Thrust Data for Original Field of Props on MM010 14-Turn Motor in 30 MPH Headwind
Propeller Comparison 14-Turn MM010, 11.1V, Phoenix 25, 25 mph 1800
1500
1200 APC 4.75x4.75 900 APC 4.7x4.25 GWS EP5043 Thrust (mN) Thrust 600
300
0 012345678 Current (Amps)
Figure A-2. Thrust Data for Selected Props on MM010 14-Turn Motor at 25 MPH
136
Propeller Comparison 14-Turn MM010, 11.1V, Phoenix 25, 30 mph 1600
1200
APC 4.75x4.75 800 APC 4.7x4.25 GWS EP5043 Thrust (mN) Thrust
400
0 012345678 Current (Amps)
Figure A-3. Thrust Data for Selected Props on MM010 14-Turn Motor at 30 MPH
Propeller Comparison 14-Turn MM010, 11.1V Phoenix 25, 35 mph 1200
900
APC 4.75x4.75 600 APC 4.7x4.25 GWS EP5043 Thrust (mN) Thrust
300
0 012345678 Current (Amps)
Figure A-4. Thrust Data for Selected Props on MM010 14-Turn Motor at 35 MPH
137
Propeller Comparison 14-Turn MM010, 11.1V Phoenix 25, 40 mph 1000
750
APC 4.75x4.75 500 APC 4.7x4.25 GWS EP5043 Thrust (mN) Thrust
250
0 012345678 Current (Amps)
Figure A-5. Thrust Data for Selected Props on MM010 14-Turn Motor at 40 MPH
A.2 Thrust Data for Mighty Micro 010 10-Turn Motor
A.2.1 7.4V Data
Propeller Comparison 10-Turn MM010, 7.4V, Phoenix 10, Zero Headwind 2000 APC 4.75x4.5 APC 5.1x4.5 1600 APC 4.5x4.1 APC 5.25x4.75 APC 4.7x4.25 1200 APC 5.0x5.0 APC 4.75x4.75 800 APC 5.5x4.5 Thrust (mN) Thrust GWS EP6030 GWS EP5043 400 GWS EP5030 GWS EP4540 0 GWS EP4530 0123456789 Current (amps)
Figure A-6. Thrust Data for Full Field of Props on MM010 10-Turn Motor at 7.4V in Zero Headwind with Phoenix 10 ESC
138
A.2.2 11.1V Data
Propeller Comparison Phoenix 25, 10 Turn MM010, 11.1V, 25 MPH headwind 2500
2000
1500 APC 4.75x4.75 GWS EP5043 GWS EP5030 1000
Thrust (mN) Thrust APC 5.1x4.5 APC 4.7x4.25 500
0 02468101214 Current (Amps)
Figure A-7. Thrust Data for Selected Props on MM010 10-Turn Motor at 11.1V and 25 MPH
Propeller Comparison Phoenix 25, 10 Turn MM010, 11.1V, 30 MPH headwind 2000
1600
APC 4.75x4.75 1200 GWS EP5043 GWS EP5030 800 APC 5.1x4.5 Thrust (mN) Thrust APC 4.7x4.25
400
0 0 2 4 6 8 10 12 14 Current (Amps)
Figure A-8. Thrust Data for Selected Props on MM010 10-Turn Motor at 11.1V and 30 MPH
139
Propeller Comparison Phoenix 25, 10 Turn MM010, 11.1V, 35 MPH headwind 1800
1500
1200 APC 4.75x4.75 GWS EP5043 900 GWS EP5030 APC 5.1x4.5 Thrust (mN) Thrust 600 4.7x4.25
300
0 02468101214 Current (Amps)
Figure A-9. Thrust Data for Selected Props on MM010 10-Turn Motor at 11.1V and 35 MPH
Propeller Comparison Phoenix 25, 10 Turn MM010, 11.1V, 40 MPH headwind 1400 1200 1000 APC 4.75x4.75 800 GWS EP5043 600 GWS EP5030 APC 5.1x4.5 Thrust (mN) Thrust 400 APC 4.7x4.25 200 0 0 5 10 15 Current (Amps)
Figure A-10. Thrust Data for Selected Props on MM010 10-Turn Motor at 11.1V and 40 MPH
140
References
[1] Saranli, U., M. Buehler, D. Koditschek, (2001), “RHex a simple and highly mobile hexapod robot,” International Journal of Robotics Research, Vol. 20, No. 7, pp. : 616-631. [2] http://www.k-team.com (website) © 2002-2006 K-Team Corporation, Y-Parc - rue Galile 9, Yverdon-les-bains 1400, Switzerland [3] Brown, B.H., M. Vande Weghe, C. Bererton, P. Khosla, (2002), “Millibot Trains for Enhanced Mobility,” IEEE/ASME Transactions on Mechatronics, Vol. 7, No. 4, pp. 452-61. [4] Koyanagi, T., R Fukui, A, Torii, A. Ueda, (2000) “Miniature Robots with Three Degrees of Freedom,” Proceedings 2000 International Symposium on Micromechatronics and Human Science, (MHS 2000), pp 207-12. [5] Birch, M.C., R.D. Quinn, R.E. Ritzmann, A.J. Pollack, S.M Philips, “Micro-robots inspired by crickets.” Proceedings of Climbing and Walking Robots Conference (CLAWAR’02), Paris, France, 2002. [6] Clark, J.E., J.G. Cham, S.A. Bailey, E.M. Froehlich, P.K. Nahata, R.J. Full, M.R. Cutkosky, (2001), “Biomimetic design and fabrication of a hexapedal running robot,” Proceedings IEEE International Conference on Robotics and Automation (ICRA), Vol.4, pp. 3643-9. [7] Kim, S., J.E. Clark, M.R. Cutkosky, (2006), “iSprawl: Design and Tuning for High Speed Autonomous Open Loop Running,” International Journal of Robotics Research, V.25, No. 9. [8] Watson, J.T., R.E. Ritzmann, S.N. Zill, A.J. Pollack (2002), “Control of obstacle climbing in the cockroach, Blaberus discoidalis: I. Kinematics,” Journal of Comparative Physiology, Vol. 188, p 39-53. [9] T.J. Allen, R.D. Quinn, R.J. Bachmann, R.E. Ritzmann, (2003), “Abstracted Biological Principles Applied with Reduced Actuation Improve Mobility of Legged Vehicles,” Proceedings of IEEE International Conference on Intelligent Robots and Systems (IROS ’03), V.2, pp. 1370–1375, Las Vegas, USA, 2003. [10] Quinn, R.D., G.M. Nelson, R.E. Ritzmann, R.J. Bachmann, D.A. Kingsley, J.T. Offi, T.J. Allen, (2003) “Parallel Strategies For Implementing Biological Principles Into Mobile Robots,” International Journal of Robotics Research, Vol. 22, No. 3, pp. 169-186. [11] Alexander, R.McN. “Three Uses for Springs in Legged Locomotion,” International Journal of Robotics Research 9:2, 1990. [12] Morrey, J.M., A.D. Horchler, N. Didona, B. Lambrecht, R.E. Ritzmann, R.D. Quinn, (2003), “Increasing Small Robot Mobility Via Abstracted Biological Inspiration,” IEEE International Conference on Robotics and Automation (ICRA’03) Video Proceedings. [13] Mueller, T.J., (1985), “The Influence of Laminar Separation and Transition on Low Reynold’s Number Airfoil Hysteresis,” Journal of Aircraft, Vol. 22, pp. 763-770. [14] Ifju, P.G. S. Ettinger, D.A. Jenkins, Y. Lian, W. Shyy, M.R. Waszak, “Flexible-Wing-Based Micro Air Vehicles,” 40th AIAA Aerospace Sciences Meeting, Reno, NV AIAA 2002-0705, Jan. 2002. [15] Frampton, K.D., M. Goldfarb, D. Monopoli, D. Cveticanin, “Passive Aeroelastic Tailoring for Optimal Flapping Wings,” Proceeding of the Fixed, Flapping and Rotary Wing Vehicles at Very Low Reynolds Numbers, pp. 26-33, 2000. [16] Jones, K.D., S.J. Duggan, M.F. Platzer, (2001), “Flapping-Wing Propulsion for a Micro Air Vehicle,” AIAA Chapter No. 2001-0126.
141
[17] Jones, K.D., C.J. Bradshaw, J. Papadopoulos, M.F. Platzer, (2004), “Improved Performance and Control of Flapping-Wing Propelled Micro Air Vehicles,” 42nd AIAA Aerospace Sciences Meeting and Exhibit, AIAA Chapter 2004-0399, Reno, Nevada, January 5-8. [18] Wood, R.J. (2008), “The First Takeoff of a Biologically Inspired At-Scale Robotic Insect,” IEEE Transactions on Robotics, Vol. 24, No. 2, pp. 341-347. [19] Lentink, D., N. Bradshaw, S.R. Jongerius, “Novel micro aircraft inspired by insect flight,” Comparative Biochemistry and Physiology, Part A, 146:S133-S134, 2007. [20] Ellington, C.P., “The Aerodynamics of Hovering Flight,” Philosophical Transactions of the Royal Society of London, Vol 305, No. 1122, 1984, pp. 1~181. [21] Grasmeyer, J.M. and M.T. Keennon, (2001), “Development of the Black Widow Micro Air Vehicle,” AIAA Paper No. 2001-0127. [22] Morris, S. and M. Holden, (2000), “Design of Micro Air Vehicles and Flight Test Validation,” Proceeding of Fixed, Flapping and Rotary Wing Vehicles at Very Low Reynolds Numbers, pp.153-176. [23] Frontiers of Engineering: Reports on Leading Edge Engineering, 2001 NAE Symposium on Frontiers of Engineering, National Academy of Engineering, pg. 12, 2002. [24] Shyy, W., Computational Modeling for Fluid Flow and Interfacial Transport, Elsevier, Amsterdam, The Netherlands, 1994 (revised printing 1997). [25] Shyy, W., S.S. Thakur, H. Ouyang, J. Liu, E. Blosch, Computational Techniques for Complex Transport Phenomena, Cambridge University Press, New York, 1997. [26] Shyy, W., H.S. Udaykumar, M.M. Rao, R.W. Smith, Computational Fluid Dynamics with Moving Boundaries, Taylor & Francis, Washington, DC, 1996 (revised printing 1997 & 1998). [27] Smith, R.W. and W. Shyy, “Computational Model of Flexible Membrane Wings in Steady Laminar Flow,” AIAA Journal, Vol. 33, No. 10, 1995, pp. 1769~77. [28] Jenkins D.A., W. Shyy, J. Sloan, F. Klevebring, M. Nilsson, “Airfoil Performance at Low Reynolds Numbers for Micro Air Vehicle Applications,” Thirteenth Bristol International RPV/UAV Conference, University of Bristol, 1998. [29] Ifju, P.G., S. Ettinger, D.A. Jenkins, L. Martinez, “Composite Materials for Micro Air Vehicles,” Proceeding for the SAMPE Annual Conference, Long Beach CA, May 6-10, 2001. [30] Waszak, M.R., L.N. Jenkins, P.G. Ifju, “Stability and Control Properties of an Aeroelastic Fixed Wing Micro Aerial Vehicle,” AIAA 2001~4005. [31] Jenkins, D.A., P.G. Ifju, M. Abdulrahim, S. Olipra, “Assessment of the Controllability of Micro Air Vehicles,” Micro Air Vehicle Conference, Bristol England, April 2001. [32] Ettinger, S.M., M.C. Nechyba, P.G. Ifju, M. Waszak, “Vision-Guided Flight Stability and Control for Micro Air Vehicles,” Proceedings of the IEEE International Conference on Intelligent Robots and Systems, vol. 3, pp. 2134~40, 2002. [33] Zufferey, J.-C., A. Klaptocz, A. Beyeler, J.-D. Nicoud, D. Floreano, “A 10-gram Vision- based Flying Robot,” Advanced Robotics, 21(14):1671-1684, 2007. [34] Georgiadis, C., A. German, A. Hogue, H. Liu, C. Prahacs, A. Ripsman, R. Sim, L.-A. Torres, P. Zhang, M. Buehler, G. Dudek, M. Jenkin, E. Milios, “AQUA: An Aquatic Walking Robot,” Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2004, Sendai, Japan, Sept. 28 – Oct. 2, 2004. [35] Lachat, D., A. Crespi, A.J. Ijspeert (2006), “Boxybot: A Swimming and Crawling Fish Robot Controlled by a Central Pattern Generator,” Proceedings of The First IEEE/RAS- EMBS International Conference on Biomedical Robotics and Biomechatronics (BioRob 2006), pp. 643-648.
142
[36] Ijspeert, A., A. Crespi, D. Ryczko, J.-M. Cabelguen, “From swimming to walking with a salamander robot driven by a spinal cord model,” Science, 315(5817):1416-1420, 2007. [37] Ayers, J., J.L. Davis, A. Rudolph (eds.), Neurotechnology for Biomimetic Robots, The MIT Press, pp 481-509. [38] Michelson, R., D. Helmick, S. Reece, C. Amareno, “A Reciprocating Chemical Muscle (RCM) for Micro Air Vehicle "Entomopter" Flight,” AUVSI’97, Proceedings of the Association for Unmanned Vehicle Systems International, July 1997. [39] Kovac, M., J.-C. Zuffrey, D. Floreano, (2009), “Towards a self-deploying and gliding robot,” in Flying Insects and Robots, Springer. [40] Pringle, J.W.S., Insect Flight. Cambridge University Press, Cambridge, 1957. [41] Ritzmann, R.E., C.R. Fourtner, A.J. Pollack, “Morphological and physiological identification of motor neurons innervating flight musculature in the cockroach, Periplaneta americana,” Journal of Experimental Zoology, 225:347-356, 1983. [42] Witushynsky, T. (2008) “Wind Tunnel Testing of Flexible Wings” Master’s Thesis, Case Western Reserve Univerity. [43] Procerus Technologies (http://www.procerusuav.com/), “Kestrel Installation and Configuration Guide,” May 2006.
143