PAGE 1

CONTR OL OF MICR O AIR VEHICLES USING WING MORPHING By HELEN MICHELLE GARCIA A THESIS PRESENTED T O THE GRADU A TE SCHOOL OF THE UNIVERSITY OF FLORID A IN P AR TIAL FULFILLMENT OF THE REQ UIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORID A 2003 

PAGE 2

T ABLE OF CONTENTS page LIST OF T ABLES . . . . . . . . . . . . . . . . . iii LIST OF FIGURES . . . . . . . . . . . . . . . . . i v ABSTRA CT . . . . . . . . . . . . . . . . . . . vi CHAPTER1 INTR ODUCTION . . . . . . . . . . . . . . . . 1 2 MICR O AIR VEHICLES . . . . . . . . . . . . . . 5 3 MORPHING . . . . . . . . . . . . . . . . . 9 4 MODELING . . . . . . . . . . . . . . . . . 16 5 24 in MICR O AIR VEHICLE . . . . . . . . . . . . . 22 5.1 V ehicle Description . . . . . . . . . . . . . . 22 5.2 Morphing . . . . . . . . . . . . . . . . 23 5.3 Flight T esting . . . . . . . . . . . . . . . 24 5.4 Modeling . . . . . . . . . . . . . . . . 28 5.5 Ev aluation . . . . . . . . . . . . . . . . 31 6 12 in MICR O AIR VEHICLE . . . . . . . . . . . . . 33 6.1 V ehicle Description . . . . . . . . . . . . . . 33 6.2 Morphing . . . . . . . . . . . . . . . . 34 6.3 Flight T esting . . . . . . . . . . . . . . . 38 6.4 Modeling . . . . . . . . . . . . . . . . 40 6.5 Ev aluation . . . . . . . . . . . . . . . . 43 7 CONCLUSION . . . . . . . . . . . . . . . . . 44 REFERENCES . . . . . . . . . . . . . . . . . . 46 BIOGRAPHICAL SKETCH . . . . . . . . . . . . . . . 48 ii 

PAGE 3

LIST OF T ABLES T able page 5–1 Range of control ef fectors . . . . . . . . . . . . . 23 5–2 Properties of the 24 in MA V . . . . . . . . . . . . 23 5–3 Poles of a linear model of the 24 in MA V . . . . . . . . . 30 6–1 Properties of the 12 in MA V . . . . . . . . . . . . 33 6–2 Poles of a linear model of the 12 in MA V . . . . . . . . . 41 iii 

PAGE 4

LIST OF FIGURES Figure page 1–1 MA Vs with 24 in (left) and 12 in (right) W ingspan . . . . . . . 3 2–1 Members of MA V Fleet at Uni v ersity of Florida . . . . . . . 7 3–1 A Gull (left) and Sno wy Owl (right) in Flight . . . . . . . . 12 3–2 Aspect Ratio of Bird W ings . . . . . . . . . . . . 13 5–1 Ov erhead V ie w of the 24 in MA V . . . . . . . . . . . 22 5–2 W ing with T orque Rod . . . . . . . . . . . . . . 23 5–3 Rear V ie w of the 24 in MA V with Undeected (left) and Morphed (right) W ing . . . . . . . . . . . . . . . . . . 24 5–4 Doublet Command to Rudder Serv o . . . . . . . . . . 25 5–5 Response to Rudder Doublet for Roll Rate(left) and Y a w Rate(right) . 25 5–6 Second Doublet Command to Rudder Serv o . . . . . . . . 26 5–7 Response to Second Rudder Doublet for Roll(left) and Y a w Rate(right) . 26 5–8 Doublet Command to Morphing Serv o . . . . . . . . . . 27 5–9 Response to Morphing Doublet for Roll(left) and Y a w Rate(right) . . 27 5–10 Second Doublet Command to Morphing Serv o . . . . . . . 28 5–11 Response to Second Morphing Doublet Command for Roll Rate(left) and Y a w Rate(right) . . . . . . . . . . . . . . 28 5–12 Simulated () and Actual (—) Roll Rate(left) and Y a w Rate(right) Responses to Morphing Doublet . . . . . . . . . . . 29 5–13 Simulated () and Actual (—) Roll Rate(left) and Y a w Rate(right) Responses to Morphing Doublets . . . . . . . . . . . 31 6–1 V ie w of the 12 in MA V . . . . . . . . . . . . . . 33 6–2 V ie w of the 10 in MA V . . . . . . . . . . . . . . 34 6–3 W ing with K e vlar Threads . . . . . . . . . . . . . 37 i v 

PAGE 5

6–4 Front V ie w of the 12 in MA V with Undeected (left) and Morphed (right) W ing . . . . . . . . . . . . . . . . . . 37 6–5 Doublet Command to Morphing Serv o . . . . . . . . . . 39 6–6 Roll Rate(left) and Y a w Rate(right) in Response to Morphing Doublet . 39 6–7 Second Doublet Command to Morphing Serv o . . . . . . . 40 6–8 Response to Second Morphing Doublet Command for Roll Rate(left) and Y a w Rate(right) . . . . . . . . . . . . . . 40 6–9 Simulated() and Actual(-) Roll Rate(left) and Y a w Rate(right) Responses to a Doublet . . . . . . . . . . . . . . 41 6–10 Simulated () and Actual (—) Roll Rate(left) and Y a w Rate(right) Responses to Morphing Doublets . . . . . . . . . . . 42 v 

PAGE 6

Abstract of Thesis Presented to the Graduate School of the Uni v ersity of Florida in P artial Fulllment of the Requirements for the De gree of Master of Science CONTR OL OF MICR O AIR VEHICLES USING WING MORPHING By Helen Michelle Garcia December 2003 Chair: Richard C. Lind, Jr Major Department: Mechanical and Aerospace Engineering A micro air v ehicle (MA V) is typically dened to ha v e a wingspan of 6 in and operates with airspeeds of less than 25 mph Recent attention has been de v oted to MA Vs because there is a v ariety of applications for which the y can be used. Specically the y are useful in missions within urban en vironments. These missions require MA Vs to be v ery small and highly agile. These characteristics are achie v ed with the use of light materials such as carbon ber airframes and plastic membrane wings; ho we v er this design also causes these v ehicles to be dif cult to operate. This construction mak es them strong with good aerodynamic properties; ho we v er hinges for con v entional control surf aces are not easily implemented on e xible wings. Therefore, these v ehicles are strong and light b ut ha v e limited control authority A dif ferent control ef fector is then necessary in order to pro vide control authority This thesis will consider using wing morphing as a control ef fector Simple techniques are used to change the shape of the wings during ight such as twisting the wings of a 24 in MA V and curling the wings of a 12 in MA V Flight tests are then performed and sho w that morphing is an ef fecti v e w ay to induce roll motion. The data are then analyzed to consider linear modeling techniques as well as control design. The use of morphing vi 

PAGE 7

results in a more ef fecti v e roll motion than the use of the rudder and impro v es the maneuv erability of the v ehicles. vii 

PAGE 8

CHAPTER 1 INTR ODUCTION Recent adv ances in technology ha v e made the use of smart v ehicles in a v ariety of applications possible. These adv ances ha v e led to the de v elopment of small unmanned air v ehicles (U A V) as well as micro air v ehicles (MA V) which ha v e mission capabilities. These v ehicles can range in size according to their specic application. Small U A Vs and MA Vs can be designed to operate within urban en vironments. Their design typically depends on the mission in which the y will be used and can range from ci vilian to military applications. F or e xample, these v ehicles might be assigned to military missions such as bomb damage assessment, which w ould in v olv e transmitting video after a bomb has been dropped at a specic location. These v ehicles could also be used in w arf are as a means to ha v e li v e video of what is occuring ahead before ground troops are sent into the battleeld. There are also some ci vilian applications to MA Vs, such as transmitting video for traf c/ne ws co v erage, and to look in specic places for search and rescue missions. Other scenarios where MA Vs w ould be useful include assessing the damage caused by chemical spills, in search and aiding in rescue missions and e v en in trafc/ne ws co v erage. Some of these applications might require MA Vs to w ork in collaboration with other MA Vs in order to co v er a lar ge area of surv eillance. Another application where a MA V w ould be useful is if for e xample, there is a biological agent released into a public area. A biohazard team w ould ha v e to suit up and prepare before being able to enter the area for testing and clearing of the biological agent. Instead, a MA V can be released from a nearby station which can ha v e 1 

PAGE 9

2 sensors on board, and be able to detect the e xtent of the contamination and the type of biological agent which w as released before an y humans ha v e arri v ed at the scene. The recent concern for terrorism being planned in urban en vironments such as small apartments leads to specic situations where MA Vs could be useful. Current technology cannot k eep track and observ e these small suspected areas. A MA V w ould be useful in this situation because it could y up to windo ws or e v en indoors and send li v e video and audio of the suspect acti vities. In order to accomplish the e v olving missions for which MA Vs can be used, their designs ha v e to be considered accordingly F or most military missions, the requirements typically demand that these v ehicles be v ery small and highly agile. The y are constructed of v ery light materials, such as carbon ber airframes and e xible membrane sheeting for wings. Ho we v er this v ehicle design does not allo w for con v entional control surf aces due to the comple xity of implementing hinges along the e xible membrane wings. The lack of con v entional control surf aces mak es these v ehicles dif cult to y Se v eral benets can be achie v ed by the use of additional control ef fectors. The implementation of wing morphing as an additional control ef fector is considered to increase controllability Simple techniques are considered to study the benets of using morphing as compared to the current control ef fectors which are used on MA Vs. This thesis will consider morphing as a control ef fector for a pair of micro air v ehicles with dif ferent dimensions. The v ehicles sho wn in Figure 1–1 with a 24 in wingspan and a 12 in wingspan will be used to demonstrate morphing for control. These v ehicles will be referred to as MA Vs, though their wings are lar ger than the denition implies, due to the similarities in design and construction. The MA V is an ideal platform for morphing because the po wer required is v ery small due to the e xibility of the wings, and there are se v eral benets including increased controllability 

PAGE 10

3 Figure 1–1: MA Vs with 24 in (left) and 12 in (right) W ingspan W ing morphing on the 12 in MA V is actuated by connecting a strand of K e vlar to a serv o inside the fuselage and then to a location on the wing. This initial design only allo wed for morphing at a single location, which w as the wing outboard. Further testing led to the consideration of more dramatic morphing. The morphing of the wing w as made more dramatic by then adjusting an e xtra K e vlar strand to the serv os. This allo ws for morphing about the span and the trailing edge at the same time. The morphing is actuated on the 24 in MA V by use of torque rods which are connected from the fuselage to the wing outboard. These torque rods are connected to the wings by being se wn along the plastic sheeting of the wing. These rods are then actuated and in turn change the shape of the wings. Data is then collected with a data acquisition system which pro vides information with 3 accelerometers and 3 gyros along an orthogonal coordinate system of the v ehicle. This data is then processed to consider the response of the aircraft to the input commands. A linear model approximation is then done using the roll rate and ya w rate responses. The approximation is done using an ARX technique in Matlab using the responses. The approximation results in a good correlation of the coef cients when compared to the data which w as collected. The linear model for the 24 in MA V can then be approximated b ut the model of the 12 in MA V requires further research. This is 

PAGE 11

4 due to the f act that the 24 in MA V is being morphed symmetrically so that a linear approximation can be made, b ut the 12 in MA V is morphed asymmetrically which requires a nonlinear study to be done. 

PAGE 12

CHAPTER 2 MICR O AIR VEHICLES Micro Air V ehicles(MA Vs) are typically dened as v ehicles with a wingspan of less than 6 in and which operate with airspeeds of less than 25 mph The idea for a MA V is to ha v e a platform which is small, ine xpensi v e and that can be used in situations which are not suitable for lar ger v ehicles [ 14 ]. The rst succesfull design of a micro air v ehicle w as achie v ed by AeroV ironment. The y designed the Black W ido w MA V with funding from D ARP A. The Black W ido w is a MA V with a 6 in wingspan, airspeed of about 30 mph and weighs under 100 gr ams [ 13 ]. This v ehicle also has the capability of carrying a video camera which transmits li v e video to the ground and has an endurance of 30 minutes. It is also equipped with an autopilot, which is capable of performing altitude, airspeed, and heading holds as well as a ya w damper The transmitter and actuators are some of the smallest and lightest systems a v ailable. This design led to further interest and research in the eld of MA Vs from se v eral countries and uni v ersities. The Uni v ersity of Florida has been v ery acti v e in the design and testing of micro air v ehicles. Dr Peter Ifju has led a successful research team at the Uni v ersity of Florida in the area of MA Vs. The y ha v e been able to win v arious aspects of the annual MA V competition, which is sponsored by the International Society of Structural and Multidisciplinary Optimization. The MA V team at the Uni v ersity of Florida has succeeded in this competition each year from 1999 to 2003 with the v arious designs that ha v e been tested and ha v e w on the o v erall rst place a w ard. The annual MA V competition typically includes entries from se v eral uni v ersities around the w orld. The 2003 competition consisted of entries from 15 uni v ersities. 5 

PAGE 13

6 Current MA V designs are based on a e xible wing design used at the Uni v ersity of Florida [ 14 ]. The most common design is an airframe constructed entirely of composite carbon ber The fuselage is typically a tw o-piece monocoque structure designed to house ight components and instrumentation. The ight components include serv os and connectors, and some of the instrumentation used in ight includes orientation systems. A con v entional empennage is af x ed to the fuselage with ele v ators and rudders hinged to the horizontal and v ertical stabilizers. The MA Vs are equipped with sensors for measurement consisting of 3-axis gyros and 3-axis accelerometers along with the serv o command. The sensing and actuation data is recorded on an on board data acquisition board which weighs 7 gr ams and w as de v eloped by N ASA Langle y Research Center specically for MA V applications [ 1 ]. This micro data acquisition board is capable of recording 27 analog channels which is suf cient for the current sensor package. The data is then a v ailable at 50 to 100 Hz and is resolv ed using a 12-bit analog-digital con v erter The data is recorded in a 4 MB ash chip on board the data acquisition board and is then do wnloaded to a PC at the end of each ight. The v ehicles use an electric motor for propulsion and the duration of ights depend on the amount of batteries which can be carried and the throttle setting on the motor On a v erage, ights ranging from 10-15 minutes are easily achie v ed for the 24 in and 12 in MA Vs which are considered. Structures used in ight, both in biological and aircraft applications, are e xible by a certain amount. F or aircraft to be able to withstand the lar ge forces obtained during ight; ho we v er the wings ha v e to be strong enough. Flight of birds also consists of e xible wings which can adapt to the changing en vironments the y y in. Birds ha v e man y layers of feathers which can be mo v ed around in order to adjust to the specic maneuv er the y need to perform [ 22 ]. The use of apping for ight, such as done by birds, has not been e xtensi v ely studied. This has 

PAGE 14

7 not been done due to the comple xity of the ight mechanics which includes changing geometry e xible surf aces and unsteady aerodynamics. The e xible membrane wings together with the size constraints of MA Vs mak e analysis and design of these v ehicles v ery challenging. Specically the aerodynamics of a MA V are complicated by lo w Re ynolds number ight, lar gely deforming structures, the ef fects of viscosity and o w separation at high angles of attack [ 22 ]. The f act that the MA Vs are challenging to design pro vides a good platform for research in the areas of dynamics and control, aeroserv oelasticity structures, microelectronics, small actuation and data acquisition systems, and other elds. The research and de v elopment of these v ehicles has progressed rapidly due to the a v ailability of smaller electronics as well as the adv ances in lighter materials. Se v eral of these v ehicles are sho wn in Figure 2–1 Figure 2–1: Members of MA V Fleet at Uni v ersity of Florida Adv ances in miniature digital electronics, communications and computer technologies ha v e made sensing capabilities on micro air v ehicles possible. A typical application of these miniature electronics is in a reconaissance mission where a small MA V w ould be preferred to a lar ger v ehicle in order to remain stealthy The use of inno v ati v e control ef fectors is an area being e xplored as an enabling technology for designing a stability augmentation system. The current generation of MA Vs uses traditional ef fectors, specically an ele v ator and rudder whose positions are commanded by the remote pilot. The ele v ator presents adequate ef fecti v eness for longitudinal control b ut the rudder presents some dif culty for lateral-directional control. 

PAGE 15

8 The rudder mainly e xcites the dutch roll mode so steering and gust rejection are really accomplished using the coupled roll and ya w motion resulting from dutch roll dynamics. Such an approach is ob viously not optimal b ut traditional ailerons are not feasible on this type of aircraft. Actuation of the MA V control surf aces is accomplished with tw o control ef fectors or serv os mounted inside the fuselage. These de vices actuate the control surf aces by rotating an arm and pushing or pulling a pushrod when a deection is commanded. Research in the design and testing of these v ehicles has been done in se v eral countries and uni v ersities. Research of micro air v ehicles has also been done e xtensi v ely at N ASA. Specically N ASA has considered a control assesment and simulation of a micro air v ehicle with aeroelastic wings which adapt to the disturbances during ight [ 23 ]. This w as done using aerodynamic data which w as obtained in pre vious testing of a MA V [ 24 ]. 

PAGE 16

CHAPTER 3 MORPHING The concept of morphing is not an idea which has been strictly dened. A morphing aircraft is generally dened to be an aircraft whose shape changes during ight to optimize performance. T ypes of shape changes include span, chord, camber area, thickness, aspect ratio and planform. The morphing can also be applied to a control surf ace in order to eliminate hinges. Morphing can be used as a control ef fector by changing the shape of the aircraft in order to alter the ight dynamics. The concept of morphing has been look ed at by D ARP A and N ASA to sho w the aerodynamic benets; ho we v er the use of morphing for control design has not been studied e xtensi v ely The wing morphing techniques for the MA Vs in this project consider using serv os which are attached to the wings. Aircraft ha v e pre viously used techniques for adapting their shape depending on the specic ight characteristics desired. The use of morphing as a control ef fector w as used initially on the Wright Flyer where the pilot used cables to twist the wings in order to achie v e the desired motion. W ing w arping did not become a common technique; ho we v er due to the po wer which is required from actuators to change the shape of the wings [ 21 ]. Morphing is also used on the F-14 which has a v ariable sweep on the wing, therefore changing the shape of the wing during ight. The wings are swept in order to balance the range and speed by slo wing do wn the increase in drag which de v elops as v elocity increases [ 21 ]. There are dif ferent w ays that an aircraft can be morphed which are appropriate for control. The current research will focus on morphing of the wings in order to consider 9 

PAGE 17

10 primarily control issues. The ight characteristics of birds will be considered as a guide since the y also change the shape of their wings to achie v e certain maneuv ers. Man y mechanisms which consider morphing ha v e been designed b ut ha v e not been tested in ight v ehicles. N ASA has designed a wing which changes the camber of the wing [ 6 ]. One of the wings considered is referred to as a Hyper -Elliptic Cambered Span (HECS) because the curv ature along the span is continuously changing. This pro vides a lar ger area which allo ws for greater lift. This v ehicle uses a hinge-less panel along the trailing edge of the wing as a form of a control surf ace for pitch and roll. The simulations demonstrated the aerodynamic benets b ut also sho w this v ehicle has unstable lateral-directional dynamics. The use of smart materials such as shape memory allo ys and piezos ha v e been considered in the design of morphing wings b ut there is still a limit in that not enough force can be produced in order to twist lar ge wings using these mechanisms. Ho we v er smart spars ha v e been b uilt which pro vide dif ferent types of morphing b ut ha v e not been tested on ight v ehicles [ 2 ]. Another mechanism for morphing which has been studied considers changing the sweep of the wings of a small unmanned air v ehicle (U A V) [ 7 ]. The morphing on this U A V is done in order to meet changing mission requirements. Actuation of the morphing is done by using inatable actuators which are po wered with compressed air One of the benets of this project is that the actuation mechanism used is much lighter than the typical hydraulic systems that are used on full scale aircraft. The ef fects of the sweep are then studied considering the change of aspect ratio, lift and drag. Similarly a study has been done considering an inatable telescopic spar which can be morphed spanwise [ 4 ]. This design allo ws for changes in the aspect ratio while still pro viding enough support from the spars for the airloads which are being applied. This is achie v ed because the telescopic spar is pressurized and the telescopic skins 

PAGE 18

11 maintain the geometry of the airfoil as well as pro vide ef fecti v e storing and deplo yment of the mechanism. Also for the purpose of considering impro v ed maneuv erability and performance, roll maneuv ers ha v e been studied using a e xible wing [ 16 ]. Numerical studies were used to consider the aerodynamic loads on a e xible wing at high speeds. W ing twist is also considered in order to reco v er the rolling moment lost b ut has not been tested in ight. This is due to the challenges in v olv ed in implementing a functioning mechanism for wing twist on a full scale aircraft. Research has also been done considering the material aspects of shape changing with a nite element model of a wing [ 17 ]. This considered roll maneuv ers using a piezoelectric material as an actuation mechanism with aerodynamic loads being applied. Piezoelectric sensors and actuators are useful for this application because the y are light, ha v e a small v olume and can achie v e v arious shapes. This technique has not been tested; ho we v er due to the lar ge deections that are needed from such small actuators. Numerical studies ha v e considered structural and aerodynamic modeling for shape changing wings [ 12 ]. These considered a generic lambda wing such as used in unmanned combat air v ehicles (UCA V). The dif ferent mode shapes were studied to consider structural modeling. This model w as then used to study the roll performance of the morphing wing. It w as sho wn that this wing with hinge-less control surf aces sho ws impro v ed roll performance because of the aerodynamic and structural benets. As mentioned pre viously there are se v eral benets such as impro v ed performance due to the use of wing morphing. Also, morphing is easily achie v ed on MA Vs because the wings are constructed of e xible membrane material. The e xible wings can be grossly deformed via mechanical actuation yet are capable of withstanding ight loads. The e xible nature of the wing also gi v es rise to the mechanism of adapti v e w ashout which permits small changes in wing shape in response to gusty wind conditions. 

PAGE 19

12 F or this project, morphing is limited to changing the shape of the wings, not of the entire airframe. This type of morphing can be studied with the use of biologically inspired techniques. The dif ferent w ays that birds change the shape of their wings during ight is studied and compared with morphing techniques. Consider for e xample, the birds in Figure 3–1 These birds typically change the shape of their wings depending on the types of maneuv ers that the y need to perform. Figure 3–1: A Gull (left) and Sno wy Owl (right) in Flight Certain techniques of morphing for aircraft can be designed by studying these birds. There are se v eral morphing techniques which are used by these birds that demonstrate ho w their ight maneuv ers can be changed, such as loitering, di ving and tak e-of f. The wings of birds are shaped similarly to airfoils and ha v e the same basic function [ 8 ]. Certain birds use their wings more often than others who just y for short periods of time. Also, the en vironment the birds are in can af fect the aerodynamics of the ight, therefore birds ha v e dif ferent shapes of wings. The aspect ratio of the wings of birds is measured as the square of the span of the wing di vided by the area of the wing. This ratio can v ary depending on the specic technique for ying of each bird. F or e xample, long wings pro vide a smoother gliding motion b ut it tak es more ener gy to ap them quickly therefore the y are not useful for increasing speed. Therefore, birds with longer wings tend to use gliding as their primary method of ight. W ing loading can also af fect ho w a bird ies since the ener gy required to ap their wings also depends on ho w hea vy the y are. 

PAGE 20

13 Consider for e xample, Figure 3–2 which sho ws the wing design for four dif ferent birds [ 8 ]. The lo wer aspect ratio wings, such as for a pheasant, typically allo w for quick tak e of f and slo w ights, b ut are not useful for gliding. The slightly lar ger aspect ratio wings, such as for eagles, are typically longer and ha v e feathers which are adjusted as a type of control surf ace for more precise maneuv ering. The wings for w aders, with a typical aspect ratio of 12.5, are useful for f aster speeds and gliding b ut do not allo w for f ast tak e of f. This limit on a f ast tak e of f is because a lot of ener gy is required to ap these longer wings. The higher aspect ratio wings, as for gulls, are typically useful for gliding close to surf aces such as sea and land and tak e adv antage of the winds in order to conserv e ener gy These are only a fe w of the man y dif ferent designs of wings which v ary depending on the migration patterns of each bird. Figure 3–2: Aspect Ratio of Bird W ings Some biologically inspired techniques can be applied to MA Vs. The span of the wings, the horizontal distance from the tip of a wing to the tip of the other can be altered to create a shorter wing for e xample. Birds and bats are also capable of changing the span of their wings to decrease the area, therefore increasing the forw ard v elocity and reducing drag. The chord, which is the distance from the leading edge to the trailing edge, can also be altered. The wing can also be morphed by twisting or rotating parts of the wing in order to af fect aerodynamic performance. 

PAGE 21

14 Another type of morphing is sweeping the wing either at an elbo w joint on the wing or at the root of the wing. This pro vides a type of wing sweep which tak es a similar shape change as seen in birds. The area of the wing can also be changed by e xtending the length or trailing edge as some birds do. The aspect ratio is also af fected by the morphing and can be used to consider lift and drag for aerodynamics. A simple form of morphing is a wing twist. This is currently being used for control on the Acti v e Aeroelastic W ing (AA W) as well as the v ehicles in this project. The morphing on the AA W causes the wings to be twisted in response to the moments induced by the control surf aces. Birds and bats also do this in order to obtain the required lift or thrust during ight. Morphing on the MA Vs is accomplished by actuation of control ef fectors located inside the fuselage. These serv os are connected to the wings by either use of a torque rod or K e vlar strand. The wing morphing is actuated by mo ving the arm which rotates the tube or pulls the strand and changes the shape of the wing. Certain maneuv ers are of interest when considering the ef fects of wing morphing on a MA V The ight test maneuv er of interest is a control doublet for both rudder and wing shaping controls. The rudder doublet is being applied only to the 24 in MA V since the 12 in MA V consists of only ele v ator and wing morphing control ef fectors. These maneuv ers are performed by commanding a constant left deection for a certain time period follo wed immediately by a right deection for the same time period and nally returning to the neutral position. Aircraft response characteristics to the control input are then determined by analysis of the serv o position and rate responses. W ing-shaping control doublets induce a dif ferent beha vior of the MA V The response of the airplane to wing shaping is similar in nature to responses from ailerons. Essentially the aircraft response to the morphing is predominantly roll motion with little ya w or pitch coupling. Thus, the doublets are performed without considerable directional or altitude de viation. 

PAGE 22

15 F ollo wing the completion of the maneuv er which resembles rocking the wings, the airplane is in a bank ed attitude. Reco v ery from the wing shaping doublet is considerably easier than that of the rudder doublet. Such a response indicates the wing shaping e xcites the roll con v er gence mode. Clearly the MA V requires a stability augmentation system to f acilitate operation and greatly e xpand its mission capability In general, lateral maneuv ers are particularly dif cult because the MA V is so responsi v e. The introduction of a controller w ould lessen pilot w orkload for trajectory tracking. The design of a controller is the ne xt step in the research of f acilitating the ability to operate these MA Vs with the aid of acti v e wing morphing. Future research will also enable de v elopment of a vision-based autopilot system currently being studied [ 9 ]. Open-loop ight tests were performed using wing morphing as an actuation mechanism. These ight tests demonstrate the v alue of morphing for consideration of a stability augmentation system. The rudder can be used to generate lateral maneuv ers b ut the tight coupling of roll and ya w complicates the control needed for trajectory tracking. Con v ersely the morphing produces almost pure roll so an associated controller for tracking roll commands will be the rst to be implemented. 

PAGE 23

CHAPTER 4 MODELING A model of a system can be described by comparing the relationship between the signals which are observ ed [ 19 ]. A model can be de v eloped with the use of data which is collected in e xperiments. System identication considers the de v elopment of the model of a system with the use of observ ed data. F or this purpose the signals typically considered are the output signals, which are measured, as well as the input signals, which consider the ef fect the observ er has on the response of a system. Other signals which can be considered are outside disturbances, which are signals that are produced from outside sources such as noise, wind gusts and sensor drift. A model is therefore a mathematical description of a system considering se v eral aspects b ut is not an e xact description of the physical system [ 19 ]. System identication is performed by rst collecting data which emphasizes the parameters that are to be considered in the model estimation. Therefore, the input and output signals as well as specic maneuv ers are selected prior to the data collection. F or some systems it is useful to describe the models using graphical interpretations. More specically the y can be described using impulse, step and frequenc y responses. Certain systems can also be described using mathematical models. These can include continuous-time and discrete-time systems as well as linear and nonlinear systems. A set of models can then be selected according to the specic application or dynamic system. A model which uses a black box approach is used for this project. This approach considers the input and output signals of the system in order to perform a t to the data without pro viding physical meaning to the v alues. This model is then 16 

PAGE 24

17 compared with the v alues obtained in the e xperiment in order to determine whether it is a good estimation of the system response. The black box model structure which considers input and output signals can be e xpressed as the linear equation sho wn in 4.1 where e(t) is the noise error term. yta 1 yt1a n a ytn anb 1 ut1b n b utn bet(4.1) Then this equation can be e xpressed in terms of the initial output signal as sho wn in 4.2 yt a 1 yt1 a n a ytn ab 1 ut1b n b utn bet(4.2) This is typically referred to as an ARX model, which denes the autore gressi v e part to be the output terms in 4.2 and the input terms in 4.2 as the e xtra input. So the initial output v alues as well as the input and output terms on the right hand side of 4.2 are collected in matrix form for each time interv al. This mak es it possible to solv e for the re gression coef cients since the initial output and the input v alues are kno wn. The initial output v alues for each time interv al can be e xpressed as in 4.3 in terms of the input and output v alues as well .r y t y t1y tn r y t 1y t n a u t 1 u t n by t1 1y t1 n a u t1 1 u t1 n b y tn 1y tn n a u tn 1 u tn n b r a 1 a n ab n b (4.3) 

PAGE 25

18 Then the re gression coef cients are obtained by using the matrix equation 4.4 and solving for the matrix of coef cients X as sho wn in 4.5 BAX (4.4) A1 BX (4.5) A transformation as sho wn in 4.6 is then applied to equation 4.1 in order to obtain a transfer function as sho wn in 4.7 In this transfer function the B term contains all the input coef cients from equation 4.6 and the A terms consists of all the coef cients in the output terms. yta 1 z1 yta n a zn a ytb 1 z1 ut b n b zn b utet(4.6) yu1B A1 (4.7) A T ustin transformation is then done using Matlab in order to create a continuous time v ersion of the discrete time system. This is done using a standard bilinear transformation such as sho wn in 4.8 z12sT22sT222sT23(4.8) The ARX model approximation is just one of se v eral types of model structures which can be used for system identication. An ARMAX model structure can similarly be used b ut w as not used in this project because an initial simple estimation w as desired. The ARMAX model considers the basic properties that the ARX model uses b ut also includes a mo ving a v erage term, which considers the noise in its coef cient calculations. 

PAGE 26

19 Another modeling technique considers recursi v e identication methods. This considers calculating a model simultaneous to obtaining data. Ho we v er this is not a requirement for this specic project b ut can be useful in dif ferent applications. Certain applications include ha ving an up to date model in order to consider these parameters when making decisions about what the system is to do ne xt. This is typically referred to as an adapti v e modeling technique because the input and output signals are calculated in order to be used as the y become a v ailable. An e xample of a recursi v e model which can be used for system identication in Matlab is the RARMAX model. This uses a recursi v e technique of an ARMAX model which considers the noise in its calculations. Ho we v er this technique only pro vides models for single-input, single-output systems. Similarly another technique is the RARX model which estimates parameters recursi v ely of a single-output system. Therefore, for this project an initial linear approximation w as done using an ARX technique. The initial step w as to design an e xperiment which consisted of specied maneuv ers such as doublets to the morphing and rudder serv os. These were done in order to consider the roll and ya w rate responses of the system. The data is then collected and processed before considering it for modeling. The data processing included using an algorithm which plotted, ltered and remo v ed the bias in the data. The ltering w as done using a lo w pass Butterw orth lter on all the parameters and the bias w as remo v ed from the parameters by subtracting the mean. This processed data is then used in the ARX modeling approximation. The roll rate and ya w rate responses are then compared to the simulation responses. This is done for both the morphing and rudder serv os. The orders and delays are selected for the parameter estimation. 

PAGE 27

20 The orders of the approximation are the orders of the polynomials A and B in equation 4.7 Therefore, the y are the orders of the polynomials in equation 4.9 and equation 4.10 Az1a 1 z1 a n a zn a (4.9) Bzb 1b 2 z1 b n b zn b1 (4.10) The delays which are referred to as nk are selected as the number of delays from input to output as sho wn in equation 4.11 Azyt Bzutnket(4.11) In multi-output systems, the orders of the polynomials ha v e as man y ro ws as outputs. This is then used to create the simulation and it is then con v erted to a continuous time system from a discrete time system. The roll rate and ya w rate responses are then compared and for this project sho w a good correlation between the estimated and the actual data for the doublet maneuv ers. The follo wing step in system identication w ould be to v alidate the model which w as chosen as the best approximation to the data. This is done by considering whether the model is a good enough approximation for what it will be used for In other w ords, whether the model can be trusted to reproduce the collected data. A model is typically not accepted as describing the actual true system, b ut simply as a good description of specic parts of the system which are of interest. The rst model obtained using these techniques typically has to be re vised because it may not describe a system considering se v eral dif ferent aspects. P articularly for this project the models were obtained by considering the inputs and outputs of the 

PAGE 28

21 system and then reproducing that data. Ho we v er the model obtained does not describe physical parameters which can be used for purposes of further control design. A model which w ould be useful for control design w ould include approximations of certain aerodynamic parameters and time constants. These parameters can then be used to design a controller as well as considering the modes. Once the system is represented in physical parameters, controllers such as roll and ya w dampers can be designed by feeding back the appropriate angles. This can be done by using a simple proportional gain in the closed loop system. 

PAGE 29

CHAPTER 5 24 in MICR O AIR VEHICLE 5.1 V ehicle Description One of the v ehicles considered is the micro air v ehicle with a 24 in wingspan sho wn in Figure 5–1 Figure 5–1: Ov erhead V ie w of the 24 in MA V The 24 in MA V consists of a carbon composite frame with a mylar membrane skin wing. The leading edge of the wings consist of carbon-ber wea v e with battens of unidirectional carbon attached to the underside and e xtending to the trailing edge. These battens pro vide the strength needed to support the airloads which are being applied while the membrane pro vides the lifting surf ace. The original control ef fectors for this MA V are the rudder and ele v ator The rudder and ele v ator each ha v e a single serv o for actuation. The control surf aces, the ele v ator and rudder are connected to the serv os using a spring steel pushrod. The approximate range of motion for each is gi v en in T able 5–1 22 

PAGE 30

23 T able 5–1: Range of control ef fectors Ef fector Range of Motion ele v ator15 o to20 o rudder25 o to25 o The basic properties of the 24 in MA V are gi v en in T able 5–2 T able 5–2: Properties of the 24 in MA V Property V alue W ingspan 24” W ing Area 100 in 2 W ing Loading 20.32 ozf t 2 Aspect Ratio 5.76 Po werplant Electric motor w/ 4.75” propeller T otal W eight 400 g 5.2 Morphing W ing morphing is used as an additional control ef fector A simple technique is used to morph the wings of the 24 in MA V The morphing is actuated by tw o serv os, one for each wing. The technique used for morphing on this v ehicle consists of the use of a torque rod which produces the deection that is commanded. This torque rod lies along each wing connected to the serv os inside the fuselage as sho wn in Figure 5–2 Figure 5–2: W ing with T orque Rod The rods are se wn into the leading edge of the membrane therefore causing mo v ement of the membrane if the rods are actuated. The ef fect of the morphing is seen 

PAGE 31

24 to act as a simple form of wing w arping. The wing deection due to the morphing actuators for the 24 in MA V is sho wn in Figure 5–3 Figure 5–3: Rear V ie w of the 24 in MA V with Undeected (left) and Morphed (right) W ing 5.3 Flight T esting Flight testing of the acti v e wing-shaping 24 in MA V is performed in the open area of a radio controlled (R/C) model eld during which wind conditions range from calm to 7 knots throughout the ights. Once the ight control and instrumentation systems are po wered and initialized, the MA V is hand-launched into the wind. This launch is an ef fecti v e method to quickly and reliably allo w the MA V to reach ying speed and be gin a climb to altitude. This airplane is controlled by a pilot on the ground who maneuv ers the airplane visually by operating an R/C transmitter There is a data acquisition system on board which be gins recording as soon as the motor is po wered. This D A Q system records accelerations and rates about the coordinate system which is centered on the MA V This aircraft design allo ws either rudder or wing shaping to be used as the primary lateral control for standard maneuv ering. The airplane is controlled in this manner through turns, climbs, and le v el ight until a suitable altitude is reached. At altitude, the airplane is trimmed for straight and le v el ight. This trim establishes a neutral reference point for all the control surf aces and f acilitates performing ight test maneuv ers. 

PAGE 32

25 Open-loop data is tak en to indicate the ight characteristics of the MA V Specically the roll and ya w rates and accelerations about a body x ed axis are measured in response to doublets commanded separately to the serv os. Se v eral sets of doublets are commanded ranging in magnitude and duration to obtain a di v erse set of ight data. The dynamics of the MA V in response to rudder commands is in v estigated to indicate the performance of the traditional conguration for this MA V A representati v e doublet command is sho wn in Figure 5–4 The roll rate and ya w rate measured in response to this command are sho wn in Figure 5–5 The roll rate is lar ge and indicates the rudder is able to pro vide lateral-directional authority; ho we v er the ya w rate is clearly lar ger than the minimal amount which can be e xpected. Actually the ya w rate is close in magnitude to the roll rate so the lateral-directional dynamics are v ery tightly coupled. The ef fect of the rudder in e xciting the dutch roll dynamics is clearly seen in this response. 0 1 2 3 4 5 6 15 10 5 0 5 10 15 Time(sec)Rudder Command Figure 5–4: Doublet Command to Rudder Serv o 0 1 2 3 4 5 200 150 100 50 0 50 100 150 Time(sec)Roll Rate (deg/sec) 0 1 2 3 4 5 200 150 100 50 0 50 100 150 Time(sec)Yaw Rate (deg/sec) Figure 5–5: Response to Rudder Doublet for Roll Rate(left) and Y a w Rate(right) 

PAGE 33

26 Another doublet is commanded to the rudder in order to consider its response. The rudder is commanded with a slightly lar ger magnitude and longer duratio doublet which is sho wn in Figure 5–6 The response to this doublet command is sho wn in Figure 5–7 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 20 15 10 5 0 5 10 15 Time(sec)Rudder Command Figure 5–6: Second Doublet Command to Rudder Serv o 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 200 150 100 50 0 50 100 150 200 Time(sec)Roll Rate (deg/sec) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 200 150 100 50 0 50 100 150 200 Time(sec)Yaw Rate (deg/sec) Figure 5–7: Response to Second Rudder Doublet for Roll(left) and Y a w Rate(right) This sho ws a similar response as with the rst doublet to the rudder serv o. It sho ws a roll rate response of similar magnitude as the ya w rate. This then describes a dutch roll motion instead of a pure roll motion e v en with a slightly lar ger command to the rudder Also, since the doublet w as of lar ger magnitude and longer duration, this w ould indicate that the v ehicle could be de viating f arther from trim than in the pre vious maneuv er which w ould result in greater nonlinearities such as increased ya w motion. Doublet commands such as sho wn in Figure 5–8 are used in order to actuate the morphing serv o. This maneuv er is done without an y input from the rudder in order to 

PAGE 34

27 consider strictly commands to the morphing actuators. The amount of deection of the morphing; ho we v er is dif cult to interpret because it is a deection of the material and it is not e xpressed in a physical dimension such as de grees. 0 0.5 1 1.5 2 8 6 4 2 0 2 4 6 8 Time(sec)Morphing Command Figure 5–8: Doublet Command to Morphing Serv o 0 0.5 1 1.5 2 2.5 200 150 100 50 0 50 100 150 200 Time(sec)roll rate (deg/s) 0 0.5 1 1.5 2 2.5 200 150 100 50 0 50 100 150 200 Time(sec)yaw rate (deg/s) Figure 5–9: Response to Morphing Doublet for Roll(left) and Y a w Rate(right) The roll rate and ya w rate in Figure 5–9 are measured in response to the doublet commanded to the morphing serv o. These measurements indicate the roll rate is considerably higher than the ya w rate. Thus, the morphing is clearly an attracti v e approach for roll control because of the nearly-pure roll motion measured in response to the morphing commands. A separate morphing doublet is commanded at a dif ferent time as sho wn in Figure 5–10 in order to consider the modeling for a dif ferent maneuv er Similarly this maneuv er consists of strictly morphing actuation and no rudder input. 

PAGE 35

28 0 0.5 1 1.5 2 2.5 8 6 4 2 0 2 4 6 8 Time(sec)Morphing Command Figure 5–10: Second Doublet Command to Morphing Serv o The roll rate and ya w rate responses to the second morphing doublet are sho wn in Figure 5–11 It can be seen that the morphing doublet commanded in the second maneuv er w as of a slightly lar ger magnitude than the rst maneuv er This results in a lar ger roll rate response due to a lar ger morphing deection. 0 0.5 1 1.5 2 2.5 200 150 100 50 0 50 100 150 200 Time(sec)Roll Rate (deg/sec) 0 0.5 1 1.5 2 200 150 100 50 0 50 100 150 200 Time(sec)Yaw Rate (deg/sec) Figure 5–11: Response to Second Morphing Doublet Command for Roll Rate(left) and Y a w Rate(right) It is also seen that there is a minimal ya w rate response from the actuation of a lar ger morphing deection. Therefore, it similarly resulted in an almost pure roll motion with a slightly f aster roll rate response than with the pre vious maneuv er 5.4 Modeling The data from open-loop ights is then used to approximate a linear time-domain model using an ARX approximation [ 18 ]. This model is generated by computing optimal coef cients to match properties observ ed in the data. 

PAGE 36

29 The maneuv ers of interest are doublets ranging in magnitude and centered around a trim condition. Therefore, the assumption of linearity is reasonable since the maneuv ers are about trim. F or the approximation, the rates which are considered are roll and ya w rate because the y are of most interest for maneuv ers such as doublets. The accelerometers were considered b ut the data w as v ery noisy therefore, not allo wing for accurate approximations to be made. The simulated and measured v alues of roll rate and ya w rate are sho wn in Figure 5–12 0 0.5 1 1.5 2 2.5 200 150 100 50 0 50 100 150 200 Time(sec)Roll Rate (deg/sec) 0 0.5 1 1.5 2 2.5 200 150 100 50 0 50 100 150 200 Time(sec)Yaw Rate (deg/sec) Figure 5–12: Simulated ( ) and Actual (—) Roll Rate(left) and Y a w Rate(right) Responses to Morphing Doublet The simulated responses sho w good correlation with the actual data. The model is thus considered a reasonable representation of the aircraft dynamics as it is e xcited by the doublet. The e xistence of such a model is important for future design of autopilot controllers b ut it is also v aluable for interpreting the morphing. When using the ARX simulation in Matlab, a linear approximation could not be made on the maneuv ers which were not centered around trim. Therefore, not only are maneuv ers around trim desired for a linear approximation, b ut the y are also necessary in order for the simulation to be done. The model which will be used w as chosen because it produced the closest match to the maneuv er as compared to four other doublets. The maneuv ers compared were from the same data set, b ut the one that w as chosen resulted in a closer match due to 

PAGE 37

30 the aircraft being closer to trim. The resulting model consists of six states and poles as sho wn in T able 5–3 The roll mode is clearly sho wn and the dutch roll mode indicates the slight oscillations which are present in the combined roll and ya w motion as sho wn in Figure 5–12 T able 5–3: Poles of a linear model of the 24 in MA V Poles V alue Dutch Roll3751384 Roll -4.03 In order to study whether this model is a good enough linear approximation of the dynamics of the 24 in MA V dif ferent inputs are considered for the same model. These inputs are doublet morphing commands at dif ferent times throughout the same set of data. The simulated and measured v alues of roll rate and ya w rate are sho wn in Figure 5–13 for se v eral inputs. It is clearly sho wn that the simulated and actual roll rate and ya w rate responses demonstrate good correlation. The simulations sho wn in Figure 5–13 sho w that the model which w as obtained from a doublet maneuv er responds well to dif ferent inputs. The rst input w as a small morphing doublet commanded from a separate data set. The follo wing tw o inputs were from a medium and lar ge morphing doublet, respecti v ely from the data set that w as used for obtaining the model. 

PAGE 38

31 0 0.5 1 1.5 2 2.5 3 3.5 4 200 150 100 50 0 50 100 150 200 Time(sec)Roll Rate (deg/sec) 0 0.5 1 1.5 2 2.5 3 3.5 4 200 150 100 50 0 50 100 150 200 Time(sec)Yaw Rate (deg/sec) 0 0.5 1 1.5 2 2.5 3 3.5 4 200 150 100 50 0 50 100 150 200 Time(sec)Roll Rate (deg/sec) 0 0.5 1 1.5 2 2.5 3 3.5 4 200 150 100 50 0 50 100 150 200 Time(sec)Yaw Rate (deg/sec) 0 0.5 1 1.5 2 2.5 200 150 100 50 0 50 100 150 200 Time(sec)Roll Rate (deg/sec) 0 0.5 1 1.5 2 2.5 200 150 100 50 0 50 100 150 200 Time(sec)Yaw Rate (deg/sec) Figure 5–13: Simulated ( ) and Actual (—) Roll Rate(left) and Y a w Rate(right) Responses to Morphing Doublets 5.5 Ev aluation If this MA V is to be used in surv eillance missions, the use of only rudder and ele v ator could pro vide suf cient control during turns. Ho we v er the resulting dutch roll motion creates dif culties if this MA V is to be used for a more demanding mission and also requires more pilot control. Therefore, this v ehicle requires further control in order to be used in situations such as in urban en vironments where more controlled turns are required. 

PAGE 39

32 Using wing twisting as an additional control ef fector for roll impro v es the performance and controllability of the aircraft. Using wing twist additionally as compared to the traditional rudder and ele v ator results in impro v ed ight path tracking, especially when considering gusty weather en vironments. The ight characteristics of the 24 in v ehicle are actually quite impressi v e to vie w The measurements of roll rate and ya w rate indicate the mathematical nature of the characteristics; ho we v er a qualitati v e e v aluation is also useful. Such an e v aluation is best achie v ed in association with step commands gi v en to each serv o. The step to the rudder causes the airplane to roll b ut the coupled ya w results in a ight path similar to a corkscre w spiral. Con v ersely the step to the morphing causes the airplane to roll with a minimum of ya w so the ight path is nearly a straight line. In other w ords, the morphing induces almost pure roll and allo ws much more accurate tracking of desired ight paths. Also, the morphing results in considerably higher roll rates than the rudder This result is quite interesting gi v en that the rudder deection is quite lar ge b ut the morphing, as sho wn in Figure 5–10 is not o v erly commanded to deect. Thus, a small amount of morphing is suf cient to cause a dramatic response from the aircraft. 

PAGE 40

CHAPTER 6 12 in MICR O AIR VEHICLE 6.1 V ehicle Description A dif ferent micro air v ehicle is also considered. This v ehicle has a wingspan of 12 in and is sho wn in Figure 6–1 Figure 6–1: V ie w of the 12 in MA V The basic properties of this v ehicle are gi v en in T able 6–1 T able 6–1: Properties of the 12 in MA V Property V alue W ingspan 12” W ing Area 44 in 2 W ing Loading 14.19 ozf t 2 Aspect Ratio 3.27 Po werplant Electric motor w/ 2.25” propeller T otal W eight 123g The 12 in MA V is designed to ha v e an ele v ator as its control ef fector The ele v ator is actuated by using a single serv o. Therefore, this v ehicle does not include a rudder and an additional control ef fector will be added. 33 

PAGE 41

34 This v ehicle is constructed using the similar designs which are used at the Uni v er sity of Florida which consist of carbon ber airframes and e xible membrane wings. The composite wing on the 12 in MA V sk eleton is co v ered with an e xtensible membrane skin of late x rubber The late x material used in the 12 in MA V is considerably more e xible than the mylar sheeting which is used in the 24 in MA V 6.2 Morphing The 12 in MA V is designed with morphing as an additional control ef fector The morphing is implemented by actuating a single serv o which is connected to each wing. The use of a more e xible material for the wing surf ace of this v ehicle w as chosen on purpose in order to consider more dramatic shape changes of the wings. This is also done on a smaller airframe without a rudder to consider strictly the ef fects of morphing. Figure 6–2: V ie w of the 10 in MA V The initial implementation of this morphing strate gy w as originally attempted on a MA V with 10 in wingspan as sho wn in Figure 6–2 The 10 in MA V is designed with a similar carbon ber airframe and e xible membrane wings. The f act that this aircraft is smaller than the 12 in MA V allo ws for a smaller wing area and shorter more closely aligned wing battons. This v ehicle, lik e the current v ehicle, has late x co v ering on the wings so w as v ery easy to morph in ight. The 10 in MA V w as chosen for the initial 

PAGE 42

35 study because it w as already constructed and could be readily adapted for the ne w study The ights of that v ehicle were quite promising and clearly indicated that the morphing pro vided an ef fecti v e form of control authority This MA V responded well to the wing morphing commands, ho we v er is restricted in the amount of ight testing which can currently be done due to its payload limitations. This led to the design of the 12 in MA V with the requirement of a lar ger aircraft in order to carry the required instrumentation to perform open-loop and closed-loop ight testing. The design of the 12 in MA V initially follo wed closely the design of the 10 in MA V with a similar airframe and wing design. The open-loop ights are similarly performed with a data acquisition board and the closed-loop ights will be done with a memory board. The 12 in v ehicle w as essentially a scaled v ersion of the original v ehicle e xcept for the wing construction. The original v ersion had a single structure for the wings that mounted atop the fuselage. The ne w v ersion had separate wings that attached to posts on each side of the fuselage. This separation of the wings allo wed for more e xibility due to the remo v al of the carbon ber structure. The leading edge of the 12 in MA V w as initially b uilt with a single layer of carbon ber This pro v ed to be f aulty during the rst attempts at ight when the leading edge w ould fold o v er when wing loads were applied. Therefore another layer of carbon ber w as applied to the wing making it stif fer and better capable of withstanding the wing loads. Another issue with the wing design w as using the same e xible late x material on this 12 in as on the 10 in MA V with the lar ger airframe requiring a lar ger wing area. The wing battons are not as closely aligned on the lar ger frame and the lar ger sheet of late x is weak er with lar ger loads. Therefore not allo wing for wing morphing as dramatic on the 12 in as on the 10 in MA V 

PAGE 43

36 Further ight testing of the 12 in MA V indicated a problem with the thread connection. The morphing of the 10 in MA V used only a single thread attached to the outboard of the trailing edge and this style w as used for the 12 in MA V Unfortunately the battens on the lar ger MA V were spaced f arther apart than on the smaller MA V so the wing w as weak er The leading edge on the 12 in MA V w ould no w remain properly shaped b ut the trailing edge w ould collapse when loaded. This problem w as addressed by attaching a second thread to the trailing edge of the wing and allo wing the morphing actuation to pro vide strength to support the loads. The 12 in MA V is designed to allo w for a more complicated type of morphing than is used for the 24 in MA V The wings of this smaller v ehicle are constructed from late x sheeting whereas the wings of the lar ger v ehicle are made of mylar sheeting. Consequently the wings of the 12 in v ehicle are considerably more e xible, and thus easier to morph, than the wings of the 24 in v ehicle. This e xibility allo ws simple mechanisms to again be appropriate for generating morphing and allo w control issues to be in v estigated. The high e xibility of the wings for this MA V allo w consideration of morphing be yond basic w arping. More specically this v ehicle is used to consider morphing that af fects the twist and span of the wings. A torque rod, as used for the 24 in MA V w ould clearly not be appropriate for such a morphing. Instead, the rod w as replaced with threads. The morphing strate gy for this MA V is sho wn in Figure 6–3 K e vlar threads are strung between a serv o in the fuselage and points near the outboard of the wings. These threads are incredibly strong and the minor stress recei v ed during ight is not suf cient to cause an y stretching. The morphing achie v ed by this strate gy is directly dependent upon the attachment points of the threads. The attachment of the threads to the fuselage is near the leading edge of the wings. The corresponding attachment to the wings is actually at separate 

PAGE 44

37 Figure 6–3: W ing with K e vlar Threads points. One attachment point is near the mid-chord point at the wing-tip outboard. Another attachment point is the trailing edge near the tw o-thirds span location. The morphing that results by actuating the serv o is sho wn in Figure 6–4 The serv o rotates and causes the threads to pull against the attachments on the wing. The morphing resulting from this strate gy is clearly be yond simple w arping. In this case, the pulling of the threads to w ard the leading-edge attachment at the fuselage causes the wing to both twist and bend. The ef fect is similar in nature to a curling of the wings. The basic parameters that are readily observ ed to change are the twist, camber chord, and span. Figure 6–4: Front V ie w of the 12 in MA V with Undeected (left) and Morphed (right) W ing The morphing is designed for a biologically-inspired ef fect. The displacement of the wing resembles shapes observ ed in birds lik e gulls. F or instance, the bending along the span is concentrated around a single point which correlates to the elbo w in 

PAGE 45

38 birds. The twist is concentrated near the trailing-edge outboard which correlates with the feathers near the wrist of birds. A more formal approach to this concept is being designed by N ASA b ut this current v ehicle is suf cient to in v estigate control issues [ 6 ]. Only a single wing is altered in Figure 6–4 The v ehicle actually contains separate serv os for each wing that allo w the morphing to act simultaneously on both wings; ho we v er this thesis will restrict attention to morphing a single wing. The current objecti v e considers roll control b ut the longitudinal issues will be in v estigated in the future. Also, this v ehicle is ideal for the focus of this thesis. Specically the morphing strate gy is quite simple b ut the morphing ef fect is complicated. This approach allo ws the control issues associated with morphing to be easily studied. The v ehicle is not designed to study the optimal strate gies for morphing; rather the v ehicle is designed to study the optimal strate gies for control. 6.3 Flight T esting Flight testing is also done on the 12 in MA V in an open area for R/C airplanes. The ight tests for this MA V are performed in similar conditions as the tests for the 24 in MA V This MA V is equiped with a data acquisition board which be gins logging when the motor is turned on. This MA V is then similarly hand launched into the incoming wind for tak eof f. The primary forms of control for this MA V are the ele v ator and wing morphing. The airplane is controlled with these surf aces for tak eof f, turns, climbs, and le v el ight. The airplane is then trimmed for straight and le v el ight. Achie ving trimmed ight is necessary as a neutral reference point for the control surf aces and in performing dif ferent ight test maneuv ers. This MA V is then tested by commanding doublets to the morphing serv os. A representati v e doublet command is sho wn in Figure 6–5 The units of this command 

PAGE 46

39 are just count commands to the serv o because the actual deection caused by morphing is dif cult to quantify 0 0.5 1 1.5 2 2.5 30 20 10 0 10 20 30 40 Time(sec)Morphing Command Figure 6–5: Doublet Command to Morphing Serv o The responses to the morphing doublets are measured by the on-board data acquisition system. The roll rate and ya w rate are presented in Figure 6–6 0 0.5 1 1.5 2 2.5 60 40 20 0 20 40 60 80 Time(sec)Roll Rate (deg/sec) 0 0.5 1 1.5 2 2.5 60 40 20 0 20 40 60 80 Time(sec)Yaw Rate (deg/sec) Figure 6–6: Roll Rate(left) and Y a w Rate(right) in Response to Morphing Doublet The roll rate is clearly correlating well with the commanded doublet and demonstrates the morphing is capable of commanding roll maneuv ers. The ya w rate is some what more dif cult to understand. Notably the aircraft b uilds up ya w rate approximately 0.5 seconds after the onset of the doublet command. This ight characteristic results from the single-sided nature of the morphing. Essentially the wing that is morphed loses lift b ut also increases drag. The loss of lift immediately causes rolling and the increase of drag causes a slight delay in b uilding up the ya w rate. 

PAGE 47

40 A separate morphing doublet is commanded at a dif ferent time as sho wn in Figure 6–7 in order to consider the modeling for a dif ferent maneuv er This maneuv er consists of strictly morphing actuation. 0 0.5 1 1.5 2 2.5 3 3.5 30 20 10 0 10 20 30 40 Time(sec)Morphing Command Figure 6–7: Second Doublet Command to Morphing Serv o The roll rate and ya w rate responses to the second morphing doublet are sho wn in Figure 6–8 It can be seen that the morphing doublet commanded in the second maneuv er w as of a slightly smaller magnitude than the rst maneuv er This results in a smaller roll rate response due to a smaller morphing deection. 0 0.5 1 1.5 2 2.5 3 3.5 20 0 20 40 60 80 Time(sec)Roll Rate (deg/sec) 0 0.5 1 1.5 2 2.5 3 3.5 20 0 20 40 60 80 Time(sec)Yaw Rate (deg/sec) Figure 6–8: Response to Second Morphing Doublet Command for Roll Rate(left) and Y a w Rate(right) 6.4 Modeling A linear model is identied from the ight data. A 6-state model w as originally identied b ut reduced to a 3-state model as sho wn in T able 6–2 The simulated responses of this model are compared with measured v alues of roll rate and ya w rate in Figure 6–9 

PAGE 48

41 0 0.5 1 1.5 20 0 20 40 60 80 Time(sec)Roll Rate (deg/sec) 0 0.5 1 1.5 20 0 20 40 60 80 Time(sec)Yaw Rate (deg/sec) Figure 6–9: Simulated( ) and Actual(-) Roll Rate(left) and Y a w Rate(right) Responses to a Doublet T able 6–2: Poles of a linear model of the 12 in MA V Poles V alue Dutch Roll3670414732 Roll -7.521 The responses of the model are reasonably close to the measured responses. The roll rate sho ws a good correlation although the ya w rate is some what less accurate. The model contains a roll con v er gence mode which, based on the accurac y of roll simulations, is accepted. The model also contains a dutch roll mode which attempts to capture the dynamics associated with ya w rate. The inability of this mode to represent the ya w dynamics may indicate some nonlinearity is associated with the v ehicle. Such nonlinear dynamics w ould not be une xpected gi v en the e xtreme nature of the morphing and the asymmetry resulting from morphing a single wing. In order to study whether this model is a good enough linear approximation of the dynamics of the 12 in MA V dif ferent inputs are considered for the same model. These inputs are doublet morphing commands at dif ferent times throughout the same set of data. The simulated and measured v alues of roll rate and ya w rate are sho wn in Figure 6–10 for se v eral inputs. It is clearly sho wn that the simulated and actual roll rate and ya w rate responses demonstrate good correlation. The simulations sho wn in Figure 6–10 sho w that the model which w as obtained from a doublet maneuv er responds well to dif ferent 

PAGE 49

42 inputs. The ya w rate; ho we v er is dif cult to model because the v ehicle is morphed asymmetrically and this introduces nonlinearities. These nonlinearities ha v e to be considered in an approximation in order to obtain an accurate model. 0 0.5 1 1.5 2 2.5 3 3.5 4 100 50 0 50 100 Time(sec)Roll Rate (deg/sec) 0 0.5 1 1.5 2 2.5 3 3.5 4 80 60 40 20 0 20 40 60 80 Time(sec)Yaw Rate (deg/sec) 0 0.5 1 1.5 2 60 40 20 0 20 40 60 80 Time(sec)Roll Rate (deg/sec) 0 0.5 1 1.5 2 10 5 0 5 10 15 20 25 30 35 Time(sec)Yaw Rate (deg/sec) Figure 6–10: Simulated ( ) and Actual (—) Roll Rate(left) and Y a w Rate(right) Responses to Morphing Doublets 

PAGE 50

43 6.5 Ev aluation When using the ARX simulation in Matlab, a linear approximation could not be made on the maneuv ers which were not centered around trim. Therefore, not only are maneuv ers around trim desired for a linear approximation, b ut the y are also necessary in order for the simulation to be done. It is particularly dif cult to nd maneuv ers to use for a linear approximation for this specic v ehicle due to the f act that it is v ery hard to trim. Therefore, before most maneuv ers there is an indication in the data that there are some oscillations about the roll axis. Similarly the data indicates that se v eral seconds after a maneuv er has been initiated, a component of ya w rate be gins to de v elop. The b uild up of ya w rate after a morphing maneuv er has been initiated can be attrib uted to the f act that the shape of the wing is being altered. The wings on this v ehicle are being curled underneath, one at a time resulting in an asymmetric morphing actuation. This leads to nonlinearities in the ight characteristics of the v ehicle which can e xplain the oscillations and the b uild up of ya w seconds after a maneuv er has been initiated. The modeling is a linear approximation of coef cients, and when considering nonlinear beha vior it might not pro vide data that can be trusted. The 24 in MA V is morphed by twisting the wings symmetrically which can be approximated as a linear beha vior Ho we v er the 12 in MA V is being morphed asymmetrically and therefore requires further studies to obtain a better approximation of the model. 

PAGE 51

CHAPTER 7 CONCLUSION Current micro air v ehicles are designed with the common features of carbon ber airframes and e xible membrane wings. The typical control surf aces which can be implemented on MA Vs consist of rudder and ele v ator The con v entional control surf aces such as ailerons can not be easily included in the design of the common MA Vs due to the f act that the wings are made of e xible material. A dif ferent form of actuation is necessary in order to impro v e maneuv erability and performance for MA Vs if the y will be assigned to e v olving missions. A particularly demanding mission is one which tak es place in an urban en vironment which requires these v ehicles to ha v e adv anced maneuv ering capabilities. A simple approach to increasing maneuv erability is to include an additional control ef fector such as wing morphing. This paper has demonstrated that morphing can be an ef fecti v e means to achie v e roll control for a micro air v ehicle. The e xible nature of the wings enables their shapes to be easily altered. Simple mechanisms, such as a torque rod and K e vlar threads, are used on a 24 in MA V and a 12 in MA V In each case, the v ehicle w as o wn using morphing as the primary ef fector for roll maneuv ers. The ight data clearly sho ws the morphing produces signicant roll rates and pro vides signicant controllability These v ehicles are o wn in an R/C eld and are equipped with a data acquisition system(D AS). This D AS consists of gyros and accelerometers for the three axis. The data is then retrie v ed and the accelerations and rates of the three axis can be studied. This data is then used for a linear modeling technique. 44 

PAGE 52

45 A linear approximation is then considered using an ARX modeling technique in Matlab This sho ws a good correlation with the data for the 24 in and the 12 in MA V Ho we v er further studies ha v e to be considered for the modeling of the 12 in MA V since it is being morphed asymmetrically by curling one wing at a time. This then introduces nonlinearities in the ight characteristics which ha v e to be considered in modeling approximations. The 24 in MA V can be approximated linearly since the morphing is being induced with symmetric wing twisting. 

PAGE 53

REFERENCES [1] M. Abdulrahim,H. Garcia,J. Dupuis, and R. Lind,“Flight Characteristics of W ing Shaping for a Micro Air V ehicle with Membrane W ings, ” International F orum on Aer oelasticity and Structur al Dynamics IF ASD-US-24, June 2003. [2] M. Amprikidis and J.E. Cooper ,“De v elopment of Smart Spars for Acti v e Aeroelastic Structures, ” AIAA-2003-1799, 2003. [3] J. Bo wman,“ Af fordability Comparison of Current and Adapti v e and Multifunctional Air V ehicle Systems, ” AIAA-2003-1713, 2003. [4] J. Blondeau,J. Richeson and D.J. Pines,“Design, De v elopment and T esting Of A Morphing Aspect Ratio W ing Using An Inatable T elescopic Spar ” AIAA-20031718, 2003. [5] C.E.S. Cesnik and E.L. Bro wn, “ Acti v e W arping Control of A Joined-W ing Airplane Conguration, ” AIAA-2003-1715, 2003. [6] J.B. Da vidson,P Chw alo wski and B.S. Lazos,“Flight Dynamic Simulation Assessment of a Morphable Hyper -Elliptic Cambered Span W inged Conguration, ” AIAA-2003-5301, 2002. [7] P de Marmier and N.M. W erele y “Morphing W ings of a Small Scale U A V Using Inatable Actuators for Sweep Control, ” AIAA-2003-1802, 2003. [8] Ecobirds, ”Putting Birds Back Into Ecology ” http://birds.ecoport.or g, 10/15/2003. [9] S.M. Ettinger M.C. Nechyba, P .G. Ifju and M. W aszak,“V ision-Guided Flight Stability and Control for Micro Air V ehicles, ” Pr oceedings of the IEEE Confer ence on Intellig ent Robots and Systems 2002, pp. 2134-2140. [10] S.W Gano,J.E. Renaud,S.M. Batill and A. T o v ar ,“Shape Optimization for Conforming Airfoils, ” AIAA-2003-1579, 2003. [11] H. Garcia,M. Abdulrahim and R. Lind,“Roll Control for a Micro Air V ehicle using Acti v e W ing Morphing, ” AIAA Guidance Navigation and Contr ol Confer ence AIAA-2003-5347, 2003. [12] F .H. Gern, D.J. Inman and R.K. Kapania, ”Structural and Aeroelastic Modeling of General Planform W ings with Morphing Airfoils, ” AIAA J ournal V ol 40,No. 4,2002, pp. 628-637. 46 

PAGE 54

47 [13] J.M. Grasme yer and M.T K eennon,“De v elopment of the Black W ido w Micro Air V ehicle, ” AIAA-2001-0127, 2001. [14] P .G. Ifju,D.A. Jenkins,S. Ettinger ,Y Lian and W Shyy ,“Fle xible-W ing Based Micro Air V ehicles, ” AIAA-2002-0705, 2002. [15] C.O. Johnston,D.A. Neal,L.D. W iggins,H.H. Robertsha w W .H. Mason and D.J. Inman,“ A Model T o Compare The Flight Control Ener gy Requirements Of Morphing And Con v entionally Actuated W ings, ” AIAA-2003-1716, 2003. [16] N.S. Khot, F .E. Eastep, and R.M. K olonay “Method for Enhancement of the Rolling Maneuv er of a Fle xible W ing, ” J ournal of Air cr aft V ol. 34, No. 5, 1997, pp. 673-678. [17] S.K. Kw ak and R.K. Y eda v alli, ”Ne w Modeling and Control Design T echniques for Smart Deformable Aircraft Structures”, J ournal of Guidance Contr ol and Dynamics V ol 24, No. 4, 2001, pp. 805-815. [18] L. Ljung, User s Manual for System Identication T oolbox The Math W orks, Inc, Natick, MA, 1991. [19] L. Ljung, System Identication: Theory for the User Prentice Hall, Engle w ood Clif fs, NJ,1987. [20] S.L. P adula,J.L. Rogers and D.L. Rane y ,“Multidisciplinary T echniques And No v el Aircraft Control Systems, ” AIAA-2000-4848, 2000. [21] B. Sanders, F .E. Eastep and E. F orster ”Aerodynamic and Aeroelastic Character istics of W ings with Conformal Control Surf aces for Morphing Aircraft”, J ournal of Air cr aft V ol 40, No. 1, 2003, pp. 94-99. [22] W Shyy M. Ber g and D. Ljungqvist, “Flapping and Fle xible W ings F or Biological And Micro Air V ehicles, ” Pr o gr ess in Aer ospace Sciences V ol. 35, No. 5, 1999, pp. 455-506. [23] M.R. W aszak,J.B. Da vidson and P .G. Ifju, “Simulation and Flight Control of an Aeroelastic Fix ed W ing Micro Aerial V ehicle, ” AIAA-2002-4875, 2002. [24] M.R. W aszak,L.N. Jenkins and P Ifju, “Stability and Control Properties of an Aeroelastic Fix ed W ing Micro Aerial V ehicle, ” AIAA-2001-4005, 2001. [25] R.W Wlezien,G.C. Horner ,A.R. McGo w an,S.L. P adula,M.A. Scott, R.J. Silcox and J.O. Simpson,“The Aircraft Morphing Program, ” AIAA-98-1927, 1998. 

PAGE 55

BIOGRAPHICAL SKETCH Helen Garcia w as born in Santo Domingo, Dominican Republic, on May 22, 1979. Her f amily mo v ed to Miami, FL, in 1990. She recei v ed her high school diploma from Miami Coral P ark Senior High School in Miami, FL. She then attended the Uni v ersity of Florida and recei v ed her bachelor' s de gree in aerospace engineering in May 2002. She has w ork ed in the dynamics and control research group under Dr Rick Lind and will recei v e her master of science de gree in aerospace engineering in December 2003. 48