See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/277326537

Mathematical modeling and control of a tilt-

rotor aircraft

Article in Aerospace Science and Technology · December 2015

DOI: 10.1016/j.ast.2015.10.012

CITATIONS READS

3 269

2 authors:

Xinhua Wang Lilong Cai

65 PUBLICATIONS 647 CITATIONS The Hong Kong University of Science and Tec…
100 PUBLICATIONS 1,138 CITATIONS
SEE PROFILE

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Design and test for a prototype of agile tail-sitter aircraft View project

Research on signal estimation of aircraft systems View project

All content following this page was uploaded by Xinhua Wang on 09 October 2017.

The user has requested enhancement of the downloaded file.

 Aerospace Science and Technology 47 (2015) 473–492

Contents lists available at ScienceDirect

Aerospace Science and Technology

www.elsevier.com/locate/aescte

Mathematical modeling and control of a tilt-rotor aircraft

Xinhua Wang a,∗ , Lilong Cai b
a
Department of Mechanical and Aerospace Engineering, Monash University, Melbourne, VIC 3800, Australia
b
Department of Mechanical and Aerospace Engineering, Hong Kong University of Science and Technology, Hong Kong, China

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents a novel model of large-size tilt-rotor aircraft, which can operate as a helicopter as
Received 3 September 2014 well as being capable of transition to fixed-wing flight. Aerodynamics of the dynamic large-size tilt-
Received in revised form 19 May 2015 rotors based on blade element method is analyzed during mode transition. For the large-size aircraft
Accepted 19 October 2015
with turboshaft engines, the blade pitch angles of the rotors are regulated to vary according to the
Available online 23 October 2015
desired level of thrust, and the following expressions are formulated explicitly: rotor thrust and blade
Keywords: pitch angle, drag torque and blade pitch angle. A finite-time convergent observer based on Lyapunov
Tilt-rotor aircraft function is developed to reconstruct the unknown variables and uncertainties during mode transitions.
Rotor thrust The merits of this design include the modeling of dynamic large-size tilt-rotor, ease of the uncertainties
Blade pitch angle estimation during the tilting and the widely applications. Moreover, a switched logic controller based on
Uncertainties estimation the finite-time convergent observer is proposed to drive the aircraft to implement the mode transition
Mode transition with invariant flying height.
© 2015 Elsevier Masson SAS. All rights reserved.

1. Introduction
Helicopters and fixed-wing airplanes have their advantages and
shortcomings. Helicopters can take off and land vertically, but they
cannot fly forward in high speed, and their payloads are very lim-
ited comparing to fixed wing plane with the same gross weight
[1–4]. On the other hand, conventional fixed-wing aircrafts can fly
forward in high speed and large payloads. However, they cannot
take off and land vertically, and the appropriate runways are re-
quired.
There are several ways to perform the vertical takeoff and land-
ing (VTOL) maneuver such as tilting rotor, tilting wing, thrust vec-
toring, tail sitter, tilting fuselage, flapping wing, multi-propeller
multifunction, etc. There are some types of thrust vectoring air-
crafts, such as manned aircrafts AV-8B Harrier [5] and F-35 [6].
These aircrafts are designed for actual combat. The reason why
these aircrafts can perform vertical takeoff and landing is due to
their jet engines with tilt jettubes. Although they are powerful, the
jet gas is very hot and harmful, and they can easily destroy the
ground environment or inflict injuries to people nearby. These air-
crafts are not suitable for many civil and rescue operations. More-
over, such VTOL aircraft with jet engines is less efficient in hover
than a conventional helicopter or a tilt-rotor aircraft of the same
gross weight [7].
* Corresponding author.
E-mail address: wangxinhua04@gmail.com (X. Wang).
One of the classical tilt-rotor aircrafts is V-22 [8]. Similarly the
Bell Eagle Eye UAV [9] which is also based on tilt-rotor technology
has a large success in the civil and military domains. However, this
design brings its own problems, since the degradation in stability
is usually observed in high speed forward flight (airplane mode)
[10]. Moreover, the involved equations of motion are highly cou-
pled and nonlinear. The fundamental characteristics of V-22, which
includes a very high roll moment of inertia, result in a tendency to
pilot-induced oscillation. Extreme attention must be paid to pre-
venting over control in the roll channel. Moreover, the roll channel
is easily affected by pitch dynamics. V-22 is a large “U”-shaped
structure with two very large masses at the ends (the engines). The
result is the need for long and flexible runs of hydraulic, electrical,
and mechanical lines that are more susceptible to tensile loads,
bending loads, abrasions, and deformations than conventional he-
licopter designs. Because of the high disk-loading (129.63 kg/m2 ),
the downwash velocity of V-22 is about twice that of any con-
ventional helicopter. And because of the side-by-side placement of
the prop-rotors, there are two distinct downwash wakes that are
transverse to the flight direction. This has several operational im-
plications that bear on safety issues.
Tail-sitters, as the name implies, sit on their tail when not in
flight [11–14]. They take off and land vertically, making them a
member of the VTOL family of aircraft. Equipped with a power-
ful engine, tail-sitters can utilize a “prop-hanging” technique to
hover in place. Additionally, they can make transition to a level
flight mode and fly in a conventional fixed-wing mode, which is
more energy efficient than the hover mode. These aircrafts have

http://dx.doi.org/10.1016/j.ast.2015.10.012
1270-9638/© 2015 Elsevier Masson SAS. All rights reserved.
474 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492

Nomenclature

me weight of each engine dT i incremental lift of rotor

Te rotor thrust under rated power dQ i incremental drag torque of rotor
ρe thrust-to-weight ratio of engine Qi drag torque of rotor
m gross weight of aircraft C Qi drag torque coefficient of rotor
Ti rotor thrust T sum of rotor thrusts
C Ti rotor thrust coefficient V βt airflow generated by rotor tilting
Ωi rotor speed V rt resultant velocity at the free wing
ρ air density αf angle of attack for free wing
σ rotor solidity df distance between rotor center to leading edge of free
p blade number wing
R rotor radius Li (i = 1, 2, 3, 4) lift force generated by free wing
ρb blade span-chord ratio Di (i = 1, 2, 3, 4) lift force generated by free wing
A area of rotor disk L i (i = 1, 2, 3, 4) uncertain lift force for free wing
g gravity acceleration Di (i = 1, 2, 3, 4) uncertain drag force for free wing
C d0 constant profile drag coefficient S fi area of free wing
κ induced power factor Cf lift coefficient brought by angle of attack α f to free
b chord length of blade wing
PL power loading of rotor CDf 0 drag coefficient when angle of attack α f is equal to
DL disk loading of rotor zero
R 2r rotor radius of aircraft with two rotors δi (i = 1, 2, 3, 4) flap bias angle of free wing
Rε Reynolds number Cr lift coefficient brought out by flap bias angle δi
M Mach number ef value of the Oswald’s Efficiency Factor for free wing
DLQT Desired disk loading of rotor Af span-chord ratio of free wing
Aw span-chord ratio of fixed wing αmax angle of attack with respect to maximal lift coefficient
WL w desired wing loading of fixed wing of free wing
Sw wing area of fixed wing Li (i = 5, 6) lift force generated by fixed wing
Lw wing span of fixed wing Di (i = 5, 6) lift force generated by fixed wing
Cw chord length of fixed wing L i (i = 5, 6) uncertain lift force for fixed wing
S front area of front fixed wing root Di (i = 5, 6) uncertain drag force for fixed wing
Sfi area of free wing α angle of attack for fixed wing
Γg right handed inertial frame S ri area of the left (right) fixed wing
Γb frame attached to the aircraft’s fuselage C wα lift coefficient brought by angle of attack α to fixed
Γβ right handed inertial frame of the tiltrotor wing
ψ yaw angle C w0 lift coefficient when angle of attack α is equal to zero
θ pitch angle C D w0 drag coefficient when angle of attack α is equal to
φ roll angle zero
R bg transformation matrix for Γb to Γ g δi (i = 5, 6) flap bias angle of fixed wing
R βb transformation matrix for Γβ to Γb C wδ lift coefficient brought out by flap bias angle δi
β
VβX air relative velocity on axis E x ew value of the Oswald’s Efficiency Factor for fixed wing
V βY
β
air relative velocity on axis E y f rl yaw force generated by vertical tail
β f rd drag force generated by vertical tail
VβZ air relative velocity on axis E z
Gβ gyroscopic moment for tilting rotors in frame Γβ
Vb air relative velocity with respect to frame Γb
Gb gyroscopic moment for tilting rotors in frame Γb
vi induced velocity of rotor disk
τT thrust vectoring moment in frame Γb
vh induced velocity in hover
Ci (i = 1, 2, 3, 4) center of rotor
r length from rotor center to blade element
Q sum of drag torques of rotors in frame Γβ
β∗ relative inflow angle at blade element
Qr sum of drag torques of rotors in frame Γb
α∗ angle of attack of blade element
τβ tilting pitch moment in frame Γb
ϕ∗ blade pitch angle
τδ moment generated by lift of fixed wings in frame Γb
ϕ7i pitch angle at r / R = 0.7
τf moment generated by free wings in frame Γb
ϕ∗ linear torsion of blade
τfd moment generated by drag of fixed wings in frame Γb
U resultant velocity at disk
τxd moment generated by drag of free wings in frame Γb
dL lift per unit span on blade element β
J1 moment of inertia with respect to axis E x
dD drag per unit span on blade element β
Cl lift coefficient of blade J2 moment of inertia with respect to axis E y
β
Cd drag coefficient of blade J3 moment of inertia with respect to axis E z
dr incremental radial distance of blade J4 tilting moment of inertia
 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492
higher rotor disk loadings. A novel unmanned tail-sitter aircraft
called SkyTote has been designed [15–18]. It was originally con-
ceived as an airborne conveyor belt that would use a VTOL capa-
bility to minimize ground handling. The concept demonstrator is
a ‘tail-sitter’ configuration and utilizes coaxial, counter-rotating ro-
tors. A relatively large cruciform tail provides directional control in
the airplane modes as well as serving as landing gear in the he-
licopter mode. However, the flying height varies when the aircraft
undergoes a vertical to horizontal transition. The aircraft climbs
upward during mode transition. Importantly, a large thrust force
should be provided to finish the mode transition. Moreover, the
aerodynamics of the complex mechanical drive device, i.e., coaxial
counter-rotating propellers, is difficult to be analyzed during mode
transition.
A tilt-fuselage aircraft was presented in [19] to keep the flying
height invariant during mode transition. It is a rotor-fixed wing air-
craft with two free wings. During mode transition, the fuselage is
tilted, and the free wings are kept at a given angle of attack. How-
ever, it is difficult to analyze the aerodynamics of the tilt fuselage
during mode transition.
In [20], a quad tilt-wing unmanned aerial vehicles (UAV) were
presented, which is capable of vertical takeoff and landing (VTOL)
like a helicopter and long duration horizontal flight like an air-
plane. The aircraft is electric powered and small size. During mode
transition, all the wings and the quad rotors are tilted. In [21],
a multi-rotorcraft with morphing capabilities was presented. The
aircraft is able to rotate its body into a vertical flying mode config-
uration. The middle base, in which the inertial sensors are fixed, is
always horizontal during the morphing stage. However, the servo
motor is difficult to keep the middle base horizontal during mode
transition. Accordingly, the attitude measurement will be affected
adversely. The two aircrafts in [20,21] are all electric powered and
small size.
The electric powers are usually suitable to drive small-size air-
crafts. For an electric powered tilt-rotor aircraft, the rotor speeds
are allowed to freely vary according to the desired level of thrust.
However, for a large-size aircrafts with turboshaft engines, the ro-
tor speeds freely varying according to the desired level of thrust is
unrealistic for current state-of-the art technology for conventional
rotorcrafts, i.e., the control of the thrust output is impossible to be
achievable by changing the rotor speed. The aerodynamics of large-
size tilt-rotor is more complex than that of small-size tilt-rotor
during mode transition. Moreover, the tilt-rotor aircrafts are un-
deractuated mechanical systems, which exhibit high nonlinear and
time-varying behaviors. Meanwhile, the influences of aerodynamic
disturbance, unmodelled dynamics and parametric uncertainties
are not avoidable in modeling. These nonlinearities and uncertain-
ties render great challenges in the design of flight control system.
Nonlinear modeling and analysis for mechanical systems have
been presented in recent years [22]. To overcome these problems
above, in this paper, a novel model of large-size tilt-rotor aircraft
is presented, which can operate as a helicopter as well as be-
ing capable of transition to fixed-wing flight. The control of the
thrust outputs is achieved by changing the blade pitch angles of
the rotors, and the rotor speeds are fixed in rated power. The
aerodynamics of the large-size tilt-rotor based on blade element
method is analyzed during mode transition, and the following re-
lations are formulated explicitly: rotor thrust and blade pitch angle,
rotor drag torque and blade pitch angle. The designed tilt-rotor
aircraft is more suitable to large-size aircraft applications. Its struc-
ture is based on quad-rotor aircrafts design [23–25]. Quad-rotor
aircraft is a system consisting of four individual rotors attached to
a rigid cross frame. It is an omnidirectionally symmetrical VTOL
vehicle ideally suited to stationary and quasi-stationary flight con-
ditions. Therefore, when the tilt-rotor aircraft hovers, takes off or
lands, the control methods of quad-rotor aircraft can be used di-
475
rectly. During the mode transition from hover to forward flight and
vice versa, the quad rotors are tilted synchronously by a gearing
driven arrangement. Moreover, a finite-time convergent observer is
developed to reconstruct the unknown variables and uncertainties
during mode transitions. Based on the observer, a switched logic
controller is proposed to drive the aircraft to implement the mode
transition with the invariant flying height. Comparing with the
electric powered tilt-rotor aircrafts, the modeling of the presented
tilt-rotor aircraft is large size, the uncertainties during mode tran-
sition are estimated easily, and it has the widely applications with
much long-time cruise and larger payload.
2. Tilt-rotor aircraft design
2.1. Configuration of aircraft
The designed tilt-rotor aircraft is shown in Figs. 1(a)–1(d).
Mechanical components of tilt-rotor aircraft in Fig. 1(a):
1: front-right rotor, 2: front-left rotor, 3: rear-left rotor, 4: rear-
right rotor, 5: right fixed wing, 6: left fixed wing, 7: right aileron,
8: left aileron, 9: vertical tail, 10: fuselage, 11: front-right free
wing, 12: front-left free wing, 13: rear-left free wing, 14: rear-right
free wing, 15: front-right gearing arrangement, 16: front-left gear-
ing arrangement, 17: rear-left gearing arrangement, 18: rear-right
gearing arrangement.
For the tilt-rotor aircraft, there are the following four flight
modes:
Hover: When the aircraft is climbing, it is put on to at a cer-
tain height, normally out ground effect where the thrusts of the
quad rotors compensate the helicopter weight mg (see Fig. 1(b)).
The attitude is regulated by changing the blade pitch angles of the
quad rotors, and the rotor speeds are fixed in rated power. Thus
the control of the thrust and torque outputs is achievable. A con-
trol method for quad-rotor aircraft can be used directly.
Vertical fight: This fight mode starts when the aircraft is at rest
on the ground in ground effect. Then takeoff is produced and the
aircraft climbs. Vertical descent precedes landing. In the absence
of perturbations the thrusts of the quad rotors are always vertical
(see Fig. 1(b)).
Mode transition: It is considered that this fight mode is out of
ground effect. At the beginning, the aircraft is assumed to be in
hover (see Fig. 1(b)). As the gearing arrangement is driven, the
quad rotors are tilted synchronously and pitch towards the hori-
zontal, which in turn causes the horizontal speed of the aircraft
to increase (see Fig. 1(c)). The attitude is controlled by changing
the blade angles of the quad rotors to make the fixed wings ob-
tain a given angle of attack in accordance with the relative wind.
The gravity of the aircraft is counteracted mainly by the vertical
force of the thrust generated by the rotors, and the height of the
aircraft is kept invariant. With the horizontal speed increasing, the
fixed wings develop lift. The aircraft is made transition into hori-
zontal flight in a fixed wing mode (see Fig. 1(d)). During this mode
transition, before the tilt angle attains a given value, roll dynamics
is controlled by the thrusts differential of the left and right rotors.
Yaw dynamics is regulated by the counteractive moments brought
by the rotation of the rotors in the air. Pitch dynamics is controlled
by the thrusts differential of the front and rear rotors. After the
tilt angle exceeds a given value, the flap bias angle of rear two
free wings is selected to control pitch dynamics. Roll dynamics are
controlled by the counteractive moments brought by the rotation
of the rotors in the air. Yaw dynamics are regulated by the thrusts
differential of the left and right rotors. For level-to-hover transi-
tion, the reverse procedure is used. As the gearing arrangement is
driven, the quad rotors are tilted backward. It causes the horizon-
tal speed of the aircraft to decrease and the vertical thrust vector
476 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492

Fig. 1. Tilt-rotor aircraft: (a) Aircraft structure; (b) Modes of hover, takeoff and landing; (c) Mode transition from hover to forward flight and vice versa; (d) Mode of forward
flight.

gradually increases to keep the height invariant. Thus, the aircraft is known that P i = T i v i , where v i is the induced velocity of rotor.
slows and performs transition to hover. Therefore, power coefficient C P i can be written as
Forward fight: We also consider that this fight mode will be out
ground effect. The thrust of the quad rotors ensures forward fight, 2T i v i 2T i vi
and the lift force generated by fixed wings keeps the height of the C Pi = = = C T i λi (3)
ρ A (Ωi R )3 ρ A (Ωi R )2 Ωi R
aircraft invariant (see Fig. 1(d)).
where λi = v i /(Ωi R ) is the induced inflow ratio in hover. From (1),
2.2. Gearing arrangement for tilt rotors we obtain
 
A synchronous gear-driven arrangement for the tilt-rotor air- vi 1 Ti 2T i 1
craft is designed to implement mode transition, as shown in λi = = = = C Ti (4)
Ωi R Ωi R 2ρ A 4ρ A (Ωi R )2 2
Fig. 2(a). The directions of the drive arrows denote the hover-to-
level transition. In the gear-driven arrangement, the symmetrical
Therefore, from (3) and (4), the ideal power can be written as
structure shown in Fig. 2(b) is adopted. With this gear-driven ar-
rangement, the quad rotors can be tilted synchronously from ver- 1 1 3/2
tical to level and vice versa. C Pi = C Ti C Ti = CT (5)
2 2 i

2.3. Gross weight of aircraft In hovering flight, the induced power predicted by the simple mo-
mentum theory can be approximately described by a modification
In this paper, let the gross weight of aircraft be m, and four to the momentum result in (5), i.e.,
same turboshaft engines are selected, the weight of each engine
is me , and its thrust-to-weight ratio is ρe = T e /(me g ), where g 1
is the gravity acceleration, T e is rotor thrust in rated power. In C P i1 = κ C T3/i 2 (6)
2
order to consider the control maneuverability ρmT during hover or
mode transition, the thrust 4T e should be larger than
√ the gross where κ is called an induced power correction factor. The profile
weight mg of aircraft. We select ρmT = 4T e /mg = 2. Therefore, power of rotor can be obtained as follow:
each rotor thrust can be carried out as T e = ρmT mg /4.
R
2.4. Rotor radius P 0 = Ωi p D ydy (7)
Assume the thrust force generated by the i-th rotor is T i (i = 0
1, 2, 3, 4), and its thrust coefficient C T i and rotor power coefficient where p is the number of blades and D is the drag force per unit
C P i are defined, respectively, as [7]
span at a section on the blade at a distance y from the rotational
  axis. The profile power consumed by the rotor requires the drag
C T i = 2T i / ρ Ωi2 R 2 A (1)
  coefficients of the airfoils that make up the rotor blades. The drag
C P i = 2P i / ρ Ωi3 R 3 A (2) force can be expressed conventionally as
where, A is the disk area, Ωi is the rotating speed of the i-th rotor,
R is rotor radius, P i is the rotor power, and ρ is the air density. It D = 0.5ρ U 2 bC d = 0.5ρ (Ωi y )2 bC d (8)
 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492 477

Fig. 2. Tilt structure and driven gears.

where b is the blade chord, and U is the resultant velocity at the optimum radius of the rotor to maximize power loading at a given
disk. The section profile drag coefficient C d is assumed to be con- gross weight. Solving for the rotor radius in Eq. (15) gives
stant, i.e., C d = C d0 . Then the profile power is deduced to 
R= mg /(π DL)/2 (16)
R
1 Similarly, for two-rotor √aircrafts, such as V-22, we obtain the rotor
P 0 = 0.5ρ p Ωi3 bC d0 y 3 dy = ρ p Ωi3 bC d0 R 4 (9) radius as follow: R 2r = mg /(2π DL√ ). Under the same disk loading
8
0 requirement, we obtain R / R 2r = 1/ 2.
With respect to the conventional two-rotor aircrafts, the rotor
Therefore, the profile power coefficient C p0 can be written as
radius of the presented tilt-rotor aircraft is shorter. For different
2P 0
1
ρ p Ωi3 bC d0 R 4 VTOL rotorcrafts, they have different disk loadings: CH-47D Chi-
4
C P0 = = nook (1 rotor, 2 blades), 14 kg/m2 ; Bell 206B3 JetRanger (1 ro-
ρ A (Ωi R )3 ρπ R 2 (Ωi R )3 tor, 2 blades), 18 kg/m2 ; CH-47D Chinook (2 rotors, 3 blades),
   
1 pb 1 pbR 1 43 kg/m2 ; Mil Mi-26 (1 rotor, 7 blades), 71 kg/m2 ; CH-53E Super
= C d0 = C d0 = σ C d0 (10)
Stallion (1 rotor, 7 blades), 72 kg/m2 ; MV22B Osprey, 129.63 kg/m2
4 πR 4 π R2 4
(2 rotors, 3 blades). A suitable disk loading DL = DLQT = 60 kg/m2
where rotor solidity is expressed as σ = ( pbR )/(π R 2 ) = ( pb)/(π R ), is required for the presented quad tilt-rotor aircraft. Therefore,
and it has been defined as the ratio of total blade area to the disk from Eq. (16), the rotor radius of the tilt-rotor aircraft can be com-
area. Armed with the estimates of the induced power coefficient puted as follow:
(6) and the profile power coefficient (10), the rotor power by using 
the modified momentum theory results that R= mg /(π DLQT )/2 (17)
1 1
C p i = C P i1 + C P 0 = κ C T3/i 2 + σ C d0 (11) 2.5. Number of blades and rotor solidity
2 4
From Eqs. (1), (2) and (11), the ratio P i / T i is given by The selection of the number of blades (for a given blade area
  or solidity) is usually based on dynamic rather than aerodynamic
Pi C Pi 1 1 σ C d0 criteria, that is, it is based on the minimization of vibratory loads,
= Ωi R = Ωi R κ C T1/i 2 + (12)
which is easier for rotors that use a larger number of blades. An
Ti C Ti 2 4 C Ti
analysis of the data [26] suggests that hover performance is pri-
By differentiating Eq. (12) with respect to C T i , it is shown that for marily affected by rotor solidity σ ; the number p of blades is
a rotor with rectangular blades the operating C T i to give the lowest secondary. Based on the requirement of the blade aspect ratio, we
T / P is can select the blade aspect ratio ρb . Then from Eq. (17), we ob-
  tain the blade chord b. Values of σ for contemporary helicopters
1 1 σ C d0
0 = Ωi R κ C T−i1/2 − (13) vary from about 0.06 to 0.12. Lager helicopters generally tend to
4 4 C T2 have upper solidity rotors. Here, we expect the rotor solidity is
i

i.e.,
σ = ( pb)/(π R ). Therefore, the number of blades is
p = σ π R /b (18)
C T i = (σ C d0 /κ )2/3 (14)
From (1) and (14), the disk loading for maximum power loading is 2.6. Fixed-wing loading
at
A large wing aspect ratio A w = 12 is selected for the fixed wing
Ti mg in order to obtain a long-time cruise. Moreover, we expect the de-
DL = = = 0.5C T i ρ (Ωi R )2
A 4π R 2 sired wing loading WL w after comparing with other fixed wing
 
σ C d0 2/3 aircrafts. Therefore, the wing area is obtained:
= 0.5ρ (Ωi R )2 (15)
κ S w = m/WL w (19)
It is assumed that for design purposes each rotor of the quad rotors Accordingly, the wing span l w and chord c w for the fixed wing are
carries a quarter of the total weight of the aircraft. Profile drag co- obtained, respectively:
efficient C d0 is assumed be constant and independent of Reynolds  
number R ε and Mach number M. This equation determines the lw = Aw Sw, cw = Sw/Aw (20)
478 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492

Fig. 3. Size of tilt-rotor aircraft.

In addition, considering of the effect of the free wings during

cruise, we expect the desired wing loading WL ws . Then the sum of
wings area is calculated as:

S = S w + S front + S f i (21)
where S w , S front and S f t are the areas of fixed wing, front wing
and free wings, respectively. Therefore, the area of each free wing
is: S f t = (m/WL ws − S w − S front )/4. Other sizes for the aircraft are
shown in Fig. 3.

3. Mathematical model

In this section, by analyzing the aerodynamics of mode transi-

tion, a 6-degree of freedom (DOF) nonlinear model of the tilt-rotor
aircraft is established. Moreover, the uncertainties in the system
dynamics are considered. Fig. 4. Forces and torques of tilt-rotor aircraft.

3.1. Mathematical model during mode transition β is the tilt angle of rotors. Γβ b is the transformation matrix rep-
resenting the orientation of the rotorcraft from frame Γβ to Γb ,
The forces and torques of the aircraft during mode transition i.e.,
are shown in Fig. 4. ⎡ ⎤
cβ 0 sβ
3.1.1. Coordinates and frames R βb = ⎣ 0 1 0 ⎦ (23)
In Fig. 4, C is the center of gravity of the aircraft. M β is the −sβ 0 cβ
pitch moment generated by gearing arrangement. Let
3.1.2. Aerodynamic analysis of large-size tilt rotor
(i) Γ g = ( E x , E y , E z ) denote the right handed inertial frame; Algebraic expressions of the aerodynamic forces and torques are
(ii) Γb = ( E bx , E by , E bz ) denote the frame attached to the aircraft’s used to deduce the generalized external forces acting on the air-
fuselage whose origin is located at its center of gravity; craft. In this subsection, using the blade element method [27,28],
β the force and torque computations are presented, i.e., the follow-
(iii) Γβ = ( E x , E y β , E zβ ) denote the right handed inertial frame of
the tilt-rotor aircraft; ing expressions are formulated, respectively: rotor thrust and blade
(iv) Γr = ( X r , Y r , Z r ) denote the right handed inertial frame of pitch angle, drag torque and blade pitch angle. Moreover, the rela-
blade element (see Fig. 5). tionship between the drag torque and rotor thrust is obtained.
The airflow velocity vector in reference system Γβ which is lo-
p b = (xb , y b , zb ) is the position of center of gravity relative cated at the center of each rotor is
to frame Γ g . (ψ, θ, φ) ∈ 3 describes the aircraft orientation ex-
T
pressed in the classical yaw, pitch and roll angles (Euler angles). V β = V β X V βY VβZ = R− 1
βb V b (24)
We use c θ for cos θ and sθ for sin θ . R bg is the transformation ma-
trix representing the orientation of the rotorcraft from frame Γb to where V b is the airflow velocity vector in frame Γb . Let
Γ g , i.e., VβX
vi r
⎡ ⎤ v̄ i = , r̄ = , μx = ,
cθ cψ c φ sψ + sφ sθ c ψ sφ sψ − c φ sθ c ψ ΩR R ΩR
R bg = ⎣ −c θ sψ c φ c ψ − sφ sθ sψ sφ c ψ + c φ sθ sψ ⎦ (22) V βY VβZ β̇ r
sθ −sφ c θ cφ cθ μy = , μz = , μβ = (25)
ΩR ΩR ΩR
 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492
Fig. 5. Coordinates and forces for tilt-rotor: (a) Blade and blade element coordinates;
(b) Forces on blade element; (c) Tilt dynamics on blade.
W̄ x = r̄ + μx sin γ + μ y cos γ ,
W̄ y = μx cos γ + μ y sin γ ,
W̄ z = v̄ i + μz + μβ cos γ , ρβ̇ = β̇/Ω (26)
where v i is the induced velocity of rotor, and Ω is the rotor speed.
Assumption 1. The following assumptions are taken into account
in this development:
(i) During mode transition, rotor speed Ω is constant such that
the outputs of the four engines are all in rated power;
(ii) The relative inflow angle at the blade element (see Fig. 5(b))
will be β ∗ = arctan( W z/ W x) ≈ W z/ W x;
(iii) Tilt rate β̇ is far smaller than rotor speed Ω , i.e.,
|β̇|  Ω = γ̇ (27)
The blades without blade cyclic pitch are selected. And for the
blade with linear torsion ϕ∗ , the blade pitch angle ϕ∗ is ex-
pressed as
ϕ∗ = ϕ7i + ϕ∗ (r̄ − 0.7) (28)
where ϕ∗ = ϕ1 − ϕ0 , and ϕ7i is pitch angle at r̄ = 0.7. The chord
length of blade with wingtip is constant, and we let b̄ = b̄1 , where
b̄ = b/ R.
Remark 1. Assumption 1 is presented for the large-size aircrafts
with turboshaft engines, which haven’t variable-speed rotors.
Small-size tilt-rotor aircrafts are usually electric powered. For an
electric powered tilt-rotor aircraft, the rotor speeds are allowed to
freely vary according to the desired level of thrust. However, the
479
rotor speeds varying freely according to the desired level of thrust
is unrealistic for current state-of-the art technology for conven-
tional large-size rotorcrafts, i.e., the control of the thrust output is
usually difficult to be achievable by changing the rotor speed. For
the large-size aircrafts with turboshaft engines, the blade pitch an-
gles of the rotors are regulated to vary according to the desired
level of thrust.
In fact, using a variable velocity joint between the rotor and
engine is a means to allow for speed changes of the rotor. By
changing the transmission ratio of gearbox or belt wheels, the
speed change of rotor is implemented. However, for the gearbox,
speed range is limited in rated power, and it only has several
speeds. Continuously variable transmission (CVT), also known as
single-speed transmission, gearless transmission, variable pulley
transmission, is a transmission in which the ratio of the rotational
speeds of two shafts, as the input (or engine) shaft and output
shaft, can be varied continuously within a given range, providing
an infinite number of possible ratios. The CVT is allowed to select
the relationship between the speed of the engine and the speed
of the wheels within a continuous range, and the transmission ac-
tion is smooth. However, it is not suitable for high maneuverability
and large-torque load, and it offers approximately 88% efficiency.
With respect to using a variable velocity joint, controlling the blade
pitch angles of the rotors by rudders is very fast and accurate, and
the rotor speeds are fixed in rated power.
Other than the Boeing A160 Hummingbird, all rotorcrafts to
date feature fixed speed rotors (the V-22 is occasionally referred to
as having a variable-speed rotor, but it really has only two speeds).
Unlike most rotorcrafts, which feature turboshaft engines for re-
quirements on high power output, the A160 has a piston engine
that allows for variable-speed rotors. A turboshaft version of the
A160 does exist, but it hasn’t overcome the technical challenge
of being variable-speed yet. Unless the intended vehicle does not
have high power requirements (which would be unlikely consid-
ering that the tilt-rotor aircrafts are highly inefficient in hovering
flight relative to pure helicopters like the A160), the vehicle would
need to have turboshaft engines, and it is impossible to be as-
sumed that control of the thrust output is achievable by changing
the rotor speed. Besides, the downside to having variable-speed ro-
tors is that a departure from the optimal rotor speed means that
engine efficiency is reduced.
Therefore, in this paper, the control of the thrust output is
achievable by changing the blade pitch angles of the rotors, and
the rotor speeds are fixed in rated power.
Rotor thrust generation
Theorem 1. For the tilt-rotor, which is satisfied with Assumption 1, the
following relationship between rotor thrust coefficient and blade pitch
angle holds:
3C T π
+ (0.3μ2x + 0.3μ2y − 0.05)ϕ∗
pb̄1 a∞
ϕ7i = 3
1+ 2
μ2x + 32 μ2y
3
( v̄ i + μz ) + 34 μ y ρβ̇
+ 2
(29)
3
1+ 2
μ2x + 32 μ2y
Especially, if the sideslip can be ignored during mode transition, i.e.,
μ y = 0, then
 
3C T i π   3
ϕ7i = + 0.3μ2x − 0.05 ϕ∗ + ( v̄ i + μz )
pb̄1 a∞ 2
 
3
1+ μ2x (30)
2
480 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492

The proof of Theorem 1 is presented in Appendix A.

Remark 2 (Relationship between rotor thrust and blade pitch angle).

From Eq. (106) in Appendix A and Eq. (1), the rotor thrust gener-
ated by each rotor can be written as

T i = h i1 ϕ7i + h i2 (31)

where
 
h i1 = ρ Ω 2 R 4 pb̄1 a∞ 1 + 1.5μ2x /6

  
h i2 = ρ Ω 2 R 4 pb̄1 a∞ 0.05 − 0.3μ2x ϕ∗ − 1.5( v̄ i + μz ) /6
Fig. 6. Force on free wing: (a) tilt dynamics of free wing; (b) tilting from hover to
(32) forward flight; (c) tilting from forward flight to hover.

and in the frame Γβ , the thrust vector is described by

Remark 3. From Eqs. (36) and (37), the drag torque can be rewrit-
T ten as
T i = 0 0 Ti (33)
Q i = di1 ϕ7i + di2 (39)

Blade drag torque where di1 = 0.5ρ Ω R 2 2

π R R d̄i1 , di2 = 0.5ρ Ω R π R R d̄i2 .
2 2 2 2

For the blade element drag torque, considering of Eq. (100) in

Appendix A and Fig. 5, we obtain Remark 4. We suppose the sideslip can be ignored during mode
transition, i.e., μ y = 0. Consequently, Eq. (38) becomes
dQ i = (dL sin β∗ + dD cos β∗ )r ≈ (dLβ∗ + dD)r
  pb̄1 a∞1
= 0.5ρ ba∞ W x2 C d /a∞ − W z2 + W x W z ϕ∗ rdr (34) d̄i1 = ( v̄ i + μz )
3π

Taking into account Eq. (34), the number of blades, the blade el- pb̄1 a∞ C d0   1
ement conditions and the total contribution in one revolution, we d̄i2 = 1 + μ2x + μ2y + ( v̄ i + μz )ϕ∗
can write:
π 4a∞ 60

1 
2πR − v̄ 2i + μ2z + 2 v̄ i μz (40)
p dM M 2
Qi = drdγ
2π dr
0 0 Remark 5. From Eqs. (31) and (39), we obtain the relationship be-
tween thrust T i and drag torque Q i as follow:
2πR  
ρ pa∞ Cd
= W x2 − W z2 + W x W z ϕ∗ brdrdγ (35) Q i = di1 T i /h i1 + (di2 − di1 h i2 /h i1 ) (41)
4π a∞
0 0
Moreover, from Fig.
4, the sum of the thrusts generated by the
4
The drag torque coefficient C Q i is defined as quad rotors is T = i =1 T i .
 
C Q i = Q i / 0.5ρ Ω 2 R 2 π R 2 R (36) 3.1.3. Lift and drag forces generated by wings
In this subsection, the aerodynamics analysis is presented for
Theorem 2 (Relationship between drag torque coefficient and blade free wings and fixed wings, and the effect of rotor tilting is con-
pitch angle). For the tilt-rotor, which is satisfied with Assumption 1, the sidered.
following relationship between drag torque coefficient blade pitch angle
Aerodynamics of free wing
holds:
For free wings, from Fig. 6, we obtain
C Q i = d̄i1 ϕ7i + d̄i2 (37)
V β t = (d f + 0.25dr )β̇,
where 1/2
V rt = ( V β X + V β t )2 + ( v i + V β Z )2 ,
 
pb̄1 a∞1 1 α f = arctan( V β X + V β t )/( v i + V β Z ) (42)
d̄i1 = v̄ i + μz + μ y ρβ
3π 2
 where V β t is the airflow generated by rotor tilting, V rt is the re-
pb̄1 a∞ C d0  2 2
 sultant velocity at the free wing, and α f is angle of attack for free
d̄i2 = 1+μ +μ x y
π 4a∞ wing, d f is the distance from the center of rotor to the leading
  edge of the free wing. Therefore, the lift and drag forces on free
1 1
+ v̄ i + μz + μ y ρβ̇ ϕ∗ wing (i = 1, 2, 3, 4) are written, respectively, as
60 2
  L i = L i0 +  L i , 2
L i0 = 0.5ρ S f i C L f V rt (43)
1 1
− v̄ 2i + μ2z + 2 v̄ i μz + ρβ̇2 (38)
2 4 and

2
The proof of Theorem 2 is presented in Appendix A. D i = D i0 +  D i , D i0 = 0.5ρ S f i C D f V rt (44)
 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492 481

where  L i and  D i are the bounded uncertainties, and 3.1.4. Forces generated by vertical tail
From Fig. 4, the yaw and drag forces generated by vertical tail
C L f = C f α f + Cr δ f i , C D f = C D f 0 + C L2 f /(π A f e f ), T T
in frame Γb are denoted by 0 − f rl 0 and − f rl 0 0 ,
  respectively.
e f = 1.78 1 − 0.045 A 0f .68 − 0.46 (45)

S f i is the area of the free wing, C f is the lift coefficient brought by
angle of attack α f to free wing, C D f 0 is the drag coefficient when
the angle of attack α f is equal to zero. C r is the lift coefficient
brought out by the flap bias angle δ f i on free wing. e f is the value
of the Oswald’s efficiency factor, and A f is the aspect ratio of free
wing.
3.1.5. Gyroscopic and counteractive moments
Tilting the rotors around the axes l1 and l2 creates gyroscopic
moments which are perpendicular to these axes and to the spin
β
axe (E z ). Indeed, these moments are defined by the cross product
β
of the kinetic moments ( J r Ωi E z ) of the rotors and the tilt velocity
vector. They are first expressed in the rotor frames as [31]

Remark 6. The case without Oswald’s efficiency factor applies only 

4
(−1)i +1 J r Ωi β̇ E x
β
to wings with elliptical lift distributions. It is possible to modify Gβ = (53)
the lift coefficient expression slightly to make it apply any wing by i =1
using an Oswald’s efficiency factor. The value of Oswald’s efficiency for the longitudinal tilting. In order to express these gyroscopic
factor is 1 for elliptical wings and between 0.5 and 1 for most moments in the fixed body frame Γb , we should multiply the
common wings shapes. above equation by the rotational matrix R β . We therefore find

For the free wing, let the angle of attack with respect to the 
4
(−1)i +1 J r Ωi β̇ E x
β
maximal lift coefficient be αmax . In order to avoid the stall phe- G b = [ G b1 G b2 G b3 ]T = R β (54)
nomenon, the following relation should hold: i =1
 
|α f | = arctan( V β X + V β t )/( v i + V β Z ) ≤ αmax (46) 3.1.6. Thrust vectoring moment
Denote C 1 , C 2 , C 3 and C 4 the application points of the thrust
Therefore, from Eq. (46) and Fig. 6(b) (i.e., tilting from hover to
T 1 , T 2 , T 3 and T 4 , respectively. From Fig. 4, we can define
forward flight), for dβ/dt ≥ 0, we obtain
C 1 C = (l4 , −l1 , 0), C 2 C = (l4 , l1 , 0), C 3 C = (−l3 , l2 , 0) and C 4 C =
 (−l3 , −l2 , 0) as the positional vectors expressed in Γb . Then, the
β̇ ≤ ( v i + V β Z ) tan αmax + V β X /(d f + 0.25dr ) (47)
moment exerted by the forces T i (i = 1, 2, 3, 4) on the airframe Γb
From Eq. (46) and Fig. 6(c) (i.e., tilting from forward flight to is
hover), for dβ/dt < 0, we obtain
 
4

−β̇ ≤ ( v i + V β Z ) tan αmax − V β X /(d f + 0.25dr ) (48) τT = T i × Ci C (55)
i =1
For the rear free wings (see Fig. 4), we select them as the con-

trollers for the pitch dynamics when the tilt angle β exceeds a where T i (i = 1, 2, 3, 4) is the thrust vector in frame Γb . After some
given value. Therefore, we rewrite Eq. (43) as follow: computations, we obtain
⎡ ⎤
L i = L 1i0 + L 2i0 +  L i , i = 3, 4 (49) ( T 2 − T 1 )l1 c β + ( T 3 − T 4 )l2 c β
⎢ ⎥
where L 1i0 = 0.5ρ S f i V rt
2
C r δi , L 2i0 = 0.5ρ S f i V rt
2
C f αf . τT = ⎣ ( T 1 + T 2 )l4 cβ − ( T 3 + T 4 )l3 cβ ⎦ (56)
The bias angle δ3 = δ4 = δ regulates the pitch dynamics. −( T 2 − T 1 )l1 sβ − ( T 3 − T 4 )l2 sβ
Aerodynamics of fixed wing
3.1.7. Reactive torques
The lift and drag forces on the fixed wings (see Fig. 4) are, re-
As the blades of the quad rotors rotate, they are subject to drag
spectively,
forces which produce torques around the aerodynamic center C i .
L i = L i0 +  L i , These moments act in opposite direction relative to Ωi . The reac-
tive torque generated in free air by the rotors due to rotor drag is
L i0 = 0.5ρ S ri C L w V b2 , i = 5, 6 (50) given by:
and

4
Q = (−1)i +1 Q i (57)
D i = D i0 +  D i , D i0 = 0.5ρ S ri C D w V b2 (51)
i =1
where
After some computations, we obtain reactive torque vector in
frame Γb as follow:
C L w = C w0 + C w α α + C w δ δi ,
 T
C D w = C D w0 + C L2w /(π A w e w ), 
4 
4
  Q r = sβ (−1)i +1 Q i 0 c β (−1)i +1 Q i (58)
e w = 1.78 1 − 0.045 A 0w.68 − 0.46 (52) i =1 i =1
 L i and  D i are the uncertainties affected by rotors tilting. A w is
aspect ratio of fixed wing. S ri is the area of the left (right) fixed 3.1.8. Adverse reactionary moment
wing, C w α is the lift coefficient brought by the angle of attack α As described previously, this moment appears when forcing the
of fixed wing, C w0 is the lift coefficient when angle of attack α rotors to tilt longitudinally. It depends especially on the propeller
is equal to zero, C D w0 is the drag coefficient when the angle of inertia J 4 and on tilt accelerations. This moment acts as a pitch
attack α is equal to zero. C w δ is the lift coefficient brought out moment and can be expressed in frame Γb as follow:
by the normal flap bias angle δi .e w is the value of the Oswald’s
efficiency factor. τβ = J 4 β̈ E by (59)
482 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492

 
3.1.9. Moment generated by fixed and free wings Γ p2 = −c θ sψ sβ + (sφ c ψ + c φ sθ sψ )c β T
From Fig. 4, the moment generated by lift forces L 5 and L 6 of  
fixed wings in frame Γb is obtained as follow: + −c θ sψ sα + (sφ c ψ + c φ sθ sψ )c α L 5,6
 
T + c θ sψ c α + (sφ c ψ + c φ sθ sψ )sα D 5,6
τδ = τδr τδ p τδ y
T  
4
= ( L 6 − L 5 )l5 c α 0 ( L 5 − L 6 )l5 sα (60) + c θ sψ c β + (sφ c ψ + c φ sθ sψ )sβ L i0
i =1
Meanwhile, in frame Γb , the moment generated by free wings is
given as follow:  
4

⎡ ⎤ ⎡ ⎤ + c θ sψ sβ − (sφ c ψ + c φ sθ sψ )c β D i0
τfφ ( L 2 − L 1 )l1 sβ + ( L 3 − L 4 )l2 sβ i =1
τ f = ⎣ τ f θ ⎦ = ⎣ ( L 1 + L 2 )l4 sβ − ( L 3 + L 4 )l3 sβ ⎦ (61) + (−c φ c ψ + sφ sθ sψ ) f rl + c θ sψ f rd
τfψ ( L 2 − L 1 )l1 c β + ( L 3 − L 4 )l2 c β
Because the moment generated by drag forces D 5 and D 6 of fixed
Γ p3 = (sθ sβ + c φ c θ c β ) T + (sθ sα + c φ c θ c α ) L 5,6
wings is very small, it is taken as an uncertain vector defined in
frame Γb , and it expressed as follow: τfd = [τ f d1 0 τ f d3 ]T . + (−sθ c α + c φ c θ sα ) D 5,6
Also, the moment generated by the drag forces D 1 , · · ·, D 4 of
free wings is very small, it is taken as an uncertain vector defined 
4

in frame Γb , and it is expressed as follow: τxd = [τxd1 τxd2 τxd3 ]T . + (−sθ c β + c φ c θ sβ ) L i0

i =1

3.1.10. Moment generated by vertical tail 

4
From Fig. 4, in frame Γb , the moment generated by vertical tail + (−sθ sβ − c φ c θ c β ) D i0
is given as follow: τt = [0 0 f rl l3 ]T . i =1

+ sφ c θ f rl − sθ f rd − mg
3.1.11. The sum moment in the fixed-body frame
From the moments analysis above, the sum of the moment in
the fixed-body frame is presented as follow: τ = τt + Q r + τβ +
L 5,6 = L 50 + L 60 , D 5,6 = D 50 + D 60 ,
τδ + τ f + τxd + τfd + τt . T 
 p =  p1  p2  p3 = ¯1 
¯2 
¯3 T (65)
3.2. The motion equation of the aircraft
and
Aerodynamic interference effects between the different compo- T
nents are taken as the bounded uncertainties and included in the ua = ua1 ua2 ua3 = c β uac + sβ uas ,
motion equations. The equations of motion written in terms of the ⎡ ⎤ ⎡ ⎤
center of mass C in the fixed axes of co-ordinate ( X , Y , Z ) are then u 11 ( T 2 − T 1 )l1 + ( T 3 − T 4 )l2
⎢ ⎥
uac = ⎣ u 21 ⎦ = ⎣ ( T 1 + T 2 )l4 − ( T 3 + T 4 )l3 ⎦ ,
ẍ p = m−1 (u p + Γ p +  p ) (62) u 31  4
(−1)i +1 Q
i =1 i
ẍa = J −1 (ua + u β + Γa + a ) (63) ⎡ ⎤ ⎡ 4
i +1
⎤
u 31 i =1 (−1) Qi
⎢ ⎥
J 4 β̈ = − M β (64) uas = ⎣ u 22 ⎦ = ⎣ ( L 30 + L 40 )l3
1 1
⎦
−u 11 −[( T 2 − T 1 )l1 + ( T 3 − T 4 )l2 ]
where, J = diag{ J 1 , J 2 , J 3 }; J 1 , J 2 , and J 3 are the three-axis mo-
ment of inertias; J 4 is the tilting moment of inertia; and
u β = [u β 1 uβ 2 u β 3 ]T = [0 Mβ 0]T ,
xp = [ X Y Z ]T , xa = [φ θ ψ]T , T
T Γa = Γa1 Γa2 Γa3
Γ p = [Γ p1 Γ p2 Γ p3 ] , 
Γa1 = ( L 6 − L 5 )l5 c α + ( L 2 − L 1 )l1 + ( L 3 − L 4 )l2 sβ ,
⎡ ⎤ ⎡ ⎤T  
u p1 (c θ c ψ sβ + (sφ sψ − c φ sθ c ψ )c β ) Γa2 = ( L 10 + L 20 )l4 sβ − L 230 + L 240 l3 sβ
u p = ⎣ u p2 ⎦ = ⎣ (−c θ sψ sβ + (sφ c ψ + c φ sθ sψ )c β ) ⎦ 
u p3 (sθ sβ + c φ c θ c β ) Γa3 = +( L 50 − L 60 )l5 sα + ( L 20 − L 10 )l1 + ( L 30 − L 40 )l2 c β

  + f rl l3
Γ p1 = + c θ c ψ sα + (sφ sψ − c φ sθ c ψ )c α L 5,6
  
+ −c θ c ψ c α + (sφ sψ − c φ sθ c ψ )sα D 5,6 a = ¯4 
¯5 
¯6 T (66)
 
4
+ −c θ c ψ c β + (sφ sψ − c φ sθ c ψ )sβ L i0 with
i =1
 ¯ φu ,
¯ 4 = τxd1 + τ f d1 + G b1 +  ¯ 5 = +τxd2 + G b2 + 
 ¯ θu,
 
4
+ −c θ c ψ sβ − (sφ sψ − c φ sθ c ψ )c β D i0 ¯ 6 = τxd3 + τ f d3 + G b3 + 
 ¯ ψu (67)
i =1
+ (−c φ sψ − sφ sθ c ψ ) f rl − c θ c ψ f rd ¯ 1, 
 ¯ 2, 
¯ 3, 
¯ φu , 
¯ θ u and 
¯ ψ u are the bounded uncertainties.
 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492 483

3.3. Measurement sensors and actuators The proof of Theorem 3 is presented in Appendix A.

Position ( X , Y ) can be obtained by Global Positioning System 5. Controller design

(GPS). The position data from GPS is sent to the processor in the
aircraft for feedback control. Altitude Z is measured by an altime- In this section, a control law is derived for the purpose of
ter. Attitude (ψ, θ, φ ) and attitude rate (dψ/dt , dθ/dt , dφ/dt) can be stabilization and trajectory tracking, controllers for the attitude,
obtained by an Inertial Measurement Unit (IMU). Most common at- position and tilting dynamics are designed, respectively. By us-
titude sensors are based on gyros. Angle of attack is measured by an ing the finite-time convergent observers proposed above, the un-
angle of attack sensor. Rotor thrusts T 1 , T 2 , T 3 , T 4 , tilting torque known states and the generalized disturbance are reconstructed.
M β and the bias angle δ3 = δ4 = δ of the rear two free wings are Suppose the reference trajectory and its finite order derivatives are
selected as the control actuators. bounded, and they are generated directly. Therefore, the proposed
control law and its performance are presented in the following
Remark 7. Angle of attack is quite simply the angle between the theorems.
wing chord and the oncoming air that the wing is flying through.
Angle of attack can be measured by an angle of attack sensor. For Controller design of attitude dynamics
instance, Dynon Avionics was the manufacturer of affordable EFIS
products (for instance, EFIS-D100) to offer a way to measure angle Theorem 4. For attitude dynamics (63), to track the reference attitude
of attack. Through extensive wind tunnel testing, Dynon is able to aa = [φd θd ψd ]T , if the controller is designed as
offer an angle of attack (AOA) pilot probe that measures both angle
 
of attack and airspeed when connected to any of the EFIS-based ua = − J (k1 ea + k2 ėa − äd ) − u β − Γa − a (72)
products. Some other products of angle of attack sensors: AMETEK
      T
Aerospace Angle-of-Attack (AOA) transducers, the 4239-01 angle of where ea = xa − ad , ėa = ẋa − ȧd , ėa = z2 − ȧd , z2 = z12 z22 z32 ,
attack (AOA) sensors, AeroControlex, Thales angle of attack (AOA)     T
a = J 1 z13 J 2 z23 J 3 z33 , k1 and k2 are positive constants, the
sensor, et al.  
observations ėa and a are obtained from observer (70), then the closed-
4. Observer design loop system rendering by controller (72) will converge asymptotically to
the origin, i.e., the tracking error ea → 0 and dea /dt → 0 as t → ∞.
In systems (62) and (63), Ẋ , Ẏ , Ż , Ẍ , Ÿ , Z̈ , φ̈ , θ̈ and ψ̈ are
unknown. Moreover,  p and a are uncertain vectors. In order to The proof of Theorem 4 is presented in Appendix A.
reconstruct the unknown states and the uncertainties of systems
(62) and (63), we present the finite-time convergent observers, and Switching logic for controller u a
a theorem is given as follow. During mode transition, the flight speed is small when tilt an-
gle β is small, and the role of the fixed wings is weak. The pitch
Theorem 3. For systems (62) and (63), the finite-time convergent ob- dynamics is regulated by the differential control of the quad ro-
servers are presented as follows: tor thrusts; the flight speed becomes large when tilt angle β is
large, and the effect of the fixed wings is strong. Moreover, the
(i) For position dynamics (62), the finite-time convergent observer is differential control of the quad rotor thrusts is week for the pitch
designed as dynamics. Therefore, the rear free wings can be selected as con-
troller of the pitch dynamics.
x˙ i1 = xi2 − k p1 |xi1 − y in |2/3 sign(xi1 − y pi )
   
Based on the analysis above, we will allocate controller ua with
a switching logic based on the selection of β . It is supposed that
x˙ i2 = xi3 + m−1 (u pi + Γ pi ) − k p2 |xi1
  
we have selected a tilt angle β w , and at this angle the controllers

− y pi |1/3 sign(xi1 − y pi ) switches. Angle β w will be decided later.
(i) When β ≤ β w , from Eq. (66), let
x˙ i3 = −k p3 sign(xi1 − y pi )
 
(68)
 
where i = 1, 2, 3, and y p1 = X , y p2 = Y , y p3 = Z , s + k p1 s2 +
3 u 22 = L 130 + L 140 l3 = 0 where δ3 = δ4 = 0 (73)
k p2 s + k p3 is Hurwitz, and k p3 > m−1 |d( ¯ i )/dt |, i = 1, 2, 3. There-
Therefore, controller (72) can be rewritten as
fore, a time t s exists, for t ≥ t s , such that
 
  
¯ 1,−1 ua = M c uac = − J (k1 ea + k2 ėa − äd ) − u β − Γa − a (74)
x11 = X , x12 = Ẋ , x13 = m 

x21 = Y ,

x22 = Ẏ ,

¯ 2,
x23 = m−1  where uac has been defined in Eq. (66), and
⎡ ⎤

x31 = Z ,

x32 = Ż ,

¯3
x33 = m−1  (69) cβ 0 sβ
(ii) For attitude dynamics (63), the finite-time observer is designed as Mc = ⎣ 0 cβ 0 ⎦, | M c | = cβ ,
−sβ 0 cβ
z˙ i1 = zi2 + J i−1 (uai + u β i + Γai ) ⎡ ⎤
 
cβ 0 sβ
M c−1 = ⎣ 0 c β−1 0 ⎦
 
− ka1 | zi1 − yai |1/3 sign( zi1 − yai ) (75)

z˙ i2 = −ka2 sign( zi1 − yai )

(70) −sβ 0 cβ

where i = 1, 2, 3, and ya1 = φ̇ , ya2 = θ̇ , ya3 = ψ̇ , ya3 = ψ , s3 + Then we obtain

ka1 s2 + ka2 s + ka3 is Hurwitz, and ka3 > J i−1 |d( ¯ i )/dt |, i = 4, 5, 6.   
Therefore, a finite time t s > 0 exists, for t ≥ t s , such that uac = − M c−1 J (k1 ea + k2 ėa − äd ) − u β − Γa − a (76)
  −1
¯ 4, 
z11 = φ̇, z12 = J 1  z21 = θ̇, (ii) When β > β w , from Eq. (66), let
 −1
¯ 5,  
¯6−1
z22 = J 2  z31 = ψ̇, z32 = J 3  (71) u 21 = ( T 1 + T 2 )l4 − ( T 3 + T 4 )l3 = 0 (77)
484 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492

and Because βd is required to attain π /2 when the hover-to-level tran-

  sition finishes, we require M t t 12 = π /2. Then M t = π /(2t 12 ). When
u 22 = L 130 + L 140 l3 , where δ3 = δ4 = δ (78) t = t 1 , dβd /dt attains its maximal value. Therefore, the following
relation should be satisfied:
Therefore, controller (72) can be rewritten as
( v id + V β Zd ) tan αmax + V β Xd 

ua = M s uas = − J (k1 ea + k2 ėa − äd ) − u β − Γa − a

(79)
β̇d (t 1 ) = M t t 1 <  (84)
d f + 0.25dr t =t 1

where uas has been defined in Eq. (66), and β β

where V β Xd and V β Zd are the projects on E x and E z , respectively,
⎡ ⎤ T
sβ 0 −c β on frame Γβ for the desired flight speed ẋd = Ẋ d Ẏ d Ż d .
Ms = ⎣ 0 sβ 0 ⎦, | M s | = sβ , Therefore, we obtain
cβ 0 sβ 
⎡ ⎤ t 1 > 0.5π (d f + 0.25dr )/ ( v i + V β Z ) tan αmax + V β X t =t (85)
sβ 0 −c β 1

M s−1 = ⎣ 0 s−
β
1
0 ⎦ (80) ii) Tilt for hover-to-level transition: (β, β̇, β̈ ) is designed as:
cβ 0 sβ ⎧
⎨ −Mt , 0 ≤ t ≤ t1
Then we obtain β̈d = M t , t 1 < t ≤ 2t 1 ,
⎩
   0, t > 2t 1
uas = − M s−1 J (k1 ea + k2 ėa − äd ) − u β − Γa − a (81) ⎧
⎨ −Mt t , 0 ≤ t ≤ t1
β̇d = − M t t 1 + M t (t − t 1 ), t 1 < t ≤ 2t 1
Selection of tilt angle β w ⎩
0, t > 2t 1
We expect the matrices M c and M s have the same gain during
the controllers switching. Therefore, we obtain c β w = sβ w . Thus ⎧
β w = π /4. ⎪
⎪ 0.5π − 0.5M t t 2 , 0 ≤ t ≤ t1
⎨
0.5π − 0.5M t t 12 − M t t 1 (t − t 1 )
Controller design for position dynamics βd = (86)
⎪
⎪ + 0.5M t (t − t 1 )2 , t 1 < t ≤ 2t 1
⎩
0.5π − M t t 12 , t > 2t 1
Theorem 5. For position dynamics (62), to track the reference trajectory
xd = [ X d Y d Z d ]T , if the controller is designed as where M t is positive and constant, and t 1 is the switching time.
  Because βd is required to attain zero when the level-to-hover tran-
u p = −m(k3 e p + k4 ė p ) − Γ p −  p + mẍd (82) sition finishes, we require π /2 − M t t 12 = 0. Then M t = π /(2t 12 ).
      T When t = t 1 , dβd /dt attains its maximal value. Therefore, the fol-
where e p = x p − xd , ė p = ẋ p − ẋd , ėa = x2 − ẋd , x2 = x12 x22 x32 , lowing relation should be satisfied:
    T
 p = mx13 mx23 mx33
( v id + V β Zd ) tan αmax − V β Xd 
, k3 and k3 are positive constants, the ob-
 
servations ė p and  p are obtained from observer (68), then the closed- −β̇d (t 1 ) = M t t 1 <  (87)
d f + 0.25dr t =t 1
loop system rendering by controller (82) will converge asymptotically to
the origin, i.e., the tracking error e p → 0 and ė p → 0 as t → ∞. Therefore, we obtain

The proof of Theorem 5 is presented in Appendix A. t 1 > 0.5π (d f + 0.25dr )/ ( v id + V β Zd ) tan αmax − V β Xd t =t
1

Controller design for tilting dynamics (88)

During the hover-to-level transition, we expect that the flight For tilting dynamics (64), let e β = β − βd , ė β = β̇ − β̇d , the system
speed attains a sufficient large value such that the fixed wings can
error is ë β = − J 4−1 M β − β̈d . The controller M β can be selected as
provide the sufficient lift force. The tilt angle β is required to be
tilted from zero degree to π /2. On the other hand, during the
M β = J 4 (k5 e β + k6 ė β − β̈d ) (89)
level-to-hover transition, we expect the flight speed attain zero,
and hover be implemented. The tilt angle β is required to be tilted where k5 , k6 > 0 are constant. The closed-loop system is
from π /2 degree to zero. Thus, the bounded controllers should be
designed to implement these tilting actions during mode transi- ë β = −k5 e β − k6 ė β (90)
tions. In following, we firstly give the desired tilting actions for
Selecting the Lyapunov function be V a = k5 e β e β /2 + ė β ė β /2, we
T T
hover-to-level and level-to-hover transitions, respectively.
can obtain that e β → 0andė β → 0 as t → ∞.
i) Tilt for hover-to-level transition: (β, β̇, β̈ ) is designed as:
⎧
⎨ Mt , 0 ≤ t ≤ t1 6. Computational analysis and simulation experiments
β̈d = − M t , t 1 < t ≤ 2t 1 ,
⎩ In this section, simulation results of mode transitions are pre-
0, t > 2t 1
⎧ sented in order to observe the performance of the proposed air-
⎨ Mt t , 0 ≤ t ≤ t1 craft model and control law. We consider two cases of hover-to-
β̇d = M t t 1 − M t (t − t 1 ), t 1 < t ≤ 2t 1 level and level-to-hove transitions. The parameters used for the
⎩
0, t > 2t 1 aircraft model are given in Table 1. These values are based on the
previous modeling, observer and control law design in this paper.
⎧ i) In the first simulation part, the hover-to-level transition
⎨ 0.5M t t 2 , 0 ≤ t ≤ t1
is considered. The desired hover-to-level transition can be de-
βd = 0.5M t t 12 + M t t 1 (t − t 1 ) − 0.5M t (t − t 1 )2 , t 1 < t ≤ 2t 1
⎩ scribed as follow: The aircraft starts its mission in hover at the
M t t 12 , t > 2t 1 desired height Z d = 100 m, and its desired starting position is
(83) ( X d (0), Y d (0)) = (0, 0); The forward flight speed changes from
 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492 485

Table 1 flight speed V = 100 m/s at the desired height Z d = 100 m,

Parameters of aircraft model and flight. and its desired starting position is ( X d (0), Y d (0)) = (0, 0); The
Parameter Value Parameter Value forward flight speed changes from 100 m/s to zero by tilt-
m 3313 kg g 9.8 m/s2 ing the rotors synchronously; During the mode transition, the
me 195 kg
√ ρe 6 height is kept invariant. The desired attitude angle is fixed at
ρmT 2 DLQT 60 kg/m2 (φd , θd , ψd ) = (0◦ , 0◦ , 0◦ ). We take the following initial conditions
R 2.0966 m p 4 ( X (0), Y (0), Z (0), φ(0), θ(0), ψ(0)) = (1, −1, 102, 5◦ , 3◦ , −5◦ ),
ρb 13 σ 0.1
Sw 43.7500 m2 S front 12.3458 m2
( T 1 (0), T 2 (0), T 3 (0), T 4 (0)) = (1917N , 1917N , 1917, 1917N ). The
S ft 4.3795 m2 l1 4.0900 m uncertainties in the aircraft dynamics are the same as in hover-
l2 2.8050 m l3 5.6800 m to-level transition. The results in Fig. 8 illustrate the performance
l4 3.4900 m b 0.1613 m of the level-to-hover transition.
a∞ 0.012 df 2.0000 m
Although the uncertainties exist in the dynamics equations of
dr 1.0445 m S ft 4.3795 m2
CDf 0 0.008 Cr 0.15
the tilt-rotor aircraft, the presented controller based on the finite-
Cf 0.5 Af 4 time convergent observer drive the aircraft to the desired trajec-
αmax 25◦ Aw 12 tories during the two mode transitions. We can find out that the
S ri 21.8750 m2 C w0 0.32 thrust (1917N) generated by each rotor during forward flight mode
C wα 0.5 C D w0 0.008
is far smaller than that (8117N) in hover. Therefore, under the
C wδ 0.15 Jr 8.5 m2 kg
J1 220 m2 kg J2 220 m2 kg same cruise speed, the presented tilt-rotor aircraft can save much
J3 400 m2 kg J4 50 m2 kg energy than the conventional helicopters. It can increase the cruise
θd 5◦ ρ 1.225 kg/m3 time and flying range. Moreover, the computational analysis and
dxd /dt 100 m/s βw 45◦ simulations exhibit the agile maneuverability of the presented tilt-
k p1 6 k p2 11
rotor aircraft with the simple control algorithm.
k p3 6 ka1 6
ka2 11 ka3 6
k1 5 k2 5 7. Conclusion
k3 2.5 k4 4.5
k5 2.63 k6 4.55
t1 5s ϕ∗ −7◦
In this paper, a novel 6-DOF model of large-size tilt-rotor air-
craft is presented. Not only the proposed aircraft can hover, take
off and land vertically, but also the high-speed forward flight can
zero to 100 m/s by tilting the rotors synchronously; During the be implemented. The aerodynamics of mode transition with rotors
mode transition, the height is kept invariant. The desired at- tilting is analyzed. During mode transitions, using the blade ele-
titude angle is fixed at (φd , θd , ψd ) = (0◦ , 5◦ , 0◦ ). We take the ment method, the mathematical modeling is proposed for the fol-
following initial conditions ( X (0), Y (0), Z (0), φ(0), θ(0), ψ(0)) = lowing expressions: rotor thrust and blade pitch angle, drag torque
(1, −1, 102, 5◦ , 3◦ , −5◦ ), ( T 1 (0), T 2 (0), T 3 (0), T 4 (0)) = (8117N , and blade pitch angle. Furthermore, using the Lyapunov function
8117N , 8117N , 8117N ). The uncertainties in the aircraft dynam- method, a finite-time convergent observer is designed to recon-
ics are assumed as follows: struct the unknown states and uncertainties in the aircraft system.
⎡ ⎤ ⎡ ⎤ Finally, the presented switched logic controller based on the finite-
 p1 ¯1
 time convergent observer can drive the aircraft to implement mode
 p = ⎣  p2 ⎦ = ⎣  ¯2⎦ transitions. Although the uncertainties exist in the dynamics of the
 p3 ¯3
 tilt-rotor aircraft during mode transitions, the strong stability and
⎡ ⎤ agile maneuverability are exhibited for the presented tilt-rotor air-
2 exp(−2t ) sin(3t ) + exp(−t ) cos(t )
craft. Our future work is to implement a prototype of the presented
= 50 ⎣ exp(−t ) sin(3t ) + 2 exp(−0.5t ) cos(t ) ⎦ ,
tilt-rotor aircraft.
0.5 exp(−t ) sin(3t ) + 3 exp(−2t ) cos(t )
⎡ ⎤
¯
4 Conflict of interest statement
a = ⎣  ¯5⎦
¯6 There is no conflict of interest for this paper.
⎡ ⎤
0.5 exp(−2t ) sin(3t ) + 0.8 exp(−t ) cos(t )
Appendix A
= 20 ⎣ 0.5 exp(−t ) sin(3t ) + 0.5 exp(−0.5t ) cos(t ) ⎦
2 exp(−2t ) sin(3t ) + 0.5 exp(−t ) cos(t )
Proof of Theorem 1. The mass flow rate, ṁ, through the actuator
The results in Fig. 7, obtained by considering the complete mode of disk is ṁ = ρ AU!
, where U is the resultant velocity at the disk and
the aircraft, illustrate the performance of hover-to-level transition: is given by U = V β2 X + V β2 Y + ( V β Z + v i )2 . The application of the
Fig. 7(a) presents the position in X direction; Fig. 7(b) describes conservation of momentum in a direction normal to the disk gives
the velocity in X direction; Fig. 7(c) describes the position in Y !
direction; Fig. 7(d) presents the velocity in Y direction; Fig. 7(e) T i = 2ρ A v i V β2 X + V β2 Y + V β2 Z + v 2i + 2V β Z v i (91)
describes the position in Z direction; Fig. 7(f) describes the veloc-
ity in Z direction; Figs. 7(g) and 7(h) describe the roll angle and Therefore,
roll rate, respectively; Figs. 7(i) and 7(j) describe the pitch angle
    2
and pitch rate, respectively; Fig. 7(k) and (l) describe the yaw an- v 4i + 2V β Z v 3i + V β2 X + V β2 Y + V β2 Z v 2i = T i / 2ρπ R 2 (92)
gle and yaw rate, respectively; Figs. 7(m)–7(o) show the tilt angle,
tilt rate and tilt acceleration, respectively; Fig. 7(p) shows the four For Eq. (92), Newton–Raphson iteration method can be adopted
thrusts generated by the quad rotors. to obtain v i . Especially, the sideslip can be ignored during mode
ii) A second part, we ran simulations for level-to-hover tran- transition, i.e., V β Y = 0. Therefore, Eq. (92) can be rewritten as
sition. The desired level-to-hover transition can be described as     2
follow: The aircraft starts its mission in forward flight with the v 4i + 2V β Z v 3i + V β2 X + V β2 Z v 2i = T i / 2ρπ R 24r (93)
486 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492

Fig. 7. Hover-to-level transition.

X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492 487

Fig. 7. (continued)
488 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492

Fig. 8. Level-to-hover transition.

X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492 489

Fig. 8. (continued)
490 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492

Normalize V β X , V β Z and v i by the mean induced velocity at hover Then from Eqs. (1), (25), (26) and (104), we obtain
vh
! 2π1
  p 1  
vh = T i / 2ρπ R2 , V β Xh = V β X / v h , C Ti = a∞ ϕ∗ W̄ x2 − W̄ x W̄ z b̄dr̄dγ
2π π
V β Zh = V β Z / v h , v ih = v i / v h (94) 0 0
  
1 p 3 3
For rotor flows outside the vortex-ring state, (2V β Zh + 3)2 + = b̄1 a∞ ϕ7i 1 + μ + μ 2
x
2
y
V β2 Xh > 1, the normalized ideal induced velocity can be deter-
3π 2 2
mined from the momentum theory [28,29]   3
+ 0.05 − 0.3μ2x − 0.3μ y ϕ∗ − ( v̄ i + μz ) 2
  
2
v 4ih + 2V β Zh v 3ih + V β2 Xh + V β2 Zh v 2ih = 1 (95) 3
− μ y ρβ̇ (105)
Inside the vortex-ring state (2V β Zh + 3)2 + V β2 Xh ≤ 1, the induced 4
velocity can be determined from an approximation proposed by From Eq. (105), ϕ7i can be carried out shown in Eq. (29). Especially,
Johnson [30]: we suppose the sideslip can be ignored during mode transition, i.e.,
  V β Y = 0. Therefore, we obtain μ y = 0. Consequently, we obtain
v ih = V β Zh 0.373V β2 Zh + 0.598V β2 Xh − 1.991 (96)
  
1 p 3  
In forward flight, the blade element velocity components are C Ti = b̄1 a∞ ϕ7i 1 + μ2x + 0.05 − 0.3μ2x ϕ∗
periodic at the rotor rotational frequency. As for the hover case, 3π 2
there is an in-plane velocity component because of blade ro-

3
tation about the rotor shaft, but now there is a further free- − ( v̄ i + μz ) (106)
2
stream (translational) part and tilt rotor velocity such that W x =
r Ω + V β X sin γ + V β Y cos γ , and the velocity perpendicular to the it follows that
disk can be written as W z = v i + V β Z + (dβ/dt )r cos γ . Therefore,  
 3C T i π 3  
W = W x2 + W z2 ≈ W x . The resultant incremental lift dL per unit = ϕ7i 1 + μ2x + 0.05 − 0.3μ2x ϕ∗
span on the blade element is given by pb̄1 a∞ 2
  3
dL = 0.5ρ W 2 bC l dr = 0.5ρ ba∞ ϕ∗ W x2 − W x W z dr (97) − ( v̄ i + μz ) (107)
2
and the incremental drag is Then,
 
2
dD = 0.5ρ W bC d dr ≈ 0.5ρ W x2 bC d dr (98) 3 3C T i π  
ϕ7i 1 + μ2x = + 0.3μ2x − 0.05 ϕ∗
where C l is the lift coefficient, C d is the drag coefficient, a∞ is 2 pb̄1 a∞
the slope of the lift curve, b is the blade chord, α∗ is the angle of 3
attack of the blade element, and dr is the incremental radial dis- + ( v̄ i + μz ) (108)
2
tance. Therefore, for a blade differential element, the incremental
lift force dT i is and Eq. (30) holds. This concludes the proof. 2

dT i = dL cos β∗ − dD sin β∗ (99) Proof of Theorem 2. From Eqs. (26), (35) and (36), we obtain
and the drag torque gives 2π1  
pa∞ 1 Cd
dQ i = (dL sin β∗ + dD cos β∗ )r (100) CQi = W̄ x2 − W̄ z2 + W̄ x W̄ z ϕ∗ b̄r̄dr̄dγ (109)
2π π a∞
0 0
The relative inflow angle at the blade section is cos β∗ = W x / W ,
sin β ∗ = W z/ W . Therefore, The section profile drag coefficient, C d , is assumed to be constant
C d0 , then after integrations we obtain Eqs. (37) and (38). This con-
dT i = dLW x / W − dDW z / W (101) cludes the proof. 2

dQ i = (dLW z / W + dDW x / W )r (102) Proof of Theorem 3. (1) For position dynamics, let
The total trust of one rotor is equal the number of blades (p) times ¯1
e 13 = x13 − m−1 
     
e 11 = x11 − X , e 12 = x12 − Ẋ ,
the average lift per bade:
 
e 21 = x21 − Y ,
 
e 22 = x22 − Ẏ ,
 
¯2
e 23 = x23 − m−1 
2πR
T i = p (2π )−1 (dT /dr )drdγ (103)
 
e 31 = x31 − Z ,
 
e 32 = x32 − Ż ,
 
¯ 3 (110)
e 33 = x33 − m−1 
0 0 The system error between (68) and (62) is
The second term in (101) is very tiny with respect to the first term,
e˙ i1 = e i2 − k p1 |e i1 |2/3 sign( e i1 )
   
thus, we obtain dT i ≈ dLW x / W ≈ dL. Therefore,
e˙ i2 = e i3 − k p2 |e i1 |1/3 sign( e i1 )
   
2πR
p dL
Ti = e˙ i3 = −k p3 sign( e i1 ) − δ 
¯i
 
drdγ (111)
2π dr
0 0 ¯ i = d(
where, δ  ¯ i )/dt, i = 1, 2, 3.
2πR Select the Lyapunov function be
p ρ  
= a∞ ϕ∗ W x2 − W x W z bdrdγ (104)
2π 2 V i = ξ T P pξ (112)
0 0
 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492 491

T  
ë p = −k3 e p − k4 ė p + m−1 ( p −  p )
   
where ξ = |e i1 |1/3 sign(e i1 ) e i2 e i3 , and P p is a positive def- (124)
inite and symmetrical matrix with the following form:
⎡ ⎤ We rewrite Eq. (124) as
2k p3 + k2p1 + k2p2 −k p1 −k p2
1 
ë p = −k3 e p − k4 ė p + k4 (ẋd − x2 ) + m−1 ( p −  p ).

Pp = ⎣ −k p1 2 0 ⎦ (113) (125)
2
−k p2 0 2 For t ≥ t s , the system error can be written as ë p = −k3 e p − k4 ė p .
The Lyapunov function V i is to study the stability of various dif- Selecting the Lyapunov function be V p = k3 e Tp e p /2 + ė Tp ė p /2, we
ferential equations and systems. Differentiating V i with respect to can obtain that e p → 0 and ė p → 0 as t → ∞. This concludes the
time yields proof. 2
 1/2  1/2
V̇ i ≤ − c p / λmax { P p } Vi (114) References
where c p is a positive constant. From the definition of finite-time [1] T. Oktay, C. Sultan, Modeling and control of a helicopter slung-load system,
stability [32,33], there exists a time t s > 0, for t ≥ t s , the system Aerosp. Sci. Technol. 29 (1) (2013) 206–222.
(111) is finite-time convergent. Alternatively, the convergent re- [2] C.C. Luo, R.F. Liu, C.D. Yang, Y.H. Chang, Helicopter H∞ control design with
sults of system (111) can be proved by high-order sliding mode robust flying quality, Aerosp. Sci. Technol. 7 (2) (2003) 159–169.
[3] X. Wang, J. Liu, K. Cai, Tracking control for a velocity-sensorless VTOL aircraft
theory [34].
with delayed outputs, Automatica 45 (12) (2009) 2876–2882.
(2) For attitude dynamics, let [4] X. Wang, Takeoff/landing control based on acceleration measurements for VTOL
aircraft, J. Franklin Inst. 350 (10) (2013) 3045–3063.
e 42 = z12 − J 1−1 
¯4
   
e 41 = z11 − φ̇, [5] H.L. Tinger, Analysis and application of aircraft departure prediction criteria
to the AV-8B Harrier II, in: AIAA Atmospheric Flight Mechanics Conference,
e 52 = z22 − J 2−1 
¯5
   
e 51 = z21 − θ̇, Monterey, CA, Technical Papers, 17–19 Aug. 1987, pp. 343–352.

e 62 = z32 − J 3−1 
¯6
    [6] H. Powrie, A. Novis, Gas path debris monitoring for F-35 Joint Strike Fighter
e 61 = z31 − ψ̇, (115) propulsion system PHM, in: 2006 IEEE Aerospace Conference, Big Sky, MT, July
2006.
The system error between (70) and (63) is
[7] J.G. Leishman, Principles of Helicopter Aerodynamics, second edition, Cam-

e˙ i1 = e i2 − ka1 |e i1 |1/2 sign( e i1 )

   bridge University Press, 2006.
[8] L.R. Meakin, Moving body overset grid methods for complete aircraft tiltrotor

e˙ i2 = −ka2 sign( e i1 ) − δ 
¯i
  simulations, in: 11th AIAA Computational Fluid Dynamics Conference, Orlando,
(116) FL, Technical Papers, Pt. 2, 6–9 July 1993, pp. 576–588.
¯ i = d(
where δ  ¯ i )/dt, i = 4, 5, 6. Select the Lyapunov function be [9] D. Wyatt, Eagle Eye Pocket Guide, Bell Helicopter Textron Inc, June 2004,
printed in USA.
[10] S.M. Barkai, O. Rand, R.J. Peyran, R.M. Carlson, Modelling and analysis of tilt-
V i = ς T Paς (117) rotor aeromechanical phenomena, in: American Helicopter Society Aerome-
1 chanics Specialists Conference, Bridgeport, October 1995.
where ς = [ |e i1 | sgn( e i1 ) e i2 ]T , and P a is a positive definite and
  
2 [11] J. Escareno, S. Salazar, R. Lozano, Modelling and control of a convertible VTOL
symmetrical matrix with the following form: aircraft, in: Proceedings of the 45th IEEE Conference on Decision & Control,
  Manchester Grand Hyatt Hotel, San Diego, CA, USA, December 13–15, 2006.
1 4ka2 + ka1
2
−ka1 [12] R.H. Stone, Control architecture for a tail-sitter unmanned air vehicle, in: 5th
Pa = (118) Asian Control Conference, 2004, pp. 736–744.
2 −ka1 2 [13] R.H. Stone, The T-wing tail-sitter unmanned air vehicle: from design concept to
research flight vehicle, in: Proceedings of the I MECH E Part G, J. Aerosp. Eng.
Differentiating V i with respect to time yields
(2004) 417–433.
 1/2  1/2 [14] N.B. Knoebel, S.R. Osborne, D. Snyder, T.W. McLain, R.W. Beard, A.M. Eldredge,
V̇ i ≤ − ca / λmin { P } Vi (119) Preliminary modeling, control, and trajectory design for miniature autonomous
tailsitters, in: AIAA Conference on Guidance, Navigation, and Control, Keystone
where ca is a positive constant. Therefore, the system error (116) CO, 2006, paper No. AIAA-2006-6713.
is finite-time convergent. This concludes the proof. 2 [15] Aviation Week & Space Technology “Aerospace SourceBook 2006”.
[16] AERL Accomplishment Report, May 2004.
[17] Office of the Secretary of Defense, Unmanned aerial vehicles roadmap
Proof of Theorem 4. In the light of Theorem 3, for t ≥ t s , the ob-
  2002–2007, December 2002.
servation signals z2 = ẋa , a = a . Considering controller (72), the [18] D.J. Taylor, M.V. Ol, T. Cord, Skytote advanced cargo delivery system, in:
closed-loop system is AIAA/ICAS International Air and Space Symposium and Exposition: The Next

100 Years, 2003.

ëa = −k1 ea − k2 ėa + J −1 (a − a ) (120) [19] X. Wang, H. Lin, Design and control for rotor-fixed wing hybrid aircraft, in:
Proceedings of the IMechE Part G, J. Aerosp. Eng. 225 (7) (2011) 831–847.
Eq. (120) is written as [20] E. Cetinsoy, S. Dikyar, C. Hancer, K.T. Oner, E. Sirimoglu, M. Unel, M.F. Aksit,
Design and construction of a novel quad tilt-wing UAV, Mechatronics 22 (2001)

ëa = −k1 ea − k2 ėa + k2 (ẋa − z2 ) + J −1 (a − a )
 723–745.
(121)
[21] C. Hintz, C. Torno, L.R. Garcia Carrillo, Design and dynamic modeling of a rotary
Therefore, for t ≥ t s , the system error can be written as wing aircraft with morphing capabilities, in: 2014 International Conference on
Unmanned Aircraft Systems (ICUAS), 27–30 May 2014, pp. 492–498.
[22] C.C. Wang, H.T. Yau, Nonlinear dynamic analysis and sliding mode control for a
ëa = −k1 ea − k2 ėa (122) gyroscope system, Nonlinear Dyn. 66 (1–2) (2011) 53–65.
[23] X. Wang, B. Shirinzadeh, M.H. Ang, Nonlinear double-integral observer and
Selecting the Lyapunov function be
application to quadrotor aircraft, IEEE Trans. Ind. Electron. 62 (2) (2015)
1189–1200.
V a = k1 eaT ea /2 + ėaT ėa /2 (123) [24] X. Wang, B. Shirinzadeh, Nonlinear multiple integrator and application to air-
craft navigation, IEEE Trans. Aerosp. Electron. Syst. 50 (1) (2014) 607–622.
we can obtain that ea → 0 and ėa → 0 as t → ∞. This concludes [25] X. Wang, B. Shirinzadeh, High-order nonlinear differentiator and application to
the proof. 2 aircraft control, Mech. Syst. Signal Process. 46 (2) (2014) 227–252.
[26] M. Knight, R.A. Hefner, Static thrust of the lifting airscrew, NACA TN 626.
[27] R.W. Prouty, Helicopter Performance, Stability and Control, Krieger, New York,
Proof of Theorem 5. In the light of Theorem 3, for t ≥ t s , the ob-
  1995.
servations x2 = ẋ p ,  p =  p hold. Considering controller (82), the [28] W.Z. Stepniewsky, Rotor–Wing Aerodynamics, Basic Theories of Rotor Aerody-
closed-loop system is namics, Dover, New York, 1984.
492 X. Wang, L. Cai / Aerospace Science and Technology 47 (2015) 473–492

[29] W. Johnson, Helicopter Theory, Princeton Univ. Press, Princeton, NJ, 1980,
p. 282, 283.
[30] W. Johnson, Helicopter optimal descent and landing after power loss, NASA TM
73244, May 1977.
[31] G.R. Gress, A dual-fan vtol aircraft using opposed lateral tilting for pitch con-
trol, in: American Helicopter Society 59th Annual Forum, Phoenix, Arizona,
May 2003.
[32] S.P. Bhat, D.S. Bemstein, Finite-time stability of continuous autonomous sys-
tems, SIAM J. Control Optim. 38 (3) (2000) 751–766.
[33] X. Wang, H. Lin, Design and frequency analysis of continuous finite-
time-convergent differentiator, Aerosp. Sci. Technol. 18 (1) (2012)
69–78.
[34] A. Levant, High-order sliding modes, differentiation and output-feedback
control, Int. J. Control 76 (9/10) (2003) 924–941.

View publication stats