04421364
TRANSCRIPT
-
7/30/2019 04421364
1/6
SICE Annual Conference 2007
Sept. 17-20, 2007, Kagawa University, Japan
Autonomous Flight Control System for Longitudinal Motion of a Helicopter
Atsushi Fujimori1 and Kyohei Miura2
1Department of Mechanical Engineering, University of Yamanashi, Kofu, Japan
(Tel : +81-55-220-8195; E-mail: [email protected])2Department of Mechanical Engineering, Shizuoka University, Hamamatsu, Japan
Abstract: This paper presents a flight control design for longitudinal motion of helicopter to establish autopilot techniquesof helicopter-type unmanned aerial vehicles (UAVs). The characteristics of the linearized equation of the helicopter is
changed during a specified flight mission because the trim values of the nonlinear equation are also varied. In this paper,
the longitudinal motion of the helicopter is modeled by a linear interpolative polytopic model whose varying parameter is
the flight velocity. A flight control systems with a gain scheduling state feedback law is designed to stabilize the vehicle
and track a reference which realizes the flight mission. The effectiveness of the proposed flight control system is evaluated
by computer simulation.
Keywords: Autonomous flight control, unmanned aerial vehicles (UAVs), helicopter, gain scheduling.
1. INTRODUCTION
Recently, unmanned aerial vehicles (UAVs) have been
developed for the purposes of scientific observations, de-tecting disasters, surveillance of traffic and army objec-
tives [1]-[3]. This paper presents a flight control design
for a small autonomous helicopter to give insights for de-
veloping helicopter-type UAVs.
A Helicopter is generally an unstable aircraft. Once
it is stalled, it is not easy to recover its attitude. A con-
trol system is therefore needed to keep the vehicle stable
during flight [4], [5]. This paper presents a flight control
design for longitudinal motion of helicopter to establish
autopilot techniques of helicopters. The flight mission
considered in this paper is that a helicopter hovers at a
start position, moves to a goal position with keeping aspecified cruise velocity and hovers again at the goal. The
characteristic of the linearized equation of the helicopter
is changed during this flight mission because the trim val-
ues of the nonlinear equation are also varied. In this pa-
per, the flight control system is designed as follows. The
longitudinal motion of a helicopter is first modeled by a
linear interpolative polytopic model whose varying pa-
rameter is the flight velocity. A flight control systems
with a gain scheduling (GS) state feedback law [6] is de-
signed to stabilize the vehicle and track a reference which
realizes the flight mission. The effectiveness of the pro-
posed flight control system is evaluated by computer sim-
ulation using Matlab / Simulink.
2. EQUATION OF LONGITUDINALMOTION OF HELICOPTER
Figure 1 shows a helicopter considered in this paper.
(x,y,z) represent the body-fixed-axes whose origin is lo-cated at the center of gravity of the vehicle. The forward
velocity is V whose x- and z-axes elements are u and w,respectively. The main rotor produces the aerodynamic
force which consists of the lift L and the thrust T. It de-pends on the tilting angle of the control plane. The rotor
blades are regarded as cantilevered elastic beams. The re-
stored force is modeled by a rotational spring attached at
tip path plane
horizontal plane
control plane
x
z
V
u wf
f
Vd
c
x
a1
c
Fig. 1 Helicopter in forward flight
the rotor hub.
The longitudinal motion of helicopter consists of the
transitional motion with respect to x- and z-axes and therotational motion around y-axis. It is represented as thefollowing equations:
m(u + qw) = Xmg sin (1)
m( w qu) = Z+ mg cos (2)
Iyy q = M (3)
where m and Iyy are respectively the mass and the mo-ment of inertia of the vehicle. g is the gravity accelera-tion. The external forces X, Z and the moment M aregiven by
X = T sin(c a1)Du
V(4)
Z = T cos(c a1)Dw
V(5)
M = T hR sin(c a1) (6)
where D is the drag. hR is the distance of hub from thecenter of gravity. The induced velocity vi through the
rotor is approximated by a first-order system. Define the
PR0001/07/0000-2258 400 2007 SICE- 2258 -
-
7/30/2019 04421364
2/6
0 20 40 600
20
40
60u [m/s]
0 20 40 6010
8
6
4
2
0w [m/s]
0 20 40 6010
8
6
4
2
0
f
[deg]
V [m/s]0 20 40 60
0
0.02
0.04
0.06
i
V [m/s]
(a) States in trim xp
0 20 40 606
7
8
9
10
0
[deg]
V [m/s]0 20 40 60
0
1
2
3
4
5
c
[deg]
V [m/s]
(b) Inputs in trim up
Fig. 2 Trim values with respect to forward velocity.
state and the input vectors as
xp= [u w q f i]
T 5 (7)
up
= [0 c]T
2
, (8)
where 0 and c are the collective and the cyclic pitchangles. The equation of longitudinal motion of helicopter
is then written as
xp = fp(xp, up) (9)
Equation (9) is referred as the nonlinear plant Pnl here-after.
Letting xe and he be horizontal and the vertical posi-tions of the helicopter from the start, they are given by
xe = u cos f + w sin f (10)
he = u sin f w cos f. (11)
Defining p as p= [xe he]T, they are compactly given
as
p = gp(xp). (12)
Since this paper considers hovering and forward flight,
the trim condition is given in level flight. Letting xpand up be the state and the input in trim, respectively,f(xp, up) = 0 holds. Figure 2 shows variations ofxpand up with respect to the flight velocity V. Almost all
trim values are changed in the range ofV [0, 60] [m/s].
Vc1
0tc1 tc3tc2 tc4 t
Vc2
Vr
tc5
Fig. 3 Flight velocity profile.
Pnl[E -F]KoutKpgp
Kin zpupvzr p
rxp
+
--+++
Fig. 4 Flight control system.
3. CONSTRUCTION OF FLIGHTCONTROL SYSTEM
Let the start position be the origin of the coordinates
(xe, he). The flight mission is to navigate the helicopterfrom the start (0,0) to the goal, denoted as (xr, hr), withkeeping attitude stable. To design a control system, the
followings are assumed to be satisfied:
(i) The motion in y-axis direction is not taken into ac-count.
(ii) xp is measurable.(iii) The trim values xp and up are known in advance.
To realize the flight mission, this paper constructs a track-
ing control system whose controlled variable is the flight
velocity. The flight region is divided into six phases as
shown in Fig. 3. They are referred as follows.
1. 0 t < tc1: initial hovering phase2. tc1 t < tc2: acceleration phase3. tc2 t < tc3: cruise phase4. tc3 t < tc4: deceleration phase5. tc4 t < tc5: low speed phase6. tc5 t: approach phase
Until the low speed phase, the reference of the flight ve-
locity is given by Vr shown in Fig. 3. In this paper, the
total time of the flight is not cared. But the integratedvalue ofVr for t [0, tc5] must be less than xr not toovertake the goal before the approach phase. In the ap-
proach phase, the reference is generated to meet the travel
distance p = [xe he]T with r = [xr hr]
T.
Taking into consideration the above, a double loop
control system [7], [8] is applied to design a flight control
system in this paper. It is given in Fig. 4. Pnl representsthe nonlinear helicopter dynamics given by Eq. (9), Kinis the inner-loop controller, Kout is the outer-loop con-troller and Kp is a gain. The controlled variable fromthe initial hovering to the low speed phase is given by
zp
= [u w]T
and its reference is given by zr
= [ur wr]T
.
- 2259 -
-
7/30/2019 04421364
3/6
In the approach phase, another loop is added outside of
(zr zp)-loop, where p= [xe he]
T is the controlled
variable and r= [xr hr]T is its reference.
Kin consists of
Kin = [E F] (13)
where E is a feedforward gain for tracking the reference,while F is a feedback gain for stabilizing the plant. Sincethe trim values are changed as shown in Fig. 2, the char-
acteristics of the linearized plant is also varied. Then, Fis designed by a GS technique in terms of LMI formula-
tion.
The reference zr from the initial hover to the low speedphase is generated by the flight velocity profile shown in
Fig. 3, and zr in the approach phase is derived from thepositional error r p. The switch of the reference isdone at t = tc5.
4. DESIGN OF CONTROL SYSTEM
4.1 Linear interpolative polytopic model
The objective of flight control in this paper is that
the controlled variable is regulated to the specified
trim condition. Linearized models along with the trim
may be therefore used for controller design. Letting
xp(V), up(V) be respectively the state and the input intrim where the flight velocity is V, the perturbed stateand the input are defined as
xp(t)= xp(t)xp(V), up(t)
= up(t)up(V) (14)
The linearized equation of Eq. (9) is then given as
xp(t) = Ap(V)xp(t) + Bp(V)up(t) (15)
Ap(V)=
fp(xp, up)
xTp, Bp(V)
=
fp(xp, up)
uTp.
(16)
Although matrices Ap and Bp are functions with respectto V, it is hard to get their explicit representations be-cause of complicated dependence of V as described inSection 2. Then, Ap and Bp are approximated by inter-polating multiple linearized models in the trim condition.
For the range of the flight velocity V [0, Vu], r points{V1, , Vr}, called as the operating points, are chosenas
0 V1 < < Vr Vu. (17)
The linearized model for V = Vi is a local LTI modelrepresenting the plant near the i-th operating point. Lin-early interpolating them, a global model over the entire
range of the flight velocity is constructed as
xp(t) = Ap(V)xp(t) + Bp(V)up(t)
zp(t) = Cpxp(t) + zp(V)
(18)
where
Ap(V) =r
i=1
i(V)Api, Bp(V) =r
i=1
i(V)Bpi,
Cp =
1 0 0 0 0
0 1 0 0 0
. (19)
i(V) satisfied the following relations.
0 i(V) 1 (i = 1, , r) (20)r
i=1
i(V) = 1 (21)
Equation (18) with Eq. (19) is called as the linear inter-
polative polytopic model in this paper.
4.2 Design ofKin
Under assumption (ii), consider a state feedback law
up(t) = F(V)xp(t) + E(V)v(t) (22)
where v is a feedforward input for tracking zr and is givenby v = zr zp when designing Kin. The closed-loopsystem combining Eq. (22) with Eq. (18) is given by
xp(t) = AF(V)xp(t) + Bp(V)E(V)v(t)zp(t) = Cp(V)xp(t) + Dp(V)E(V)v(t) + zp(V)(23)
AF(V)= Ap(V)Bp(V)F(V).
The steady-state controlled variable is given by
zp() = CpA1
F BpEv + zp. (24)v is then given so as to meet zp() with the reference zr;that is, zp() zr . E is designed as
E = (CpA1F Bp)
1. (25)
Next, F(V) is designed so that the closed loop systemis stable over the entire flight range and H2 cost is glob-ally suppressed [6]. The controlled plant is newly given
by
xp(t) = Ap(V)xp(t) + Bp(V)up(t) + B1(V)w1(t)
z1(t) = C1(V)xp(t) + D1(V)up(t)(26)
where z1 and w1 are respectively the input and the out-put variable for evaluating H2 cost. B1(V), C1(V) andD1(V) are matrices corresponding to z1 and w1. Substi-tuting Eq. (22) without v into Eq. (26), the closed-loopsystem is
xp(t) = AF(V)xp(t) + B1(V)w1(t)
z1(t) = C1F(V)xp(t)(27)
C1F(V)
= C1(V) D1(V)F(V).
- 2260 -
-
7/30/2019 04421364
4/6
In this paper, F(V) is designed so as to minimize theintegration ofH2 cost over V [0, Vu]. That is, theobjective is to find F(V) such that [9].
infF,P,W
Vu0
trW(V)dV subject to
P(V) P(V)B1(V)() W(V)
> 0, (28a)He(P(V)AF(V)) + V dPdV ()
C1F(V) Iq
< 0 (28b)
He(A) is defined as He(A)= A + AT where ()
means the transpose of the element located at the diag-
onal position. P(V) > 0 is the parameter dependentLyapunov variable. To derive finite number of inequal-
ities from Eq. (28), the following procedures are per-
formed. First define new variables X(V)= P1(V) and
M(V)
= F(V)X(V). Ap(V) and Bp(V) are given bythe polytopic forms Eq. (19). B1(V), C1(V), D1(V),X(V), M(V) and W(V) are also given by similar poly-topic forms. Furthermore, dP/dV are approximated as
dP
dV= X1
dX
dVX1 (29)
dX
dV
Xi+1 XiVi+1 Vi
=
XiVi
(30)
Pre- and post-multiplying diag{X, I} to Eq. (28) and us-ing the polytopic forms, the following LMIs are derived
as a sufficient condition.
infMi,Xi,Wi
r1i=1
tr(WiVi) subject to
Xi B1i
() Wi
> 0 (i = 1, , r), (31a)
He(ApiXi BpiMi) XiVi ()
C1iXiD1iMi Iq
< 0,
(31b)
He(Apj Xj Bpj Mj) XiVi ()C1j Xj D1j Mj Iq
< 0,
(31c)
He(ApiXj BpiMj
+Apj Xi Bpj Mi) 2XiVi
()
C1iXj D1iMj
+C1j Xi D1j Mi2Iq
< 0
(31d)
(i = 1, , r 1, j = i + 1), = ,
IfMi, Xi and Wi (i = 1, , r) satisfying the above
LMIs, F(V) is given by
F(V) =
r
i=1
i(V)Mi
r
i=1
i(V)Xi
1. (32)
4.3 Design ofKout
Since v in Eq. (22) is a feedforward input from theflight velocity reference, the tracking error will be oc-
curred by model uncertainties and/or disturbances. Let usevaluate this in the LTI representation. Let Tzpv(s) be atransfer function from v to zp. zp converges to a constantzr if there are no model uncertainties in Tzpv(s) becauseTzpv(0) = I. IfTzpv(0) is varied as Tzpv(0) = I + due to model uncertainties, we have the following steady-
state error:
e0= zr zp() = zr (33)
To reduce the error, a feedback from zp; that is, an outer-loop is added as shown in Fig. 4. The transfer function
from zr to zp is given by
Tzpzr(s) = (I+Tzpv(s)Kout(s))1Tzpv(s)(I+Kout(s))
(34)
The steady-state error is then
e1= zr zp() = zr lim
s0sTzpzr(s)
1
szr
= (I + Tzpv(0)Kout(0))1zr.
(35)
This means that the steady-state error e1 with the outer-loop is reduced by (I + Tzpv(0)Kout(0))
1. Summariz-
ing the above, the design requirements ofKout are givenas follows:
(i) Kout must stabilize Tzpv .(ii) The amplitude of (I + Tzpv(j)Kout(j))1
should be small in the low frequency region.
5. SIMULATION
To evaluate the proposed flight control system, a flight
simulator is built on MATLAB/Simulink. For design and
discussion hereafter, the notations about plant models are
given as follows: Plpv(V) is a linear parameter varying(LPV) model obtained by linearizing Pnl. Ppoly(V) isthe linear interpolative polytopic model given by Eq. (18)
with Eq. (19). Plti(Vd) is an LTI model where the flightvelocity is fixed at Vd.
Two cases of flight control system with respect to Fare compared in simulation. One is that F was designedby GS where the plant model was Ppoly(V). Another isthat F was designed by LQR where the plant model wasPlti(Vd). The former is called as GS-SF, while the latteris called as Fixed-SF. The parameter values of the flight
velocity profile in Fig. 3 were given as follows:
(xr, hr) = (3000, 0) [m],
Vc1 = 50 [m/s], Vc2 = 15 [m/s],
tc1 = 5 [s], tc2 = 30 [s], tc3 = 60 [s],
tc4 = 80 [s], tc5 = 100 [s].
- 2261 -
-
7/30/2019 04421364
5/6
Table 1 Operating points of polytopic models.
Model Vci [m/s] GS-SF
Ppoly1 { 0, 50 } Fgs1
Ppoly2 { 0, 25, 50 } Fgs2
Ppoly3 { 0, 10, 15, 40, 50 } Fgs3
Table 2 Design points of LTI models.
Model Vd [m/s] Fixed-SF
Plti1 0 Ff ix1
Plti2 25 Ff ix2
Plti3 50 Ff ix3
5.1 Evaluation of design models
According to Section 4.1, three linear interpolative
polytopic models were obtained. Table 1 shows the op-
erating points chosen for the models. The design pointsVd of three LTI models are shown in Table 2. The -gap metric is one of criteria measuring the model error
in the frequency domain. It had been introduced in ro-
bust control theories associated with the stability margin
[10]. The -gap metric between two LTI models, P1(s)and P2(s), is defined as
(P1, P2)= (I+P2P
2 )1/2(P1P2)(I+P1P
1 )1/2
(36)
The range is [0, 1]. A large means that themodel error is large. The -gap metric is used forevaluating the model Ppoly(V) and Plti(Vd). Figure 5shows -gap metric between Plti(Vd) and Plpv (V) andbetween Ppoly(V) and Plpv(V). (Plti(V), Plpv(Vd))was rapidly increased when V was shifted from Vd. Onthe other hand, the maximum of(Ppoly(V), Plpv (V))was reduced according to the number of the operating
points. It was seen that Ppoly(V) appropriately approxi-mated Plpv (V) over the entire flight range.
5.2 Design ofF and H2 cost
The design parameters for designing F in Eq. (26)were given as follows.
B1 =
6.039 10.977
154.03 49.188
3.954 7.187
0 0
0.38495 0.12395
,
C1 =
0.001I5
025
, D1 =
052
I2
They were used for both of GS-SF and Fixed-SF. GS-
SF was designed according to Section 4.2, while Fixed-
0 10 20 30 40 500
0.2
0.4
0.6
0.8
1
V [m/s]
gap metric between Plti
and Plpv
Plti1
Plti2
P
lti3
(a) LTI models
0 10 20 30 40 500
0.1
0.2
0.3
0.4
0.5
V [m/s]
gap metric between Ppoly
and Plpv
Ppoly1P
poly2
Ppoly3
(b) polytopic models
Fig. 5 -gap metric
0 10 20 30 40 501.7
1.8
1.9
2
2.1
2.2
2.3
2.4H
2cost
V [m/s]
Ffix3
Fgs1
Fgs2
Fgs3
Fig. 6 H2 cost
- 2262 -
-
7/30/2019 04421364
6/6
0 50 100 15020
0
20
40
60
u[m/s]
GSSF 3
0 50 100 150
30
20
10
0
10
w
[m/s]
0 50 100 1505
10
15
20
25
t [s]
0
[deg]
0 50 100 1502
0
2
4
6
t [s]
c
[deg]
u
ur
w
wr
(a) Controlled variables and inputs
0 20 40 60 80 100 120 1400
1000
2000
3000
GSSF 3
t [s]
x[m]
0 20 40 60 80 100 120 140300
200
100
0
100
t [s]
h[m]
(b) Positions
Fig. 7 Time responses using Fgs3
SF was designed by LQR technique whose weights were
given by CT1 C1 and DT1 D1.
Figure 6 shows the H2 cost of the closed-loop sys-tem which the designed F is combined with Eq. (26).The H2 cost by Ff ix3 was minimized at V = 40 [m/s]which was near the design point Vd = 50 [m/s], but wasincreased in the low flight region. The H2 cost by Ff ix1and Ff ix2 showed the similar result. On the other hand,the H2 cost by Fgs2 and Fgs3 was kept small over theentire flight region. The H2 cost by Fgs1 was small in
the low flight velocity but was increased in the high flightregion.
5.3 Tracking evaluation
The flight mission given in Fig. 3 was performed in
Simulink. Figure 7 shows the time histories of the closed
loop system whose F was Fgs3. The responses byFgs3 showed improved tracking and settling propertiescompared to other cases.
6. CONCLUDING REMARKS
This paper has presented a flight control design for
longitudinal motion of helicopter to establish autopilot
techniques of helicopter-type UAVs. The characteristics
of the linearized equation of the helicopter was changed
during a specified flight mission because the trim values
of the nonlinear equation were also varied. In this paper,
the longitudinal motion of the helicopter was modeled by
a linear interpolative polytopic model whose varying pa-
rameter was the flight velocity. The model error was eval-
uated by the -gap metric. The effectiveness of the pro-
posed flight control system was evaluated by computersimulation using MATLAB / Simulink. The model er-
ror of the polytopic model was smaller than that of LTI
models which were obtained at specified flight velocity.
Flight control systems with GS state feedback showed
better control performances than those with fixed-gain
state feedback. The double loop flight control structure
was useful for performing flight mission considered in
this paper.
REFERENCES
[1] J. R. Wilson, UAV Worldwide Roundup 2007,
Aerospace America, May, 2007, pp. 30-38.[2] J. Langelaan and S. Rock, Navigation of Small
UAVs Operating in Forests, Proc. CD-ROM of
AIAA Guidance, Navigation, and Control Confer-
ence, San Francisco, 2005.
[3] S.-J. Cho, D.-S. Jang and M.-L. Tahk, Application
of TCAS-II for Unmanned Aerial Vehicles, Proc.
CD-ROM of JSASS 19th International Sessions in
43rd Aircraft Symposium, Nagoya, 2005.
[4] G. D. Padfield, Helicopter Dynamics: The Theory
and Application of Flying Qualities and Simulation
Modeling, AIAA, Reston, 1996.
[5] E. N. Johnson and S. K. Kannan, Adaptive Trajec-
tory Control for Autonomous Helicopters, Journal
of Guidance, Control and Dynamics, Vol. 28, No. 3,
pp. 524-538, 2005.
[6] Atsushi Fujimori, Kyohei Miura and Akira Mat-
sushita, Active Flutter Suppression of a High-
Aspect-Ratio Aeroelastic Using Gain Scheduling,
Transactions of JSASS, Vol. 55, No. 636, pp. 34-42,
2007.
[7] A. Fujimori, M. Kurozumi, P. N. Nikiforuk and M.
M. Gupta, A Flight Control Design of ALFLEX
Using Double Loop Control System, AIAA Pa-
per, 99-4057-CP, Guidance, Navigation and Con-
trol Conference, pp. 583-592, 1999.[8] A. Fujimori, P. N. Nikiforuk and M. M. Gupta,
A Flight Control Design of a Reentry Vehicle Us-
ing Double Loop Control System with Fuzzy Gain-
Scheduling, IMechE Journal of Aerospace Engi-
neering, Vol. 215, No. G1, pp. 1-12, 2001.
[9] S. Boyd, L. E. Ghaoui, E. Feron and V. Balakrish-
nan, Linear Matrix Inequalities in System and Con-
trol Theory, SIAM, Vol. 15, 1994.
[10] G. Vinnicombe, Uncertainty and Feedback (Hloop-shaping and the -gap metric), Imperial Col-lege Press, 2001.
- 2263 -