You are on page 1of 4

SICE-ICASE International Joint Conference 2006

Oct. 18-2 1, 2006 in Bexco, Busan, Korea

Minimum Energy Bang-Bang Control for Vertical Launching Missiles


Sungjin Cho, Young-In Lee and Ickho Whang
3-1-3, Agency for Defense Development, Daejeon, Korea
(Tel: +82-42-821-441 1; E-mail: sjchogadd.re.kr)

Abstract: In order to pass over large uncertainties as quickly as possible in initial launching phase of vertical launching
missile systems, bang-bang control strategy is widely used to produce high pitch over maneuver. Once fast pitch down
is achieved, conventional PID compensator can be used for fine pitch regulation. The new controller combining open
loop control and closed loop control is designed through two steps. Firstly, PI attitude controller combining with
conventional rate compensation is designed. Secondly, the optimal switching time of bang-bang control is determined
in the sense of minimum energy. Linear and nonlinear 3-DOF simulation results show that the proposed controller is
very effective for fast pitch down of vertical launching missiles.

Keywords: Minimum Energy Control, Bang-Bang Controller, PID Controller, Vertical Launching Missiles

1. INTRODUCTION In equation (1), m is the mass of missile and I, is


Vertical launching missiles experience the rapid the moment of inertia of missile. g is gravity. u and
change of attitude to reach the desired velocity and w are velocities in body axis X and Z. q is a pitch
attitude in the initial flight phase[1]. In the high rate and o is a pitch angle. F F and MYA are
pitch-over maneuver stage, bang-bang control is aerodynamic forces and moment. T is a thrust force
effective for turning the vertical attitude into horizontal and d is a distance between center of gravity and
attitude. However, when open loop bang-bang controller
is switched to closed loop attitude controller, control center of thrust force. 5 is the control input of missile.
input of the attitude controller can be excessive due to Linearized missile model of nonlinear model can be
unsatisfactory attitude at final time of bang-bang represented by
control.
This paper presents a minimum energy control for the xg = Agxg + Bgu (2)
strategy of optimal bang-bang control followed by PI
attitude controller combining with rate compensation. where,
The optimal bang-bang controller plays a role in ZW Zq + Uo -g sinO Z w .
approaching the desired attitude within a restricted time. Ag= MWv Mq 0 Bg = MJ xg =
Also, this strategy provides an effective way to combine 0 1
bang-bang control with conventional PID compensator.
The paper is organized as follows. Firstly, nonlinear
and linearized models are described. Secondly, In equation (2), u and 0 are operating points of
minimum energy bang-bang control design method velocity and pitch angle. z", Zq and M is longitudinal
which includes both open loop control and closed loop derivatives of aerodynamic force and moment and
control is presented. The next section contains linear
and nonlinear 3-DOF simulation results about this Z. and M are control derivatives of thrust force and
method. Finally, some conclusions are provided in the moment. Although equation (2) is generally used for
last section. longitudinal autopilot design, linear missile model
ignoring aerodynamic effect is considered for designing
bang-bang controller and PID controller. The rationale
behind this is that aerodynamic pressure is negligible
2. MISSILE MODELS due to low velocity of missile during working time of
Nonlinear 3-DOF model considered in this paper can bang-bang control. However, as velocity builds up,
be found in [1, 2]. aerodynamic effect is increased. Therefore, the
proposed PID controller structure is selected for coping
with this aerodynamic uncertainty.
m(t qw) =
mg sin 0+ Fx, Tcosd5
+ +

(1) The model is described by


m(w qu)
- = mg cos 0 + F,, - T sin
Iy4=Mya -Tdsind x ap =AapXap +B ap U (3)

89-950038-5-5 98560/06/$10 2006 ICASE


105
where, cost function jl. Equation (7) which means closed loop
IO UO, -g sin 0, Z. w control input is concerned with cost function j2. In
Aap L 0 J Bap LMaJ Xap q
equation (7), the state vector xa includes a new state,
77, which is an integrator state of attitude controller.
Based on the reasoning used for the simplification of And then, closed loop system matrix can be represented
(3), in initial phase, the effect of w on pitch attitude by
dynamics can be negligible. Accordingly, it is possible
to consider only the following simple pitch dynamics of A= Aa +BaK (8)
(4) for the pitch attitude regulation. The model is
described by where,
FO 0 0 MFM
x =Ax+Bu (4) A,= I1 ° ° Ba= °
0 1 0 0O
where,
A= 0[ B= M.5
X=[q]. By using equation (6), (7) and (8), equation (5) is
derived by

3. MINIMUM ENERGY BANG-BANG (rUmaxdt + xa


f (t)KTKxa(t)dt
CONTROL [ ( t2)K KMC (t-t2 )Xa (t2 )dt
UMax (t2 to) + a f Xa ( (2
3.1 Optimal Bang-Bang Controller Design Umax (t2 to) + axjj (12) f 4D[ (')KTK4I (r)dr] Xa (t2 )
The working time of bang-bang controller is of
importance in the two viewpoints. Firstly, the (9)
controller's working time plays a vital role in velocity where, (t)=eAt.
46

build-up. In order to obtain missile's desired velocity,


the gravity effect, major factor of drag, must be Total cost function is
decreased by changing 90 deg of pitch angle to desired
angle for minimum time. Secondly, after the bang-bang j=U2 ( t2 -to)+aXT (t2)WX, (t2)
max a
(10)
controller's working is over, missile's attitude has effect
on the performance of closed loop PID attitude
controller. If the attitude of missile is not approached to where, W= f D[T (r)KTK(D,(r)dr.
desired attitude through bang-bang control, excessive
control input diminishes the performance of the attitude In here, W is a constant value. In order to calculate
controller. state vector xa (t2), state variables at t=t2 are defined
The cost function is considered as
by
j = J +J2
(5) [X(t2) 7]T (1 1)
f~ u2 dt+aj7u~2dt
Xa(t2) =
12
=
By using Equation (6), (7), state vector x(t,) is derived
According to the working time of a proposed by
controller, the control input u of cost function j is
defined by
x(t2 = ztX0 +, 4(2- )Bu] dz-
-

0, t<to, 2)= (D(t2 )X(0) +0


FL(DJ (t2-zdzJo@t-) 8a
(6) = D(t2 )X(0) + fD ( I(t )dz- 2I(t )d] Bumax
u= Umax, to < t < ti,
-umax, tl < t < t2
u =Kxa to <t<o0 (7) (12)
where, In equation (12), note that iD(t) is different from
K =[k, k2 k3j : PI and rate compensator gains q.(t). The definition of iD(t) is as follows.
Xa = [q 0
77]T state vector
4D(t) =eAt (13)
In equation (6), open loop control input is related to

106
The optimal solution of t, and t, is obtained by
solving the equation of partial derivatives of total cost D = (s+a )(s2+27s+2)
+ (18)
function as the following.
Therefore, by using equation (16), (17) and (18), the
controller gains are determined by
dJ
2xT (t2 )Waa (t2
at1 (14)
-2a (O2W Bm
Kp=2q + t2 X KI = a
M K
CO2 KD
D
2 ;no + a,
M

a
(19)
T
X,(t2)
2

Umax +2Xa(t2)W (15) In equation (19), PID gains are selected by design
2 )TWt Ax(t2 BUmaxl specification of controller. Once theses gains are
Umax+2 a,)oW2
=0
determined, each gains are used for equation (7). This
means that W can be obtained in equation (10)

where, through closed loop gains. Also, bang-bang time ti and


ax(t2) =2F(t2 -tl)Bum x(t2) Ax(t2) -BUmax
t2 are given in equation (14) and (15).
1tl
=

at2
a 4. SIMULATION RESULTS
Xa(t2) atl ] Xa(t2) x(t2)
0 In equation (5), weighting factor a is one. The
@(t) =eAt =I+At (A2 =0). optimal solution of ti and t2is as the following table.

The reason why A2 is equal to zero is that the Table 1 Optimal solution and cost
elements of first row are all zeros in the A matrix of Variable t t2 Jn
equation (4). Value 1.86s 2.48s 0.049

For linear and nonlinear 3-DOF simulation, initial


3.2 PID attitude Controller Design pitch angle is considered as 90 deg and pitch angle
An attitude controller structure is depicted in Fig. 1. command is zero deg. Also, during 1sec, control input
holds to zero for avoiding interference with launcher or
its outer structure. Linear simulation results and
nonlinear 3-DOF simulation results are as follows.

Pitch angle
100 -
90 ------- Linear
3
DOF
80-

70-

Fig. 1 PI controller and rate compensator _ 60- .......... ..\..... ........ ....................... . . .

n 50

O 40-
In Fig. 1, the attitude controller is as follows.
10 ......

(16) 30

a (Kp +s )(t -0)+KDq


-10
0 1 2 3 4 5 6

When bang-bang control is finished, parameters of Time(sec)


matrix A in equation (4) will be changed due to the
increase of missile velocity. Therefore, the controller Fig. 2 Pitch angle
structure in Fig. 1, especially integrator, has the
advantage of coping with aerodynamic uncertainty.
Plant transfer function of equation (4) is represented by

-M (17)
35

In order to use pole placement technique, desired


characteristic polynomial is defined by

107
Control input
influenced by aerodynamic effect related to velocity
increase. In Fig. 5, missile trajectory is generated by
Linear
Umax -
3-DOF pitch down through minimum energy optimal
bang-bang controller.

.-!---------------.-----------.--.................
- 5. CONCLUSIONS
A proposed controller is designed by optimal
bang-bang controller and pitch attitude controller. Once
-Umax PID attitude controller is designed by specification of
o 1 2 3 4 5 6
system requirement, bang-bang time t1 and t, are
Time(sec) determined by optimal solutions of two cost functions.
Fig. 3 Control input Through the linear and nonlinear 3-DOF simulation
results, the controller is very effective for high
pitch-over maneuver.
Pitch rate

REFERENCES
-20 -
U
----Linear
~~~~~~~~3-DOF [1] R. Solis, "An Analysis of the Vertical Launch
-40-
Phase of a Missile Concept," AIAA 21" Aerospace
'a4) Sciences Meeting, AIAA-83-0569, pp. 1-15, 1983.
60 [2] J. H. Blakelock, Automatic Control of Aircraft
-80
and Missiles, John Wiley & Sons, Inc., Newyork,
1991.
-100 -....
[3] Arthur L. Greensite, Analysis and design of
-120 -
space vehicle flight control systems, Spartan,
o 1 2 3 4 5 6 Newyork, 1970.
Time(sec)

Fig. 4 Pitch rate

Trajectory

I)

Downrange

Fig. 5 Trajectory
In Fig 2, vertical attitude of missiles is turned into
horizontal attitude through optimal minimum energy
bang-bang control. Especially, nonlinear 3-DOF
simulation results coincide with linear simulation results
for the most part. In Fig 3, simulation results illustrate
two characteristics. Firstly, when closed loop controller
is worked, the control input of linear and nonlinear
simulation reaches positive angle. This means that P
gain is larger than rate compensator gain because of fast
bandwidth. Secondly, since 5 seconds, the control input
of nonlinear 3-DOF simulation which is different from
linear simulation goes away from zero. This is

108

You might also like