2011 IEEE GCC Conference and Exhibition (GCC), February 19-22, 2011, Dubai, United Arab Emirates

FUZZY SLIDING MODE CONTROL FOR MISSILE AUTOPILOT

DESIGN
Negar Elmi sadr, Hamid Reza Momeni

Faculty of Electrical Engineering, Tarbiat Modares University,Tehran,Iran.

ABSTRACT position vector (of course, this ratio depends on the

missile range, too).
In this paper, we have discussed about the Sliding mode In other words, the accuracy of the hitting a ballistic
Controller (SMC) for missile autopilot. The chattering missile to target is a function of velocity magnitude and
phenomenon has been studied as the main problem of direction vector of it when the engine stops. Since, the
these systems; we use boundary layer method to attack and side slip angles are small; we can control the
overcome this problem. Therefore, the system is robust orientation angles, object or their variation rate in order to
under various uncertainties and external disturbances. control the direction of the velocity vector [3]. The aim of
Also, the stability of the system is proved by the this paper is an autopilot design to control effectively the
Lyapunov theory. In following, Fuzzy control is applied
angle rates of Roll, Pitch and Yaw.
to calculate the controller gain (k), which is responsible
Now, we must find inputs to control the above mentioned
for stability.
outputs. In this regard, we consider three inputs: the
Keywords— Siding mode control, Ballistic missile, angles of the three flaps, i.e. aileron, elevator, and rudder.
Autopilot, Fuzzy control, chattering phenomenon. The paper is organized as follows: plant dynamics is
described in section.2. In section.3, the SMC theory is
explained and in following a fuzzy control is presented to
1. INTRODUCTION determine the control gain. In section.4, we study the
simulation results. And finally, section.5, presents some
Guided missile is a solution for the problem of hitting a conclusions remarks.
specific point at the space called target. Researches in this
field have provided many experimental and theoretical 2. PLANT DYNAMICS
studies with new opportunities in recent decades. The
most obvious, not the easiest method to achieve this goal The nonlinear differential equations for the missile model
is: are obtained as, [4].
1. Specify the immediate position of the moving target F ( x) = m(u& + qw − rv) = Qsc x
along the missile trajectory (navigation). d
2. Justify the missile trajectory based on the position of F ( y) = m(v& + ru ) = QS (c yβ β + c yr r + c yδrδr )
V
the missile as well as the moving object (guidance).
d
3. Conduct the guidance system's commands using a F ( z ) = m( w& − qu ) = QS (c zα α + c zq q + czδeδe) (1)
control link through justification of the flaps angles or V
thrust force depending on the control method (autopilot) L = I x p& = QSD(clδaδa + cl )
[1], [2]. d
Generally, the methods applied to guide short and M = I y q& = QSD(cMα α + cMq q + cMδeδe)
V
medium range missiles toward the target use computer
d
simulations and calculations. Then, it is tried to minimize N = I z r& = QSD(c Nβ + cNr r + c Nδr δr )
deviation from the designated trajectory. These guidance V
systems are generally called preset guidance. In ballistic Where α= Angle of attack, β= Sideslip angle, p= Roll
missiles which pass a short distance of their route from rate, q=Pitch rate, r= Yaw rate, L= Roll moment, F=
dense atmosphere and experience low atmospheric force, IX= Roll moment of inertia, Q= Dynamic pressure,
turbulences at the rest of their way, the control is S= Reference area, D= Reference diameter, δa= Aileron
performed at the early steps of the flight. Then the engine deviation, δe= Elevator deviation, δr= Rudder
stops and a simple throw motion continue under the deviation,V= Speed of missile, Cx= drag coefficient, CY=
gravity force. This motion is a function of the velocity side force coefficient, Cz=Normal force coefficient, Cl
and position vectors of the flying objects at the moment =rolling moment coefficient, CM=Pitching moment
of engine stops. It can be shown that the importance of coefficient .
the velocity vector is hundred times more than the

978-1-61284-119-9/11/$26.00 ©2011 IEEE 453

 3. PROBLEM STATEMENT to reach the sliding mode is equivalent to stabilization of
the sliding surface when s=0.
3.1. Problem Formulation Therefore, consider the positive definite lyapunove
function as
Sliding mode control is widely used in linear and non- 1
V ( t ) = SS T (6)
linear systems in spite of parameter uncertainties and 2
external disturbances. In order to move the trajectories toward a sliding surface,
Consider the multi input-multi output (MIMO) control the time derivative of the function must be negative
system of the form definite.
p
si s&i ≤ −η i si
= f i ( x ) + ∑ g ij ( x )u i + d i (t )
( ni )
xi (7)
j =1 (2) s&i = x&di − λi ei − f i( x) − g i ( x)ui
y i = xi (i = 1,..., p ) Therefore we choose the control law as
)
−1
( ni −1)
are the state vector, f(x)and g(x) are an unknown
ui = gi ( x)(ui − ki sgn( si ). i = 1,2,3
x i ) (8)
continous functions, d is the external disturbance and u is ui = x&di − λi ei − f i ( x ).
the control input. Where k is the gain controller, and by using (7)
determined as
Assumption I: f and g are are divided in two parts as: )
k i ≥ Gi ( Fi + ηi + γ i ) + (Gi − 1) ui (9)
f i ( x) = f i ( x) + Δf i ( x)
(3)
gi ( x) = gi ( x) + Δgi ( x) 3.3. Chattering Elimination
f (x) and g ( x ) are the nominal values of f and g.
Sliding mode control has chattering along the sliding
Δf(x) and Δg(x) are uncertain parts of f(x) and g(x) that surface; in general, chattering must be eliminated for the
are bounded as control to perform properly. This can be achieved by
Δf i ( x ) ≤ Fi ( x ), Δg i ( x) ≤ Gi ( x) using continuous function instead of discontinuous
Assumption II: d i (t ) ≤ γ i ,where γ i are positive control laws inside of boundary layer.
⎧ s
functions.
⎪⎪sgn( φ ) s 〉φ
Let, Define the output, and the input vectors as s
sat ( ) = ⎨ (10)
Y = [ p q r ]T , U = [δa δe δr ]
T
φ ⎪
s
s 〈φ
Therefore, by using (1), the third order system equations ⎪⎩ φ
are obtained as Where, Φ is the boundary layer thickness. And the
⎡ ⎤ boundary layer width is assumed as
⎡ ⎤ ⎢
•
q sDcl / I x ⎥
⎢ p• ⎥ ⎢
⎢ q ⎥ = ⎢q sD(c α + c d ⎥
q) / I y ⎥ + ε =φ (11)
⎢•⎥ ⎢ Mα Mq
V ⎥
(4) λn −1
⎢r⎥ ⎢ d ⎥ 3.4. Fuzzy Control
⎣⎢ ⎦⎥ ⎢ q sD (c Nβ + c Nr r ) / I z ⎥
⎣ V ⎦
⎡q sDclδa / I x 0 0 ⎤ ⎡δa ⎤ Sliding mode with boundary layer may cause steady state
+ ⎢⎢ ⎥ ⎢δe ⎥ error in to the sliding surface. In this section, fuzzy
0 q sDcMδe / I y 0 ⎥⎢ ⎥ sliding mode control is applied to achieve the controller
⎢⎣ 0 0 q sDc Nδr / I z ⎥⎦ ⎢⎣δr ⎥⎦ gain (k). In the fuzzy sliding mode control (FSMC), the
membership functions of the control gain and therefore k
The objective of the research is to determine a robust are updated online and the stability of the system is
control u which guarantees that the tracking error guaranteed. Fuzzy control is an intelligent method to use
converges to zero as t → ∞ . tentative rules to calculate the output variable [6], [7]. In
order to use a fuzzy control, we have to know the domain
3.2. Sliding Mode Control of the parameter, and the fuzzy rule base.
Figure.1 shows the configuration of a fuzzy system.
The design of sliding mode control consists of two steps:
The first step is the selection of a proper sliding surface.
As discussed in [5], the sliding surfaces can be selected as
si = e&i + λi ei i = 1,2,3 (5)
Where, λ ∈ ℜ is positive definite matrix.
n×n
Figure 1. Block diagram for a fuzzy system
In second step, we design the control law that direct the
system states to the sliding surface s(x) =0, the problem Fuzzy rule base consists of linguistic sets in the form of

454
“IF x1 is A1 and … And xN is AN, Then y1 is B1 and … k i max ≤ αλ i
2
(15)
And yN is BN.”
Therefore, we can determine the controller gain on-line
Where xi are the inputs to the fuzzy system, and yi are the
by using a fuzzy algorithm with single fuzzification,
output variables of the fuzzy system. Ai and Bi are fuzzy
product inference engine, and center average
sets. Here, we choose s and s& as the inputs of fuzzy defuzzification.
system to calculate the control gain (k).
The range of s(t) and s&(t ) are covered by seven different 4. SIMULATION RESULTS
fuzzy sets as follows:
We implement the designed controller for Roll, Pitch, and
Yaw channels in the model of 6 freedom degree (6-DOF)
under the presence of disturbances and uncertainties. We
have to choose some commands as the outputs of
guidance system.
Note that the Roll channel command, i.e. pd must be
always zero to prevent the missile's rotation along its
length axis, As you see in Figure.3, Figure.4 and
Figure.5. It reduces the system interference and increases
the accuracy. Therefore, it is possible to control the
Figure 2. Membership functions for s and s& missile's motion in each three channels independently.
Figure.6 and figure.9 Show the tracking of the system
Where, PB= positive big, PM= positive medium, PS= outputs to the unit step function for the sliding mode
positive Small, ZE= zero, NS= negative small, NM= control without chattering elimination. It is seen that
negative medium, NB= negative big. Therefore, there are although the system is stable but there exists chattering
49 fuzzy tuning rules as shown in Table 1. and there is no guarantee that lim s=0 as t→∞. To
eliminate the chattering, we replace discontinuous
Table 1. Fuzzy rule table function sgn(s) by a saturation function. The simulation
results with chattering elimination are shown in Figure.7
and figure.10. As seen, there is no chattering but the
accuracy is decreased and there exists a steady state error.
In order to overcome this problem, we design a fuzzy
sliding mode control (FSMC) law. Figure.8 and figure.11
show the tracking of the system outputs under the fuzzy
sliding mode controller. We can see that the outputs track
the desired trajectory and the steady state errors converge
to zero. This result concludes that the FSMC proposed in
this paper eliminates the chattering without steady state
errors.
Theorem: Assume ε ≤α <1 , we can obtain kmax as The parameters in the simulation study are as follows:
follow: V, the flight speed of the missile, is 10 Ma; g, the
k i max ≤ αλ i
2 gravitational acceleration, is 9.8; M, the mass of the
(12)
missile, is 700 kg; d, the length of the missile, is 6m; D,
Proof. As discussed in [5], by replacing saturation the diameter of the missile, is 0.4m; λi, the coefficients of
function instead of sign function, we can rewrite (7) as sliding surface, are λ1=5, λ2= λ3=15; φ, the thickness of
following: boundary layer, are φ1=1, φ2=φ3=0.35; α, the coefficient
s ) of the controller gain, is 0.2; Note that the aerodynamic
s& = k − u + x&d − λe− f (x ) (13)
φ coefficients changes according to the flight condition.
Equation(13)is the first order filter, that it’s input is
5. CONCLUSION
uncertainties and the variable s is the output of it, which
in turn is the input of(5),Since λ is the break frequency of
In this paper the none-linear sliding mode control for
filter(5). Because k is depended on uncertainties and
ballistic missiles was studied. The three main
disturbances, if the maximum of it replace in (13), the
characteristics for an autopilot are: rapid response with
maximum of λ will be obtained. Where, λ must be
the minimum sustainable error, stability, and robustness
smaller than high frequency unmodeled dynamics.
against uncertainties. The simulation results show that the
Therefore, we can tune Φ as
controller proposed in this paper is proper and
k i max satisfactory for meeting the favorable characteristics.
λi = (14)
φi
6. REFERENCES
For n=2, by using (11) and (14) we obtain:

455
[1] Y. J. Shi, G. F, Ma and H. Z. Ma, “Study of the compound
control of cruise missile based on variable c Ballistics, in
Cambridge university press, vol.49,pp.573-583,2008.
[2] Y. J. SHI, G. F. MA, “Missile control design based on the
linear Multiple sliding mode recursive method”, Anziam J,
vol.492 ,pp.573-587,2008.
[3] A. Khalid, and A. Sabanovic, “Sliding-mode control for
high- precisionmotion of a piezostage,” IEEE Trans. Ind.
Electron., vol. 54, pp. 629– 637,2007
[4] P. Garnell and D. J. East, Guided Weapon Control System.
Figure 7. Tracking of pitch rate with a disturbance
Pergamon Press,2nd Ed.,Oxford,1980. in the SMC with chattering elimination
[5] J.Slotine, and W.Li, “Applied Nonlinear Control”,
EnglewCliffs, NJ: Prentice Hall, 1991.
[6] S. Kawaji, W.Li, ” Fuzzy control of vss type and its
robustness”. Proc 1FSA world congress,
Engzneerzng, pp.81-84, 1991.
[7] Y. S. Zhao, Y. Q. Zhang, J. Yang and L. P. Chen,
“Enhanced fuzzy sliding mode controller for robotic
manipulators”, Int. J. Robot, vol.22, pp.170–183, 2007.

Figure 8. Tracking of pitch rate with a disturbance

in the FSMC

Figure 3. Tracking of roll rate with a disturbance

in the SMC without chattering elimination

Figure 9. Tracking of yaw rate with a disturbance

in the SMC without chattering elimination

Figure 4. Tracking of roll rate with a disturbance

in the SMC with chattering elimination

Figure 10. Tracking of yaw rate with a disturbance

in the SMC with chattering elimination

Figure 5. Tracking of roll rate with a disturbance

in the FSMC

Figure 11. Tracking of yaw rate with a disturbance

in the FSMC

Figure 6. Tracking of pitch rate with a disturbance

in the SMC without chattering elimination

456