World Academy of Science, Engineering and Technology

International Journal of Aerospace and Mechanical Engineering

Vol:6, No:12, 2012

State Dependent Riccati Equation Based Roll

Autopilot for 122mm Artillery Rocket
Muhammad Kashif Siddiq, Fang Jian Cheng, and Yu Wen Bo

applications, now demand stringent accuracy and cost

Abstract—State-dependent Riccati equation based controllers are requirements in nonlinear control systems. This has expedited
becoming increasingly popular because of having attractive the development of nonlinear control theory for application to
properties like optimality, stability and robustness. This paper focuses challenging, complex, dynamical real-world problems,
on the design of a roll autopilot for a fin stabilized and canard
particularly those that bear major practical significance in
controlled 122mm artillery rocket using state-dependent Riccati
equation technique. Initial spin is imparted to rocket during launch military industries. Researchers are striving to develop control
and it quickly decays due to straight tail fins. After the spin phase, the algorithms that are simple, and yet produce optimal
performance in the sense of control effort and state errors.
International Science Index, Aerospace and Mechanical Engineering Vol:6, No:12, 2012 waset.org/Publication/2792

roll orientation of rocket is brought to zero with the canard deflection

commands generated by the roll autopilot. Roll autopilot has been State-dependent Riccati equation (SDRE) control is a highly
developed by considering uncoupled roll, pitch and yaw channels. promising and very attractive practical tool for obtaining
The canard actuator is modeled as a second-order nonlinear system.
approximate solutions to infinite-time horizon nonlinear
Elements of the state weighing matrix for Riccati equation have been
chosen to be state dependent to exploit the design flexibility offered optimal control problems in feedback form. The SDRE
by the Riccati equation technique. Simulation results under varying method provides an attractive alternative to solving the
conditions of flight demonstrate the wide operating range of the Hamilton Jacobi Bellman partial differential equation,
proposed autopilot. allowing for the systematic and effective design of nonlinear
feedback controllers for a variety of applications. The
Keywords—Fin stabilized 122mm artillery rocket, Roll potential of this method is characterized by possessing the
Autopilot, Six degree of freedom trajectory model, State-dependent crucial features of stability, optimality, real-time
Riccati equation.
implementation, and inherent robustness with respect to
parametric uncertainties.
I. INTRODUCTION
Although application potential of the SDRE nonlinear
R OLL autopilot design for guided artillery rockets that can
ensure stable performance over the full flight envelope is a
control technique in practical nonlinear control problems is
well recognized, the industry acceptance of the technique has
challenging task primarily because of flexible nature of the not been appreciable. The main reasons for this being the
airframe, cross coupling, uncertainty in aerodynamic SDRE approach requires advanced numerical methods for its
parameters, external disturbances, and inaccuracies in implementation, and the perception that this technique may
measurements obtained from onboard sensors. As per the not be computationally feasible for real-time implementation
conventional practice of classical linear control techniques, on commercial off-the-shelf processors. P. K. Menon et al [1],
dynamic models of rockets are linearized around several using software based on the Schur algorithm and the Kleinman
operating points in flight envelope and then the gain scheduled method, showed that SDRE control laws can be implemented
autopilots are designed. However, the performance of these at speeds up to 2 kHz sample rates using commercial off-the-
classical autopilots can only be guaranteed within a narrow shelf processors, for problems of the size commonly
range of variations in flight conditions and model encountered in missile flight control applications. The
uncertainties. This has necessitated the application of potential of SDRE technique for flight control applications has
nonlinear control techniques in order to maintain the been demonstrated by researchers in [2]–[9]. Cimen provided
operability of controllers over a wider range of operating a comprehensive overview of the present state of the art of
conditions. SDRE control technique in [10] and [11], and addressed the
The advent of powerful low-cost micro-processors has systematic design of nonlinear controllers via SDRE method
equipped the designers with an effective tool to meet the in [12]. Controllability and stability issues of SDRE technique
challenges in applications of nonlinear control. The most are well addressed in [13]–[17].
recent applications, particularly in aerospace and military This paper focuses on the design of a SDRE based roll
autopilot for a canard controlled 122mm artillery rocket. The
Muhammad Kashif Siddiq is Ph.D student in Beijing University of rocket having front canards and folded straight tail fins is
Aeronautics and Astronautics, Beijing, 100083 China (phone: 0086-
13436479484; e-mail: kashif_s4u@ hotmail.com).
given initial spin at the time of launch. Tails fins are deployed
Fang Jian Cheng is Professor in Instrumentation Science and Opto- immediately after launch and offer high roll damping moment
electronics Engineering Department of Beijing University of Aeronautics and thereby reducing the spin rate to zero within six seconds of
Astronautics, Beijing, 100083 China (e-mail: fangjiancheng@ buaa.edu.cn).
Yu Wen Bo is Associate Professor in Instrumentation Science and Opto-
flight. The canards are then deployed and the roll orientation
electronics Engineering Department, Beijing University of Aeronautics and of rocket is regulated to zero with the canard deflection
Astronautics, Beijing, 100083 China (e-mail: [email protected]).

International Scholarly and Scientific Research & Innovation 6(12) 2012 2814 scholar.waset.org/1307-6892/2792
 World Academy of Science, Engineering and Technology
International Journal of Aerospace and Mechanical Engineering
Vol:6, No:12, 2012

commands generated by the SDRE based roll autopilot. The infinite number of ways to factor f(x) into A(x)x and that
designed controller produces promising results for the subject A(x) can be parameterized as A(x,α), where α is a vector
application. Extensive simulations have been carried out and of free design parameters. In order to obtain a valid
the results corrugate the efficacy of the proposed autopilot solution of the SDRE, the pair {A(x, α), B(x)} must meet
over a wide range of flight conditions. the condition of point wise stability in the linear sense for
all x in the domain of interest.
II. SDRE CONTROL METHODOLOGY 3) Solve the algebraic state-dependent Riccati equation
A. SDRE Problem Formulation
Consider the autonomous, infinite-horizon, nonlinear AT ( x ) P ( x ) + P ( x ) A ( x )
(8)
regulator problem for minimizing the performance index − P ( x ) B ( x ) R −1 ( x ) B T ( x ) P ( x ) + Q ( x ) = 0

∫(x ( t ) Q ( x ) x ( t ) + uT ( t ) R ( x ) u ( t ) ) dt
1 T to obtain P(x) ≥ 0. P(x) is the unique, symmetric, positive-
J= (1)
2 definite solution of the algebraic state-dependent Riccati
0
equation i.e. (8), and hence the name SDRE control.
4) The nonlinear feedback controller equation is given by
International Science Index, Aerospace and Mechanical Engineering Vol:6, No:12, 2012 waset.org/Publication/2792

with respect to the state x and control u subject to the 5)

nonlinear differential constraints:
u ( x ) = − R −1 ( x ) BT ( x ) P ( x ) x                          (9)

x& ( t ) = f ( x ) + B ( x ) u ( t ) , x ( 0) = x0 (2)  
and the resulting SDRE controlled trajectory is the solution of
                      
the quasi-linear closed-loop dynamics
where Q(x) ≥ 0 (positive definite) and R(x) > 0 (semi-positive
definite) for all x and where
x& ( t ) = ⎡ A ( x ) − B ( x ) R−1 ( x ) BT ( x ) P ( x ) ⎤ x ( t )         (10)
Condition 1. f(x) is a continuously differentiable function of x, ⎣ ⎦
i.e.  
f ( x )∈C1                                         (3) The SDC matrix for the closed loop dynamics is
 
Condition 2 ACL ( x ) = A ( x ) − B ( x ) K ( x )                       (11)
f ( 0) = 0                                          (4)  
  and the state feedback gain for minimizing the cost function
Under the specified conditions, a control law (1) is

u ( x ) = k ( x ) =−K ( x ) x, k ( 0) = 0 (5)   K ( x ) = R −1 ( x ) BT ( x ) P ( x )                          (12)

                      
is sought that (approximately) minimizes the cost function in The SDRE solution to (1) and (2) is a true generalization of
(1) subject to the input affine nonlinear differential constraint the infinite-horizon time-invariant linear quadratic regulator
in (2) while regulating the system to the origin for all x, such (LQR) problem, where all of the coefficient matrices are state-
that limt→∞x(t) = 0. This is the basic idea of the SDRE method dependent. At each instant of computing the control action,
for nonlinear regulation [10]. the method treats the state-dependent coefficient matrices as
being constant, and computes a control action by solving a
B. SDRE Controller Structure linear quadratic optimal control problem. As is evident from
The SDRE approach as outlined in [10] and [18] for (8), the resulting controller relies on a solution, point wise in
obtaining a suboptimal, locally asymptotically stabilizing R n , of an algebraic Riccati equation thereby leading to the
solution of (1) and (2) is: SDRE terminology.
1) Use direct parameterization to factorize nonlinear system If the coefficient and weighting matrices are selected as
dynamics into a linear like structure which contains the constant, the nonlinear regulator problem becomes the LQR
state-dependent coefficient (SDC) matrices problem and the SDRE control method matches the steady-
2) state linear regulator.
x& ( t ) = A ( x ) x ( t ) + B ( x ) u ( t ) , x ( 0) = x0 (6)   In order to perform tracking / command following, the
SDRE controller can be implemented as an integral servo-
                   
mechanism as explained and demonstrated in [3],[10], and
where
[18]. The procedure is outlined here briefly. The state x is
f ( x ) = A ( x ) x                                      (7) decomposed as
 
If the condition, f(x) ∈ C1, is satisfied then there is an

International Scholarly and Scientific Research & Innovation 6(12) 2012 2815 scholar.waset.org/1307-6892/2792
 World Academy of Science, Engineering and Technology
International Journal of Aerospace and Mechanical Engineering
Vol:6, No:12, 2012

⎡x ⎤ vaalues. Reference [12] propoosed a proceduure for selectiing the

x =⎢ T ⎥ (13) staate dependent elements of Q and R matriices. Althoughh there
⎣ xN ⎦ hese matrices,, some
aree no rigorouss methods forr selecting th
guuidelines for their selectioon based on good practiices as
where it is desiired for the vvector componnents of xT to track a
w meentioned in [19] are useful.
reeference comm mand rc. Augmment the state vector x withh xI, the
o xT :
inntegral states of I
III. ROCKET CONFIGURATTION
Rocket confiiguration beinng studied haas following salient
⎡ xI ⎤ feaatures:
⎢ ⎥
x% = ⎢ xT ⎥ (14) 1) Rocket com mes out of laaunch tube withw spin rate of 5.8
⎢⎣ xN ⎥⎦ revolutions per second.
2) Fixed tail having
h straighht fins as shoown in Fig. 1.1 Rear
view of thee rocket is shoown in Fig. 2. 2 The tail finns open
Thhe augmentedd system is givven by
up once thhe rocket com mes out of laaunch tube/caanister.
These straight fins provide a high roll damping moment m
x% = A% ( x% , α ) x% + B% ( x% ) u (15) and retard the
t spin rate of rocket body.
International Science Index, Aerospace and Mechanical Engineering Vol:6, No:12, 2012 waset.org/Publication/2792

3) Deployablee canards for control purpo oses. During launch,

l
w
where canards reemain foldedd inside rock ket body annd are
⎡0 IM 0 ⎤ ⎡ 0 ⎤ deployed onnly after 6 secconds of launcch.
x% = A% ( x% , α ) = ⎢ B% ( x% ) = ⎢
A ( x, α ) ⎥⎦
, ⎥ (16)
⎣ B ( x )⎦
The main phyysical parameeters of the sub bject artillery rocket
⎣0 aree summarizedd in Table I. T The rocket leaaves the muzzlle with
iniitial spin rate of 5.8 revoluttions per secoond. This initiial spin
annd the SDRE integral
i servo controller is given
g by miitigates the efffects of masss and configuuration asymm metries.
Thhe straight taill fins offer high roll dampiing moment and a the
⎡ x − r dt ⎤
⎢ I c
⎥ ∫ spiin rate of rocket reduces too zero within first six secoonds of
fligght. Canards are deployedd at the time of six second ds after
u = − R ( x% ) B ( x% ) P ( x% ) xT − rc ⎥
% −1 % T % ⎢ (17)  lauunch and are used to bringg the roll orieentation of roccket to
⎢ ⎥
⎢ xN ⎥ zerro, later theese canards can be usedd for aerodyynamic
⎢⎣ ⎥⎦ maaneuvering to facilitate guiddance and con ntrol of the roccket.
C. SDC Paraameterization
There are maany systems thhat do not con
nform to the sttructure
orr conditions given
g in (2) to (4) and thhe SDRE tecchnique
caannot be direcctly applied. In these casees, there is ann art in
coonverting the given
g system to a conformiing system so that an
efffective SDRE b carried out. References [11] and
E design can be
[118] present several casees where thhe system iss non-
coonforming, annd show in eacch case how tot convert the system
too a conformingg one. Fig. 1 Canardd controlled 1222mm rocket witth straight tail fins
f

D. Selection of Q and R MMatrices

The solution to the Riccatii Equation deppends on selecction of
thhe state and coontrol weightinng matrices Q and R, respeectively. nozzle exit
Q is a matrix of o weighting coefficients used
u to penalize any
state from becooming too largge. Similarly, R is used to penalize
p
thhe control action to remain w within boundss. These matriices are
deesign parametters that also affect the overall performaance of
thhe closed looop system. A remarkable design fleexibility foldeed fin rests on th
his
offfered by SDRE approaach is that by penalizinng the flat surface
apppropriate staates or contrrol action, thhe designer can c set
coonstraints on the system. Elements off Q and R matrices m foldable tail
t fin
shhould be seleected such thhat the corressponding stattes and
innputs which should be resttrained from becoming arbbitrarily Fig. 2 Rear vview of the rockket
laarge, are penallized the mostt. From simplee to complicaated, the
ellements of maatrices Q and R can simply be b constant orr can be
chhosen as funnctions of sttates with thheir values varyingv
thhroughout thee control proocess dependding upon thhe state

International Scholarly and Scientific Research & Innovation 6(12) 2012 2816 scholar.waset.org/1307-6892/2792
 World Academy of Science, Engineering and Technology
International Journal of Aerospace and Mechanical Engineering
Vol:6, No:12, 2012

TABLE I
⎛
QD S r ω y DC N β& ⎞
PHYSICAL PARAMETERS OF 122MM ARTILLERY ROCKET w& = –ω x v +ω y u + ⎜ CN β β + ⎟
m ⎜ 2Vab ⎟⎠
Rocket Parameters Value ⎝ (20)
Q SC
caliber 122mm + D c N δ (δY − β ) + Gzb
overall length 2.87m m
total mass 66.0kg
propellant mass 20.5kg
propellant burning time 1.67s where δP is pitch channel canard deflection angle, δR is yaw
mean thrust 23600N channel canard deflection angle, CNα is normal aerodynamic
initial center of gravity from nose tip 1.374m force coefficient for rockety body, CNδ is normal aerodynamic
final center of gravity from nose tip 1.264m
initial axial moment of inertia 0.1499kg.m2 force coefficient due to canards, D is reference diameter of
final axial moment of inertia 0.1238kg.m2 rocket body, Sr is reference area of rocket, Sc is surface area of
initial lateral moment of inertia 41.58kg.m2 canard. Dynamic pressure, QD, is calculated by the expression
final lateral moment of inertia 33.83kg.m2
launch velocity 26.7m/s
initial spin rate 2088°/s 1
QD = ρVab2   (21)  
2
IV. SIX DEGREE OF FREEDOM TRAJECTORY MODEL
International Science Index, Aerospace and Mechanical Engineering Vol:6, No:12, 2012 waset.org/Publication/2792

                     
A computer code is developed which models the flight where Vab is the magnitude of aerodynamic velocity expressed
dynamics of fin stabilized and canard controlled 122mm in body frame. Vab is calculated as following
projectile. The code considers the projectile as a rigid, six
degree-of-freedom body and solves the equations of motion in
r r r
Vab =Vb −Vwb                                    (22)
a body coordinate system. The muzzle conditions are used as
 
initial conditions in the calculations. A body-fixed reference r
frame is chosen for this study since the dynamic behavior of where Vb is the rocket velocity in body frame, with respect to
the vehicle is the main object of study. The equations of r
earth, and Vwb is the wind velocity in body frame. Wind
motion are developed using a body-fixed coordinate system as r
shown in Fig. 3. The origin of the body axes is the vehicle velocity is Vwb is usually given in local vertical frame(North,
center of mass, O. The angular velocity of the body relative to East, Down). It has to be transformed to body frame before
an inertial frame is ωb(ωbx, ωby, ωbz) and the components of the being used in (22).
translational velocity, Vb in body frame are (u, v, w). Angle of attack α and angle of side slip β, are calculated by
r
using components of Vab
Yb
Xb r
Vab = Vab = uab
2
+ vab
2
+ wab
2
(23)

O
α = tan −1 ( −vab / uab ) (24)
Zb
β = sin −1 ( wab / Vab ) (25)
Fig. 3 Orientation of rocket body axes

The earth is treated as spherical and non-rotating in this With the assumptions in (26) and (27), angular accelerations
study since the time of flight for such type of vehicles is of the are given in their simplest form in (28) to (31).
order of few tens of seconds. Equations used for generating six
degree of freedom trajectory model are summarized in the I xy = I yx = I xz = I zx = I yz = I zy = 0 (26)
following lines.
I yy = I zz (27)
Q S C QS C T
u& =−ωby w + ωbz v − D r A − c Ac + xb + Gxb         (18)
m m m
QD S r D 2Clp QD S r D 2Clr
ω& bx =
2Va I xx
(ωbx ) +
2Va I xx
( )
ωby
Q S ⎛ ω z DC Nα& ⎞            (28) 
v& = –ωbz u + ωbx w + D r ⎜ C Nα α + ⎟ QD Sc d c C Nδ
m ⎝ 2Vab ⎠
(19)
+ (δ R )
I xx
Q SC
+ D c Nδ (δ P + α ) + G yb
m

International Scholarly and Scientific Research & Innovation 6(12) 2012 2817 scholar.waset.org/1307-6892/2792
 World Academy of Science, Engineering and Technology
International Journal of Aerospace and Mechanical Engineering
Vol:6, No:12, 2012

QD Sr D ⎛ ⎛ ωby D ⎞ ⎞ ⎡ x& ⎤ ⎡u ⎤
ω& by =
I yy ⎜⎝
(
⎜ Cmβ β + C Nr + Cmβ& ) ⎜⎜⎝ 2V ⎟⎟ ⎟⎟
a ⎠⎠
uur ⎢ l ⎥
Vl = ⎢ y&l ⎥ = TBL ⎢⎢ v ⎥⎥ (35)
(29) ⎢⎣ z&l ⎥⎦ ⎢⎣ w⎥⎦
QD Sc xc C Nδ (δY − β ) ( I xx − I zz ) ω
− − bxωbz x&l = Cψ Cθ ( u ) + ( Sφ Sψ Cθ − Cφ Sθ )( v )
I yy I yy
(36)
+ ( Cφ Sψ Cθ + Sφ Sθ )( w )
Q S D⎛ ⎛ ω D ⎞⎞
I zz ⎜⎝
(
ω& bz = D r ⎜ Cmα α + Cmq + Cmα& ⎜ bz ⎟ ⎟
⎟ ) y&l = Cψ Sθ ( u ) + ( Sφ Sψ Sθ + Cφ Cθ )( v )
⎝ 2Va ⎠ ⎠
           (30) (37)
+ ( Cφ sinψ Sθ − Sφ Cθ )( w )
QD Sc xc C Nδ (δ P + α ) ( I yy − I xx ) ω
+ − bxωby z&l =−Sψ ( u ) + SφCψ ( v ) + CφCψ ( w)
I zz I zz (38)

Ixx, Iyy, Izz, Ixy, Ixz, Iyx, Iyz, Izx, and Izy are moments of inertia V. TRAJECTORY SIMULATIONS
about respective axis. Clp is roll damping moment coefficient, Nominal trajectories for the rocket under study have been
Clr is roll moment coefficient derivative with yaw rate, Cmβ is simulated for launch elevation angles ranging from 30° to 70°.
yawing moment coefficient, CNr is, Cmα is pitching moment
International Science Index, Aerospace and Mechanical Engineering Vol:6, No:12, 2012 waset.org/Publication/2792

Aerodynamic data for 122mm rocket given in [20] has been

coefficient due to angle of attack, Cmq is pitching moment used for simulations. A plot of roll rate versus time for launch
coeeficient due to pitch rate , δR is roll channel canrd angle of 50° is shown in Fig. 4(a). Figure 4(b) depicts the plot
deflection, dc is lateral distance between center of pressure of of rocket altitude versus time, and the plot of altitude versus
canard and rocket’s roll axis, xc is is longitudinal distance downrange is shown in Fig. 4(c). It has been observed for all
between center of pressure of canard and rocket’s centre of cases of launch angles that the roll rate of rocket damps out to
gravity. Attitude of the body frame with respect to launch zero within first six seconds of flight owing to the high roll
frame is determined by Euler angles θ , ψ and φ i.e pitch, damping moment offered by straight tail fins, to be published
yaw and roll respectively. Euler angles are related to angular [21].
velocity components of body frame according to the following
differential equations. 2500

φ& = ωbx + (ωby sinφ +ωbz cosφ ) tanψ                    (31)

2000
Roll rate (°/s)

1500

ψ& = ωby cosφ − ωbz sinφ (32)

1000

θ& =
(ωby sinφ + ωbz cosφ ) (33) 500

cosψ
0
0 1 2 3 4 5 6 7 8 9 10
Time (s)

Equations (18) to (20) are integrated to obtain the velocity Fig. 4 (a) Roll rate versus time for 50° launch elevation
components of the rocket referred to the body frame. The 8000

coordinate transformation is used to determine velocity 7000

components of the rocket referred to the launch frame, and 6000

since the launch frame is not rotating, these velocity 5000

Altitude (m)

components can be integrated directly to obtain the 4000

displacements (rocket position) referred to the launch frame. 3000

Following is the matrix for coordinate transformation from 2000

body frame to launch frame. 1000

0
0 10 20 30 40 50 60 70 80

⎡ Cψ Cθ Sφ Sψ Cθ − Cφ Sθ Cφ Sψ Cθ + Sφ Sθ ⎤
Time (s)

= ⎢⎢ Cψ Sθ Cφ Sψ Sθ − Sφ Cθ ⎥⎥ (34)
Fig. 4 (b) Rocket altitude versus time for 50° launch elevation
TBL Sφ Sψ Sθ + Cφ Cθ 8000

⎢⎣ − Sψ Sφ Cψ Cφ Cψ ⎥⎦ 7000

6000

5000

C stands for cosine, and S stands for sine function. Velocity

Altitude (m)

4000

vector in launch frame and three differential equations for 3000

rocket position in launch frame are given by (35) to (38). 2000

1000

-1000
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
4
Downrange (m) x 10

Fig. 4 (c) Rocket altitude versus downrange for 50° launch elevation

International Scholarly and Scientific Research & Innovation 6(12) 2012 2818 scholar.waset.org/1307-6892/2792
 World Academy of Science, Engineering and Technology
International Journal of Aerospace and Mechanical Engineering
Vol:6, No:12, 2012

Spinning of rocket in the initial phase helps to average out ⎡0 1 0 ⎤0

⎡ φ& ⎤ ⎢ ⎥⎡ φ ⎤ ⎡ 0 ⎤
the effects of configurational and mass asymmetries of the ⎢ ⎥ ⎢ QD Sr D 2Clp QD Sc dcC N δ ⎥⎢ ⎥ ⎢ ⎥
rocket, and aids in reducing dispersion. Once the roll rate is ⎢ φ&& ⎥ ⎢ 0 0 ⎥ ⎢ φ& ⎥ ⎢ 0 ⎥ (44)
⎢& ⎥=⎢ 2Va I xx I xx ⎥ ⎢δ ⎥ + ⎢ 0 ⎥ δ Rc
zero, the front canards are deployed which can be used for ⎢δ R ⎥ ⎢ 0 0 0 1 ⎥⎢ ⎥ ⎢ 2⎥
R

aerodynamic maneuvering in order to execute guidance and ⎢δ&& ⎥ ⎢ ⎥ ⎢⎣δ&R ⎥⎦ ⎢⎣ωa ⎥⎦

⎣ R ⎦ ⎢0 −ωa2 −2ζωa ⎥⎦
control function. Late deployment of canards is adopted in ⎣ 0

order to avoid unnecessary drag during early part of trajectory

that may reduce the effective range of the rocket. After the The saturation limits of canard deflection angle and rocket
deployment of canards their foremost function is to bring the angular rate during the control phase are set to be 30° and
roll orientation of rocket to zero degree so that pitch and yaw 800°/s, respectively. All state variables are assumed onboard
controls can be appropriately managed at later stages. The measurable. The state weighting matrix Q is chosen to be
next section describes the SDRE model for roll autopilot diagonal as in (45), and the elements are initially selected on
which generates actuator commands for canard deflection the basis of Bryson’s rule i.e. every diagonal element of Q
required to orientate the rocket to zero roll position. should be reciprocal of the square of maximum permissible
value of corresponding state, as it provides a good starting
VI. SDRE BASED DESIGN MODEL FOR ROLL AUTOPILOT point. However, the elements of Q matrix are tailored by trial
International Science Index, Aerospace and Mechanical Engineering Vol:6, No:12, 2012 waset.org/Publication/2792

and error to obtain the appropriate response over the desired

Roll dynamics of the rocket under study are governed by
operating range which is the flight envelope for the subject
(39) and (40).
rocket in our case. A wide operating range is achieved by
QD S r D 2 Clp making the elements of Q matrix to be state dependent. This
QD Sr D 2 Clr
ω& bx =
2Va I xx
(ωbx ) +
2Va I xx
( )
ωby makes the state weightings keep on varying at every instance
(39) of calculating control action.
Q S dC
+ D c c N δ (δ R )
I xx ⎡ q11 0 0 0 ⎤
⎢0 q22 0 ⎥⎥
φ& = ωbx + (ωby sinφ +ωbz cosφ ) tanψ
0
(40) Q =⎢ (45)
⎢0 0 q33 0 ⎥
⎢ ⎥
In order to develop the SDRE based roll autopilot these ⎣0 0 0 q44 ⎦
equations have to be reformulated to become consonant with
the requisite SDRE structure. First we assume decoupled roll, Following choice of elements of Q matrix gave the control
pitch and yaw dynamics leading to following simplified performance within acceptable bounds.
equations for roll dynamics of the rocket.
5 φ
q11 = − (46)
φ& = φ& (41)
φi2 (5φ ) i
2

2
QD S r D Clp
φ&& = (φ&) + QD ScIdcCNδ (δ R ) (42) q22 =
1
+
φ&2
(47)
2Va I xx xx
(800 )2 (10 ) 9

The canard actuator is modeled as a second-order nonlinear

1
system with natural frequency of ωa = 150rad/s and damping q33 = (48)
ratio ζ = 0.7. Actuator dynamics are governed by following (30 ) 2

equations
1
q44 = (49)
δ&R = δ&R (43) (800 ) 2

δ&&R = −ωa2δ R − 2ζωaδ&R + ω a2δ Rc

(44) i corresponds to the roll orientation of rocket when the roll
autopilot starts regulating the roll orientation. The first term of
where δRc is the commanded deflection and δR is the actual q11 remains constant throughout the control process, however
canard deflection for roll channel. the second term keeps on decreasing as the roll error
Complete equations for roll dynamics taking into account decreases. Keeping the second term of q11 as negative helps to
actuator dynamics is expressed in state space form consonant avoid large deflection angle of the canard at the start of control
with (6) as following process, thereby eliminating the chance of canards going to
their saturation limits. Similarly the second term of q22
decreases as the roll rate generated due to deflection of
canards decreases. Matrix elements q33 and q44 are chosen as
constants to avoid unnecessary computational burden. At the

International Scholarly and Scientific Research & Innovation 6(12) 2012 2819 scholar.waset.org/1307-6892/2792
 World Academy of Science, Engineering and Technology
International Journal of Aerospace and Mechanical Engineering
Vol:6, No:12, 2012

end of spin phase, the foremost action to be taken by roll 100

autopilot is to nullify the residual roll error/orientation to zero. 80

During remaining phase of flight the task of the roll

autopilot is to keep the roll orientation stabilized to zero roll 60

Roll angle (°)

position and continuously nullify any roll error occurring due 40

to disturbing roll moments. Usually these roll errors are not of 20

high order and remain within one to two tens of degrees. In

order to nullify disturbing roll errors, we selected diagonal 0

elements of Q matrix as in (50) to (53). -20

6.5 6.6 6.7 6.8 6.9 7 7.1 7.2 7.3 7.4 7.5
Time (s)

Fig. 5 (a) Roll error decay profile for 90° initial roll error
1 QD
q11 = − (50)
φi2 (3 ×10 )
10
7
5

Canard deflection angle (°)

0

1 φ&2
q22 = + (51)
(10 )
-5

( 360 )2 6
International Science Index, Aerospace and Mechanical Engineering Vol:6, No:12, 2012 waset.org/Publication/2792

-10

-15

q33 = 0 (52) -20

q44 = 0 (53) -25

6.5 6.6 6.7 6.8 6.9 7
Time(s)
7.1 7.2 7.3 7.4 7.5

Fig. 5 (b) Canard deflection for correcting 90° initial roll error
The matrix element q11 has been chosen to be a function of 100

initial roll error and dynamic pressure. Since the dynamic 0

pressure depends on altitude and velocity of the rocket, the -100

element q11 being function of dynamic pressure enables the -200

Roll rate (°/s)

autopilot to automatically adjust its performance while -300

performing at different altitudes. As the rocket traverses its -400

trajectory in ascending phase, its velocity and atmospheric -500

density decrease, thereby, reducing the dynamic pressure. This -600

leads to increase in the value of q11 at low dynamic pressures -700

6.5 6.6 6.7 6.8 6.9 7
Time (s)
7.1 7.2 7.3 7.4 7.5

and comparatively larger canard deflection is produced to Fig. 5 (c) Roll rate generated for correcting 90° initial roll error
compensate the effect of low aerodynamic force available at 140

low dynamic pressure. 120

100

VII. AUTOPILOT SIMULATION RESULTS 80

Roll angle (°)

With the above mentioned choice of Q matrix (46) to (49), 60

simulations for SDRE based roll autopilot have been 40

performed for initial roll angles (or residual roll error of spin 20

phase) 90°, 122°, and 150° considering the rocket launch 0

elevation angle to be 50°. During standard trajectory -20

6.5 6.6 6.7 6.8 6.9 7 7.1 7.2 7.3 7.4 7.5
Time (s)
simulations it has been observed that residual roll error of spin
phase is about 115° to 125° for launch elevation angles Fig. 6 (a) Roll error decay profile for 122° initial roll error
ranging from 30° to 70°, and it is 122° for launch elevation of 10

50°. We have considered the cases of 90° and 180° initial roll 5

errors as worst case scenarios. Performance of the autopilot is

Canard deflection angle (°)

depicted in simulation results shown in Fig. 5 to Fig. 7. The -5

results show that the proposed roll autopilot eliminates 90° -10

roll error in about 0.45 seconds, 122° roll error in 0.6seconds, -15

and 180° initial roll in 0.7seconds. Moreover, the canard

-20
deflection angles and roll rates remain well within the
prescribed limits. Thus, the proposed roll autopilot -25
6.5 6.6 6.7 6.8 6.9 7
Time(s)
7.1 7.2 7.3 7.4 7.5

successfully performs the task of orientating the rocket to zero Fig. 6 (b) Canard deflection for correcting 122° initial roll error
roll position soon after the initial spin phase is over.

International Scholarly and Scientific Research & Innovation 6(12) 2012 2820 scholar.waset.org/1307-6892/2792
 World Academy of Science, Engineering and Technology
International Journal of Aerospace and Mechanical Engineering
Vol:6, No:12, 2012

100 10

0
8

-100

6
-200

Roll angle (°)

Roll rate (°/s)

-300 4

-400
2

-500

0
-600

-700 -2
6.5 6.6 6.7 6.8 6.9 7 7.1 7.2 7.3 7.4 7.5 10.3 10.4 10.5 10.6 10.7 10.8 10.9
Time (s) Time (s)

Fig. 6 (c) Roll rate generated for correcting 122° initial roll error Fig. 8 (a) Roll error decay profile for 10° roll error at 4000m altitude
200 4

150
0

-2

Canard deflection angle (°)

Roll angle (°)

100
-4
International Science Index, Aerospace and Mechanical Engineering Vol:6, No:12, 2012 waset.org/Publication/2792

-6
50
-8

-10
0

-12

-50 -14
6.5 6.6 6.7 6.8 6.9 7 7.1 7.2 7.3 7.4 7.5 10.3 10.4 10.5 10.6 10.7 10.8 10.9
Time (s) Time(s)

Fig. 7 (a) Roll error decay profile for 180° initial roll error Fig. 8 (b) Canard deflection for correcting 10° roll error
10
at 4000m altitude
20
5

0
Canard deflection angle (°)

-20
-5
-40
Roll rate (°/s)

-10
-60

-15
-80

-20 -100

-25 -120
6.5 6.6 6.7 6.8 6.9 7 7.1 7.2 7.3 7.4 7.5
Time(s)
-140
10.3 10.4 10.5 10.6 10.7 10.8 10.9
Fig. 7 (b) Canard deflection for correcting 180° initial roll error Time (s)

100
Fig. 8 (c) Roll rate generated for correcting 10° roll error
0 at 4000m altitude
-100
10
-200
Roll rate (°/s)

-300 8

-400
6
-500
Roll angle (°)

-600
4
-700

-800 2

-900
6.5 6.6 6.7 6.8 6.9 7 7.1 7.2 7.3 7.4 7.5
Time (s) 0

Fig. 7 (c) Roll rate generated for correcting 180° initial roll error -2
18.1 18.2 18.3 18.4 18.5 18.6 18.7 18.8 18.9 19
Time (s)

After eliminating the residual roll error of spin phase the Fig. 9 (a) Roll error decay profile for 10° roll error at 6000m altitude
roll autopilot keeps the rocket in roll stabilized state at zero 4

degree roll. Performance of the proposed autopilot has also 2

0
been simulated for 10° roll error at flight altitudes of 4000m,
Canard deflection angle (°)

-2

6000m, and 7500m. The simulation results are presented in -4

Fig. 8 to Fig. 10. The results show that the roll errors are -6

-8
successfully eliminated at different altitudes by the designed -10

autopilot while remaining within performance bounds, thus -12

demonstrating the efficacy of the proposed scheme. -14

-16
18.1 18.2 18.3 18.4 18.5 18.6 18.7 18.8 18.9 19
Time(s)

Fig. 9 (b) Canard deflection for correcting 10° roll error

at 6000m altitude

International Scholarly and Scientific Research & Innovation 6(12) 2012 2821 scholar.waset.org/1307-6892/2792
 World Academy of Science, Engineering and Technology
International Journal of Aerospace and Mechanical Engineering
Vol:6, No:12, 2012

20
[4] P. K. Menon, “Integrated Design of Agile Missile Guidance and Control
Systems,” in 7th Mediterranean Conference on Control and Automation,
0
pp. 1469–1494, 1999.
-20
[5] A. Ratnoo and D. Ghose, “SDRE Based Guidance Law for Impact
-40 Angle Constrained Trajectories,” in AIAA Guidance, Navigation and
Roll rate (°/s)

-60
Control Conference and Exhibit, no. August, pp. 1–16, 2007.
[6] S. S. Vaddi and P. K. Menon, “Numerical SDRE Approach for Missile
Integrated Guidance-Control,” in AIAA Guidance, Navigation and
-80

-100
Control Conference and Exhibit, pp. 1–17, 2007.
-120 [7] E. W. Bogdanov Alexander, “State-Dependent Riccati Equation Control
-140
for Small Autonomous Helicopters,” Journal of Guidance, Control, and
Dynamics, vol. 30, no. 1, pp. 47–60, January 2007.
18.1 18.2 18.3 18.4 18.5 18.6 18.7 18.8 18.9 19
Time (s)

Fig. 9 (c) Roll rate generated for correcting 10° roll error at 6000m [8] B. A. Steinfeldt and P. Tsiotras, “A State-Dependent Riccati Equation
Approach to Atmospheric Entry Guidance,” in AIAA Guidance,
altitude Navigation, and Control Conference, Toronto, 2010, pp. 1–20.
10 [9] F. Tyan and J. F. Shen, “SDRE Missile Guidance Law,” in 8th IEEE
International Conference on Control and Automation, Xiamen, 2010,
8
pp. 866–870.
6
[10] Tayfun Cimen, “State-Dependent Riccati Equation (SDRE) Control: A
Survey,” in 17th World Congress, IFAC, Seoul, 2008, pp. 3761–3775.
Roll angle (°)

4 [11] Tayfun Cimen, “Survey of State-Dependent Riccati Equation in

International Science Index, Aerospace and Mechanical Engineering Vol:6, No:12, 2012 waset.org/Publication/2792

Nonlinear Optimal Feedback Control Synthesis,” Journal of Guidance,

2
Control, and Dynamics, vol. 35, no. 4, pp. 1025–1047, 2012.
0
[12] T. Çimen, “Systematic and effective design of nonlinear feedback
controllers via the state-dependent Riccati equation (SDRE) method,”
-2
30.6 30.7 30.8 30.9 31 31.1 31.2 31.3 31.4 31.5
Annual Reviews in Control, vol. 34, no. 1, pp. 32–51, Apr. 2010.
Time (s)
[13] J. S. Shamma, J. R. Cloutier, E. Air, and F. Base, “Existence of SDRE
Fig. 10 (a) Roll error decay profile for 10° roll error at 7500m Stabilizing Feedback,” Aerospace Engineering, vol. 1, no. 3, pp. 1–14,
altitude 2002.
[14] K. D. Hammett, C. D. Hall, and D. B. Ridgely, “Controllability Issues in
10
Nonlinear State-Dependent Riccati Equation Control,” Journal of
5
Guidance, Control, and Dynamics, vol. 21, no. 5, pp. 767–773, 1998.
[15] A. Bracci, M. Innocenti, and L. Pollini, “Estimation of the Region of
Canard deflection angle (°)

0 Attraction for State-Dependent Riccati Equation Controllers,” Journal of

Guidance, Control, and Dynamics, vol. 29, no. 6, pp. 1427–1430, 2006.
-5
[16] Q. M. Lam, “SDRE Control Stability Criteria and Convergence Issues:
-10
Where Are We Today Addressing Practitioners’ Concerns?,” in
Infotech@Aerospace, California, 2012, pp. 1–20.
-15 [17] S. Elloumi, I. Sansa, N. B. Braiek, and L. Syst, “On the Stability of
Optimal Controlled Systems with SDRE Approach,” in 9th International
-20
30.6 30.7 30.8 30.9 31
Time (s)
31.1 31.2 31.3 31.4 31.5 Multi-Conference on Systems, Signals and Devices, 2012, pp. 1–5.
[18] J. R. Cloutier and D. T. Stansbery, “The Capabilities and Art of State-
Fig. 10 (b) Canard deflection for correcting 10° roll error at 7500m Dependent Riccati Equation-Based Design,” in American Control
altitude Conference, Anchorage, 2002, pp. 86–91.
20
[19] S. Katsev, “Streamlining of the State-Dependent Riccati Equation
Controller Algorithm for an Embedded Implementation,” Rochester
0 Institute of Technology, 2006.
[20] M. Khalil, H. Abdalla, and O. Kamal, “Trajectory Prediction for a
-20
Typical Fin Stabilized Artillery Rocket,” in 13th International
Roll rate (°/s)

Conference on Aerospace Sciences & Aviation Technology, Cairo, 2009,

-40
pp. 1–14.
-60
[21] M. K. Siddiq, J. C. Fang, W. B. Yu, “Monte Carlo Trajectory
Simulations of a Roll Stabilized 122mm Artillery Rocket” in
-80 International Conference on Computer and Electrical Engineering,
Hong Kong, 2012.
-100
30.6 30.7 30.8 30.9 31 31.1 31.2 31.3 31.4 31.5
Time (s)

Fig. 10 (c) Roll rate generated for correcting 10° roll error at 7500m
altitude

REFERENCES
[1] P. K. Menon, T. Lam, L. S. Crawford, V. H. L. Cheng, and L. Altos,
“Real-Time Computational Methods for SDRE Nonlinear Control of
Missiles,” in American Control Conference, pp. 232–237, 2002.
[2] C. P. Mracek and J. R. Cloutier, “Full Envelope Missile Longitudinal
Autopilot Design Using the State-Dependent Riccati Equation Method,”
in AIAA Guidance, Navigation and Control Conference, pp. 1697–1705,
1997.
[3] J. R. Cloutier and D. T. Stansbery, “Nonlinear, Hybrid Bank-to-
Turn/Skid-to-Turn Missile Autopilot Design,” in AlAA Guidance,
Navigation, and Control Conference, vol. 298, no. 0704, pp. 1–11, 2001.

International Scholarly and Scientific Research & Innovation 6(12) 2012 2822 scholar.waset.org/1307-6892/2792