This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3117704, IEEE Access

Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/[Link] Number

Design and Verification of Attitude Control

System for a Boost-Glide Rocket
XIAOSHUAI FAN, XIBIN BAI, ZHENYU JIANG, LONGBIN LIU, AND SHIFENG ZHANG
College of Aerospace Science and Engineering, National University of Defense Technology, Changsha 410073, China
Corresponding author: Longbin Liu (longbuaa@[Link])
This work is supported in part by the National Natural Science Foundation of China under Grant 11802334.

ABSTRACT The boost-glide rocket needs to maneuver in the atmosphere with complex flight conditions,
which is a big challenge for the attitude control system. In this paper, the design process of the control
system is analyzed. With specific points of the flight trajectory, the design of improved PID controller
based on the small perturbation theory is carried out. The design approach of attitude control system is
verified by the six-degree-freedom model simulation. Rocket flight control program is written to the
onboard computer so that hardware-in-the-loop simulation experiment is completed. The experiment shows
that mean and standard deviation of error between simulation value and experiment value are below 0.5°, so
the control parameters and design method are reasonable with great attitude tracking effect. This paper
proposes the design and verification method of a boost-glide rocket attitude control system, which can
provide reference for design of the boost-glide rocket around the world.

INDEX TERMS Boost-glide, attitude control, PID, hardware-in-the-loop simulation.

I. INTRODUCTION disturbance will make the control process generate errors

Boost-glide rocket is the hotspot object of current research, and result in serious flight accidents.
which can maneuver in the atmosphere to complete specific Attitude control system is the basis for a boost-glide
flight missions. It can achieve great success with joint rocket to complete flight missions. Attitude divergence is a
execution of multiple rockets [1]-[2]. The boost-glide high frequency problem of rockets, which is also the focus
rocket has many advantages such as high speed, long range of research by scholars in related fields [4]-[9]. Common
and high accuracy. Design and application of the control control methods include adaptive control [10]-[12], control
system is a big challenge for high maneuver flight. The allocation [13]-[15], sliding mode control [16]-[18], multi-
boost-glide rocket changes attitude by rudder deflection, model switching control [19]-[21], nonlinear control, etc.
which requires fast tracking ability of the command attitude [22]. The design of controller with stability and strong
angle. It could hit the target and complete missions by robustness is the focus of current research. Wei et al. [23]
correcting attitude deviation constantly. proposed a formation keeping controller. In [24], Wei et al.
The boost-glide rocket obtains guidance instructions designed an adaptive control controller. In [25], Wang et al.
from measurement data of sensors, then generates control proposed a method to deal with tracking and attitude
torque so as to change the angular velocity and attitude control problems. However, most of the current advances
angle. Attitude control system is a complex mechatronics only remain in the theoretical level with pure numerical
system, which not only needs to ensure correctness of the simulation, lack of experimental verification, so it is
calculation algorithm, but also needs to ensure the normal difficult to obtain engineering application.
work of flight control program and the reliability of Taking a small boost-glide rocket as an example shown
hardware. The control system is composed of a variety of in Fig. 1, we will analyze the control system working
software and hardware, any problem in the loop will lead to principle and hardware composition, propose the transfer
failure of the whole mission [3]. However, the rocket is in a function calculation method and the control law design
bad flight environment for a long time, disturbance and method. The three-channel control law will be designed and
complex electromagnetic environment will lead to failure verified through nonlinear six-degree-freedom trajectory
frequently, the control chip’s voltage inversion caused by simulation and hardware-in-the-loop simulation experiment.

VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see [Link]
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3117704, IEEE Access

The experimental results have high reliability and II. DESIGN OF THE ATTITUDE CONTROL SYSTEM
universality, which can provide reference for attitude
control system design of the boost-glide aircraft. A. OVERALL CONCEPT DESIGN
There are three parts of overall concept design.
1) Composition of the attitude control system.
Block diagram of the attitude control system is shown in
Fig. 2.
While the rocket is flying in the atmosphere, it is affected
by various disturbances, which cause the actual attitude
angle to deviate from the command attitude angle. The
actual attitude angle is calculated by the integrated
navigation system. The attitude controller calculates rudder
deflection instruction according to attitude deviation and
transfers it to the rudder system to have deflection angle,
then it will generate control torque to change attitude of the
rocket.

FIGURE 1. A physical view of the propulsion glide technology test

aircraft.

Disturbance

Attitude deviation
Command Actual rudder deflection Actual attitude angle

Attitude controller Actuator Rocket

Rudder transfer function G(s)

Measured attitude angle

Attitude measurement system

FIGURE 2. Block diagram of the attitude control system.

The components of attitude control system are shown in
Fig. 3.

(e) Ground station (f) Telemetry hardware

FIGURE 3. Components of the control system for a boost-glide rocket.

Simulation is required for control system of the boost-
glide rocket. The flight process of rocket is simulated by
(b) Integrated navigation and rudder Real Time Model Solution Platform and integrated
(a) Body of the boost-glide rocket
controller navigation, flight control computer and rudder controller,
which are integrated into the integrated flight control
machine, as shown in Fig. 4.

(c) Flight control computer (d) The rudder system

(a) Integrated flight control (b) Real Time Model Solution Platform
machine

FIGURE 4. Simulation equipment of the attitude control system.

VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see [Link]
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3117704, IEEE Access

2) Block diagram simplification.  dV

When to design the control law, it is considered that there m dt  P cos  cos   X  mg sin  cos   Fcxv1

is no disturbance and the measuring sensor and actuator are mV d cos   P (sin  cos  cos  sin  sin )
ideal equipment. The block diagram of control law is  dt

simplified as shown in Fig. 5. Y cos  Z sin  mg cos   R cos  R sin
 
Attitude deviation Rudder deflection Actual attitude angle  I z1 z1  M z1  M z1  ( I x1  I y1 ) x1 y1
 (1)
Attitude controller Rocket  dx  V cos cos 
 dt

 dy  V sin  cos 
 dt

FIGURE 5. Simplified block diagram of the attitude control law.
 z1   cos cos    sin 
3) Design process of the boost-glide rocket attitude  = +
control law.
Variables of the lateral motion model include track yaw
First, we should calculate three-channel transfer function
angle, sideslip angle, yaw angle, yaw angular velocity, roll
of the boost-glide rocket, then complete transfer function of
angle, roll angular velocity, and lateral distance etc., which
the control system. Finally, adjusting control parameters to
can be expressed as
make performance of the control system meet requirements
 d
of indicators, the process is shown in Fig. 6. mV dt  P (sin  sin  cos  sin  cos )
Three-channel transfer function of the boost-glide rocket 
Y sin  Z cos  mg sin  sin   R sin  R cos
 
 I x1 x1  M x1  M x1   ( I y1  I z1 ) y1 z1
 
Transfer function of the attitude control system  I y1 y1  M y1  M y1  ( I z1  I x1 ) z1 x1
 dz (2)
  V sin 
 dt
Control parameters revision
 x1     sin

 y1   cos    cos sin 
 = +

 =
Meeting the No 2) Linearization of the boost-glide rocket movement
requments?
model.
The longitudinal and lateral motion models of rocket are
Yes
nonlinear, parameter time-varying with modeling errors, so
models need to be linearized to facilitate the derivation of
Finish transfer function. The perturbation equation of rocket flight
is derived by small perturbation method on theoretical
FIGURE 6. Design process of the boost-glide rocket attitude control
law. trajectory.
The longitudinal perturbation model is expressed as
 d V
m dt   P cos   X  V  ( P sin   X )  mg cos
B. ROCKET THREE-CHANNEL TRANSFER FUNCTION V V 

There are five steps to get rocket three-channel transfer 

 d   V d 
mV dt   P sin   Y  m dt  V   P cos   Y  

function. V

1) Movement model. 
 
 mg sin   Y  
The movement model of rocket is established, specific 
symbol meaning is referred to [26]. The model is divided  d  z1  z  (3)
 I z1 dt  M z1V  M z1  M z1  z  M z1  M zB
V

into longitudinal movement model and lateral movement 

    
model. Variables of the longitudinal model include velocity,   
velocity inclination angle, attack angle, pitch angle, pitch  z1
x  cosV  V sin 
angular velocity, transmitting orientation displacement and 
altitude etc., which can be expressed as y  sin V  V cos

VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see [Link]
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3117704, IEEE Access

The longitudinal perturbation model is also expressed as where

 d V   P cos cos   Z 
 mg sin  g sin 
 dt  a11V  a12   a13 b11   , b12  ,
 mV V
 d   a V  a   a   a  P sin   Y Z

 dt 21 22 23 25  b13 =  1, b14   , b15  0, b16  , b23  1,
 (4) mV mV
 d   a V  a   a   a   a  M

M y1 M x1 M 
 (8)
 dt 31 32 33 34 z1 35 
b31  y1 , b33  y1 , b35  y1 , b36  y1 , b45  1,
 d  I y1 I y1 I y1 I y1
 z1
 a41V  a42   a43  a44  z1  a45  
 dt M x1 M y1 M  x1 M 
b51  , b53  x1 , b55  x1 , b57  x1
where I x1 I x1 I x1 I x1
PV cos   X V The variable calculation method of aerodynamic
a11  , a12   g cos ,
m derivative is as follows
 P sin   X  Y  =C y1  q  S m
a13  , a14  0,  
m Y  =C y1  q  S m
PV sin   Y V  m g sin    
a21  , a22  ，  M z1 =CM z1  q  S m  Lk +Y   X cg =CM z1  q  S m  Lk +C y1  q  S m  X cg
mV V      
  M z1 =CMz1  q  S m  Lk +Y   X cg =CMz1  q  S m  Lk +C y1  q  S m  X cg
P cos   Y  Y  
a23  ，a24  0，a25  ，  M z =C z  q  S  L  Lk
mV mV  z1 M z1 m k
V

a31   a21  
P V
sin   Y V  m ， (5)   
 Z =Cz1  q  Sm
  
mV  Z =Cz1  q  S m (9)
g sin  P cos   Y       
a32   a22   ，a33   a23   ,  M y1 =CM y1  q  Sm  Lk  Z  X cg =CM y1  q  S m  Lk  C z1  q  S m  X cg
V mV      
  M y1 =CM y1  q  S m  Lk  Z  X cg =CM y1  q  S m  Lk  C z1  q  S m  X cg
Y  MV M 
a34  1, a35   , a41  z1 , a43  z1 ,  M y1y1 =CM y1  q  S m  Lk  Lk
mV I z1 I z1  y1
V
M z1z M  Lk
a44  , a45  z1  M x1 =CM  q  S m  Lk 
x1 x1

I z1 I z1  x1
V
   L
The lateral perturbation model is expressed as  M x1 =CM x1  q  S m  Lk  k
 V
 d 
mV dt    P cos  cos   Z    mg sin 

where Lk represents the characteristic length and S m
 represents the characteristic area.
+  P sin   Y    Z  cos  Fdz

 3) Derivation of three-channel transfer function for

 I d  x1  M    M x1   M  y1   M    M the boost-glide rocket.
 x1 dt x1 x1 x x1 y x1  xB
The velocity variables are long period movement, which

 I d  y1  M    M x1   M  y1   M    M (6) change slowly relative to angular variables. We usually
 y1 y1 y1 x y1 y y1  YB
ignore the small speed variable while designing the attitude
 dt
 x1     control law. Therefore, the longitudinal perturbation model

 y1    
can be rewritten as
      d 
  dt  a22   a23  a25 
   
z  V   d 
   a22   a23   z1  a25  (10)
 dt
The lateral perturbation model is also expressed as  d  z1
  a43  a44  z1  a45
 d   dt
 dt  b11  b12   b13 y  b14   b16 
 The model above can be obtained by Laplace transform
 d   b   s  a22 a23 0    ( s)   a25 
 dt 23 y1
 a
  s  a33 a34     ( s)  =  a35    ( s) (11)
 d  y1
32

  b31  b33 y1  b35  x1  b36  (7)  a42 a43 s  a44   z1 ( s)  a45 
 dt
 d  The transfer function from pitch rudder angle  (s) to
  b45  x1 velocity inclination angle  (s) can be obtained as follows
 dt
 d  x1
 dt  b51  b53 y1  b55  x1  b57  


VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see [Link]
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3117704, IEEE Access

  s  The roll channel model can be obtained as follows

G ( s ) 
  s   s b45     s    0 
(12)    =       s  (19)
a25 s 2  (a44 a25  a45 ) s  a22 a45 0 s  b55   x1  s   b57 

s  (a23  a22  a44 ) s 2   a44 (a22  a23 )  a43  s  a43a22
3
The transfer function from yaw rudder angle   s  to
The transfer function of pitch rudder angle  (s) to yaw angle   s  can be obtained
attack angle   s    s  b36 s  b11b36  b31b16
G  s    (20)
  s    s  s 3  (b11  b33 )s 2  (b31  b11b33 )s
G ( s ) 
  s  The transfer function from roll rudder angle   s  to
a25 s  a44 a25 s  a45a23  a43a25
2 (13) roll angle   s  can be obtained
  (s) 1 x1
s 3  (a23  a22  a44 ) s 2   a44 (a22  a23 )  a43  s  a43a22
1 b b57
G ( s)      57  (21)

 ( s) s   ( s) s s  b55 s 2  b55s
The transfer function from pitch rudder angle  (s) to
pitch angle   s  can be obtained Therefore, the transfer function of the three-channel of
 ( s)  ( s)  ( s) rocket can be obtained
G ( s )      a45 s  a45 (a23  a22 )  a43a25
 ( s)  ( s)  ( s) G ( s )  s 3  (a  a  a ) s 2  a (a  a )  a s  a a
(14)  23 22 44  44 22 23 43  43 22
a45 s  a45 (a23  a22 )  a43a25  
 b36 s  b11b36  b31b16
s  (a23  a22  a44 ) s 2   a44 (a22  a23 )  a43  s  a43a22
3 G  s   3 (22)
 s  (b11  b33 ) s 2  (b31  b11b33 ) s
The lateral perturbation model is obtained by Laplace   b
G ( s)  2 57
transform  s  b55 s
 s  b11 b12 b13 b14 0     s   where the coefficient aij , bij are called dynamic coefficient
 0  
 s b23 0 0     s   calculated according to theoretical trajectory.
 b31 0 s  b33 0 b35     y1  s   The transfer function of the three-channel can be
   
 0 0 0 s b45     s   rewritten as
 b b53 s  b55    x1  s     n1  s  n 0
 51 0 0
G ( s ) 
(15)  d 3  s  d 2  s 2  d1  s  d 0
3
 b16 0
 0 
0    n 1  s  n 0
   s   G ( s )  (23)
  b36 0   d 3  s  d 2  s 2  d 1  s
3

      s    n 0
 0 0 G ( s ) 
 0 b57   
d 2  s 2  d 1  s

The boost-glide rocket described in Fig. 1 is skid-to-turn where
(SST). The interaction between yaw channel and roll n1 =a45，n 0 =a45 (a23  a22 )  a43a25

channel is very small. Therefore, the cross partial d 3 =1，d 2 =a23  a22  a44，d1 =a44 (a22  a23 )  a43，d 0 =a43a22
derivatives of yaw and roll channels can be ignored, namely n 1 =b36，n 0 =b11b36  b31b16
 M y1x1  (24)
b35  0 d 3 =1，d 2 =b11  b33，d 1 =  (b31  b11b33 )
 I y1
n 0 =b57
 
 M x1
b51  0 (16) d 2 =1，d 1 =  b55
 I x1
4) Derivation of transfer function for the boost-glide
 
M x1y1
b53  0 rocket attitude controller.
 I x1 The improved PID controller is adopted for derivation.
Under the action of attitude control system, roll angle of The traditional PID controller is
t de(t )
rocket is small, so the projection of vertical normal force of u (t )  K pe(t )  Ki  e(t )  K d (25)
rocket is very small so that could be ignored in the lateral 0 dt
side, namely In a boost-glide rocket attitude control system, the
P sin   Y symbol e represents attitude angle error, which is
b14        0 (17)
mV difference between command attitude angle and actual
According to above simplified conditions, the yaw attitude angle. The differential term is improved, angular
channel model can be obtained as follows velocity of the boost-glide rocket is directly feedback.
 s  b11 0 b13     s    b16  Angular velocity is directly measured, which is easily used
 0  
b23      s   =  0    s 
to the control system. The three-channel controller can be
 s  (18) designed as
 b31 0 s  b33    y1  s    b36 

VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see [Link]
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3117704, IEEE Access

1 To select the specific points of flight trajectory, including

  K p   K i   K d s
s engine shutdown point (maximum dynamic pressure point),
1 highest trajectory point, pull starting point, parachute
  K p   K i   K d s (26)
s opening point, etc.
1 2) Calculation of the boost-glide rocket three-channel
   K p   K i   K d  s
s transfer function.
5) Transfer function of the boost-glide rocket attitude According to dynamic characteristics of specific points,
control system. the dynamics coefficient is calculated, then three-channel
Taking pitch channel as an example, the control loop is transfer function of the boost-glide rocket is obtained.
shown in Fig. 7. Yaw channel and roll channel are designed 3) To select the controller parameters.
in the same method as pitch channel To select the initial control parameters, K p , Ki , Kd .
According to the open-loop transfer function, Bode
Rudder deflection Pitch angle
Command pitch angle
+ diagram is drawn to calculate the frequency domain
+ characteristics.
-
According to the closed-loop transfer function, step
response curve is drawn to calculate the time domain
characteristics
To judge whether the performance indicators meet
FIGURE 7. Block diagram of pitch channel control loop. requirements. If not, adjusting the control parameters and
To select control parameters, the transfer function of recalculate performance indicators until performance
pitch channel can be obtained. Firstly, the open-loop indicators meet requirements. The control law of yaw
transfer function of pitch channel is calculated. The open- channel and roll channel can be designed as the same
loop control block diagram is shown in Fig. 8 method.
4) Verification and modification of attitude control
Command pitch angle
Rudder deflection Pitch angle law for the boost-glide rocket.
+
Rocket control law is designed by some hypotheses,
+
- including rocket linear movement model hypothesis, no
disturbance hypothesis, the ideal sensors and actuators
hypothesis, etc. The actual movement model is nonlinear
and time varying, control parameters should be verified and
FIGURE 8. Open-loop control block diagram of pitch channel.
corrected in six-degree-freedom simulation. If the attitude
According to control block diagram in Fig. 8, the open- diverges, we should adjust the control parameters according
loop transfer function of pitch channel is to simulation.
The selected control parameters of each specific point are
 Ki  
 K p   G ( s) taken as the initial value, the simulation time is set from
 ( s)  s   (27)
G0c ( s)   zero to this point. The control parameters are corrected
c ( s ) 1  K d sG ( s)
according to tracking derivation of the attitude angle along
To select a certain moment, the transfer function of pitch time. When attitude control correction of this specific point
channel can be calculated is finished, simulation time is extended to the next specific
K p  n1  s 2   K p  n 0  Ki  n1   s  Ki  n 0 point. The correction process is shown in Fig. 9.
G0c ( s)  (28)
d 3  s 4   d 2  K d  n1   s 3   d1  K d  n 0   s 2  d 0  s Initial specific point

Closed-loop transfer function is calculated by the open-

loop transfer function
Control parameters

G0c ( s)
G ( s)=
c
1+G0c ( s)
(29)
Revision
The open-loop transfer function and closed-loop transfer Next specific point Simulation

function of yaw channel and roll channel can be derived the

same as above method.
C. VERIFICATION PROCESS OF ATTITUDE CONTROL
LAW Meeting the No
requments?
There are four steps to verify attitude control law.
1) Selection of specific points for the boost-glide Yes
rocket attitude control system.
FIGURE 9. The process of control parameters correction.

VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see [Link]
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3117704, IEEE Access

The coordination relationship of the design, verification shown in Fig. 10.

and modification of the boost-glide rocket control system is

Initial specific point

Initial specific point control parameters

Dynamic coefficient Initial control parameters Yes

Adjusting control
parameters Other specific points
control parameters
Meeting the
Rocket transfer function Controller transfer function
requments?
No

(1)Amplitude margin
Open-loop transfer function
(2)Phase margin
Nonlinear six- degree-freedom
trajectory simulation

(1)Steady-state error
Closed-loop transfer function (2)Rising time
(3)Overshoot

Transfer function calculation method Control law design method Control law validation and revision method

FIGURE 10. The control law design process of rocket pitch channel.
is in the transmitting plane, therefore, the command yaw
III. NUMERICAL SIMULATION VERIFICATION angle and command roll angle of attitude control system are
zero. In order to realize the trajectory in Fig. 11, the boost-
A. NUMERICAL SIMULATION VERIFICATION glide rocket will get lift force by adjusting pitch angle and
The boost-glide rocket needs to glide in the atmosphere, so attack angle. Then the rocket will glide in the atmosphere,
a glide trajectory is designed. For designed trajectory, the theoretical pitch angle corresponding to trajectory is shown
boost-glide rocket will glide in the longitudinal plane and in Fig. 12.
suppresses yaw and roll movement. The designed glide
trajectory in the longitudinal plane is shown in Fig. 11.
Command yaw angle and command roll angle of the lateral
movement are both zero. It is a boost-glide trajectory so
attack angle of the rocket is zero in ascending stage and
changes in descending stage with a pull-up glide trajectory
to increase the range.

FIGURE 12. Theoretical pitch angle of the boost-glide rocket.

The theoretical pitch angle in Fig. 12 is as command
pitch angle of the rocket, the mission is to track theoretical
boost-glide trajectory. The goal of attitude control system is
to keep yaw angle and roll angle at 0, and the pitch angle
need to track command pitch angle. The rocket will realize
the boost-glide trajectory with great effect by tracking
FIGURE 11. Theoretical trajectory of the boost-glide rocket. command attitude angle.
Design of the small boost-glide rocket does not require
very strict trajectory tracking without considering the
guidance problem, so it only needs to control the three-
channel attitude. Trajectory of the boost-glide rocket theory

VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see [Link]
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3117704, IEEE Access

B. SPECIFIC POINTS SELECTION FOR THE BOOST- According to the attitude control law design method
GLIDE ROCKET ATTITUDE CONTROL SYSTEM proposed above, the time of engine ignition is regarded as
Design of control law based on three-channel transfer zero point. The time of pull starting point is 27s, the
function of the boost-glide rocket is aimed at a certain transfer function of pitch channel can be obtained as
moment in the flight trajectory. If whole flight process follows
should be controlled stably, the representative moment in 2.6636s  2.9725
flight trajectory should be selected. The small boost-glide G (s)  (30)
s3  2.2483s 2  59.4734s  3.4251
experimental rocket should have the function of full rocket where K p =-20，Ki =-20，Kd  3 , the open-loop transfer
recovery to analyze flight process efficiently. According to function of pitch channel is
parachute opening time, movement of the boost-glide 3052.2608s 2 +6458.4949s + 3406.2341
rocket before parachute opening point is mainly simulated, G0c ( s)  (31)
s 4 +460.0874s3 +570.4085s 2  3.4251s
and movement after parachute opening point has little value.
The closed-loop transfer function of pitch channel is
The control starting point, maximum dynamic pressure
3052.2608s 2 +6458.4949s  3406.2341
point, highest trajectory point, pull starting point and Gc (s)= (32)
s +460.0874s3 +3622.6693s 2 +6461.92s  3406.2341
4
parachute opening point are selected here. In order to
ensure coordination of the whole rocket movement, another Nyquist diagram, Bode diagram and step response curve
specific point is taken between the starting point and the according to transfer function above is shown in Fig. 13-
engine shutdown point, and another specific point is taken Fig. 15. Performance indicators according to transfer
before the rocket falls to the ground, so 7 specific points are function above is shown in Table 2.
selected total. The time of each specific point is determined
according to the theoretical trajectory as shown in Table 1.
TABLE 1. Selection of specific points.

The time of specific points Meaning

0.5s near control starting point
2s assistant point
3.8s near maximum dynamic pressure point
18.45 near highest trajectory point
27 near pull starting point
52 near parachute opening point
62 near fall down point

In table 1, the boost-glide rocket has maximum speed and

maximum dynamic pressure at engine shutdown point, and
pull starting point is the moment while the boost-glide
rocket starts to pull up. The parachute opening point is the
moment while attitude of the boost-glide rocket is near 0 in FIGURE 13. Nyquist diagram of pitch channel.
descending stage. The control parameters corresponding to
time between each specific point are obtained by linear
interpolation of specific points.

C. CONTROL LAW VERIFICATION

The performance indicators of attitude control system:
1) The amplitude margin is greater than 10dB.
2) The phase margin is greater than 45°.
3) Overshoot is less than 15%.
4) Rising time is less than 3 seconds.
5) Steady-state error is less than 0.01.
The boost-glide rocket needs to glide in the atmosphere
by changing attack angle and attitude angle. The moment of
pull starting point is valuable, which faces great difficulties
in attitude control system. Taking pitch channel attitude
control law design of pull starting point as an example to FIGURE 14. Bode diagram of pitch channel.
analyze. The design of three-channel attitude control
system at other specific points is similar to pitch channel at
pull starting point.

VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see [Link]
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3117704, IEEE Access

The six-degree-freedom model of rocket is established in

Real Time Simulation Platform, designed trajectory is
discretized according to specific points, then the linear
interpolation method is used to calculate the command
attitude angle. The selected PID control parameters at each
specific point are brought into the six-degree-freedom
model of rocket to obtain simulation results, as shown in
Fig. 16- Fig. 18.

FIGURE 15. Step response curve of pitch channel.

TABLE 2. Performance indicators of pitch channel at pull starting point.

Whether meeting
Names of indicators Magnitude
requirements
amplitude margin Gm inf dB
√
phase margin Pm 81.9082 °
√
overshoot σ% 8.3153%
√ FIGURE 16. Comparison between simulation value and command value
rising time tr 0.25068 s
√ of pitch angle curve.

steady-state error ess 0

√
open-loop poles s=0, -388.7166,
-0.9944, -0.01
√
closed-loop poles s=-381.9629, -5.7719,
-1.0282, -0.9579
√

PID parameters selected of other specific points are

shown in Table3.
TABLE 3. PID parameters of each specific point.
Time(s)
Channels Courses
[0.5 2 3.8 18.45 27 52 62]
P [1 1 1 1 2 2 2]*(-1)
roll channel I [1 2 2 1 1 1 1]* (-1)
D [0.1 0.2 0.2 0.1 0.5 0.5 0.5]* (-1)
P [1 0 2 2 3 5 5.8]* (-1)
yaw
I [10 15 25 5 1 1 1]* (-1) FIGURE 17. Comparison between simulation value and command value
channel
D [0.1 0.2 0.2 2 3 2 1.6]* (-1) of yaw angle curve.
P [1 2 5 10 20 10 6]* (-1)
pitch
I [1 1 1 2 2 2.5 2.7]* (-10)
channel
D [0.3 0.3 0.2 0.5 3 1.5 0.9]* (-1)

D. ANALYSIS OF SIMULATION RESULTS

Designed parachute opening time is 51s after engine
ignition, then the boost-glide rocket falls free for recovery.
It is enough to ensure that the control system works to the
parachute opening time. So the movement process from
engine ignition to parachute opening is simulated.
It is assumed that the boost-glide rocket is launched with
initial angle and the initial pitch angle is set as 65°, the
initial yaw angle is set as 5°, and the initial roll angle is set
as 10 °, so as to verify the control effect of the designed
attitude control system. FIGURE 18. Comparison between simulation value and command value
of roll angle curve.

VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see [Link]
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3117704, IEEE Access

In Fig. 16- Fig. 18, the x-coordinate is time, the y- that the design method of attitude control system is
coordinate is attitude angle, phi represents pitch angle, psi reasonable and ideal control effect can be achieved.
represents yaw angle, and gamma represents roll angle.
Command represents command attitude angle, Simulation IV. HARDWARE-IN-THE-LOOP SIMULATION
represents simulation attitude angle. VERIFICATION
We can draw a conclusion that selected PID control Real Time Simulation Platform, integrated flight control
parameters of specific points have good control effect machine, integrated inertial measurement machine, ground
according to simulation. The simulation value and the station, rotation platform and other hardware are used for
command value basically coincide. In the case of a large closed-loop hardware-in-the-loop simulation according to
initial error of yaw angle and roll angle, it can quickly the schematic diagram shown in Fig. 19.
converge and track the command attitude angle. It is proved
Zero point device
Parachute
opening device
Rudder feedback data
Zero point flag Control command
Parachute opening flag
Real Time Integrated flight
Simulation Platform Ground station
control machine
Simulated data
Simulated data and Telemetry data
measured angular velocity

Attitude angle Integrated inertial

measurement machine
Rotation platform

FIGURE 19. Schematic diagram of hardware-in-the-loop simulation.

Real Time Simulation Platform uses the rudder
deflection angle calculated by integrated flight control
machine to finish closed-loop trajectory simulation, which
drives rotation platform to simulate rocket flight attitude.
The simulated flight data is transmitted to integrated inertial
measurement machine, then the mixed data with simulated
data and measured angular velocity data is transmitted to
integrated flight control machine to complete guidance,
navigation and control calculation.
The simulation results of Real Time Simulation Platform
are used as standard values to verify hardware-in-the-loop
simulation control effect, as shown in Fig. 20- Fig. 23

FIGURE 21. Comparison between simulated values and experimental

values of yaw angle curve.

FIGURE 20. Comparison between simulated values and experimental

values of pitch angle curve.
FIGURE 22. Comparison between simulated values and experimental
values of roll angle curve.

VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see [Link]
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3117704, IEEE Access

Hardware-in-the-loop simulation experiment scheme is

proposed, and hardware of the small solid propellant boost-
glide rocket is connected to closed-loop experiment to
simulate real flight environment. Taking simulation results
of Real Time Simulation Platform as standard value and
taking simulation results of hardware-in-the-loop
simulation as experimental value, contrast between results
of simulation and experiment shows that they are close very
much. The mean and standard deviation of three-channel
attitude deviation are less 0.5°, it achieves ideal control
effect and verifies proposed design method for a small solid
propellant boost-glide rocket attitude control system.
Based on analysis with design process of the attitude
control system for a small solid propellant boost-glide
FIGURE 23. Deviation curve between simulated attitude angle and rocket, the hardware-in-the-loop simulation verification
experimental attitude angle.
In Fig. 20- Fig. 22, the x-coordinate is time, the y- scheme is designed, which can provide reference for the
coordinate is attitude angle, phi represents pitch angle, psi design and verification of relevant aircrafts, then more
represents yaw angle, and gamma represents roll angle. advanced control methods can be adopted in the future to
Simulation represents simulation attitude angle of Real achieve better control effect.
Time Simulation Platform. FlightControl represents
REFERENCES
experimental attitude angle of hardware-in-the-loop [1] I.-S. Jeon, J.-I. Lee, and M.-J. Tahk, ‘‘Homing guidance law for
simulation experiment. In Fig. 23, the x-coordinate is time, cooperative attack of multiple missiles,’’ J. Guid., Control, Dyn.,
the y-coordinate is deviation of attitude angle. vol. 33, no. 1, pp. 275–280, Jan./Feb. 2010.
[2] J. Zeng, L. Dou, and B. Xin, ‘‘A joint mid-course and terminal
To calculate the mean and standard deviation of error for course cooperative guidance law for multi-missile salvo attack,’’
hardware-in-the-loop simulation, as shown in Table 4. Chin. J. Aeronaut., vol. 31, no. 6, pp. 1311–1326, Jun. 2018.
[3] M. Chen, G. Tao, and B. Jiang, ‘‘Dynamic surface control using
TABLE 4. The mean and standard deviation of error for hardware-in-
the-loop simulation.
neural networks for a class of uncertain nonlinear systems with
input saturation,’’ IEEE Trans. Neural Netw. Learn. Syst., vol. 26,
Standard deviation of no. 9, pp. 2086–2097, Sep. 2015.
Attitude angle Mean of error(°)
error(°) [4] H.-J. Uang and B.-S. Chen, ‘‘Robust adaptive optimal tracking
pitch angle 0.2513 0.3560 design for uncertain missile systems: A fuzzy approach,’’ Fuzzy
yaw angle -0.1836 0.2731 Sets Syst., vol. 126, no. 1, pp. 63–87, 2002.
roll angle 0.0345 0.4101 [5] S.-H. Kim, Y.-S. Kim, and C. Song, ‘‘A robust adaptive nonlinear
It can be seen from Fig. 20- Fig. 23 that the numerical control approach to missile autopilot design,’’ Control Eng. Pract.,
vol. 12, no. 2, pp. 149–154, 2004.
simulation value is as the ideal control effect, and the [6] M. Hou, X. Liang, and G. Duan, ‘‘Adaptive block dynamic surface
hardware-in-the-loop simulation value is as the actual control for integrated missile guidance and autopilot,’’ Chin. J.
control effect. The deviation of pitch angle is kept within 1°, Aeronautics, vol. 26, no. 3, pp. 741–750, 2013.
and the deviation of yaw angle and roll angle converges [7] X. Lidan, Z. Ke’nan, C. Wanchun, and Y. Xingliang, ‘‘Optimal
control and output feedback considerations for missile with blended
quickly and approaches zero. The control effect is good and aero-fin and lateral impulsive thrust,’’ Chin. J. Aeronautics, vol. 23,
desired effect is achieved. It is proved that the design no. 4, pp. 401–408, 2010.
method of attitude control system for a small solid [8] G. Yuan and X. Shi, ‘‘Nonlinear adaptive controller design for
missile system with special uncertainties,’’ (in Chinese), Elect.
propellant boost-glide rocket is feasible and the control Mach. Control, vol. 14, no. 5, pp. 104–108, 2010.
effect is good. [9] Y. Zhang, Z. Li, Z. Cheng, L. Liu and Y. Wang, ‘‘Attitude
Tracking Control Reconfiguration for Space Launch Vehicle With
V. CONCLUSION Thrust Loss Fault, ’’ in IEEE Access, vol. 7, pp. 184353-184364,
2019, doi: 10.1109/ACCESS.2019.2959836.
The design method of attitude control system is proposed [10] Q. Shen, C. Yue, C. H. Goh, and D. Wang, ‘‘Active fault-tolerant
based on dynamic model of a small solid propellant boost- control system design for spacecraft attitude maneuvers with
glide rocket. Through linearization of movement model, actuator saturation and faults,’’ IEEE Trans. Ind. Electron., vol. 66,
no. 5, pp. 3763–3772, May 2019.
three-channel transfer function of the boost-glide rocket is [11] Z. Gao, B. Jiang, P. Shi, J. Liu, and Y. Xu, ‘‘Active fault-tolerant
derived. An improved PID controller is used to design the tracking control for near-space vehicle attitude dynamics with
attitude control law, and designed attitude control law is actuator faults,’’ Proc. Inst. Mech. Engineers, I, J. Syst. Control
Eng., vol. 225, no. 3, pp. 413–422, 2011.
brought into six-degree-freedom model of the boost-glide [12] Q. Yang, S. S. Ge, and Y. Sun, ‘‘Adaptive actuator fault tolerant
rocket for verification. The attitude deviation converges control for uncertain nonlinear systems with multiple actuators,’’
quickly and control effect is great. It shows that the attitude Automatica, vol. 60, pp. 92–99, Oct. 2015.
control system for a small solid propellant boost-glide [13] T. A. Johansen and T. I. Fossen, ‘‘Control allocation—A survey,’’
Auto-matica, vol. 49, no. 5, pp. 1087–1103, 2013.
rocket proposed in this paper is very effective with ideal [14] M. E. N. Sørensen, S. Hansen, M. Breivik, and M. Blanke, ‘‘Perfor-
control effect. mance comparison of controllers with fault-dependent control

VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see [Link]
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3117704, IEEE Access

allocation for uavs,’’ J. Intell. Robotic Syst., vol. 87, no. 1, pp. 187–
207, Jul. 2017.
[15] Q. Shen, D. Wang, S. Zhu, and E. K. Poh, ‘‘Control allocation
based fault-tolerant control design for spacecraft attitude tracking,’’
in Proc. 53rd IEEE Conf. Decis. Control, Dec. 2014, pp. 4983–
4988.
[16] Y. Zhang, S. Tang, and J. Guo, ‘‘Adaptive-gain fast super-twisting
sliding mode fault tolerant control for a reusable launch vehicle in
reentry phase,’’ ISA Trans., vol. 71, pp. 380–390, Nov. 2017.
[17] C. Jing, H. Xu, X. Niu, and X. Song, ‘‘Adaptive nonsingular
terminal sliding mode control for attitude tracking of spacecraft
with actuator faults,’’ IEEE Access, vol. 7, pp. 31485–31493, 2019.
[18] X. Yin, B. Wang, L. Liu, and Y. Wang, ‘‘Disturbance observer-
based gain adaptation high-order sliding mode control of
hypersonic vehicles,’’ Aerosp. Sci. Technol., vol. 89, pp. 19–30, Jun.
2019.
[19] B. Jung, Y. Kim, and C. Ha, ‘‘Fault tolerant flight control system
design using a multiple model adaptive controller,’’ Proc. Inst.
Mech. Eng., G, J. Aerosp. Eng., vol. 223, no. 1, pp. 39–50, 2009.
[20] B. Jung, S.-K. Jeong, D.-H. Lee, and Y. Kim, ‘‘Adaptive
reconfigurable flight control system using multiple model mode
switching,’’ IFAC Proc. Volumes, vol. 38, no. 1, pp. 115–120, 2005.
[21] M. Rodrigues, D. Theilliol, and D. Sauter, ‘‘Fault tolerant control
design for switched systems,’’ Anal. Des. Hybrid Syst., vol. 39, no.
5, pp. 223–228, 2006.
[22] X. Lv, B. Jiang, R. Qi, and J. Zhao, ‘‘Survey on nonlinear reconfig-
urable flight control,’’ J. Syst. Eng. Electron., vol. 24, no. 6, pp.
971–983, Dec. 2013.
[23] C. Wei, Y. Shen, X. Ma, X. Ma, J. Guo, and N. Cui, ‘‘Optimal
formation keeping control in missile cooperative engagement,’’
Aircr. Eng. Aerosp. Technol., vol. 84, no. 6, pp. 376–389, Oct. 2012.
[24] C. Wei, J. Guo, B. Lu, Y. Shen, and L. Zhang, ‘‘Adaptive control
for missile formation keeping under leader information
unavailability,’’ in Proc. 10th IEEE Int. Conf. Control Automat.
(ICCA), Hangzhou, China, Jun. 2013, pp. 902–907.
[25] X. Wang, H. Fang, L. Dou, B. Xin, and J. Chen, ‘‘Integrated
distributed for-mation flight control with aerodynamic constraints
on attitude and control surfaces,’’ Nonlinear Dyn., vol. 91, no. 4, pp.
2331–2345, Mar. 2018.
[26] K.J. Chen, L.H. Liu and Y.H. Meng, “Flight dynamics and
guidance of long range rockets,” Beijing, China, NDIP., 2013, pp.
66-89.

VOLUME XX, 2017 1

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see [Link]