2018 IEEE 7th Data Driven Control and Learning Systems Conference

May 25-27, 2018, Enshi, Hubei Province, China

Application of ADRC in Stability Control of Tank Gun System

Yang Lei1, Jing Xu1, Qiang Hao1
1. China North Vehicle Research Institute, Beijing, 100072, China
E-mail: [email protected]

Abstract: Overcoming the disturbance of vehicle body is an important function of the tank gun control system. Therefore, the
tank artillery can be stabilized to a given angle in space. However, due to the change of load factors such as inertia, friction and
clearance, the stability precision of the gun control system cannot be further improved, and hitting probability will drop under the
condition of tank high speed maneuverability. In this paper, the tank gun control system and the disturbance model were analyzed,
and the active disturbance rejection controller was designed. The control performance of this active disturbance rejection
controller was verified by MATLAB simulation and the experiments of analog prototype. The results showed that the application
of the ADRC is more effective than the traditional PID controller in improving the stability precision of gun control system.
Key Words: active disturbance rejection control (ADRC), tank gun control system, stable control, stability precision

However, many kinds of modern control theories, to a

1 Introduction great extent, are not realistic. Because they demand an
As an offensive weapon, the firepower performance accurate mathematical model which is not possible to be
occupies a decisive position in the whole tank system. The obtained in some specific conditions [8]. Recognizing the
tank gun control system plays a key role on firepower vulnerability of the reliance on accurate mathematical
performance which can achieve rapid artillery, accurate model, there has been a gradual recognition over the years
targeting and gun stability. However, tanks are facing with that active disturbance estimation is a viable alternative to an
the demands for hitting targets fleetly and accurately under accurate plant model [9]. The focal point is how external
the condition of high maneuverability in the future. So, it is disturbance and unknown dynamics can be estimated. ADRC
necessary to propose higher stability precision for tank gun was firstly proposed by Han in [10], explained carefully by
control system. It can keep the stability of artillery in vertical Han in [11, 12], analyzed in depth by Gao in [13] and Huang
and azimuth when the tank gun control system is disturbed in [14]. ADRC is a control method that does not depend on
by the movement of the vehicle body. At present, the the accurate mathematical model of the unknown object. By
traditional PID control method is adopted in the tank gun real-time estimation and compensations of the internal and
control system. This method would lead to a certain limit on external disturbances of system, combining with nonlinear
the performance improvement and adaptability of the gun control strategy, it can get better static and dynamic
control system, because the control parameters of the system performances, strong robustness and adaptability. Since
were not changed correspondingly. The corresponding ADRC does not depend on the accurate model of the system,
compensation correction processing was also not carried out it is very robust against parameter variations, disturbances
when the load was changed. For example, the vehicle body and noises, not only in some operation areas but also in the
changes 5 degrees angle in vertical, the turret rotation inertia whole working area [15-20]. Through the above literature, we
varies in 8%, the number of ammunition and optional armor can see that the active disturbance rejection control (ADRC)
are changed. These would effect on the mass of the turret and is a good way to improve stabilization precision of tank gun
cause the corresponding changes of inertia and friction. With control system.
the continuous application of the weapon control system, the In this paper, the whole composition of tank gun control
friction force and mechanical clearance have also changed system was explained firstly. For convenient simulation,
greatly with the factory state. corresponding models of gyroscope, permanent magnet
In view of the demands of gun control system, many synchronous motor (PMSM), motor driver were established.
scholars have tried to do a series of research at home and For simulating disturbance, variable disturbance models for
abroad. They used the variable structure control [1, 2], optimal friction, inertia and mechanical clearance of artillery
control [3, 4], fuzzy control [5] and robust control [6]. Ref. [1] actuator were designed. Then, the active disturbance
showed variable structure control (VSC) with a sliding mode rejection controller was designed to perform MATLAB
to solve robustness problem. Ref. [2] introduced sliding simulation and physical testing. It achieves good control
mode control combined with adaptive fuzzy control of gun effect when compared with the traditional PID controller.
control system for tank tracking and disturbance rejection. Finally, the application of active disturbance rejection
Ref. [7] studied the disturbance rejection problem for a control technology for tank gun control system was analyzed
linear model of the weapon using model predictive control and summarized.
due to its ability to handle the constraints. 2 Model Building
The block diagram of tank gun control system is shown in
Fig.1. The gun plays the control object in Fig.1. It can be
*This work is supported by the army’s plan for 13th Five-Year of China
stabilized in the position of space angle by the stability of the
5620000.

978-1-5386-2618-4/18/$31.00 ©2018 IEEE DDCLS'18

670
control system when the vehicle body produces azimuth and φf
Where p = d K u and Ku is induction
, =
vertical angle vibrations during the driving process. dt La La
electromotive force coefficient.
The mathematical model above can be described by the
block diagram, as shown in Fig.2.
1
Uq
La

•
Fig.1: The block diagram of tank gun control system Ku − + iq iq
ωr
La

−
2.1 Mathematical Model of Gyroscope
It shows that gyroscope is a sensitive measuring element Rs
for gun control system from Fig.1. The azimuth and vertical La
angles and angular velocities of the artillery can be measured
by the gyroscopes. The measurement data is used to produce Fig.2: Relationship of voltage, speed and current of PMSM
an electrical signal that is stable and aimed. The transfer
function is, The state equation for permanent magnet synchronous
kp (1) motor can be written as follows,
T ( s) = e−Ts
1  Rs φf  uq 
s +1  •  − −   
2π f i La   q   La 
i (5)
 q  =  La +
where kp is the magnification of the gyroscope, and its •   K ω   

B  r T
actual demarcation value is 0.995. f is the closed loop ω r   t −  − l 
 J J   J 
bandwidth of the gyroscope, and its value is 100Hz. T is the So we get the following block diagram, as shown in Fig.3.
output delay time of the gyroscope, and its value is 1ms. Tl

2.2 Mathematical Models of PMSM and Motor Driver 1

J

For the permanent magnet synchronous motor which is

shown in Fig.1, the direction of the fundamental magnetic
• •
1 + iq Kt + − ωr ωr
 
field of the motor permanent magnet is selected as the d axis. La
−
− J −

Along the d axis direction of rotation angle, the q axis is

Rs
ahead of the d axis for 90 degrees of electric angle. The La
B
J
counterclockwise direction is set as a rotating positive
direction. So, the mathematical model of the permanent Ku
magnet synchronous motor is obtained as, La

 uq   Rs + La p ωr La   iq  ωrφ f 
u  =  −ω L + (2) Fig.3: The structure of PMSM
 d  r a Rs + La p  id   0 
The instantaneous torque equation is, 2.3 Establishment of Disturbance Model
T em = Kt iq = Tl + BΩr + JpΩr (3)
Due to the influence of nonlinear friction, the dynamic and
where Rs is the stator resistance, La is armature inductance, static characteristics of tank gun control system become
J is the moment of inertia of the motor, φ f is magnetic flux worse. The stabilization precision of the system is down and
the limit cycle of the system is even concussion. Friction
produced by permanent magnets( φ f is a constant value), force is considered as the main form of external disturbance
to the gun control system when the vehicle body is moving.
ωr is angular frequency of rotor, B is viscous friction In this paper, Coulomb friction and viscous friction are
coefficient, p is differential operator, Tl is load torque, simplified and used to establish a disturbance friction model.
The expression is as follows,
K t is torque coefficient, ud , uq , id and iq are the voltage  Fk
Vr , Vr < Dv
and current of the d axes and q axes, Ω r is angular velocity Ff =  Dv (6)
of rotor. F , Vr > Fk
 k
For the motor driver which is shown in Fig.1, the
where F f is friction force, Vr is relative velocity, Dv is
controlling method of SVPWM which id equals 0 is used to
describe the model. Then, we can get the equation, the critical point of velocity, Fk is the maximal friction
• uqR ωφ force.
iq = − s iq − r f (4)
La L a La

DDCLS'18
671
 In this model, the characteristics of the friction force at the control system. In this way, the disturbance of the gun is
zero point of the relative sliding velocity are linearly reduced and the adaptability is enhanced, which is more
represented by a straight line with a certain slope. The slope conducive to the aiming and shooting in high-speed
of the straight line is determined by Dv and Fk which is the maneuvering. b) Using TD to extract differential signal, and
improve the speed of tank gun control system, and realize
maximum static frictional force. In this way, the friction
small overshoot and small state error control. Thus, the gun
model can be simplified to a single function model of relative
control system has better dynamic performance and higher
velocity.
control precision, which can quickly track and attack target.
Mechanical clearance is an unavoidable nonlinear
c) Using NLSEF, it can arrange the composition of control
disturbance in the transmission of power. There are two main
quantity, which can improve the efficiency of the feedback
types of nonlinear perturbations from the gun control system.
and the system performance at low speed.
One is the relative motion between the system driving part
Based on above ideas, the ADRC controller is designed
and the driven part in the mechanical clearance. The other is
for the speed loop and position loop of the gun control
the instantaneous impact and elastic deformation process
system. The structure model is shown in Fig.5.
caused by the difference of the speed between the driving p _ ref =0
and the driven parts when they contact again at the end of the β1
mechanical clearance. The first kind of disturbance would
cause the drive delay of the system. The elastic deformation
of the second kinds of disturbances would change the
v _ ref
β2 1 u y
b0
oscillation of the system. The model, as shown in Fig.4, is z2
established for the nonlinear disturbance of mechanical b0
clearance.

τ
u 1 1 ω 1 Δθ 1 ωm
Kt 1/ s
R + Ls Js s Jms
Fig.5: Internal structure model of ADRC
Ke
In Fig.5, ESO represents the second order expansion state
observer, TD represents the tracking differentiator, and other
Fig.4: The mathematical model of mechanical clearance parts are nonlinear state error feedback (NLSEF). The design
structures of ESO, TD, and NLSEF are described in detail
As shown in Fig.4, Error! Reference source not below.
found. τ ( s ) is the output torque of the clearance link. The 3.1 Extended State Observer
model can be described as follows,
The input of extended state observer (ESO) is the actual
k[Δ(θ ) − a] Δ(θ ) > a speed of the load and the current command. The output of
 (7)
τ ( s ) = 0 −a ≤ Δ(θ ) ≤ a ESO is the error estimate of the system. The order of ESO is
k[Δ(θ ) + a] Δ(θ ) < a second in the model, and the discrete form is as follows,

where Δ(θ ) is the relative displacement of gear clearance, e = z1 − y

2a is the width of gear clearance, k is the coefficient of  z1 (k + 1) = z1 (k ) + h [ z2 (k ) − β 01 fal (e1 , a1 , δ1 ) + b0u ] (9)
rigidity. According to the mechanical design manual, the 
calculation formula of torsional rigidity of the solid axis of  z2 (k + 1) = z2 (k ) + h [ − β 02 fal (e2 , a2 , δ 2 ) ]
unit length is as follows. where y is the speed of the controlled object, u is the
T observation of actual current, z1 is the state of the ESO,
k = 7350 4 (8)
d z 2 is the output of the ESO, b0 、 β 01 、 β 02 are the gain of
Where T is the torque transmitted by a shaft, d is the
diameter of the axis. ESO, h is the sampling step length, fal is the nonlinear
function, which is shown as below,
3 ADRC Controller Design
 e
For the disturbance of load change, ADRC controller is  1− a , e ≤ δ (10)
fal (e, a, δ ) =  δ
designed to compensate. The designed ADRC controller  e a sign(e), e > δ
consists of three parts. First, the tracking differentiator (TD) 
is used to correctly extract differential signals. The second is where e is the input error, a is the nonlinear coefficient,
the extended state observer (ESO), which is used to estimate δ is the critical point of error.
system disturbance. The third is the nonlinear state error
feedback (NLSEF), which is used for the rational 3.2 Tracking Differentiator
combination of system signals. Tracking differentiator (TD) is designed to handle the
The main idea of designation is as follows. a) Using ESO, transition process of reference signals. The input of the TD is
it has a good transition process and the ability to detect the the speed command. The output of the TD is the speed
undetermined disturbance, which can solve the influence of command after the transition process. The discrete form of
non-uniform friction torque and gear clearance in the gun TD is designed as follows,

DDCLS'18
672
 e = z1 − y
 (11)
 z1 (k + 1) = z1 (k ) + h [ − β * fal (e, a, δ )]
where y is the input of differentiator, z1 is the state of
differentiator, h is the sampling step length.
3.3 Nonlinear State Error Feedback (NLSEF)
The structure of nonlinear state error feedback (NLSEF) is
reflected in Fig.5. The signal was combined by NLSEF. The
output of NLSEF is the q axis current command, and the
discrete form is as follows,
e1 = p _ ref − position
 (12)
e2 = v _ ref − v
iq _ ref = β * fal (e , a , δ ) + β * fal (e , a , δ ) − d Fig.7: Azimuthal disturbance (30km/h)
 1 1 1 1 2 2 2 2

where, e1 is the position error, e2 is the speed error, When the tank is affected by the disturbance which is
shown in Fig.7, the stabilization effect of the PID controller
p _ ref is the position command, position is the actual and ADRC is shown in Fig.8. It can be calculated that the
position, v _ ref is the speed command, v is the actual stability precision of PID controller is 0.8mil, while ADRC
speed, d is the output of ESO, β1 is the position error gain, controller is 0.72mil.

β2 is the speed error gain, a1 is the nonlinear coefficient of

the position controller, a2 is the nonlinear coefficient of the
speed controller, δ 1 is the error critical parameter of
position controller, δ 1 is the error critical parameter of
speed controller, iq _ ref is the output of NLSEF, which is
the current command of the system.
4 Verification Based on MATLAB Simulation
Fig.8: The stabilization effect (PID VS ADRC, 30km/h)
The simulation based on MATLAB is carried out
according to the model established in section 2 and the The clearance was increased by 50% on the basis of the
ADRC controller designed in section 3. Using the simulation perturbation of Fig.8. The stabilization effect of the PID and
verified the effectiveness of ADRC controller. Firstly, the ADRC controller is shown in Fig.9. It can be calculated that
sine disturbance signal was given to the azimuth of the tank. the stability precision of PID controller is 1.16mil, while
The disturbance frequency is 0.8 Hz and the amplitude is 1.5 ADRC controller is 1.04mil.
degrees. Then, the stabilization effect of the PID controller
and ADRC controller is shown in Fig.6. It can be seen from
Fig.8 that the stability precision of PID controller is 1mil,
while ADRC controller is 0.89mil.

Fig9: The stabilization effect (clearance increased by 50%)

The inertia was increased by 50% on the basis of the

perturbation of Fig.8. The stabilization effect of the PID and
ADRC controller is shown in Fig.10. It can be calculated that
Fig.6: The stabilization effect (PID VS ADRC)
the stability precision of PID controller is 0.83mil, while
ADRC controller is 0.7mil.
The disturbing curve of the azimuth is shown in Fig.7
when the tank is traveling at 30 kilometers per hour on the
moderate rolling road.

DDCLS'18
673
 loaded on the simulation prototype. The practical
photograph is shown in Fig.12.
Prototype of tank
gun control system Target sheet

SDOF
Platform

Fig10: The stabilization effect (inertia increased by 50%)

The frictional force was increased by 50% on the basis of

the perturbation of Fig.8. The stabilization effect of the PID
and ADRC controller is shown in Fig.11. It can be calculated
that the stability precision of PID controller is 0.85mil, while
Fig.12: practical photograph of stability control of SDOF
ADRC controller is 0.74mil.
Sine wave with frequency 0.3-2.1Hz, amplitude of 0.8
degree (about 14 densities) is loaded to the azimuth direction
of the simulation prototype by the six degree of freedom
(SDOF) platform. The target scales hitting by the laser pen
are observed under two different control strategies. At the
same time, the gyroscope sensor information of the
simulation prototype is collected. The measured stability
effects of the two control strategies are shown in Fig.13 and
Fig.14. The frequencies are 0.6Hz and 1.5Hz.

Fig11: The stabilization effect (frictional force increased by 50%)

From the above simulation results, it can be seen clearly

that the gun control system with ADRC controller has the
advantages of small overshoot and adjusting time, when
compared with the traditional PID. This is mainly because
that the ADRC technology uses ESO to observe the
disturbances accurately and makes accurate compensation.
The stabilization effect of PID controller and ADRC
controller is shown in table 1 under different disturbances. Fig.13: Stability effects of the two control strategies in 0.6Hz
Table 1: The stabilization effect (PID VS ADRC controller)
Stabilization Stabilization
Azimuthal
effect effect
disturbance
(PID) (ADRC)
0.8 Hz 1.5o
1mil 0.89mil
sine signal
1.5 Hz
1.25mil 1.02mil
sine signal
tank traveling
0.8mil 0.72mil
30km/h
clearance increased
1.16mil 1.04mil Fig.14: Stability effects of the two control strategies in 1.5Hz
by 50%
inertia increased by
0.83mil 0.7mil The next experiment was carried out on the basis of the
50%
frictional force above experiments. Seven groups of sinusoidal waves with
0.85mil 0.74mil different frequencies which are 0.3Hz, 0.6Hz, 0.9Hz, 1.2Hz,
increased by 50%
1.5Hz, 1.8Hz and 2.1Hz and same amplitude of 0.8 degree
5 Practicality Test (about 14 densities) are loaded to the simulation prototype
The experimental prototype of tank gun control system is by SDOF platform. The following results are obtained
fixed to the six degree of freedom (SDOF) platform for through the experiment, as shown in table 2.
doing stability control test. The horizontal and vertical
Table 2: The disturbance results of seven different sine waves
sinusoidal waves can be produced by the SDOF and be

DDCLS'18
674
 PID ADRC [5] D. H. Lin, Z. Teng, T. Chen, S. Yang, H. Zhang, H. Chi,
Improved fuzzy control method for temperature in water tank
stabilization stabilization
Frequency of intelligent viscometer, Proceedings of Information and
precision precision
(Hz) Automation, Changsha, China, 2008: 1106–1110.
(mil) (mil)
[6] D. Wang, H. Yan, and J. H. Qiao, Robust control research on
0.3 0.3274 0.3059 adaptive wavelet neural netwaor identification for tank
0.6 0.4173 0.3262 all-electric gun control system. Inrelligent Control and
sinusoidal
Information Processing(ICICIP), 2011: 375-379.
disturbance 0.9 0.4416 0.4031 [7] D. G. Kumar, P.Y. Tiwari, V. Marcopoli, M.V. Kothare, A
in azimuth
1.2 0.5192 0.4395 study of a gun-turret assembly in an armored tank using
model predictive control, in Proceedings of American
1.5 0.5605 0.4877
Control Conference, St. Louis, MO, 2009: 4848–4853.
1.8 0.6156 0.5403 [8] D. Z. Gao, R. R. Rhinehart, Theory vs. practice forum, in
2.1 0.6388 0.5414 Proceedings of American Control Conference, Boston, MA,
2004: 1341–1349.
From the above data, it is not hard to know that the control [9] Y. Q. Xia, Q. Yuan, F. U. Meng-Yin, Z. H. Zhi, X. M. Ren,
Recent development in sliding mode control and active
effect of the ADRC controller is better than the traditional
disturbance rejection control, Control Theory& Applications,
PID controller in the frequency 0.3-2.1Hz of sinusoidal vol.30(2): 137-145, 2013
interference. It is proved that the ADRC algorithm works [10] J. Han, Control theory: is it a theory of model or control,
better in the stabilization precision of tank gun control Systems Science and Mathematical Sciences, 1(4): 328 –
system. 335,1989.
[11] J. Han, From PID to active disturbance rejection control.
6 Conclusions and Further Remarks IEEE Transactions on Industrial Electronics, 31(3):
900–906, 2009.
Aiming at the stabilization precision reduction of gun [12] J. Han, Active Disturbance Rejection Control Technique - the
control system, the active disturbance rejection control Technique for Estimating and Compensating the
method was adopted when the controlled device changes in Uncertainties. Beijing: National Defense Industry Press,
this paper. In this method, the load fluctuation such as 2008, 35-52.
friction, inertia and mechanical clearance of gun were [13] Z. Gao, Active disturbance rejection control: a paradigm shift
attributed to ‘unknown disturbance’. Then, the input and in feedback control system design, in Proceedings of tje
American Control Conference. Piscataway, NJ: IEEE, 2006:
output data of the model were used to estimate and 2399–2405.
compensate for the unknown disturbance. The control [14] Y. Huang, W. Xue, and C. Gao, Active disturbance rejection
method was proved to be able to improve the stabilization control: methodology and theoretical analysis, Journal of
precision of gun control system greater than the traditional Sysems Science and Mathematical Sciences, 31(9):
PID control through the MATLAB simulation and physical 1111–1129, 2011.
testing. It has showed good robustness and adaptive ability [15] T. G. Pedro, G. Germain, Design of an Optimal PID
and provides a strong guarantee for the precise strike of tanks Controller for a Coupled Tanks System employing ADRC,
in the future. IEEE LATIN AMERICA TRANSACTIONS. Vol 15(2),
189-196, 2017.
References [16] J. F. Sun, L. Shan, Z. D. Qi, Y. X. Qu, J. Li, Fraction order
ADRC algorithm to ship servo system, in Proceedings of the
[1] D. R. Dana, E. Kreindler, Variable structure control of a tank 35th Chinese Control Conference. 2016: 1086-1091.
gun, in Proceedings of Control Applications, Dayton, 1992: [17] C. Z. Zhao, Y. L. Xue, C. E. Huang, Tuning ADRC for
928–933. Multi-Variable System Based on Existing PID Parameters,
[2] D. L. Feng, X. J. Ma, Z. F. Yan, H. Li, Method of adaptive Internation Conference on Control, 2016:1272-1277.
fuzzy sliding mode control of gun control system of tank, [18] J. Li, X. H. Qi, Y. Q. Xia, D. L. Ma, S. F. Li, Y. W. Xu, On
Dian ji yu Kong Zhi Xue bao/ Electric Machines and Control the absolute stability of nonlinear ADRC for SISO systems,
11(1), 2007: 65–69. in Proceedings of the 34th Chinese Control Conference.
[3] D. Son, Kuswadi, M. N. Tamara, H. W. Dwi, Gun turret 2015:1571-1576.
automatic weapon control system design and realizatuin, [19] Z. H. Wang, Z. H. Liu, Application of active disturbance
Electronics and Smart Devices(ISESD), International rejection controller in wheeled mobile robot servo system,
Symposium on, 2016: 30-34 Navigation and Control Conference (CGNCC), 2014:
[4] J. Ma, D. D. Zhang, and W. Q. Yuan, et al, Electromagnetic 324-329.
gun loading control system based on cloud model, Industrial [20] S. J. Ma, M. W. Sun, Z. Q. Chen, Interactive ADRC design
Electronics and Applications (ICIEA), 2016 IEEE 11th for flight attitude control. Data Driven Control and Learning
Conferenca on, 2016: 957-962. System (DDCLS), 2017:611-616.

DDCLS'18
675