See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/290305212

A Method for Line of Sight Angle Rate Reconstruction Based on Strapdown

Seeker

Article · January 2014

DOI: 10.2991/meic-14.2014.147

CITATIONS READS

4 765

4 authors, including:

Ke Zhang
East China University of Science and Technology
70 PUBLICATIONS 890 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ke Zhang on 31 August 2019.

The user has requested enhancement of the downloaded file.

 International Conference on Mechatronics, Electronic, Industrial and Control Engineering (MEIC 2014)

A Method for Line of Sight Angle Rate

Reconstruction Based on Strapdown Seeker
Cong Nie*, Ke Zhang, Pei Wang, Meibo Lv
School of Astronautics
Northwestern Polytechnical University
Xi’an 710072 China
E-mail:[email protected]

Abstract—The line of sight (LOS) information which detected by model[4]. Pei Wang proposed using Nonlinear Tracking-
strapdown seeker is coupled with missile attitude information. Differentiator to estimate the angle rate[5].
On the other hand, since the instantaneous field of view is wider,
the noise levels of strapdown seeker are higher than those of LOS angle rate was obtained by using optimal filtering
conventional gimbals seeker. According to the relationship theory based on the model mentioned above. Since the range
among coordinates which are used to describe motion of missile, and relative velocity are unobservable to strapdown seekers, it
analysis of the attitude decoupling algorithm was deduced in is difficult to obtain a satisfactory result.
detail. Based on the proposed algorithm, the LOS reconstruction
In this paper, we focus on the strapdown seeker which can
filter was designed to obtain more precise line of sight angle,
where, the theory of Central Difference Kalman Filter (CDKF) only get LOS in body coordinates. Based on the relationship of
which based on Central Difference (CD) method was applied to LOS, LOS in body frame (BLOS) and rate gyros’ information,
deal with the filtering problem of nonlinear system. Compared the LOS first-order dynamic model for strapdown seeker was
with EKF, the solving of Jacobi matrixs can be avoided and only constructed, and a LOS reconstruction filter was designed by
one adjustable parameter h was introduced which leads to a using the model and CDKF method.
reduction of the computational cost. Based on this, the algorithm CDKF have some advantages such as, three order precision
was suitable for engineering application. The results of computer
for Gauss distributed random variable, less cost of
simulation confirmed the validity of the proposed inertial LOS
reconstruction filter, which can be used to improve the precision
computations compared with UKF theory. The results of
of inertial LOS and LOS rate estimation. The method meanwhile simulation in MATLAB environment have showed that the
can be used as a new method to estimate inertial LOS rate. method has a good performance in estimating the LOS angle.
The precision of the estimation can be improved based on this
Keywords—Strapdown Seeker, LOS Rate Reconstruction, method.
Central Difference Kalman filter, Fadin- MemoryFilter
II. THE LOS RECONSTRUCTION METHODS
I. INTRODUCTION
A. the relationship among the required coordinates
Proportional navigation (PN) guidance law based on line of
sight (LOS) angle and its rate was applied in most modern We take the ground coordinate as the inertial frame. The
homing seeker missile. All hardware structure of the detectors coordinate systems which needed to reconstruct LOS angle rate
were directly fixed on the strapdown seeker. The structure of were showed as Fig .1. The relationships of coordinate
guidance system is simple and reliable by removing transformation among these coordinate systems were showed
stabilization and tracking platform. The problems of as Fig .2.
application for some tactical missiles which was limited by The yaw angle qIH and the pitch angle qIV in the inertial
space were solved. Adopting Strapdown seeker makes the
integrated guidance and control design become possible. reference frame S I are essential to implement guidance law
However, it also brings two serious disadvantages. Firstly, only design. Since the optical system was fixed on the missile’s
LOS angle information relative to body coordinate system can body, only the yaw angle qBH and pitch angle qBV of LOS in
be measured, thus the metrical LOS angles will couple with body frame S B can be obtained. Missile attitudes in inertial
body attitude motion which finally leads to high nonlinearity.
Secondly, because of the wider instantaneous field of view, space are changing all the time during the process of guidance.
more serious measurement noise especially non-Gaussian noise Hence, it is essential to establish the “Mathematic Platform” to
will be induced into the strapdown guidance system compared decouple the missile attitudes information from strapdown
with the conventional system[1]. Thus, Jan K proposed using seeker’s detecting information.
filter method[2] to obtain the target line of sight angle rate.
Tingting Sun establish the mathematical model of stapdown
optical image seeker[3]. Guojiang Zhang establish the second-
order of the LOS angle based on the target maneuvering
This work is supported by National Natural Science Foundation (NNSF)
of China under Grant 61101191 and the Fund of Aeronautics Science of China
under Grant 20110153003.

© 2014. The authors - Published by Atlantis Press 658

 xC  xL
yB yC  a1  cos  cos cos qBV cos qBH  sin  cos cos  sin qBV
yI   sin sin  sin qBV  sin  cos sin  cos qBV sin qBH

  sin cos  cos qBV sin qBH
xB

 sin  cos qBV cos qBH  cos  cos  sin qBV
yL
a2 (5)

  cos  sin  cos qBV sin qBH
O  a3   cos  sin cos qBV cos qBH  sin  sin cos  sin qBV
xI 
  cos sin  sin qBV  sin  sin sin  cos qBV sin qBH
  cos cos  cos qBV sin qBH
zI

So the decoupling equation was established. We can
zB zC zL
reconstruct the qIH and qIV according to missile attitudes
which was measured by strapdown seeker. The inertial LOS
Figure 1 the relationship of coordinate systems
rate can be obtained by differential network. The method
L( , , )
proposed above ignored the second-order dynamic
Earth Body
characteristics of BLOS angle and missile attitudes. The
coordinate coordinate
extraction of the LOS angle rate dynamic performance was
restricted by the seeker sampling frequency. We can reduce
L(qIH , qIV ) L(qBH , qBV ) the LOS angle rate’s lag by introducing missile attitude which
was detected by rate gyro in inertial navigation system.
L(qc )
LOS BLOS The inertial coordinate system conversion to LOS
coordinate coordinate coordinate can be realized through the following methods. We
assumption that the angle rate of LOS coordinate relative to
Figure 2. the relationships of coordinate transformation the BLOS coordinate is Ω3
Ω3   q&c 0 0 (6)
T

B. the LOS decoupleing method

As showed in Fig .1, the target has the same position The angle rate vector of the frame Sc relative to the frame
( Rr , 0, 0) both in LOS coordinate and BLOS coordinate. S B represented in coordinate frame S B is
According to the space position invariance principle, the target
Ω 2   q&BV sin q&BH q&BV q&BV cos q&BH  (7)
T
position in inertial frame S I is
where, q&BH and q&BV are the yaw and pitch angle in body frame
x z   T  qIH , qIV g Rr 0 0 (1)
T T
y
respectively, which can be obtained by derivative network.
The target position in body frame S B is The angle rate vector of the frame S B relative to the frame
S I represented in coordinate frame S B is
 xb zb   T  qBH , qBV g Rr 0 0 (2)
T T
yb
T
Accord to the transform relation between body frame S B Ω1  x  y z  (8)
and inertial frame S I , we can deduce that
According to the Eq. 6-8, we can get the angle rate vector
of the frame SC relative to the frame S I represented in
x y z   T ( , ,  )gT (qBH , qBV )g Rr 0 0
T T
(3)
coordinate frame S I is
The qIH and qIV can be obtained by the following equation (9)
Ω  TBI (Ω1  Ω2 )  TLOS
I
Ω3

 qIH qIV     arctan  a3 a1  arcsin(a2 ) 

T
In addition, the angle rate vector of the frame S LOS relative
T
(4)
to the frame S I represented in coordinate frame S I can be
where,
written as follows

Ω   q&IV sin qIH q&IH q&IV cos qIH  (10)

T

Thus, the relation between the rate q&IV , q&IH and q&BV , q&BH
are established. But the rate q&c can’t be obtained by
strapdown seeker. So decoupling matrix can be constructed to
deal with this problem. The LOS rate vector can be written as
follows

659
  q&IH q&IV   M q Ω
T
(11) covariance R v , obtained by using CD, are used as input and
the covariance of zero-mean process noise v respectively.
where, Ignored the high-order information of LOS angle, the Eq. 11
can be rewritten in discrete-time form as follows
  tan qIV cos qIH 1 tan qIV sin qIH 
Mq   (12)  qIH (k  1) 
 q (k  1)   F qIH (k ), qIV (k ), ΩC Im (k ), v(k )
sin qIH 0 cos qIH  I
 
 IV 
According to Eq. 4, the observation equation can be written (15)
 q (k ) 
as   IH   T gM q (k )gΩCI Im (k )  v(k )
 qIV (k ) 
 qBH   arctan  b3 b1  
 q    arcsin b (13) where, T is the strapdown seeker’s sampling time. It is the
 BV    2   same that Eq. 11 can be rewritten in discrete form as follows
where,  qBH (k ) 
 q (k )   H qIH (k ), qIV (k ), (k ), (k ),  (k ), n z (k ), n q (k )
 b1  cos cos cos qIV cos qIH  sin  sin qIV  BV 
(16)
  cos  sin s cos q sin q   arctan b3 (k ) b1 (k )
 IV IH    n q (k )
b2   sin  cos cos  cos qIV cos qIH  sin sin  cos qIV cos qIH  arcsin b2 (k ) 

  cos  cos  sin qIV  sin  sin cos  cos qIV sin qIH Because of the nonlinearity in process Eq. 15, the CDKF
 (14)
  cos sin  cos qIV sin qIH algorithm is applied and the structure of LOS reconstruction
 b3  sin  cos sin  cos qIV cos qIH  sin cos  cos qIV cos qIH filter is depicted in Fig .3.

  cos sin  sin qIV  sin  sin sin  cos qIV sin qIH v(k )
  cos cos  cos q sin q qˆ IH (k ), qˆ IV (k )
( , ,  ) nz
 IV IH Ω I
c Im

G (k ) z 1 F (k  1, k ) H (k )
III. THE DESIGN OF LOS RECONSTRUCTION FILTER BASED ON qBH (k )
-
CDKF METHOD qBV (k )

CDKF is nonlinear filter, Compared with UKF algorithm, it

has some advantages such as moderate calculation cost, high nq
precision, less adjust parameter and good robustness.
The background noise of strapdown seeker, and high Figure. 3 the structure frame of LOS reconstruction filter
nonlinearity of decoupling information make it difficult to
calibrate the system noise. A simple method is define the Ω as The details of LOS reconstruction filter by using CDKF
Ω cI Im plus Gaussian white noise. Weighted statistical algorithm are as follows
linearization method provides a way to solve the problem that
1) Initialization
the random variable will transmit its nonlinear characteristics
to the next one via the nonlinear function. The method will Let the adjust parameter h  3 . Considering that the
select sever deterministic sampling points, and calculate, with process noise vk and noise nz are not additive noises, and
these sample points as the function inputs, the value of assumes that the process noise is not related with state in the
nonlinear function. The random variable’s statistical initial estimate. The augmented state vector is selected as
linearization can be realized by using regression techniques.
xak   xkx  x 
T
The foundation of CDKF is central difference approximation,    x T
T
 vk  
T v T T

 k
  k 
which is a method for calculating the statistics of random
variable undergone a nonlinear function[6-8]. We suppose that The initial mean and covariance are
states are Gaussian random variables, therefore, only mean
and covariance are essential to be approximated. xˆ0a  [qˆ IH (0) qˆ IV (0) 0 0 0]T , P0a  diag{Px 0 , R v }
According to the calibrations of seeker and navigation
system, the statistic of BLOS noise vector n q , the BLOS rate Calculate the initial input ΩcI Im (0) and covariance of
input noise Rv (0) .
noise vector n q , the missile attitude noise vector n z , the rate
gyro output noise vector n rg . The statistic of Ω can be 2) For sampling time k  1, 2,L ,  , Calculate the sigma
calculated by Central Difference approximation. By using CD points
method, the mean and the covariance of output variables can χ ak 1   xˆ ak 1 xˆ ak 1  h Pka1 xˆ ka 1  h Pka1 
be captured precisely up to the second order, and higher order  
information can be partially incorporated in the mean and the
3) Time updates
covariance, which leads to even higher precision. The
implementation is extremely rapid because it is unnecessary to χ kx|k 1  F  χ kx 1 , ΩCI Im (k  1), χ vk 1 
calculate the Jacobian matrixs[9,10]. The mean ΩCI Im and

660
 2L with the LOS calculated by solving directly and low-pass filter,
xˆ k   w i( m ) χ ix, k |k 1 the LOS estimation curves are shown as Fig .4. Good
i 0
performances are obtained by using LOS reconstruction filter,
χ  χ ix L , k |k 1  χ ix, k |k 1  χ ix L , k |k 1  
L
P   [w and then the precision of LOS estimation is improved by
 ( c1) x T
xk i i , k | k 1
i 1 applying the output of the proposed filter. Fifty Monte Carlo
w i( c 2)  χ ix, k |k 1  χ ix L , k |k 1  2χ 0,x k |k 1   simulations are performed. The mean and standard deviation
of estimation error of LOS reconstruction angle rate by using
χ  χ ix L , k |k 1  2χ 0,x k |k 1  ] EKF and CDKF are showed as Table 2. We can see that the
x T
i , k | k 1
proposed method has a higher precision to enhance the
where,
performance of guidance system compared with EKF.
 w0( m )  (h 2  L) / h 2 i0 TABLE 1. Simulation Condition

 (m) 1 1 2
h 1
( c1)
 wi  2 , wi  2 , wi 
( c 2)
i  1, 2,L , 2 L Subsystem Assumption
 2h 4h 4h 4
4) Measurement updates equations Missile Initial
In the process of measurement updates, let the system state [0,7000,0]T (m)
and measurement noise merge into one augmented state vector. Position

xˆ km|k 1   xˆ k  n T  Pkm|k 1  diag{Pxk , R n } Missile Velocity 306 (m/s)

T T

 
Target Initial
χ mk |k 1   xˆ mk |k 1 xˆ mk | k 1  h Pkm| k 1 xˆ km| k 1  h Pkm| k 1  [8000,0,1000]T (m)
 
Position
Thus, the measurement updates equations are
2L
Target Velocity  27 0 20sin(0.05 t )  (m)
γ k |k 1  H   kx|k 1 ,  kn|k 1  yˆ k   w i( m ) γ i , k |k 1
i 0
PN Guidance 4
 γ i,k |k 1  γ i  L,k |k 1  γ i,k |k 1  γi  L,k |k 1  
L
P   [w
 ( c1) T
yˆ k i Jx 0.250956 kg gm 2
i 1

w i( c 2)  γ i , k |k 1  γ i  L , k |k 1  2 γ 0, k |k 1   J y , Jz
6.606011 kg gm 2
nq
γ  γ i  L , k |k 1  2γ 0, k |k 1  ]
T
i , k | k 1 0.0001rad
n q
Pxk yˆk  w1( c1) Pxk  γ1:L , k |k 1  γ L 1:2 L , k |k 1  0.02 rad/s
K k  Pxk yˆ k Pyˆk1
n rg
0.001rad/s
xˆ k  xˆ  K k (y k  yˆ )
 
nz
k k 0.002rad

Pxk  P  K k Pyˆ K Tk
xk

5) Calculation the input ΩcI Im (0) and covariance of input TABLE 2. MEAN OF RMSE
noise Rv (k ) in measurement values of seeker and navigation Elevation rate(rad/s) Azimuth rate(rad/s)
system by the method of CD. EKF 0.0066 0.0061
6) Back to 2) CDKF 0.0016 0.0019
Since the square root of the covariance matrix must be
calculated when time updates, the method may has large -0.095
computation. The square root of the covariance matrix can be EKF

calculated by Cholesky decomposition to deal with the -0.1

CDKF
Real signal
problem of large computation[9]. Compared with the standard
form, the square root form not only improves the -0.105
computational efficiency, but also to ensure the covariance
LOS yaw angle(rad)

matrix to be a positive definite matrix. Thus, more stable

numerical calculation results can be obtained.
-0.11

-0.115

IV. SIMULATION RESULTS

In this section, the performance of LOS reconstruction filter
-0.12

is validated by a simple homing missile intercept simulation.

The simulation condition is depicted in Table 1. DFMF is used
-0.125
0 1 2 3 4 5 6 7 8 9 10

as low-pass filter and derivative network [10,11]. The memory

t(s)

length parameter of DFMF is 0.8 in simulations. Compared Figure. 4 the LOS pitch angle reconstruction

661
 V. CONCLUSIONS
-0.6
EKF By the reason of the high nonlinearity and serious
measurement noise contained in the guidance system based on
CDKF
-0.62 Real signal
strapdown seeker, the CDKF is employed to estimate inertial
-0.64 LOS rate in this paper. Estimation results are compared with
LOS pitch angle(rad)

that of EKF by Monte Carlo simulations. These simulations

-0.66 show that the CDKF is superior to the EKF both in precision
of estimation and velocity of convergence for the strapdown
-0.68 guidance system. Besides, calculation of Jacobian matrices
can be avoided for the CDKF, which is usually a complicated
-0.7 process, and may suffer singularity to make the EKF
infeasible.
-0.72
0 1 2 3 4 5 6 7 8 9 10

REFERENCES
t(s)

Figure. 5 the LOS yaw angle reconstruction

[1] Ozkan, Bulent. Comparison of the strapdown and gimbaled seekers
0.02
utilized in aerial applications[C]//Proceedings of SPIE the international
EKF Society for Optical Engineering. Bellingham:[s.n.]. 2012.
0.015 CDKF [2] Jan K. A Simple Study of the Information Requirements for Missile
Real signal
Guidance[C]//SPIE Conference on Acquisition, Tracking, and Point
0.01 XVII. Orlando,United states: SPIE. 2003, pp.77-83
LOS yaw angle rate(rad/s)

[3] Tingting Sun, Hairong Chu, Hongguang Jia. Line of sight angle rate
0.005
estimation of strapdown optical image seeker[J]. Chinese Optical
Society. vol.34, June 2014.
0
[4] Guojiang Zhang, Yu Yao, Kemao Ma. Line of Sight Rate Estimation of
-0.005 Strapdown Imaging Guidance System based on Unscented Kalman
Filter[C]//Proceedings of the Fourth International Conference on
-0.01 Machine Learning and Cybernetics. United States:Institute of Electrical
and Electronics Engineers Computer Society, 2005, pp.1574-1578.
-0.015 [5] Pei Wang,Ke Zhang, Cong Nie. Research on line of sight rate estimation
of strapdown seeker[C]//2014 International Conference on Mechatronics
-0.02
0 1 2 3 4 5 6 7 8 9 10 Engineering and Computing Technology. United States:Trans Tech
t(s) Publications. 2014, pp.556-562
[6] Jihua Zhu, Nanning Zheng. A SLAM algorithm based on the central
Figure. 6 the LOS pitch rate estimate
difference Kalman filter[C]//Intelligent Vehicles Symposium. IEEE.
United States:Institute of Electrical and Electronics Engineers Inc. 2009,
pp.123-128
0.02
EKF [7] Yunyi Hou, Guangfu Ma. A Novel central difference Kalman filter with
CDKF application to spacecraft control[C]//Control and Decision
0.01 Real signal Conference(CCDC), 2013 25th Chinese. Washington:IEEE Computer
Society. 2013, pp.2319-2322
LOS pitch angle rate(rad/s)

0 [8] Jun Li, Huazhong Lu, Zhou Yang, Feng Pei. State-of-charge estimation
and charge equalization for electric agricultural machinery using Square-
Root Central Difference Kalman Filter[C]//2011 Internation Conference
-0.01
on Transportation,Mechanical, and Electrical Engineering. United States:
IEEE Computer Society. 2011, pp.966-969
-0.02 [9] Hufei Zhu, Chen Wen, Bin Li, Feifei Gao. An Improved Square-Root
Algorithm for V-BLAST Based on Efficient Inverse Cholesky
Factorization. Wireless Communications[J]. Piscataway:IEEE
-0.03
Transcations on. 2011, pp.43-48
[10] Kim, Woo Hyun. An optimal FIR filter with fading memory. IEEE
-0.04
0 1 2 3 4 5 6 7 8 9 10 Signal Processing Letters[J]. Piscataway: Institute of Electrical and
t(s) Electronics Engineers Inc. 2011, pp.327-330
Yucheng Liu, Yubin Liu. Adaptive fading memory unscented kalman
Figure.7 the LOS yaw rate estimate filter algorithm for passive target tracking. Sensor Letters[J]. California:
American Scientific Publishers. 2013, pp.2110-2113

662

View publication stats