Generalized Formulation of Weighted Optimal Guidance Laws with Impact Angle Constraint

Correspondence
The generalized formulation of weighted optimal guidance laws with impact angle constraint is investigated here. From the generalized formulation, we explicitly find the feasible set of weighting functions that lead to analytical forms of weighted optimal guidance laws. This result has potential significance because it can provide additional degrees of freedom in designing a guidance law that accomplishes the specified guidance objective.

I. INTRODUCTION For antiship and antitank missile systems, the guidance laws achieving the desired impact angle have been considered in order to maximize the warhead effect and attack a targets weak spot. Over the past decades the optimal control theory [1] has been extensively used to design the impact angle control guidance laws because of its benefits: it can easily provide a guidance law that satisfies the terminal constraints and some performance requirements as well as the analytical form and state feedback form of guidance laws. In the application of optimal control for deriving the impact angle control guidance laws, the minimization of the control effort has been widely considered for the cost function [28] as follows: Z tf J= u2 ( )d (1)
t0

where u and tf represent the missiles acceleration command and the time of interception, respectively. In this method, how to design the cost function is an important issue for guidance law designers because the selection of the cost function can affect the response of state variables and then decide the guidance performance. Accordingly, to improve the guidance performance, the control energy costs with weighting functions have also been used to derive the impact angle control guidance laws. A power of time-to-go function [911] and an exponential function [1214] were considered for the weighting function of the energy cost. In such previous works,
Manuscript received June 15, 2012; revised September 12, 2012; released for publication October 4, 2012. IEEE Log No. T-AES/49/2/944538. Refereeing of this contribution was handled by L. Kaplan. This research was supported by Agency for Defense Development under Contract UD110031CD. c 2013 IEEE 0018-9251/13/$26.00 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 49, NO. 2 APRIL 2013 1317

angle frame can be written as _ =V y M sin M _ = a =V : M M M (4)

It is assumed that V M is constant and M is small enough to linearize the engagement kinematics as follows: _ =V y M M = v (5) _ = aM v where y and v represent the lateral distance and velocity perpendicular to the desired impact course. In the linearized engagement kinematics, the flight path angle and LOS angle with respect to the impact angle frame can be determined, respectively, as follows: v M = V M (6) y y = = : R V M tgo By using (3) and (6), the lateral distance and velocity perpendicular to the impact course can be rewritten in terms of the flight path angle and LOS angle, respectively. y=V M tgo (f ) (7) v=V M (M f ): The linearized engagement kinematics as given in (5) can be rewritten in the matrix form as _ = Ax + Bu x where x =[y , v]T , A=

Fig. 1. Homing engagement geometry and parameter definitions.

for achieving the specified guidance purpose, the weighted cost functions were introduced to shape the missiles trajectory or to distribute the acceleration demand during the engagement. These previous studies now raise a question: Could any weighting function be used for accomplishing such guidance objectives? The purpose of this paper is to find answers to this question. In this paper we first investigate the generalized formulation of the weighted optimal control problems with the terminal constraints (i.e., zero miss distance and the desired impact angle) as follows: Z tf W ( )u2 ( )d: (2) J=
t0

(8)

Then, we determine the feasible set of W (t) that leads to the analytical solutions because the analytical forms of solutions are more desirable for practical uses. From a practical standpoint, the potential significance of this result is that through appropriate selections in the set of the weighting functions we have determined, the designer can achieve the guidance purpose as desired. II. PROBLEM FORMULATION Let us consider the engagement geometry for a stationary target as described in Fig. 1, where (XI , Y I) and (xf , yf ) denote the inertial reference frame and the impact angle frame, respectively. The impact angle frame is defined to be rotated from the inertia reference frame by f , which is the desired impact angle. The flight path angle and the line-of-sight (LOS) angle are denoted by M and . In the impact angle frame, these angles are expressed as follows: M = M f = f : (3)

u = aM , B=

0 1 0 0

0 1

(9) :

In order to satisfy the zero miss distance and the desired impact angle at the terminal time, the following boundary conditions should be achieved. x1 (tf ) = x2 (tf ) = 0: (10)

Now, let us set the following optimal control problem which minimizes the control effort weighted by general functions of W (t). Z 1 tf W ( )u2 ( )d where W ( ) > 0 min J = u 2 t0 for 2 (t0 , tf ): (11) The conditions of feasible weighting functions are determined in the next section. III. GENERALIZED FORMULATION OF WEIGHTED OPTIMAL SOLUTIONS In this paper we use Schwarzs inequality approach as studied in [2] in order to solve the

In Fig. 1 other variables are self-explanatory, and the engagement kinematics with respect to the impact
1318

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 49, NO. 2 APRIL 2013

optimal problem. First, according to the linear control theory, the general solution of (8) can be expressed as Z tf x(tf ) = (tf t)x(t) + (tf )B ( )u( )d
t

where K is a constant to be determined. Equation (22) can be regarded as the acceleration command that minimizes the cost function. Substituting (22) into (17) gives the following equation: K = R tf
t 2 ( )W 1 ( )d h1

(12) where (tf t) represents the state transition matrix and is determined. 1 tf t : (13) (tf t) = eA(tf t) = 0 1 Then, (12) provides the expression of state variable at the final time. Z tf x1 (tf ) = f1 h1 ( )u( )d (14) x2 (tf ) = f2 where f1 = x1 (t) + (tf t)x2 (t), f2 = x2 (t),

f1 : Rt t f h1 ( )h2 ( )W 1 ( )d (23)

For convenience we introduce shorthand notations as follows: Z tf 2 g1 = h1 ( )W 1 ( )d (24)

t

g12 = g2 =

Z Z

tf

tf

t tf

h1 ( )h2 ( )W 1 ( )d
2 h2 ( )W 1 ( )d:

(25) (26)

h2 ( )u( )d

(15)

h1 ( ) = (tf ) h2 ( ) = 1:

(16)

Using these notations provides a simplified expression of K as f1 K= : (27) g1 g12 Substituting (27) into (22), we have the following equation: u( ) = f1 [h1 ( ) h2 ( )]W 1 ( ) : g1 g12 (28)

By imposing the boundary conditions (i.e., x1 (tf ) = x2 (tf ) = 0), (14) and (15) can be rewritten as Z tf f1 = h1 ( )u( )d (17) f2 = Z
t tf

h2 ( )u( )d:

(18)

From (21) the minimum value of the cost function can be expressed using shorthand notations: J= (f1 f2 )2 : 2(g1 2g12 + 2 g2 ) (29)

Hereafter, let us introduce a new variable denoted by . Then, we can combine (17) and (18) as follows: Z tf [h1 ( ) h2 ( )]u( )d: (19) f1 f2 =
t

The above equation can be rewritten by introducing a slack variable with respect to W( ). Z tf [h1 ( ) h2 ( )]W1=2 ( )W 1=2 ( )u( )d: f1 f2 =
t

Because the undetermined value exists in (29), this expression is incomplete. From the calculus we can find which further minimizes J by taking the derivative of J with respect to and then by imposing dJ=d = 0 as follows: = f1 g12 f2 g1 : f1 g2 f2 g12 (30)

(20) Then, applying Schwarzs inequality to (20) and rearranging the obtained result yields the following inequality condition: Z 1 tf (f1 f2 )2 W ( )u2 ( )d: Rt 2 t 2 t f [h1 ( ) h2 ( )]2 W 1 ( )d (21) Note that the right hand side of (21) is equal to the cost function defined in (11). It can predict that when the equality (i.e., =) holds, the left hand side is identical to the minimum value of the cost function. According to Schwarzs inequality, the acceleration command that holds the equality can be expressed as u( ) = K [h1 ( ) h2 ( )]W
CORRESPONDENCE
1

Substituting (30) into (28), we have the optimal acceleration command.

u ( ) = [f1 h1 ( )g2 g12 (f1 h1 ( ) + f1 h2 ( )) + f2 h2 ( )g1 ]W 1 ( ) : 2 g1 g2 g12

(31) In the time domain, the optimal acceleration command can be further simplified by substituting (9) and (16) into (31) and by introducing newly defined variables called the equivalent guidance gains. a M = k1 y v k2 2 tgo tgo (32)

( )

(22)

1319

The notations k1 and k2 represent the equivalent guidance gains and are defined as follows: ! 3 2 g2 tgo g12 tgo k1 = W 1 (t) 2 g1 g2 g12 ! 3 2 g1 tgo + g2 tgo 2g12 tgo k2 = W 1 ( t ) 2 g1 g2 g12 where g1 = g12 = g2 = Z Z Z
tf t tf t tf t

where tgo = tf t is the remaining time of interception.

(33)

PROPOSITION 1 If the weighting function W ( ) satisfies the condition of (11) and the integrations of W ( ) as shown in (42) are analytically given, then this weighting function can lead to the analytical solution. Z 1 W ( ) = W ( ), W ( ) = W 0 1 0 ( )d (42) Z Z ( ) = W ( ) d , W ( ) = W ( ) d W 2 1 3 2 where W 1 ( ), W 2 ( ), and W 3 ( ) represent the indefinite integral, double integral, and triple integral of the inversed weighting function.

(34)

(tf )2 W 1 ( )d (tf )W 1 ( )d W 1 ( )d:

(35) (36) (37)

From (33) and (34) the denominator of k1 and k2 should not be zero. It is proven by the following lemma. LEMMA 1 Regardless of choices in the weighting functions, the following relation is always guaranteed.
2 g1 g2 g12

PROOF From (33) and (34), in order to obtain analytical solutions, the composition terms of the equivalent guidance gains should be analytically given. Therefore, the inverse of weighting function W 1 ( ) basically has an analytical form and the terms of g1 , g12 , and g2 are also given in analytical forms. From the calculus, the terms of g1 , g12 , and g2 can be further expanded based on the method of integration by parts.
2 g1 = tgo W 1 (t) 2tgo W 2 (t) + 2(W 3 (tf ) W 3 (t))

(43) g12 = tgo W 1 (t ) + W 2 (tf ) W 2 (t ) g2 = W 1 (tf ) W 1 (t ): (44) (45)

> 0:

(38)

PROOF From (36) the expression of g12 can be reformulated as: Z tf (tf )W 1=2 ( )W 1=2 ( )d: g12 =
t

(39)

Accordingly, the analytical result of W 1 ( ), W 2 ( ), and W 3 ( ) introduces analytical expressions of g1 , g12 , and g2 , which completes the proof. Note that the condition in (42) represents the feasible set of weighting functions to obtain the analytical weighted optimal guidance laws. Additionally, to ensure that the guidance command does not blow up during the engagement, the equivalent guidance gain should be bounded: The functions of W 1 ( ) is bounded as W 1 ( ) < 1 for 2 (t0 , tf ). From a practical standpoint, these results are helpful to derive a new guidance law that improves the guidance performance and attains the specific guidance objective through appropriate choices in the feasible set of weighting functions, which satisfy the previously determined conditions. Hereafter, we illustrate our results with two simple cases as the weighting functions W 1 ( ) = 1 and W 1 ( ) = (tf )N , which comply with the conditions as we have discussed. For W1 ( ) = 1, which means the control effort is equally weighted during the engagement, we have the following results: Z tf 1 3 g1 = (tf )2 d = tgo (46) 3 t Z tf 1 2 g12 = (tf )d = tgo (47) 2 t Z tf g2 = 1d = tgo : (48)
t

Applying Schwartzs inequality to (39) yields the following results: Z tf Z tf 2 g12 (tf )2 W 1 ( )d W 1 ( )d: (40)
t t

The equality sign of (40) does not hold because of (tf )W 1=2 ( ) 6= W1=2 ( ), where is a constant. Therefore, the final result is written using the terms of g1 and g2 as follows:
2 g1 g2 g12 >0

(41)

which completes the proof. IV. FEASIBLE SET OF WEIGHTING FUNCTIONS The expressions in (32)(37) represent the generalized formulation of weighed optimal acceleration commands satisfying zero miss distance as well as the impact angle constraint. According to selections of weighting functions, the state feedback form of optimal guidance law can be determined by computing (35)(37) and then by substituting these results into (33) and (34). Accordingly, in order to obtain analytical forms of weighted optimal guidance laws, we require the weighting functions as provided in the following proposition.
1320

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 49, NO. 2 APRIL 2013

We substitute (46), (47), and (48) into (33) and (34) under the condition of W 1 ( ) = 1. Then, the optimal acceleration command is obtained in that case as follows: y v a : (49) M = 6 2 4 tgo tgo Since the lateral distance and velocity are defined the impact angle frame, the terminal values of these variables are zero as yf = vf = 0. Rearranging (49) and using the terminal condition yf = vf = 0 yields the following alternative form of this acceleration command: 6 2 a [v v ]: (50) M = 2 [yf y vtgo ] tgo tgo f In addition, substituting (7) into (49), the guidance command can be rewritten as follows: V M a [6 + 4M + 2f ]: (51) M = tgo Note that these results are identical to the optimal control guidance law with terminal impact angle constraint as studied in [13, 8]. If we choose W1 ( ) = (tf )N , which increases the weight of acceleration demand as t ! tf , then g1 , g12 , and g2 can be computed as Z tf 1 N +3 (tf )N +2 d = (52) g1 = t N + 3 go t Z tf 1 N +2 (tf )N +1 d = (53) t g12 = N + 2 go t Z tf 1 N +1 g2 = (tf t)N d = : (54) t N + 1 go t Then, we have the following optimal acceleration command: y v a : (55) M = (N + 3)(N + 2) 2 2(N + 2) tgo tgo In a similar way alternative forms of this command can be obtained as (N + 3)(N + 2) [yf y vtgo ] a M = 2 tgo a M = (N + 1)(N + 2) [vf v] tgo (56)

weighted control energy costs. The results indicated that any weighting function can provide the analytical form of optimal solution if up to triple integrations of the inverse of the weighting functions are analytically given. The potential significance of these results is that the feasible set of weighting functions as we have determined can provide additional degrees of freedom for designing a guidance law that achieves the guidance purpose as desired and enhances the guidance performance.
CHANG-HUN LEE MIN-JEA TAHK Department of Aerospace Engineering Korea Advanced Institute of Science and Technology (KAIST) 335 Gwahangno, Yuseong-gu Daejeon, 305-701 Korea E-mail: (chlee@fdcl.kaist.ac.kr) JIN-IK LEE Agency for Defense Development Team 1-1 PO Box 35-3, Yuseong Daejeon, 305-600 Korea REFERENCES [1] Bryson, Jr., A. E. and Ho, Y-C. Applied Optimal Control. Hoboken, NJ: Wiley, 1975. Zarchan, P. Tactical and Strategic Missile Guidance (5th ed.). Washington, D.C.: AIAA, 2007, pp. 541547. Ben-Asher, J. Z. and Yaesh, I. Advances in Missile Guidance Theory (Progress in Astronautics and Aeronautics), vol. 180. Reston, VA: AIAA, 1998, pp. 2532. Kim, M. and Girder, K. V. Terminal guidance for impact attitude angle constrained flight trajectories. IEEE Transactions on Aerospace and Electronic Systems, AES-9, 6 (1973), 852859. Idan, M., Golan, O., and Guelman, M. Optimal planar interception with terminal constraints. Journal of Guidance, Control, and Dynamics, 18, 6 (1995), 12731279. Song, T. L., Shin, S. J., and Cho, H. Impact angle control for planar engagements. IEEE Transactions on Aerospace and Electronic Systems, 35, 4 (Oct. 1999), 14391444. Shaferman, V. and Shima, T. Linear quadratic guidance laws for imposing a terminal intercept angle. Journal of Guidance, Control, and Dynamics, 31, 5 (2008), 14001412. Ryoo, C. K., Cho, H., and Tahk, M. J. Optimal guidance laws with terminal impact angle constraint. Journal of Guidance, Control, and Dynamics, 28, 4 (2005), 724732. Ryoo, C. K., Cho, H., and Tahk, M. J. Time-to-go weighted optimal guidance with impact angle constraints. IEEE Transactions on Control Systems Technology, 14, 3 (2006), 483492. 1321

[2]

[3]

[4]

[5]

[6]

[7]

V M [(N + 3)(N + 2) + 2(N + 2)M tgo + (N + 1)(N + 2)f ] (57)

[8]

where N is the power of time-to-go. Note that these results are equal to the time-to-go weighted optimal control guidance laws as studied in [9][10].
[9]

V. CONCLUSION In this paper, optimal guidance laws with terminal impact angle constraint are generalized for the
CORRESPONDENCE

[10]

[11]

[12]

[13]

[14]

Ohlmeyer, E. J. and Phillips, C. A. Generalized vector explicit guidance. Journal of Guidance, Control, and Dynamics, 29, 2 (2006), 261268. Jeon, I. S. and Lee, J. I. Optimality of proportional navigation based on nonlinear formulation. IEEE Transactions on Aerospace and Electronic Systems, 46, 4 (2010), 20512055. Speyer, J. L., Deyst, J., and Jacobson D. A. Optimization of stochastic linear systems with additive measurement and process noise using exponential performance criteria. IEEE Transactions on Automatic Control, AC-19, 4 (1974), 358366. Rusnak, I. Guidance law based on an exponential criterion for acceleration constrained missile and a maneuvering target. Journal of Guidance, Control, and Dynamics, 19, 3 (1996), 718721. Ben-Asher, J. Z., Farber, N., and Levinson, S. New proportional navigation law for ground-to-air system. Journal of Guidance, Control, and Dynamics, 26, 5 (2003), pp. 822825.

Ground Moving Target Detection and Velocity Estimation Based on Spatial Multilook Processing for Multichannel Airborne SAR

A robust method is addressed here for ground moving target detection and velocity estimation. This method is proposed by forming multilook Doppler beam sharpening (DBS) images with different illuminated angles which are divided by slow time segment in the azimuth direction. Compared with only a single looking angle, the detection probability with multiple looking angles can be improved under the same condition. For the clutter rejection procedure, we describe an approach which utilizes the fast time information and therefore has the robustness to image coregisrtration error. After clutter rejection and target detection, a new series of single-look and multilook velocity estimation approaches are developed in the presence of channel mismatch and image coregistration error. The robustness and practicability of the proposed algorithm are verified by processing the measured data of the experiment with three-aperture airborne radar.

I. INTRODUCTION Ground moving target detection is an important capability of the modern airborne/spaceborne radar. In practice, due to the motion of the platform, the Doppler shift of stationary clutter seen by the radar varies with the looking direction, which leads to moving targets being masked in the clutter background. Therefore, before target detection, the key and challenging task is the clutter rejection. The multi-channel spatial processing is the main approach for clutter suppression [1-2]. For the synthetic aperture radar (SAR) together with ground moving target indication (GMTI) system [3], the long dwell time results in the internal clutter motion (ICM). Clutter patches will fluctuate during the whole synthetic aperture time, which leads to the difficulty of clutter suppression. Furthermore, the Doppler chirp rate of the moving target is distinguished from that of stationary clutter in the presence of along-track velocity or cross-track acceleration. As a result, it leads to moving targets defocusing and thus great degradation of the signal-to-clutter ratio (SCR) [45]. Moreover, the computational complexity of the SAR imaging, range walk of the target and various errors
Manuscript received June 26, 2009; revised October 1, 2010 and May 3, 2011; released for publication September 19, 2012. IEEE Log No. T-AES/49/2/944539. Refereeing of this contribution was handled by P. Lombardo.

c 2013 IEEE 0018-9251/13/$26.00 1322 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 49, NO. 2 APRIL 2013