Proceedings of 2014 IEEE Chinese Guidance, Navigation and Control Conference August 8-10, 2014 Yantai, China

Trajectory Tracking for RLV Terminal Area Energy Management

Phase based on LQR
Bowen Zheng, Zixuan Liang, Qingdong Li, Zhang Ren

Abstract— A new longitudinal trajectory tracking law for the equation of RLV in TAEM phase is to be calculated out. For
Terminal Area Energy Management (TAEM) phase of the acquiring feedback control, performance index is also
Reusable Launch Vehicle (RLV) is presented in this paper. The computed. Finally the feedback matrix is obtained by
conventional PID method controls the height and the velocity by resolving the Riccati equation. Input of the system is the
the angle of attack and the angle of airbrake respectively, and feedback of the state.
no coupling is considered. To improve the tracking precision,
the new tracking law is designed based on the linear quadratic II. TRAJECTORY TRACKING METHODS
regulator (LQR) theory where the coupling is taken into account.
Finally, the trajectory tracking law based on the LQR is Assuming that the earth is a flat surface and the RLV is a
simulated and compared with the conventional method. particle, the motion equation of RLV is as follows [5]:
Simulation results indicate the effectiveness and the robustness
of the new tracking law.
­ mv = - D - mg sin γ
I. INTRODUCTION ° mvγ = L cos φ − mg cos γ
°°
Reusable Launch Vehicle (RLV) is a type of reusable L sin φ (1)
® ψ =
aircrafts that can make a round trip between the earth and the °
outer space [1]. It is forecasted that RLV is the trend of space mv cos γ
°
technology in the following years [2]. The reentry flight of °¯ h = v sin γ
RLV is consisted of three phases: initial entry phase, Terminal
Area Energy Management (TAEM) phase and landing phase where m and v represents the mass and velocity of RLV; L and
[3]. The TAEM phase starts from the point with a height of D represents the lift force and resistance force acting on RLV;
30km and a velocity of 2.5Ma, and ends at the point with a , ,  represents flight-path angle, velocity azimuth angle
height of 4km and a velocity of 0.5Ma. In the TAEM phase, and bank angle of RLV respectively. Assuming that bank
the task is to dissipate surplus energy and adjust the heading angle equals to zero, we can obtain the simplified motion
direction towards the runway, so the RLV will land safely with equation of RLV is as follows:
a suitable velocity and height [4], which makes this phase a
vital phase of the reentry flight. However, a large variation ­ mv = − D − mg sin γ
range of dynamic pressure and velocity, significant changes of °
aerodynamic characteristics along with the uncertain states of ® mvγ = − L − mg cos γ (2)
°
the RLV after initial entry phase also make the task of the ¯ h = v sin γ
TAEM phase a highly difficult work.
In order that RLV can land safely with a suitable velocity
The general method of energy management in the TAEM and height onto the runway, the standard trajectory and
phase is to divide the motion into the longitudinal motion and corresponding reference guidance commands are calculated in
the lateral motion, then design the standard trajectory and advance. In flight procedure, deviations of actual height and
corresponding reference guidance commands. In actual flight, velocity from reference values are obtained, and then angle of
guidance commands are adjusted to track the standard attack and angle of airbrake are adjusted from reference values
trajectory. As for longitudinal motion, two variables are to be to eliminate these deviations [6]. In a common method, angle
controlled: height and velocity. Assuming these two variables of attack is adjusted according to the deviation of height while
are independent of each other, angle of attack is used to angle of airbrake is adjusted according to the deviation of
control the height and angle of airbrake is used to control the velocity. The following equations use PID methods as an
velocity of the RLV respectively. But actually, height and example[6]:
velocity are coupled and the separate design causes
inaccuracies in the process of tracing the reference trajectory. d Δh
Δα = k1* Δh + k 2 * + k 3* ³ Δh * dt (3)
This paper aimes to improve the accuracy of height and dt
velocity together by means of trajectory tracking law based on
the linear quadratic regulator (LQR) theory. Firstly, the state d Δv
Δδ b = k 4 * Δv + k 5* + k 6 * ³ Δv * dt (4)
* Research supported by the National Nature Science Foundation of
dt
China (No.91116002, No.91216034, No.61333011 and No.61121003). where  represents the difference between actual angle of
B. Zheng, Z, Liang, Q. Li, Z. Ren are with the Science and Technology on
Aircraft Control Laboratory, Beihang University, Beijing, 100191 China
attack and reference value; b represents the difference
(phone: 86-010-82314573-11; fax: 86-010-82313265; e-mail: between actual angle of airbrake and reference value; h
changfeng@asee.buaa.edu.cn; muziqingdong@126.com) represents the deviation of height; v represents the deviation
978-1-4799-4699-0/14/$31.00©2014 IEEE
2520
of velocity; k1, k2, k3 represent proportional gain, derivative 1 ∞ T
2 ³0
gain, integral gain of deviation of height; k4, k5, k6 represent J= ( x Qx + U T RU )dt (10)
proportional gain, derivative gain, integral gain of deviation of
velocity. Airbrake cannot be used when the velocity is too where X=[  h  v  ]T, U=[    b]T, Q and R are
high to avoid damage to it. In most cases, airbrake is open weighting matrix of deviation of state variables and input
when Mach number is under 0.8, so b equals to zero when variables. According to ‘Bryson’s rule’ they are constant
velocity is higher than 0.8Ma. symmetric positive definite matrix, defining as follows:

III. TRAJECTORY TRACKING LAW BASED ON LQR

ª 1 º
As coupling exists in the height and velocity tracking laws « 0 0 »
« Δhmax
2
in (8) and (9), the LQR method is used when both the angle of »
attack and the angle of airbrake are adjusted. « 1 »
Q=« 0 0 » (11)
« Δvmax 2 »
Before using trajectory tracking law based on LQR, the
state equation of RLV in TAEM phase has to be calculated « 1 »
« 0 0 »
out. For longitudinal motion, height, velocity, flight-path
¬« Δγ max 2 ¼»
angle are selected as state variables, angle of attack, angle of
airbrake are selected as input variables and height, velocity are
selected as output variables. Using means of small ª 1 º
« Δα 2 0 »
perturbation linearization, the state equation of RLV in TAEM
R=« » (12)
max
phase are obtained as follows [7]: « 1 »
« 0 »
¬ Δδ bmax 2 ¼
ª Δh º ª Δh º
« » « » ª Δα º where hmax, vmax and max represent maximum allowable
« Δv » = A « Δv » + B « Δδ b » (5) deviations of height, velocity and flight-path angle;  and
« Δγ » «¬ Δγ »¼ ¬ ¼ max
bmax represent maximum rate of change of angle of attack
¬ ¼
and angle of airbrake. In order to obtain the minimum value of
where the performance index J defined in (10), theory of optimal
linear regulator is used which provides the feedback matrix by
ª º resolving the Riccati equation:
« 0 sin γ v cos γ »
« » − KA − AT K + KBR −1 BT K − Q = 0 (13)
A = « − Dh − g sin γ − Dv − g cos γ » (6)
« L − g cos γ vLv − L + g cos γ g sin γ » Form of feedback control is obtained as follows:
« h »
¬ v v2 v ¼
ª Δh º
ª Δα º « »
ª 0 0 º « » = K « Δv » (14)
«−D − Dδ b »» ¬ Δδ b ¼ «¬ Δγ »¼
B=« α (7)
« Lα Lδ b » where K represents the state feedback matrix of the system.
« »
¬ v v ¼ The complete guidance commands are the sum of increments
h, v,  represent errors of height, velocity, flight-path and reference values of angle of attack and angle of airbrake:
angle of RLV; , b represent increments of angle of attack,
angle of airbrake on reference values; g represents ª α º ª α ref º ª Δα º
gravitational acceleration; Lh, Lv, L, Lb represent derivatives «δ b » = «δ b »+« » (15)
of lift force on height, velocity, angle of attack, angle of ¬ ¼ ¬ ref ¼ ¬ Δδ b ¼
airbrake and given by (8); Dh, Dv, D, Db represent derivatives where  and b represent complete guidance commands, ref
of resistance force on height, velocity, angle of attack, angle of and bref represent reference commands for angle of attack and
airbrake and given by (9); v represents the velocity of RLV; L angle of airbrake.
and D represents the lift force and resistance force acting on
RLV. IV. SIMULATION

∂L ∂L ∂L ∂L To test the proposed trajectory tracking law, a RLV model

Lh = , Lv = , Lα = , Lδ b = (8) is used in simulations. The initial and terminal conditions for
∂h ∂v ∂α ∂δ b the TAEM phase of a RLV are as follows:

∂D ∂D ∂D ∂D Initial velocity: 760m/s

Dh = , Dv = , Dα = , Dδ b = (9)
∂h ∂v ∂α ∂δ b Initial height: 30000m
Considering lowering the values of state variables and the Initial flight-path angle: -5°
input, performance index of the system is defined as [8]:

2521
 Remaining range: 100.66km
25
angle of attack
Terminal height: 4000m
angle of airbrake
Terminal velocity: 157m/s 20

Angle of attack or airbrake(deg)

A. Abbreviations and Acronyms 15
Assuming that there is no dispersion, the state variables
height and velocity and the input variables angle of attack and
10
airbrake are shown in Figs. 1-3:

30 5
reference trajectory
actual trajectory

25 0

20 -5
0 20 40 60 80 100 120
Range(km)
Height(km)

15

Figure 3. Angle of attack and airbrake without dispersion

10
Figs.1 and 2 display the reference height and velocity
designed in longitudinal plan and actual height and velocity
5 generated in actual flight. The two figures illustrate that the
accuracy is sufficient when there is no dispersion. In Fig2, the
0
velocity is dispelled late in flight because the airbrake is open
0 20 40 60 80 100 120 only when the velocity is below 0.8Ma.
Range(km)

B. Dispersed cases
Figure 1. Reference and actual height trajectories without dispersion
When dispersions are considered in the TAEM phase,
effects of PID method and trajectory tracking law based on
800
LQR are compared. To display this issue clearly, the
reference trajectory dispersion is sorted into two categories: initial condition
actual trajectory
700
dispersion and aerodynamic parameter dispersion. Adding
fluctuation on each item and the results of the two methods are
calculated.
600
Initial condition dispersion includes initial error of height,
velocity and flight-path angle. Aerodynamic parameter
Velocity(m/s)

500
dispersion includes error of atmosphere density, lift
400 coefficient and drag coefficient.

300 TABLE I. CASE DISPERSION VALUE

Case Dispersion items Dispersion value

200
1 Initial height +3km
100 2 Initial height -3km
0 20 40 60 80 100 120
Range(km)
3 Initial velocity +120m/s

4 Initial velocity -120m/s

Figure 2. Reference and actual velocity trajectories without dispersion
5 Initial flight-path angle +5e
6 Initial flight-path angle -5e
7 Range +20%

8 Range -20%

2522
 500
LQR method
TABLE II. CASE DISPERSION VALUE 400 PID method

Case Dispersion items Dispersion value 300

1 Atmosphere density +30% 200

2 Atmosphere density -30% 100

Height(m)
3 Lift coefficient +20% 0

4 Lift coefficient -20% -100

5 Drag coefficient +20% -200

6 Drag coefficient -20% -300

7 Mass +5% -400

8 Mass -5% -500

-20 -15 -10 -5 0 5 10 15 20
Velocity(m/s)

35 Figure 6. Height and velocity error of PID and LQR method

30
Figs.4-6 illustrate that the accuracy of height and velocity
under dispersion of initial condition are sufficient with means
of trajectory tracking law based on LQR. The accuracy is
25
promoted compared with PID method which does not consider
the copula between height and velocity.
Height(km)

20

15

V. CONCLUSION
10
The theoretical derivation and simulation results above
5 illustrate that the copula between height and velocity shall be
taken into consideration to improve the accuracy of
0 longitudinal trajectory tracking. Means of trajectory tracking
0 20 40 60 80 100 120
Range(km) law based on LQR is authenticated to be an effective method
to fufill this target, especially under dispersion of
Figure 4. Height trajectories for LQR in dispersed cases aerodynamic parameter.

REFERENCES
900

800 [1] Y. H. Tang, M. L. Yu,Y. Yang, Z. B. Xie, “Second generation of

reusable launch vehicle and its reentry guidance technology,” Missiles
700 and space vehicles, pp.26-41, Jan. 2010.
[2] Sarigul-Klijn M, Sarigul-Klijn N, “A study of air launch methods for
600
RLVs,” American Institute of Aeronautics and Astronautics Paper, vol.
4619, pp. 1-16, 2001.
Velocity(m/s)

[3] Thomas E. Moore, Lyndon B. Johnson Space Center, Houston, Texas,

500
“Space Shuttle Entry Terminal Area Energy Management,” NASATM,
vol. 104744, pp. 1-46, 1991.
400 [4] Brinda V, Arora R K, Janardhana E, “Mission analysis of a reusable
launch vehicle technology demonstrator (RLV-TD)”. AIAA/CIRA 13th
300 International Space Planes and Hypersonics Systems and
Technologies, vol. 3291, 2005.
200 [5] Ridder S D, Mooij E, “Optimal Longitudinal Trajectories for Reusable
Space Vehicles in the Terminal Area,” Journal of Spacecraft and
100
Rockets, pp. 642-653, 2011.
0 20 40 60 80 100 120 [6] Z. J. Bu, “Guidance law and control law design of terminal area energy
Range(km) management for reusable launch vehicle,” MAE. dissertation, Nanjing
University of aeronautics and astronautics, MA, 2012.
[7] Q. Wang, Y. Chen, Y. X. Zhang and Y. Z. Hou, Optimal
Figure 5. Velctity trajectories for LQR in dispersed cases
control-theory, method and application. Beijing: Higher education
press, 2011, ch.4.

2523
[8] Greg A. Dukeman, “Profile-following entry guidance using linear
quadratic regulator theory,” AIAA Guidance, Navigation, and Control
Conference and Exhibit, vol. AIAA 2002-4457, pp.1-10, Aug. 2002

2524