Iowa State University

Digital Repository @ Iowa State University

Graduate Theses and Dissertations

Graduate College

2010

Dynamic modeling and ascent flight control of

Ares-I Crew Launch Vehicle
Wei Du
Iowa State University

Follow this and additional works at: http://lib.dr.iastate.edu/etd

Part of the Aerospace Engineering Commons
Recommended Citation
Du, Wei, "Dynamic modeling and ascent flight control of Ares-I Crew Launch Vehicle" (2010). Graduate Theses and Dissertations.
Paper 11540.

This Dissertation is brought to you for free and open access by the Graduate College at Digital Repository @ Iowa State University. It has been accepted
for inclusion in Graduate Theses and Dissertations by an authorized administrator of Digital Repository @ Iowa State University. For more
information, please contact [email protected].

Dynamic modeling and ascent flight control of Ares-I Crew Launch Vehicle
by
Wei Du

A dissertation submitted to the graduate faculty

in partial fulfillment of the requirements for the degree of
DOCTOR OF PHILOSOPHY

Major: Aerospace Engineering

Program of Study Committee:
Bong Wie, Major Professor
Ping Lu
Thomas J. Rudolphi
Zhijian Wang
John Basart

Iowa State University

Ames, Iowa
2010
c Wei Du, 2010. All rights reserved.
Copyright

ii

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vi

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vii

ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xiii

ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xiv

NOMENCLATURE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xvi

CHAPTER 1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . .

1.1

Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.2

Ares-I Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.3

Ares-I Mission Profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.4

Interaction Between Structures and Flight Control System . . . . . . . . . . . .

1.5

Underactuated Control Problem . . . . . . . . . . . . . . . . . . . . . . . . . .

CHAPTER 2. 6-DEGREE-OF-FEEDOM DYNAMIC MODELING . . . . .

2.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.2

Reference Frames and Rotational Kinematics . . . . . . . . . . . . . . . . . . .

2.2.1

Earth-Centered Inertial Reference Frame

. . . . . . . . . . . . . . . . .

2.2.2

Earth-Fixed Equatorial Reference Frame . . . . . . . . . . . . . . . . . .

2.2.3

Body-Fixed Reference Frame . . . . . . . . . . . . . . . . . . . . . . . .

10

2.2.4

Structural Reference Frame . . . . . . . . . . . . . . . . . . . . . . . . .

10

2.2.5

Earth-Fixed Launch Pad Reference Frame . . . . . . . . . . . . . . . . .

11

2.2.6

Euler Angles and Quaternions

. . . . . . . . . . . . . . . . . . . . . . .

11

2.2.7

Initial Position of Ares-I CLV on the Launch Pad . . . . . . . . . . . . .

13

iii
2.3

2.4

The 6-DOF Equations of Motion . . . . . . . . . . . . . . . . . . . . . . . . . .

14

2.3.1

Aerodynamic Forces and Moments . . . . . . . . . . . . . . . . . . . . .

15

2.3.2

Gravity Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

16

2.3.3

Rocket Propulsion Model . . . . . . . . . . . . . . . . . . . . . . . . . .

17

2.3.4

Guidance and Control . . . . . . . . . . . . . . . . . . . . . . . . . . . .

18

2.3.5

Flexible-Body Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

19

Simulation Results of the Rigid Body Ares-I Crew Launch Vehicle . . . . . . .

21

CHAPTER 3. ANALYSIS AND DESIGN OF ASCENT FLIGHT CONTROL SYSTEMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

33

3.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

33

3.2

Pitch Control Analysis of Rigid Launch Vehicles . . . . . . . . . . . . . . . . .

34

3.3

Pitch Control of a Rigid-Body Model of the Ares-I CLV . . . . . . . . . . . . .

39

3.4

Flexible-Body Control of an Ares-I Reference Model . . . . . . . . . . . . . . .

41

3.5

NMP Structural Filter Design . . . . . . . . . . . . . . . . . . . . . . . . . . . .

46

3.6

Robust Analysis for Structural Filters Design . . . . . . . . . . . . . . . . . . .

49

3.6.1

Uncertainty Description of Rigid-Body Model . . . . . . . . . . . . . . .

49

3.6.2

Uncertainty Description of Flexible-Body Model . . . . . . . . . . . . .

51

3.6.3

Robust Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . .

54

CHAPTER 4. UNCONTROLLED ROLL DRIFT . . . . . . . . . . . . . . . .

58

4.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

58

4.2

Pitch/Yaw Closed-Loop Instability Caused by Uncontrolled Roll Drift . . . . .

59

4.3

Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

61

4.3.1

Simplified Nonlinear Closed-Loop Pitch/Yaw Dynamics . . . . . . . . .

62

4.3.2

Linear Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . .

63

4.3.3

Nonlinear Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . .

68

4.4

New Pitch/Yaw Control Logic with Modified Commanded Quaternions

. . . .

79

4.5

Simple Adjustment of Control Gains . . . . . . . . . . . . . . . . . . . . . . . .

85

4.5.1

86

Rigid Body 6-DOF Nonlinear Simulation Results . . . . . . . . . . . . .

iv
CHAPTER 5. UNDERACTUATED CONTROL PROBLEM OF AN AXISYMMETRIC RIGID BODY

. . . . . . . . . . . . . . . . . . . . . . . . .

89

5.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

89

5.2

Steady-State Oscillations

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

91

5.3

Modified Attitude Quaternion Feedback Control Law . . . . . . . . . . . . . . .

98

5.4

Nonlinear Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

98

5.5

Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

99

5.6

A Special Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

CHAPTER 6. CONCLUSIONS

. . . . . . . . . . . . . . . . . . . . . . . . . .

110

APPENDIX A. A SUMMARY OF THE 6-DOF EQUATIONS OF MOTION 112

APPENDIX B. ADDITIONAL FIGURES FROM 6-DOF SIMULATION .

114

B.1 Atmospheric Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

B.2 Aerodynamic Coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
APPENDIX C. LINEARIZATION RESULTS . . . . . . . . . . . . . . . . . .

122

C.1 Nonlinear 6-DOF Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

C.2 Linear Rigid-Body Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
C.3 Linear State-Space Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
C.4 Linear Flexible-Body Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
APPENDIX D. ATTITUDE ERROR QUATERNION KINEMATIC DIFFERENTIAL EQUATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . .

129

APPENDIX E. LINEAR MODEL OF UNCONTROLLED ROLL DRIFT

WITH AERODYNAMIC DISTURBANCE . . . . . . . . . . . . . . . . . .

131

APPENDIX F. DERIVATION OF A STEADY-STATE OSCILLATION . .

133

F.1 A Steady-State Oscillation of the Autonomous System . . . . . . . . . . . . . . 133

F.2 Solution of the Quadratic Matrix Equation . . . . . . . . . . . . . . . . . . . . 135
F.3 State Equations of the Steady-State Oscillation . . . . . . . . . . . . . . . . . . 137

v
APPENDIX G. DERIVATION OF THE DERIVATIVE OF A LYAPUNOV
FUNCTION CANDIDATE . . . . . . . . . . . . . . . . . . . . . . . . . . . .

140

BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

142

vi

LIST OF TABLES

Table 2.1

Initial conditions at liftoff . . . . . . . . . . . . . . . . . . . . . . . . .

22

Table 3.1

Ares-I reference parameters at t = 60 sec . . . . . . . . . . . . . . . . .

39

Table 3.2

Ares-I structural bending modes for the pitch axis . . . . . . . . . . . .

41

Table 3.3

Ares-I structural bending modes for the yaw axis . . . . . . . . . . . .

41

Table 3.4

Ares-I rigid-body parametric uncertainty . . . . . . . . . . . . . . . . .

50

Table 4.1

Reference Ares-I CLV parameters at t = 60 sec. . . . . . . . . . . . . .

62

Table 4.2

Routh arrays. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

64

Table 4.3

Three cases for root locus stability analysis. . . . . . . . . . . . . . . .

65

Table 4.4

Three simulation cases for nonlinear stability analysis. . . . . . . . . .

70

Table 5.1

Ares-I reference parameters at t = 60 sec . . . . . . . . . . . . . . . . .

92

Table 5.2

Parameters of steady-state oscillation M2 . . . . . . . . . . . . . . . . .

93

Table 5.3

Simulation cases with p = 0.005 rad/sec . . . . . . . . . . . . . . . . .

99

vii

LIST OF FIGURES

Figure 1.1

Comparison of Space Shuttle, Ares-I, Ares-V, and Saturn V launch

vehicles [1]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 1.2

Ares-I CLV configuration [2].

Figure 1.3

Flexible mode shapes and sensor locations of the Ares-I Crew Launch

. . . . . . . . . . . . . . . . . . . . . . .

Vehicle [1]. Currently, rate-gyro blending is not considered for the Ares-I.

Figure 1.4

Ares-I CLV mission profile [2].

Figure 1.5

Interaction between the ascent flight control and the structural bending
mode.

. . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 2.1

Ares-I CLV 6-DOF simulation block diagram. . . . . . . . . . . . . . .

Figure 2.2

~ J,
~ K},
~ EarthIllustration of Earth-centered inertial reference frame {I,
~ e }, structural reference frame {~is , ~js , ~ks },
fixed reference frame {I~e , J~e , K
and body-fixed reference frame {~i, ~j, ~k}. . . . . . . . . . . . . . . . . .

Figure 2.3

Launch Complex 39B at Kennedy Space Center . . . . . . . . . . . . .

Figure 2.4

Earth-fixed launch pad reference frame with a local tangent plan at

Launch Complex 39B at Kennedy Space Center.

Figure 2.5

. . . . . . . . . . . .

11

~ J,
~ K},
~
Illustration of the Earth-centered inertial reference frame with {I,
the Earth-fixed launch pad (up, east, north) reference frame, and the
Ares-I orientation with {~i, ~j, ~k} on Launch Complex 39B. . . . . . . . .

Figure 2.6

Figure 2.7

13

A nominal ascent trajectory of Ares-I in the Earth-fixed launch pad

reference frame. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

24

A nominal ascent trajectory of Ares-I in the pitch plane. . . . . . . . .

24

viii
Figure 2.8

Time histories of conventional roll, pitch, and yaw angles, (, , ), of

the Ares-I CLV. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

25

Figure 2.9

Trajectory in ECI. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

25

Figure 2.10

Center of pressure and center of gravity. . . . . . . . . . . . . . . . . .

26

Figure 2.11

Center of gravity offset. . . . . . . . . . . . . . . . . . . . . . . . . . . .

26

Figure 2.12

Relative velocity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

27

Figure 2.13

Altitude. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

27

Figure 2.14

Mach number. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

28

Figure 2.15

Dynamic pressure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

28

Figure 2.16

Angle of attack and angle of sideslip. . . . . . . . . . . . . . . . . . . .

29

Figure 2.17

Bending load Qtotal . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

29

Figure 2.18

RCS torque. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

30

Figure 2.19

Angular velocity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

30

Figure 2.20

Euler angles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

31

Figure 2.21

Attitude quaternion. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

31

Figure 2.22

Attitude-error quaternion. . . . . . . . . . . . . . . . . . . . . . . . . .

32

Figure 2.23

Gimbal angles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

32

Figure 3.1

Reference trajectory and an operation point of Ares-I CLV in the pitch

plane.

Figure 3.2

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

34

A simplified dynamic model of a launch vehicle for preliminary pitch

control design. All angles are assumed to be small. . . . . . . . . . . .

35

Figure 3.3

Poles and zeros of Ares-I CLV rigid-body model transfer function. . . .

40

Figure 3.4

Root locus vs overall loop gain K of the pitch control system of a rigid
Ares-I model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

40

Figure 3.5

Block diagram of attitude control loop with flexible-body dynamics. . .

42

Figure 3.6

Flexible structure in the pitch plane. . . . . . . . . . . . . . . . . . . .

43

Figure 3.7

Block diagram of the flexible-body part of the pitch transfer function.

44

Figure 3.8

Pitch transfer function model of a reference model of the Ares-I CLV. .

44

ix
Figure 3.9

Root locus of the pitch control system without structural filters. . . . .

45

Figure 3.10

Root locus of the pitch control system with two NMP structural filters.

47

Figure 3.11

Impulse responses for the pitch attitude (in degrees) of the flexible
Ares-I. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 3.12

47

Impulse responses for the pitch gimbal angle (in degrees) of the flexible
Ares-I. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

48

Figure 3.13

General control configuration. . . . . . . . . . . . . . . . . . . . . . . .

50

Figure 3.14

M- structure for robust stability analysis.

. . . . . . . . . . . . . . .

51

Figure 3.15

Plant with multiplcative uncertainty. . . . . . . . . . . . . . . . . . . .

51

Figure 3.16

Bode plot of parameter uncertainty plant and perturbed plant samples.

52

Figure 3.17

Bode plot magnitude. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

53

Figure 3.18

Block diagram of perturbed transfer function Gpi (s). . . . . . . . . . .

54

Figure 3.19

Bode plot samples of Gf lex (s) with frequencies uncertainty and the
P
boundary of perturbed models 3i=1 Gpi (s). . . . . . . . . . . . . . . .

55

Figure 3.20

Block diagram of perturbed attitude control system. . . . . . . . . . .

56

Figure 3.21

-plot for RS of structural filters design. . . . . . . . . . . . . . . . . .

57

Figure 4.1

Attitude quaternion for an unstable closed-loop system caused by uncontrolled roll drift. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 4.2

Euler angles for an unstable closed-loop system caused by uncontrolled

roll drift. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 4.3

60

Gimbal angles for an unstable closed-loop system caused by uncontrolled roll drift. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 4.4

59

60

A simplified block diagram representation of the quaternion based ascent flight control system of Ares-I CLV. . . . . . . . . . . . . . . . . .
2
1q4e
q4e .

62

Figure 4.5

Plot of the function B =

. . . . . . . . . . . . . . . . . . . . . .

65

Figure 4.6

Root locus plot for Case 1. . . . . . . . . . . . . . . . . . . . . . . . . .

66

Figure 4.7

Root locus plot for Case 2. . . . . . . . . . . . . . . . . . . . . . . . . .

67

x
Figure 4.8

Root locus plot for Case 3, showing closed-loop instability with a nominal loop gain. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

67

Figure 4.9

Trajectory on the spherical surface q12 + q42 + z 2 = 1. . . . . . . . . . .

70

Figure 4.10

The positive invariant set M . . . . . . . . . . . . . . . . . . . . . . . .

71

Figure 4.11

The stable and unstable regions in M . . . . . . . . . . . . . . . . . . .

71

Figure 4.12

Angular velocity for Case 1. . . . . . . . . . . . . . . . . . . . . . . . .

72

Figure 4.13

Attitude quaternion for Case 1. . . . . . . . . . . . . . . . . . . . . . .

72

Figure 4.14

Phase portrait of q2 and q3 for Case 1. . . . . . . . . . . . . . . . . . .

73

Figure 4.15

Control inputs for Case 1. . . . . . . . . . . . . . . . . . . . . . . . . .

73

Figure 4.16

Angular velocity for Case 2. . . . . . . . . . . . . . . . . . . . . . . . .

74

Figure 4.17

Attitude quaternion for Case 2. . . . . . . . . . . . . . . . . . . . . . .

74

Figure 4.18

Phase portrait of q2 and q3 for Case 2. . . . . . . . . . . . . . . . . . .

75

Figure 4.19

Control inputs for Case 2. . . . . . . . . . . . . . . . . . . . . . . . . .

75

Figure 4.20

Angular velocity for Case 3. . . . . . . . . . . . . . . . . . . . . . . . .

76

Figure 4.21

Attitude quaternion for Case 3. . . . . . . . . . . . . . . . . . . . . . .

76

Figure 4.22

Phase portrait of q2 and q3 for Case 3. . . . . . . . . . . . . . . . . . .

77

Figure 4.23

q1 -q2 -q3 plot for Case 3.

. . . . . . . . . . . . . . . . . . . . . . . . . .

77

Figure 4.24

Control inputs for Case 3. . . . . . . . . . . . . . . . . . . . . . . . . .

78

Figure 4.25

q2 -q3 plot, from left to right q1 = 0,q1 = 0.2,q1 = 0.4,q1 = 0.6. . . . . .

78

Figure 4.26

A block diagram representation of a proposed method for computing a

new set of commanded attitude quaternion. . . . . . . . . . . . . . . .

79

Figure 4.27

Comparison of new and original attitude Euler angles command. . . .

81

Figure 4.28

Comparison of new and original attitude quaternion command. . . . .

82

Figure 4.29

Quaternion for a closed-loop system stabilized by the proposed control

logic employing modified commanded quaternion. . . . . . . . . . . . .

Figure 4.30

83

Euler angles for a closed-loop system stabilized by the proposed control

logic employing modified commanded quaternion. . . . . . . . . . . . .

83

xi
Figure 4.31

Gimbal angles for a closed-loop system stabilized by the proposed control logic employing modified commanded quaternion.

Figure 4.32

. . . . . . . . .

Comparison of ascent trajectories: with roll control system vs. without

roll control system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 4.33

84

84

Root locus plot for Case 3 but with a new derivative gain with = 4
in the pitch channel. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

86

Figure 4.34

Attitude quaternion of 6-DOF nonlinear simulation with Kd = 4Kd . . .

87

Figure 4.35

Euler angles of 6-DOF nonlinear simulation with Kd = 4Kd . . . . . . .

87

Figure 4.36

Gimbal angles of 6-DOF nonlinear simulation with Kd = 4Kd . . . . . .

88

Figure 4.37

Comparison of ascent trajectories. . . . . . . . . . . . . . . . . . . . . .

88

Figure 5.1

Steady-state oscillations M1 and M2 on the spherical surface q12 + q42 +

Figure 5.2

z 2 = 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
p
Steady-state oscillations M1 and M2 on the surface of cone z = q22 + q32 . 94

Figure 5.3

Angular velocity of steady-state oscillation M2 . . . . . . . . . . . . . .

94

Figure 5.4

Attitude quaternion of steady-state oscillation M2 . . . . . . . . . . . .

95

Figure 5.5

Euler angles of steady-state oscillation M2 . . . . . . . . . . . . . . . . .

95

Figure 5.6

Phase portrait of q and r of steady-state oscillation M2 . . . . . . . . .

96

Figure 5.7

Phase portrait of q2 and q3 of steady-state oscillation M2 . . . . . . . .

96

Figure 5.8

The relation between vectors (q, r)T and (q2 , q3 )T of steady-state oscillation M2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

97

Figure 5.9

Gimbal angles of steady-state oscillation M2 . . . . . . . . . . . . . . . .

97

Figure 5.10

Trajectory on the spherical surface q12 + q42 + z 2 = 1 for Case 1. . . . . 100

Figure 5.11

Detailed trajectory on the spherical surface q12 + q42 + z 2 = 1 for Case 1. 100

Figure 5.12

Angular velocity components q and r for Case 1. . . . . . . . . . . . . 101

Figure 5.13

Quaternion q1 , q2 and q3 for Case 1.

Figure 5.14

Euler angles for Case 1.

Figure 5.15

Gimbal angles for Case 1. . . . . . . . . . . . . . . . . . . . . . . . . . 102

Figure 5.16

Phase portrait of q and r for Case 1. . . . . . . . . . . . . . . . . . . . 103

. . . . . . . . . . . . . . . . . . . 101

. . . . . . . . . . . . . . . . . . . . . . . . . . 102

xii
Figure 5.17

Phase portrait of q2 and q3 for Case 1. . . . . . . . . . . . . . . . . . . 103

Figure 5.18

Phase portrait of q1 and q4 for Case 1. . . . . . . . . . . . . . . . . . . 104

Figure 5.19

Trajectory on the spherical surface q12 + q42 + z 2 = 1 for Case 2. . . . . 104

Figure 5.20

Detailed trajectory on the spherical surface q12 + q42 + z 2 = 1 for Case 2. 105

Figure 5.21

Angular velocity components q and r for Case 2. . . . . . . . . . . . . 105

Figure 5.22

Quaternion q1 , q2 and q3 for Case 2.

Figure 5.23

Euler angles for Case 2.

Figure 5.24

Gimbal angles for Case 2. . . . . . . . . . . . . . . . . . . . . . . . . . 107

Figure B.1

Speed of sound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

Figure B.2

Density of air . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

Figure B.3

Wind profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

Figure B.4

Base force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

Figure B.5

CA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

Figure B.6

CY

Figure B.7

CN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

Figure B.8

CM p . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

Figure B.9

CM y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

. . . . . . . . . . . . . . . . . . . 106

. . . . . . . . . . . . . . . . . . . . . . . . . . 106

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

Figure B.10 Rocket weight. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

Figure B.11 Rocket thrust. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Figure B.12 Moments of inertia. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

xiii

ACKNOWLEDGEMENTS

I would first like to thank my supervisor, Dr. Bong Wie, for all his inspiration, guidance
and financial support throughout my study at Iowa State University. Dr. Wie provided me
a chance to make my dream, studying flight control system of a launch vehicle, come true. I
was influenced by his research philosophy, which placed great emphasis on the fundamental
knowledge and deep understanding of control theory. His words there is no magic and what
is the next logical question motivated me to finish my Ph.D. program step by step.
Meeting Dr. Ping Lu was another great fortune for me. His organized course on optimal control and aerospace vehicle guidance reshaped my knowledge in those fields. Through
discussions with him, his intuition and sense of potential research direction opened my mind.
My appreciation also goes to all my committee members Dr. Thomas Rudolphi, Dr. Zhijian
Wang and Dr. John Basart for their valuable help and suggestions to my research. I would also
like to thank Dr. Nicola Elia who is a faculty in the Department of Electrical and Computer
Engineering. His big picture teaching style and enthusiasm in theoretical research made tedious
theorems visible and colorful.
I am indebted to my colleagues at ISU, including Anand Gopa Kumar, Insu Chang, Morgan
Baldwin, Matthew Hawkins, Zhongyuan Qian, Ying Zhou, De Huang, Haiyang Gao and a lot of
people in the Department of Aerospace Engineering. And I am especially grateful to Matthew
Hawkins for his help on polishing my writing.
My sincere thanks go to my parents for their advice and guidance throughout all my
education. And I am also thankful to my parents-in-law for their spiritual support.
Lastly, I wish to thank my wife Zhimei for all her love, support and sacrifice. She has
always stood beside me and encouraged me. To her I dedicate this dissertation.

xiv

ABSTRACT

This research focuses on dynamic modeling and ascent flight control of large flexible launch
vehicles such as the Ares-I Crew Launch Vehicle (CLV). A complete set of six-degrees-offreedom dynamic models of the Ares-I, incorporating its propulsion, aerodynamics, guidance
and control, and structural flexibility, is developed. NASAs Ares-I reference model and the
SAVANT Simulink-based program are utilized to develop a Matlab-based simulation and linearization tool for an independent validation of the performance and stability of the ascent
flight control system of large flexible launch vehicles. A linearized state-space model as well
as a non-minimum-phase transfer function model (which is typical for flexible vehicles with
non-collocated actuators and sensors) are validated for ascent flight control design and analysis.
This research also investigates fundamental principles of flight control analysis and design
for launch vehicles, in particular the classical drift-minimum and load-minimum control
principles. It is shown that an additional feedback of angle-of-attack can significantly improve
overall performance and stability, especially in the presence of unexpected large wind disturbances. For a typical non-collocated actuator and sensor control problem for large flexible
launch vehicles, non-minimum-phase filtering of unstably interacting bending modes is also
shown to be effective. The uncertainty model of a flexible launch vehicle is derived. The
robust stability of an ascent flight control system design, which directly controls the inertial
attitude-error quaternion and also employs the non-minimum-phase filters, is verified by the
framework of structured singular value () analysis. Furthermore, nonlinear coupled dynamic
simulation results are presented for a reference model of the Ares-I CLV as another validation
of the feasibility of the ascent flight control system design.
Another important issue for a single main engine launch vehicle is stability under mal-

xv
function of the roll control system. The roll motion of the Ares-I Crew Launch Vehicle under
nominal flight conditions is actively stabilized by its roll control system employing thrusters.
This dissertation describes the ascent flight control design problem of Ares-I in the event of
disabled or failed roll control. A simple pitch/yaw control logic is developed for such a technically challenging problem by exploiting the inherent versatility of a quaternion-based attitude
control system. The proposed scheme requires only the desired inertial attitude quaternion
to be re-computed using the actual uncontrolled roll angle information to achieve an ascent
flight trajectory identical to the nominal flight case with active roll control. Another approach
that utilizes a simple adjustment of the proportional-derivative gains of the quaternion-based
flight control system without active roll control is also presented. This approach doesnt require the re-computation of desired inertial attitude quaternion. A linear stability criterion
is developed for proper adjustments of attitude and rate gains. The linear stability analysis
results are validated by nonlinear simulations of the ascent flight phase. However, the first
approach, requiring a simple modification of the desired attitude quaternion, is recommended
for the Ares-I as well as other launch vehicles in the event of no active roll control.
Finally, the method derived to stabilize a large flexible launch vehicle in the event of
uncontrolled roll drift is generalized as a modified attitude quaternion feedback law. It is used
to stabilize an axisymmetric rigid body by two independent control torques.

xvi

NOMENCLATURE

speed of sound = 1117 ft/s at sea level in the standard atmosphere

Ae

nozzle exit area = 122.137 ft2

reference body length = 12.16 ft

(cx , cy , cz )

components of the center of mass location in the structural

reference frame with its origin at the top of vehicle
= (220.31, 0.02, 0.01) ft at t = 0

CA

axial force coefficient

CY

side force curve slope with respect to

CN 0

normal force coefficient at zero angle of attack

CN

normal force curve slope with respect to

CM r

rolling moment curve slope

CM p0

pitching moment coefficient at zero angle of attack

CM p

pitching moment curve slope

CM y

yawing moment curve slope

C B/I

direction cosine matrix of the frame B with respect to the frame I

lateral (side) force

drag (axial) force

Fbase

base force

(Faero.xb , Faero.yb , Faero.zb )

body-axis components of aerodynamic force

(Frkt.xb , Frkt.yb , Frkt.zb )

body-axis components of solid rocket booster force

(Frcs.xb , Frcs.yb , Frcs.zb )

body-axis components of RCS force

(Ftotal.xb , Ftotal.yb , Ftotal.zb )

body-axis components of total force

xvii

(Ftotal.xi , Ftotal.yi , Ftotal.zi )

inertial components of total force

(gx , gy , gz )

inertial components of the gravitational acceleration

(~i, ~j, ~k)

basis vectors of the body-fixed reference frame B

(~is , ~js , ~ks )

basis vectors of the structural reference frame S

~ J,
~ K)
~
(I,

basis vectors of the Earth-centered inertial reference frame I

~ e)
(I~e , J~e , K

basis vectors of the Earth-fixed equatorial rotating reference frame E

J2

Earths second-order zonal coefficient = 1.082631 103

J3

Earths third-order zonal coefficient = 2.55 106

J4

Earths fourth-order zonal coefficient = 1.61 106

Kp

proportional gain

Kd

derivative gain

vehicles mass

Mach number

normal force

p0

local atmospheric pressure

pe

nozzle exit pressure

(p, q, r)

body-axis components of
~

(q1 , q2 , q3 , q4 )

attitude quaternion of the vehicle with respect to an inertial

reference frame

(q1c , q2c , q3c , q4c )

desired attitude quaternion command from ascent guidance system

(q1e , q2e , q3e , q4e )

attitude-error quaternion

(
q1c , q2c , q3c , q4c )

a modified set of desired attitude quaternion associated

with (1 , 2c , 3c )

dynamic pressure

Re

Earths equatorial radius = 20,925,646 ft

Rp

Earths polar radius = 20,855,486 ft

~r

vehicles position vector

magnitude of ~r

xviii

reference area = 116.2 ft2

total thrust

T0

total vacuum thrust

thrust at any lower level in the atomosphere

(Taero.xb , Taero.yb , Taero.zb )

body-axis components of aerodynamic torque

(Trkt.xb , Trkt.yb , Trkt.zb )

body-axis components of solid rocket torque

(Trcs.xb , Trcs.yb , Trcs.zb )

body-axis components of RCS torque

Earths gravitational potential

(u, v, w)

~
body-axis components of V

~
V

vehicles inertial velocity vector

Ve

exit velocity of the solid rocket booster

~rel
V

vehicles velocity vector relative to the Earth-fixed reference frame

~w
V

velocity vector of the wind

~m
V

air stream velocity vector

Vm

magnitude of the air stream velocity

(Vm.xb , Vm.yb , Vm.zb )

body-axis components of vehicles air stream velocity vector

(x, y, z)

inertial components of vehicles position vector ~r

Xa

aerodynamic reference point location in the structural frame = 275.6 ft

Xg

gimbal attach point location in the structural frame = 296 ft

angle of attack

angle of sideslip

pitch gimbal angle (rotation about the body y-axis)

yaw gimbal angle (rotation about the body z-axis)

Earths gravitational parameter = 1.407644176 1016 ft3 /s2

Earths geocentric latitude

(1 , 2 , 3 )

Euler angles associated with (q1 , q2 , q3 , q4 ) for

C1 (1 ) C2 (2 ) C3 (3 )

(1c , 2c , 3c )

Euler angles associated with (q1c , q2c , q3c , q4c )

xix

(1 , 2c , 3c )

a modified set of Euler angles

density of the air

flexible-mode state vector

damping ratio of the flexible modes = 0.005

= diag(i )

undamped natural frequency matrix of the flexible modes

angular velocity vector of the vehicle

~ e = e K

angular velocity vector of the Earth where e = 7.29 105 rad/s

CHAPTER 1.

1.1

Figure 1.1

INTRODUCTION

Overview

Comparison of Space Shuttle, Ares-I, Ares-V, and Saturn V

launch vehicles [1].

The Ares-I Crew Launch Vehicle (CLV), being developed by the National Aeronautics and
Space Administration (NASA) [1], is a large, slender, and aerodynamically unstable vehicle.
It will be used to launch astronauts to Low Earth Orbit and rendezvous with the International
Space Station (ISS) or NASA Exploration System Mission Directorates earth departure stage
for lunar or other future missions beyond Low Earth Orbit. In Figure 1.1, its overall configuration is compared to other launch vehicles, including the Ares-V Cargo Launch Vehicle
and Saturn V. The Ares-I CLV is a two-stage rocket with a solid-propellant first stage derived
from the Shuttle Reusable Solid Rocket Motor/Booster and an upper stage employing a J-2X

2
engine derived from the Saturn J-2 engines.

1.2

Figure 1.2

Ares-I Configuration

Ares-I CLV configuration [2].

Ares-I CLV has an in-line configuration as illustrated in Figure 1.2, as opposed to the
Shuttle, which has the orbiter and crew placed beside the External Tank. In the event of an
emergency, the Orion Crew Module can be blasted away from the launch vehicle using the
Launch Abort System (LAS), which will fly directly upward, out of the way of the launch
vehicle.
The first stage is a new 5-segment solid rocket booster (SRB), derived from a 4-segment
space shuttle reusable solid rocket motor (RSRM). It will also include separation and recovery
systems, and SRB nozzle gimbal capability for thrust vector control (TVC). The second stage
or upper stage is powered by a liquid oxygen/liquid hydrogen constant-thrust J-2X engine. It
also contains avionics and other subsystems. The upper stage and first stage are connected by
the interstage, which also contains roll control system (RCS) [3] to prevent the vehicle from
spinning as it accelerates upward from the thrust of the SRB.
In addition to the LAS, upper stage, and first stage, the stack includes a forward skirt

3
and instrument unit, which connects the Orion to the Ares-I and contains the flight computer
for controlling the launch vehicle. The Ares-I navigation hardware will be located within an
instrumentation ring near the top of the second stage and just behind crew exploration vehicle
(CEV). An Inertial Measurement Unit (IMU) located in the instrument unit (at the top of
upper stage) will provide inertial position and velocity information to the navigation system
[4], and attitude quaternion and angular velocity to the Flight Control System (FCS). Pitch
and yaw body rates are obtained from two Rate Gyro Assemblies (RGAs) located near the
interstage and the first stage aft skirt (Figure 1.3).

Figure 1.3

Flexible mode shapes and sensor locations of the Ares-I Crew

Launch Vehicle [1]. Currently, rate-gyro blending is not considered for the Ares-I.

1.3

Ares-I Mission Profile

As seen in Figure 1.4, ascent flight trajectory begins at lift-off and lasts until first stage
separation. It takes approximately 120 seconds. This dissertation will focus on flight simulation
and ascent flight control system analysis and design during the ascent flight phase. During this
phase, Ares-I will experience velocities up to Mach 4.5 at an altitude of about 130, 000 feet,

Figure 1.4

Ares-I CLV mission profile [2].

and a maximum dynamic pressure (Max Q) of approximately 800 pounds per square foot.
The ascent flight trajectory can be separated into three phases: vertical flight, transition
turn and gravity turn. After SRB ignition, the launch vehicle flies vertically until it has
cleared the launch tower. The vertical, stationary attitude flight of the launch vehicle lasts
approximately 6 seconds, and then it commences a combined pitch/roll maneuver in order
to head the crew window to the launch azimuth, which is defined as the angle between the
vertical ascent trajectory plane (or the so-called pitch plane) and a vector pointing from the
launch pad toward the North Pole. As a result, the required heads-down orientation of the
crew can be maintained during the ascent flight phase [5]. This maneuver is also known as the
transition turn [6]. The vehicle transitions from vertical rise to the gravity turn condition. It
will fly a gravity turn trajectory until burn out of the SRB and separation [7]. The gravity turn
maneuver is used to achieve an ascent trajectory with zero angle of attack and zero sideslip
angle (e.g. flying into the relative wind) by minimizing structural bending loads.
Since a detailed discussion of the launch vehicle guidance and trajectory optimization of
Ares-I CLV is beyond the scope of this dissertation, the reader is referred to the literature for

5
a more complete treatment [7, 8, 9].

1.4

Interaction Between Structures and Flight Control System

Figure 1.5

Interaction between the ascent flight control and the structural

bending mode.

A launch vehicle is essentially a long slender beam,thus it is structurally very flexible. IMUs
are placed along the vehicle body to sense angular displacement or rate for feedback control.
The IMU measures local elastic distortions as well as rigid body motion. As a result, one
significant risk for a large flexible launch vehicle ascent flight control system is the potential
for interaction between the ascent flight control and the structural bending mode. Because the
first bending mode frequency is usually close to the crossover regime of the rigid body control
system, the control system has the potential to excite the bending mode and destabilize the
vehicle dynamics [11].
This structural feedback problem can be illustrated by Figure 1.5. TVC actuators and

6
attitude sensors of launch vehicles are not collocated. The sensors pick up both rigid-body
motion of the vehicle and motion caused by structural deformations at the location of the
sensors. These deformations affect the command to the actuator (usually gimbaled rocket
engines). Since engines apply forces to the launch vehicles structure, energy can be fed into
the structure at various frequencies. This will reinforce elastic oscillations, leading ultimately
to structural failure of the vehicle.
Conventional roll-off filters and/or notch filters were often used in practice for the stabilization of such unstably interacting bending modes of large flexible launch vehicles [10, 12, 13, 14,
15]. However, in [16, 20, 21], use of non-minimum-phase (NMP) structural filters was shown
to be very effective and robust in controlling flexible structures with non-collocated actuators
and sensors. In Chapter 3 it will be shown that the NMP filters can stabilize, effectively and
robustly, the bending modes of the Ares-I CLV.

1.5

Underactuated Control Problem

The active RCS of Ares-I CLV provides rotational azimuth control to perform a roll orientation maneuver after lift-off and to mitigate against adverse roll torques [22]. It was harvested
from the Peacekeeper missiles fourth stage axial thruster system. The challenge for the roll
control system is to be able to control large rolling moments, with continuously decreasing
principal moment of inertia during flight. RCS failure is a potential threat to the safety of
astronauts and launch vehicles. In Chapter 4, the problem of ascent flight control in the event
of uncontrolled roll drift will be discussed. Furthermore, it could be generalized as a typical
underactuated control problem [23, 24, 25, 26, 27]. In Chapter 5, methods developed to stabilize Ares-I will be generalized as a method to stabilize an axisymmetric rigid body using two
control inputs.

CHAPTER 2.

6-DEGREE-OF-FEEDOM DYNAMIC MODELING

2.1

Introduction

A complete set of coupled dynamic models of the Ares-I CLV, incorporating its propulsion,
aerodynamics, guidance and control, and structural flexibility will be described in this chapter.
The Ares-I CLV is a large, slender, and aerodynamically unstable vehicle. NASAs reference
model and SAVANT Simulink-based program [11, 28, 29], as well as various dynamic models
of launch vehicles developed previously in the literature [10, 16, 30, 31, 32], were utilized to
develop a Matlab-based simulation and linearization tool for an independent validation of the
performance and stability of the ascent flight control system of the Ares-I CLV. The block
diagram of the Matlab-based simulation program is shown in Figure 2.1.

Figure 2.1

Ares-I CLV 6-DOF simulation block diagram.

2.2

Reference Frames and Rotational Kinematics

Various reference frames, which are essential for describing the six-degrees-of-freedom dynamic model of launch vehicles, are discussed in this section.

2.2.1

Earth-Centered Inertial Reference Frame

V
K Ke

= (p, q, r)

Structure Frame

ks

is
js

cg

Body Frame (x, y, z)

or (1, 2, 3)

Equator
Inertial Frame
Vernal
Equinox
Direction

Greenwich
longitude = 0
latitude = 0

Je
J

I e Earth Frame

Figure 2.2

~ J,
~ K},
~
Illustration of Earth-centered inertial reference frame {I,
~
~
~
Earth-fixed reference frame {Ie , Je , Ke }, structural reference
frame {~is , ~js , ~ks }, and body-fixed reference frame {~i, ~j, ~k}.

~ J,
~ K}
~ has its origin at the
The Earth-centered inertial frame with a set of basis vectors {I,
Earth center as illustrated in Figure 2.2. The z-axis is normal to the equatorial plane and the
x- and y-axes are in the equatorial plane. The x-axis is along the vernal equinox direction.
Because the Earths orbital motion around the sun is negligible in the trajectory analysis of
launch vehicles, this frame is often considered as an inertial reference frame.
The position vector ~r of a launch vehicle is then described as
~
~r = xI~ + y J~ + z K

(2.1)

The inertial velocity and the inertial acceleration of a launch vehicle become, respectively,
~ = x I~ + y J~ + z K
~
V

(2.2)

~ = x
~
V
I~ + yJ~ + zK

Figure 2.3

(2.3)

Launch Complex 39B at Kennedy Space Center

For example, if the inertial position vector of the Ares-I at liftoff from Launch Complex
39B (Figure 2.3) at Kennedy Space Center (with longitude 80.6208 deg west and latitude
28.6272 deg north) is given by
~ = 8.7899E4I~ 1.8385E7J~ + 9.9605E6K
~ (ft)
~r(0) = x(0)I~ + y(0)J~ + z(0)K

(2.4)

then, the inertial velocity vector of the vehicle at liftoff is obtained as

~ (0) = x(0)
~ =
V
I~ + y(0)
J~ + z(0)
K
~ e ~r(0) = 1340.65I~ 6.41J~ (ft/sec)

(2.5)

~ is the angular velocity vector of the Earth and e = 7.2921 105 rad/s,
where
~ e = e K
which corresponds to 360 deg per sidereal day of 23 h 56 min 4 s.

2.2.2

Earth-Fixed Equatorial Reference Frame

~ e }, with its
The geocentric equatorial rotating frame with a set of basis vectors {I~e , J~e , K
origin at the Earth center, is fixed to the Earth (Figure 2.2). Its Z-axis is normal to the

10
equatorial plane and its x- and y-axes are in the equatorial plane. However, its x-axis is along
~ e which is the rotational
the Greenwich meridian. This Earth frame has an angular velocity
velocity of the Earth.

2.2.3

Body-Fixed Reference Frame

The body-fixed frame with basis vectors {~i, ~j, ~k} is fixed to the vehicles body as illustrated
in Figure 2.2. Its origin is the center of mass. The ~i-axis is along the vehicles longitudinal axis.
The ~k-axis perpendicular to the ~i-axis points downward while the ~j-axis points rightward.
~ is then expressed as
The inertial velocity vector V
~ = u~i + v~j + w~k
V

(2.6)

The angular velocity vector

~ of the launch vehicle is also expressed as

~ = p~i + q~j + r~k

(2.7)

The inertial acceleration vector is then described as

~ = (u~i + v~j + w ~k) +
~
V
~ V

2.2.4

(2.8)

Structural Reference Frame

A structural reference frame with basis vectors {~is , ~js , ~ks } and with its origin at the top of
vehicle is also employed in the SAVANT program. The locations of center of gravity, gimbal
attach point, aerodynamic reference point, and other mass properties, are defined using this
structural frame. However, Eulers rotational equations of motion will be written in terms of
the body-fixed frame with its origin at the center of gravity. Because ~is = ~i, ~js = ~j, and
~ks = ~k, we have

B/S

1 0 0

=
0 1 0

0 0 1

where C B/S is the direction cosine matrix of frame B with respect to frame S.

(2.9)

11
2.2.5

Earth-Fixed Launch Pad Reference Frame

Figure 2.4

Earth-fixed launch pad reference frame with a local tangent

plan at Launch Complex 39B at Kennedy Space Center.

In order to visualize the ascent flight trajectory in an intuitive way, another reference
frame, called the Earth-fixed launch pad (up, east, north) reference frame is introduced here.
Its origin is at the Launch Complex 39B at NASAs Kennedy Space Center with the latitude
28.6272 deg and longitude 80.6208 deg as illustrated in Figures 2.4.

2.2.6

Euler Angles and Quaternions

The coordinate transformation to the body frame B from the inertial frame I is described
by three Euler angles (1 , 2 , 3 ). For a rotational sequence of C 1 (1 ) C 2 (2 ) C 3 (3 ), we
have

C B/I

cos 2 cos 3
cos 2 sin 3
sin 2

=
sin

sin

cos

cos

sin

sin

sin

sin

+
cos

cos

sin

cos

1
2
3
1
3
1
2
3
3
1

cos 1 sin 2 cos 3 + sin 1 sin 3 cos 1 sin 2 sin 3 sin cos 3 cos 1 cos
(2.10)

12
which is the direction cosine matrix of the body frame B relative to the inertial frame I.
However, the three Euler angles (1 , 2 , 3 ) do not actually represent the vehicles roll, pitch,
and yaw attitude angles to be used for attitude feedback control.
The rotational kinematic equation for three Euler angles (1 , 2 , 3 ) is

1
cos 1 sin 2

cos 2 sin 1 sin 2

1
=

cos 1 cos 2 sin 1 cos 2

2 cos 2 0

3
0
sin 1
cos 1

given by

(2.11)

The inherent singularity problem of Euler angles can be avoided by using quaternions [33].
The rotational kinematic equation in terms of quaternion

q1
0
r q p

q2 1 r 0
p q

2
q3
q p 0 r

q4
p q r 0

(q1 , q2 ,

q
1

q2

q3

q4

q3 , q4 ) is given by

(2.12)

where the quaterions are related to the three Euler angles as follows:
q1 = sin(1 /2) cos(2 /2) cos(3 /2) cos(1 /2) sin(2 /2) sin(3 /2)
q2 = cos(1 /2) sin(2 /2) cos(3 /2) + sin(1 /2) cos(2 /2) sin(3 /2)

(2.13)

q3 = cos(1 /2) cos(2 /2) sin(3 /2) sin(1 /2) sin(2 /2) cos(3 /2)
q4 = cos(1 /2) cos(2 /2) cos(3 /2) + sin(1 /2) sin(2 /2) sin(3 /2)
The coordinate transformation matrix to the body frame from the inertial frame in terms
of quaternions is

B/I

2
2
1 2(q2 + q3 ) 2(q1 q2 + q3 q4 ) 2(q1 q3 q2 q4 )

2
2
=
2(q1 q2 q3 q4 ) 1 2(q1 + q3 ) 2(q2 q3 + q1 q4 )

2(q1 q3 + q2 q4 ) 2(q2 q3 q1 q4 ) 1 2(q12 + q22 )

(2.14)

We also have

I/B

B/I 1

= [C

2(q22

q32 )

+
2(q1 q2 q3 q4 ) 2(q1 q3 + q2 q4 )
1

2
2
= [CB/I ]T =
2(q1 q2 + q3 q4 ) 1 2(q1 + q3 ) 2(q2 q3 q1 q4 )

2(q1 q3 q2 q4 ) 2(q2 q3 + q1 q4 ) 1 2(q12 + q22 )

(2.15)

13

Figure 2.5

2.2.7

Illustration of the Earth-centered inertial reference frame with

~ J,
~ K},
~ the Earth-fixed launch pad (up, east, north) refer{I,
ence frame, and the Ares-I orientation with {~i, ~j, ~k} on Launch
Complex 39B.

Initial Position of Ares-I CLV on the Launch Pad

In NASAs SAVANT program [28, 29], the inertial attitude quaternion of the Ares-I are
computed with respect to the ECI frame. For the Ares-I orientation on the launch pad,
the x-axis of body-fixed reference frame points up to the sky, the y-axis points northward,
and the z-axis points westward, as illustrated in Figure 2.5. Consequently, the initial Euler
angles (1 , 2 , 3 ) at t = 0 are (89.9881, 28.6090, 90.2739) deg for the rotational sequence
of C1 (1 ) C2 (2 ) C3 (3 ) [16]. It is emphasized that these Euler angles are not the
traditional (roll, pitch, yaw) attitude angles which describe the orientation of a launch vehicle
with respect to the boost trajectory plane or the so-called pitch plane.

14

2.3

The 6-DOF Equations of Motion

The six-degrees-of-freedom (6-DOF) equations of motion of a launch vehicle consist of the

translational and rotational equations. The translational equation of motion of the center of
gravity of a launch vehicle is simply given by
~ = F~
mV

(2.16)

where F~ is the total force acting on the vehicle. Using Equation. (2.3) and Equation. (2.8), we
obtain the translational equation of motion of the form
~
~ = F
x
I~ + yJ~ + zK
m

(2.17)

~
~ = F
u~i + v~j + w ~k +
~ V
m

(2.18)

or

The Eulers rotational equation of motion of a rigid vehicle is

~ = T~
H

(2.19)

~ is the angular momentum vector and T~ is the total external torque about the center
where H
of gravity. The angular momentum vector is often expressed as
~ = I
H
~

(2.20)

where
~ = p~i + q~j + r~k is the angular velocity vector and I is the vehicles inertia dyadic
about the center of gravity of the form [16]

I =

 Ixx Ixy Ixz

~i ~j ~k
Ixy Iyy Iyz

Ixz Iyz Izz

~i


~j


~k

(2.21)

The rotational equation of motion is then given by

I
~ +
~ I
~ = T~
where
~ = p~i + q~j + r~k is the angular acceleration vector.

(2.22)

15
2.3.1

Aerodynamic Forces and Moments

Aerodynamic forces and moments depend on the vehicles velocity relative to the surrounding air mass, called the air speed. It is assumed that the air mass is static relative to the Earth.
That is, the entire air mass rotates with the Earth without slippage and shearing. A hybrid
approach of CFD and wind tunnel data have been developed for Ares-I [34]. The air stream
~m is then described by
velocity vector V
~m = V
~rel V
~w = V
~
~w
V
~ e ~r V

(2.23)

~rel is the vehicles velocity vector relative to the Earth-fixed reference frame, V
~w is
where V
~ is the inertial velocity of the vehicle,
the local disturbance wind velocity, V
~ e is the Earths
rotational angular velocity vector, and ~r is the vehicles position vector from the Earth center.
The matrix form of Eq. (2.23) in the body frame is

V
u
0

0
e
m.xb

x Vw.xb

B/I

V
0
0
m.yb = v C
e
y Vw.yb

Vm.zb
w
0
0
0
z
Vw.zb

(2.24)

where (Vm.xb , Vm.yb , Vm.zb ) are the body-axis components of the vehicles air stream velocity
~ and
~
vector. Note that ~r = xI~ + y J~ + z K
~ e = e K.
The aerodynamic forces are expressed in the body-axis frame as
D = CA QS Fbase

(2.25a)

C = CY QS

(2.25b)

N = (CN 0 + CN )QS

(2.25c)

where the base force Fbase is a function of the altitude, the aerodynamic force coefficients are
functions of Mach number, and
M=

Vm
= Mach number
a

1
Q = Vm2 = dynamic pressure
2

(2.26)

(2.27)

16


= arctan


= arcsin

Vm.zb
Vm.xb

Vm.yb
Vm

= angle of attack

(2.28)

= sideslip angle

(2.29)

The speed of sound a and the air density are functions of the altitude h.
Furthermore, we have
Faero.xb = D

(2.30a)

Faero.yb = C

(2.30b)

Faero.zb = N

(2.30c)

The aerodynamic moments about the center of gravity are also expressed in the body-axis
frame as

Taero.xb

T
aero.yb

Taero.zb

cz
cy
0

= c
0
Xa + cx
z


cy Xa cx
0

Faero.xb

F
aero.yb

Faero.zb

CM r QSb

+ (C

+
C
)QSb
M p0
M p

CM y QSb
(2.31)

where Xa = 275.6 ft is the aerodynamic reference point in the structure frame, (cx , cy , cz )
is the center of gravity location in the structure reference frame with its origin at the top of
vehicle. At t = 0, we have (cx , cy , cz ) = (220.31, 0.02, 0.01) ft. The aerodynamic moment
coefficients are functions of Mach number.

2.3.2

Gravity Model

The J4 gravity model used in the SAVANT program is given as






Re2
3
1
Re 5
3
2
3
C1 = 1 + 2 3J2
sin
+ 4J3
sin sin
r
2
2
r 2
2


2
Re 35
15
3
4
2
sin
sin +
(2.32)
+5J4 2
r
8
4
8
where is the Earths geocentric latitude.




Re
15
3
R2
35
15
C2 = J2 (3 sin ) +
J3
sin2
+ 2e J4
sin3
sin
r
2
2
r
2
2

(2.33)

17
The inertial components of the gravitational acceleration are

z/r y/r
g
x/r

0
x

g =

0
x/r
y r2 C1 y/r C2 z/r

y/r x/r
0
gz
z/r

y/r

x/r

(2.34)

The mathematical models used in the SAVANT program for computing the vehicles altitude h are summarized as
f=

Re R p
1
=
Re
298.257

tan =


A=

cos
Re

tan
(1 f )2

2


+

(2.35)

(2.36)

sin
Rp

2
(2.37)

x2 + y 2 cos z sin

Re2
Rp2

B=


C=

z
Rp

2

B
h=
A

x2 + y 2
1
Rp2

s

B
A

2

C
A

(2.38)

(2.39)

(2.40)

where f is the Earths flatness parameter, is the geocentric latitude, and is the geodetic
latitude (which is commonly employed on geographical maps).

2.3.3

Rocket Propulsion Model

The rocket thrust is simply modeled as

T = T0 + (pe p0 )Ae

(2.41)

where T is the total thrust force, T0 = |m|V

e the jet thrust, pe the nozzle exit pressure, p0 the
local atmospheric pressure (a function of the altitude), m
the propellant mass flow rate, Ve the
exit velocity, and Ae the nozzle exit area (= 122.137 ft2 ).

18
If the thrust in the vacuum of space above the atmosphere is called T , then the thrust at
any lower level in the atmosphere is [8]
T = T p 0 Ae

(2.42)

where T = T0 + pe Ae .
The body-axis components of the thrust force are

F
rkt.xb T

F

rkt.yb = T z

Frkt.zb
T y

(2.43)

where y and z are the pitch and yaw gimbal deflection angles, respectively. Gimbal deflection
angles are assumed to be small (with max = 10 deg).
The body-axis components of the rocket thrust-generated torque are

cz
cy
Trkt.xb 0
Frkt.xb

0
Xg + cx
rkt.yb = cz
Frkt.yb

Trkt.zb
cy Xg cz
0
Frkt.zb

(2.44)

where Xg = 296ft is the gimbal attach point location in the structural frame.
The body-axis components of the roll control torque from the RCS are

Trcs.xb Trcs

rcs.yb = 0

Trcs.zb
0
2.3.4

(2.45)

Guidance and Control

The commanded quaternion (q1c , q2c , q3c , q4c ) computed by the guidance system are used
to generate the attitude-error quaternion (q1e , q2e , q3e ,

q
q
q3c q2c
1e 4c

q2e q3c q4c
q1c

q3e q2c q1c q4c

q4e
q1c
q2c
q3c

q4e ) as follows [16]:

q1c
q
1

q2c
q2

q3c q3

q4c
q4

(2.46)

19
where the attitude quaternion (q1 , q2 , q3 , q4 ) are computed by numerically integrating the
kinematic differential equation, Equation. (2.12).
The guidance command used in the simulation is for the ISS mission at an orbital inclination
of 51.6 deg [7].
The simplified control laws of the ascent flight control system are then described as
Trcs = Kpx (2q1e ) Kdx p
Z
y = Kpy (2q2e ) Kiy (2q2e )dt Kdy q
z = Kpz (2q3e ) Kdz r

(2.47a)
(2.47b)
(2.47c)

An integral control is added to the pitch control channel. The terms (2q1e , 2q2e , 2q3e ) are the
roll, pitch, and yaw attitude errors, respectively. This quaternion-error feedback control is in
general applicable for arbitrarily large angular motion of vehicles [16, 17, 18, 19]. Feedback
of Euler-angle errors (1 1c , 2 2c , 3 3c ) is not applicable here because the Euler
angles employed in this paper (also used in the SAVANT program) are defined with respect to
the Earth-centered inertial reference frame, not with respect to the so-called pitch plane or a
navigation reference frame of launch vehicles [10, 16, 30, 31, 32, 35].

2.3.5

Flexible-Body Modes

For the purposes of ascent flight control system stability analysis, the lateral vibration
modes are important, since this motion is sensed by the IMU [10]. Usually, a forced vibration
of a free-free beam model can be expressed mathematically by Euler-Bernoulli beam model,
neglecting shear distortion and rotational inertia, as follows:
m(l)

2 (l, t)
2
2 (l, t)
+ 2 [EI(l)
] = T (t)
2
t
l
l2

(2.48)

where m is mass per unit length, EI is bending stiffness and is beam deflection. Note that
for the case of free vibration, the term on the right side of the equal sign in Equation. (2.48)
is zero.

20
For the free-free case where the shear

2
l2

and bending moment

3
l3

at the ends of the beam

are zero, the boundary conditions are given by

2 (0, t)
2 (L, t)
=
=0
l2
l2

(2.49)

3 (0, t)
3 (L, t)
=
=0
l3
l3

(2.50)

Assuming that there is one solution of the free vibration, it can be written in the form
(l, t) = (l)(t)

(2.51)

where (l) presents the shape of a natural vibration mode and (t) is the modal coordinate of
this mode.
Substituting Equation. (2.51) in Equation. (2.48) leads to

1 d2 (t)
d2
d2 (l)
=
[EI(l)
]
(t) dt2
dl2
dl2

(2.52)

The left side is a function of time t only, and the right side is a function of l. This equation
is valid only if the function on either side is equal to some constant, say 2 . Thus the partial
differential equation Equation. (2.48) becomes two ordinary differential equations as follows:
d2 (t)
+ 2 (t) = 0
dt2

(2.53)

d2 (l)
d2
[EI(l)
] 2 m(l)(l) = 0
dl2
dl2

(2.54)

where is the vibration frequency corresponding to the mode (l). Here, for Equation. (2.54),
numerical methods must be used to calculate the natural frequencies and mode shapes corresponding to specific boundary conditions. Once these are known, the complete solution in the
case of free vibrations may be written as
(l, t) =

i (x)i (t)

(2.55)

i=1

where i (t) is the ith normal mode shape, i (t) is the ith modal coordinate. It is very straightforward to express the forced motion in these terms and to take account of structural damping.

21
Detailed derivations for the Euler-Bernoulli model and forced vibrations of nonuniform
beam can be found in text books [10, 16, 36]. Structural dynamics of Ares-I were modeled
as linear second order systems with a damping ratio of 0.5%. The value 0.5% used for flight
control analysis is considered conservative. A Finite Element Model (NASTRAN/PATRAN)
was used to obtain bending mode frequencies and shapes [34].
A flexible-body model of Ares-I is expressed as
+ 2 + 2 = T F rkt

(2.56)

where F rkt = (Frkt.xb , Frkt.yb , Frkt.zb )T and is the flexible mode influence matrix at the
gimbal attach point.
Sensor measurements including the effects of the flexible bending modes are modeled as

2q
1e

+
eattitude =
(2.57)
2q
2e

2q3e

erate

=
q +

r

(2.58)

where is the flex-mode influence matrix at the instrument unit location (Figure 1.3).
A summary of the 6-DOF equations of motion can be found in Appendix A.

2.4

Simulation Results of the Rigid Body Ares-I Crew Launch Vehicle

A set of initial conditions for the Ares-I is provided in Table 2.1. The corresponding the
initial Euler angles (1 , 2 , 3 ) at t = 0 are (89.9881, 28.6090, 90.2739) deg. The inertia
matrix about

Ixx

I
xy

Ixz

the body frame with its origin

Ixy Ixz 1.2634E6

Iyy Iyz
= 1.5925E3

Iyz Izz
5.5250E4

at the center of gravity at t = 0 is

1.5925E3 5.5250E4

2
2.8797E8 1.5263E3
slug-ft

1.5263E3 2.8798E8

(2.59)

22

Table 2.1

Initial conditions at liftoff

State variables
x
y
z
x
y
z
p
q
r
q1
q2
q3
q4

Initial values
1340.65
6.41
0
8.7899 104
1.8385 107
9.9605106
3.4916105
6.4018105
0
0.3594
0.6089
0.3625
0.6072

Units
ft/s
ft/s
ft/s
ft
ft
ft
rad/s
rad/s
rad/s

Flexible-body mode shape matrices (with 6 flexible modes) are

0.000000272367963 0.000000174392026 0.000000347086527

=
0.000364943105155 0.006281028219530 0.000491932740239

0.006281175443849 0.000364891432306 0.006260333099131

0.000000369058169

0.006259406451949 0.000542750533582 0.007673360355205

0.000491798506301 0.007676195145027 0.000542218216634

0.000000266173427

(2.60)

0.000000329288262

0.002287263504447 0.003936428406260 0.002315093057592

=
0.193164818571118 0.011222878633268 0.253963647069578

0.011222390194941 0.193169876159476 0.019955470041040

0.003866041325542
0.002898204936058
0.005033045198158

3
0.019956097043432 0.130518411476224 0.009224310239736
10

0.253971718426341 0.009226496087855 0.130493158175449

(2.61)
The first three bending mode frequencies are: 6 rad/s, 14 rad/s, and 27 rad/s. The damping
ratio is assumed as = 0.005.

23
The simulation results of a test case for a Matlab-based simulation program are shown in
Figures 2.9- 2.23. These results are identical to those obtained using the SAVANT program for
the same test case. However, these simulation results were for a preliminary reference model
of the Ares-I available to the public, not for the most recent model of the Ares-I with properly
updated, ascent flight guidance and control algorithms. The purpose of this chapter was to
develop a Matlab-based simulation tool for an independent validation of the performance and
stability of NASAs ascent flight control system baseline design for the Ares-I rigid body model.
The center of pressure (cp) location shown in Figure 2.10 was computed as
Xcp = Xa

(CM p0 + CM p )b
CN 0 + CN

(2.62)

where Xcp is the distance to the cp location from the top of vehicle and Xa is the distance
to the aerodynamic reference point from the top of vehicle (i.e., the origin of the structure
reference frame).
A nominal ascent flight trajectory of the Ares-I obtained using the Matlab-based program
is shown in Figure 2.6. The nominal ascent trajectory on the pitch plane is shown in Figure 2.7.
The launch azimuth can be seen to be about 42 deg. The launch azimuth is defined as the
angle between the vertical trajectory plane (or pitch plane) and a vector pointing from the
launch pad toward the North Pole. Time histories of a different set of Euler angles of the
Ares-I CLV, often called (roll, pitch, yaw) attitude angles, with respect to the vertical pitch
plane, are shown in Figure 2.8. A 48 deg roll maneuver, prior to the start of the gravity turn
pitch maneuver, can be seen in this figure. Because the crew are oriented with their heads
pointing east on the launch pad, the 48 deg roll maneuver is designed to maintain the required
heads-down orientation of the crew [5]. Because the International Space Station mission has a
higher inclination (51.6 deg) than the lunar mission (28.5 deg), the larger roll angle maneuver
has been the primary focus in the roll control system design for Ares-I CLV [7].
Additional figures from the simulation of the Ares-I can be found in Appendix B.

24

x 10

Trajectory
Projection

12
10
Up

(ft)

8
6
4
North

2
0
10
East

x 10

0
(ft)

Figure 2.6

10

x 10

(ft)

A nominal ascent trajectory of Ares-I in the Earth-fixed launch

pad reference frame.

14

x 10

12

Up (ft)

10

Figure 2.7

6
8
10
Pitchingplane position (ft)

12

14

16
4

x 10

A nominal ascent trajectory of Ares-I in the pitch plane.

25

roll (deg)

100
50
0
50

20

40

60

80

100

120

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

pitch (deg)

100
50
0
50

yaw (deg)

2
0
2
4

Figure 2.8

Time histories of conventional roll, pitch, and yaw angles,

(, , ), of the Ares-I CLV.

x 10
1.015

Z Position (ft)

1.01

1.005

0.995
1.838
1.84
7

x 10

2
1.842

1
1.844

Y Position (ft)

0
1.846

Figure 2.9

x 10
X Position (ft)

Trajectory in ECI.

26

cx

cp
50

Location (ft)

100

150

200

250

20

Figure 2.10

40

60
Time (sec)

80

100

120

Center of pressure and center of gravity.

0.06

cy (ft)

0.05
0.04
0.03
0.02

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

0.005

cz (ft)

0.01
0.015
0.02
0.025
0.03

Figure 2.11

Center of gravity offset.

27

4500
4000

3000
2500
2000
1500
1000
500
0

20

40

Figure 2.12

60
Time (sec)

80

100

120

100

120

Relative velocity.

14

x 10

12

10

Altitude (ft)

Relative Velocity (ft/sec)

3500

20

40

60
Time (sec)

Figure 2.13

80

Altitude.

28

4.5
4
3.5

Mach Number

3
2.5
2
1.5
1
0.5
0

20

40

Figure 2.14

60
Time (sec)

80

100

120

100

120

Mach number.

700

Dynamic Pressure (psf)

600

500

400

300

200

100

20

40

Figure 2.15

60
Time (sec)

80

Dynamic pressure.

29

Angle of Attack (deg)

Angle of Sideslip (deg)

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

Figure 2.16

Angle of attack and angle of sideslip.

2000
1800

Bending Load Qtotal (psfdeg)

1600
1400
1200
1000
800
600
400
200
0

20

40

Figure 2.17

60
Time (sec)

80

Bending load Qtotal .

100

120

30

x 10

RCS Torque (lbft)

2
1
0
1
2
3
4

20

40

60
Time (sec)

Figure 2.18

80

100

120

RCS torque.

p (rate/sec)

0.2
0
0.2

20

40

60

80

100

120

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

q (rate/sec)

0.04
0.02
0
0.02

r (rate/sec)

0.02
0.01
0
0.01

Figure 2.19

Angular velocity.

31

Actual output
Command

1 (deg)

150
100
50

20

40

60

80

100

120

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

2 (deg)

20
40
60

3 (deg)

0
50
100

Figure 2.20

Euler angles.

q1

0.6
Actual output
Command

0.4
0.2

20

40

60

80

100

120

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

q2

0
0.5
1

q3

0.5
0
0.5

Figure 2.21

Attitude quaternion.

32

0.04

q1e

0.02
0
0.02

20

40

60

80

100

120

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

0.02

q3e

0.01
0
0.01
0.02

q3e

0.01
0
0.01

Figure 2.22

Attitude-error quaternion.

y (deg)

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

z (deg)

0.5

0.5

Figure 2.23

Gimbal angles.

33

CHAPTER 3.

ANALYSIS AND DESIGN OF ASCENT FLIGHT

CONTROL SYSTEMS

3.1

Introduction

In analyzing and designing the attitude control system, the short period dynamics of the
launch vehicle is used for expressing the rigid-body and flexible-body motion. It is assumed
that the motion of the launch vehicle consists of small deviations from a reference trajectory.
Another important assumption is that time varying mass, inertial, and other physical properties
are changing slowly during the flight. As a result, all parameters of the launch vehicle can be
frozen over a short period of time. In this way, analysis and design techniques for Linear
Time-Invariant (LTI) systems can be exploited most fully.
In this section, a Matlab-based program is used to generate the reference trajectory of the
Ares-I CLV. In this program, the Ares-I is considered to achieve attitude quaternion command
perfectly and data for the reference trajectory is calculated in the ECI frame. Another Matlabbased program is developed to compute an LTI model at any operation point as shown in
Figure 3.1. Linearization results in the ECI and linear state-space equations of both rigidbody and flex-body model can be found in Appendix C.

34

Figure 3.1

3.2

Reference trajectory and an operation point of Ares-I CLV in

the pitch plane.

Pitch Control Analysis of Rigid Launch Vehicles

For the preliminary analysis and design of a pitch-axis flight control system of a launch
vehicle as illustrated in Figure 3.2, an inertial reference frame (X, Y, Z) with its origin at
the vehicles center of gravity is assumed with its X-axis along the vertical axis and its Z-axis
along the horizontal direction. Body-fixed (x, y, z) axes with origin at the center of gravity
are also shown in Figure 3.2.
A set of simplified pitch-axis dynamical models with small angular motions can be found
as [10]
mV = (F D) mg

(3.1)

mZ = (F D) N + Tc

(3.2)

= M + M

(3.3)

= + + w = effective angle of attack

(3.4)

= Z/V
= flight-path (drift) angle

(3.5)

F = T0 + Tc = total thrust force

(3.6)

35

Inertial Reference

Wind Disturbance

Vw
w
Effective
Wind
Velocity

cp
N = N

z
D

x cp

cg

x cg

mg
To
Tc

Figure 3.2

A simplified dynamic model of a launch vehicle for preliminary

pitch control design. All angles are assumed to be small.

where m is the vehicle mass, V is the vehicle velocity, g is the local gravitational acceleration,
T0 is the ungimballed sustainer thrust, Tc is the gimbaled control thrust, D is the aerodynamic
axial (drag) force, Z is the inertial Z-axis drift position of the center-of-mass, Z is the inertial
drift velocity, N = N is the aerodynamic normal (lift) force acting on the center-of-pressure,
is the gimbal deflection angle, is the small pitch attitude from a vertical inertial reference
axis X, w = Vw /V is the wind-induced angle of attack, Vw is the wind disturbance velocity.
We also have
M = xcp N /Iy

(3.7)

M = xcg Tc /Iy

(3.8)

36
1
N = V 2 SCN
2

(3.9)

where Iy is the pitch moment of inertia. For effective thrust vector control of a launch vehicle,
we need
M max > M max

(3.10)

where max is the gimbal angle constraint and max is the maximum wind-induced angle of
attack.
Combining Equations. (3.2), (3.3), and (3.4), we obtain a

0
1
0

d
=

M
0
M
/V

dt

Z
Z
(F D + N )/m 0 N /(mV )

state-space model of the form

0
0

w
+ M + M

N /m
Tc /m
(3.11)

and = + Z/V
+ w . Assuming all constant coefficients in the state-space model, we obtain
the open-loop transfer functions from the control input (s) as
1
(s)
=
(s)
(s)


M

N
s+
mV

M T c
+
mV


(3.12)




Z(s)
1
Tc 2
M (F D + N )
=
(s M )
(s)
(s) m
m
(s)
1
=
(s)
(s)

Tc 2
M (F D)
s + M s
mV
mV

(3.13)


(3.14)

where
(s) = s3 +

N 2
M (F D)
s M s +
mV
mV

(3.15)

In 1959, Hoelkner [37] introduced the drift-minimum and load-minimum control concepts as applied to a launch vehicle flight control system. The concepts have been further
investigated in [10, 12, 13, 31, 38, 39, 40, 41, 42]. Basically, Hoelkners controller utilizes a
full-state feedback control of the form

= K1 K2 K3 where = + Z/V
+ w

(3.16)

37
The feedback gains are to be properly selected to minimize either lateral drift velocity Z
or the bending moment caused by the angle of attack.
Substituting Equation. (3.16) into Equations. (3.2)-(3.3), we obtain the closed-loop transfer

function from the wind disturbance w (s) to the drift velocity Z(s)
as

Z(s)
A2 s2 + A1 s + Ao
= 3
V w (s)
s + B2 s2 + B1 s + Bo

(3.17)

where
Tc
B2 = M K2 +
mV

N
K3 +
Tc

K2 Tc
B1 = M (K1 + K3 ) M +
mV
Tc K1
Bo =
mV

M N
M +
Tc
Tc
A2 =
mV

A1 =

K2 Tc
mV

M N
M +
Tc

(3.18)

F D
(M K3 M )
mV

N
K3 +
Tc


M +

(3.19)

(3.20)

M N
Tc

(3.21)

Ao = Bo

(3.22)

(3.23)

For a step wind disturbance with a magnitude of Vw , the steady-state value of Z can be
found as
(A2 s2 + A1 s + Ao )
Ao
Z ss
= lim 3
=
= 1
2
s0 s + B2 s + B1 s + Bo
Vw
Bo

(3.24)

The launch vehicle drifts along the wind direction with Z ss = Vw and with = = =
= 0 as t . It is interesting to notice that the steady-state drift velocity (or the flight
path angle) is independent of feedback gains for an asymptotically stable closed-loop system
with Bo 6= 0.

38
If we choose the control gains such that Bo = 0 (i.e., one of the closed-loop system roots is
placed at s = 0), the steady-state value of Z becomes
Z ss
(A2 s + A1 )
A1
1
= lim 2
=
=
s0 s + B2 s + B1
Vw
B1
1+C

(3.25)

where
C=

mV [M (K1 + K3 ) M ]
M K2 Tc + M N /Tc

(3.26)

For a stable closed-loop system with M (K1 + K3 ) M > 0, we have C > 1 and
|Z ss | < Vw

(3.27)

when Bo = 0. The drift-minimum condition, Bo = 0, can be rewritten as

M K 3 M
N
=
M K 1
F D

xcp
1+
xcg


(3.28)

Consider the following closed-loop transfer functions:

(s)
s(s2 + M K2 s + M K1 )
= 3
w (s)
s + B2 s2 + B1 s + Bo

(3.29)

s(K3 s2 + M K2 s + M K1 )

=
w
s3 + B2 s2 + B1 s + Bo

(3.30)

For a unit-step wind disturbance of w (s) = 1/s, we have = = 0 as t . However,

for a unit-ramp wind disturbance of w (s) = 1/s2 , we have
lim (t) = M K1

(3.31)

lim (t) = M K1

(3.32)

Consequently, the bending moment induced by and can be minimized by choosing

K1 = 0, which is the load-minimum condition introduced by Hoelkner [37]. The closed-loop
system with K1 = 0 is unstable because
Bo =

F D
(M K3 M ) < 0
mV

(3.33)

However, the load-minimum control for short durations has been known to be acceptable
provided a deviation from the nominal flight trajectory is permissible.

39
A set of full-state feedback control gains, (K1 , K2 , K3 ), can be found by using a poleplacement approach or the linear-quadratic-regulator (LQR) control method, as follows:
Z
min

(xT Qx + 2 )dt

(3.34)

subject to x = Ax + B and = Kx where x = [ ]T and K = [K1 K2 K3 ]. Some

simulation results and comparisons of those designs can be found in [43].

3.3

Pitch Control of a Rigid-Body Model of the Ares-I CLV

Table 3.1

Ares-I reference parameters at t = 60 sec

Parameters
Iy
m
Tc
V
CN
g
N
M
M
xcg
xcp

Values
2.186 108
38, 901
2.361 106
1347
0.1465
26.10
686, 819
0.3807
0.5726
53.19
121.2

Unit
slug-ft2
slug
lb
ft/s
ft/s2
lb/rad
s2
s2
ft
ft

(s)
0.5726(s + 0.04309)
=
(s)
(s + 0.6330)(s 0.01942)(s 0.6005)

(3.35)

40

PoleZero Map
1
0.8
0.6

Imaginary Axis

0.4
0.2
0
0.2
0.4
0.6
0.8
1
0.8

Figure 3.3

0.6

0.4

0.2

0
Real Axis

0.2

0.4

0.6

0.8

Poles and zeros of Ares-I CLV rigid-body model transfer function.

Root Locus

0.5
0.4

Imaginary Axis

0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
2

Figure 3.4

1.5

0.5
Real Axis

0.5

Root locus vs overall loop gain K of the pitch control system

of a rigid Ares-I model.

41

3.4

Flexible-Body Control of an Ares-I Reference Model

Now we consider the flexible-body control of the Ares-I CLV. Its flexible mode shapes and
sensor locations are shown in Figure 1.3. For the Ares-I having a high degree of axial symmetry,
there is negligible coupling between the pitch and yaw lateral modes. This statement can be
verified by inspecting Equation. (2.60) and Equation. (2.61). After neglecting relatively small
values in and , we obtain

0
0
0
0
0
0

=
0
0.006281
0
0.006259
0
0.007673

0.006281
0
0.006260
0
0.007676
0

(3.36)

0
0
0
0
0
0

=
0
0.2540
0
0.1305
0
0.1932

0
0.1932
0
0.2540
0
0.1305

103

(3.37)

The pitch and yaw lateral modes are decoupled in Equation. (2.56). The first three bending
modes data of the pitch and yaw lateral modes are given in Table 3.2 and Table 3.3 respectively.

Table 3.2

Ares-I structural bending modes for the pitch axis

Mode number
1
2
3

Table 3.3

i , rad/sec
6.0469
14.2206
27.1667

i
0.006281
-0.006260
0.007676

i
0.1932 103
0.2540 103
0.1305 103

Ares-I structural bending modes for the yaw axis

Mode number
1
2
3

i , rad/sec
6.0470
14.2213
27.1712

i
0.006281
0.006259
-0.007673

i
0.1932 103
0.2540 103
0.1305 103

Thus the over all Ares-I reference model can be divided into two parts (Figure 3.5), the

42
rigid-body part and the flexible-body part,
(s)
= Grigid (s) + Gf lex (s)
(s)

(3.38)

where Grigid (s) is expressed in Equation. (3.35).

Figure 3.5

Block diagram of attitude control loop with flexible-body dynamics.

The flexible-body part of the pitch transfer function model as shown in Figure 3.7 is
3

Gf lex (s) =

X
b (s) X
i i T
=
Gi (s) =
2
(s)
s + 2i + i2
i=1

(3.39)

i=1

where b is the additional angle due to the bending vibration measured by the IMU (Figure 3.6).
For the ith bending mode
Gi (s) =

s2

i i T
+ 2i + i2

(3.40)

The pitch transfer function model of the Ares-I CLV can be written as
(s)
0.9036(s + 0.041)(s + 3.68)(s 3.75)(s2 35s + 510)(s2 + 35s + 512)
=
(s)
(s + 0.63)(s 0.019)(s 0.60)(s2 + 0.06s + 36.56)(s2 + 0.14s + 202.2)(s2 + 0.27s + 738)
(3.41)
where is the pitch attitude error measured by the instrument unit (Figure 1.3) and is the
pitch gimbal angle.

43

Figure 3.6

Flexible structure in the pitch plane.

The poles and zeros of this pitch transfer function are illustrated in Figure 3.8. Such a
pole-zero pattern is typical for flexible vehicles with non-collocated actuator and sensor.
The root locus vs overall loop gain K of a simple PD control system is shown in Figure 3.9.
The instability of the first and third bending modes, caused by unstable interactions with the
rigid-body control, is evident from the root locus.

44

Figure 3.7

Block diagram of the flexible-body part of the pitch transfer

function.

PoleZero Map
30

Imaginary Axis

20
10
0
10
20
30
20

Figure 3.8

15

10

0
Real Axis

10

15

20

Pitch transfer function model of a reference model of the Ares-I

CLV.

45

Root Locus

30

Imaginary Axis

25

20

15

10

0
15

Figure 3.9

10

5
10
Real Axis

15

20

25

30

Root locus of the pitch control system without structural filters.

46

3.5

NMP Structural Filter Design

The root locus, shown in Figure 3.9, clearly indicates that those two unstably interacting
bending modes can be effectively stabilized by using two NMP filters. Detailed discussions of
the classical gain-phase stabilization approach using NMP filters can be found in [16, 20, 21]. It
is important to notice that nowadays, one can easily perform an inherently iterative, classical
control design using the interactive root locus tool of the SISO Design Toolbox of Matlab.
After several design iterations, the structural filter for the first bending mode is found as
F1 (s) =

1.0036(s2 4.295s + 48.76)

s2 + 11.9s + 48.94

(3.42)

and the structural filter for the third bending mode is

F3 (s) =

0.91123(s2 19.48s + 708.7)

s2 + 17.93s + 645.8

(3.43)

The impulse responses provided in Figures 3.11 and 3.12 also indicate that the bending
modes are more actively controlled by using such NMP filters although standard roll-off and
notch filters [12, 13, 14] can also be employed to stabilize such unstably interacting bending
modes. This design has met the standard rigid-body stability margin requirements (6 dB
gain margin and 30 deg phase margin). Nonlinear coupled dynamic simulation results of
validating the stability of the NMP filters as well as the baseline attitude-error quaternion
feedback control scheme for the Ares-I are the similar to Figures 2.19-2.23.

47

Root Locus
30

25

Imaginary Axis

20

15

10

0
20

Figure 3.10

15

10

0
Real Axis

10

15

20

Root locus of the pitch control system with two NMP structural filters.

Impulse Response
8

Amplitude

Figure 3.11

4
5
Time (sec)

Impulse responses for the pitch attitude (in degrees) of the

flexible Ares-I.

48

Impulse Response
1.2
1
0.8

Amplitude

0.6
0.4
0.2
0
0.2
0.4
0.6

10

15

20

25

Time (sec)

Figure 3.12

Impulse responses for the pitch gimbal angle (in degrees) of

the flexible Ares-I.

49

3.6

Robust Analysis for Structural Filters Design

In this section, the robustness of the ascent flight control system will be analyzed based on
the structured singular value . A general configuration which includes all the different systems
resulting from variations in the form and location of the controller and the system uncertainty
is illustrated in Figure 3.13. P is called the generalized plant and includes the nominal plant
together with dynamics associated with the weighting transfer functions used to model system
uncertainty. K is the generalized controller. In this dissertation, it is a PD-type baseline
controller with NMP structural filters designed in the last section. is an unknown but
norm-bounded uncertainty. The signal w is called the exogenous input. Typically, it includes
external disturbances, measurement noise and command signals. The signal z is exogenous
output. Usually, it is the error signal to be minimized. The signal u is called the actuator
input to P . For a launch vehicle, u could be gimbal angle command. The signal y is controller
input and is composed of all the measured plant output which are available for feedback.
For the problem of analyzing the robustness of a given controller K, we can rearrange the
system into the M- structure of Figure 3.14 where M is the transfer function from the output
to the input of the perturbations. We need to determine whether the system remains stable for
all the uncertainty set. The basic conceptions and framework of linear robust control theory,
especially the -condition for robust stability, can be found in [44, 45, 46]. MATLAB Robust
Control Toolbox (-Analysis) is used in this section.

3.6.1

Uncertainty Description of Rigid-Body Model

Typical rigid-body parametric uncertainties for a launch vehicle can be found in Table 3.4.
The parametric uncertainties are quantified by a ratio. Although the sources of uncertainties
are known, it is hard to represent perturbed plants by a structured set with a finite number
of scalar parameters. For this case, dynamic (frequency-dependent) uncertainty is particularly
well-suited. This leads to a normalized complex perturbations kk 1.
Let the set of possible plants be
Gp (s) = Grigid (s)(1 + WI (s)I (s)); |I (j)| 1

(3.44)

50

Figure 3.13

Table 3.4
Parameters
Iy
Tc
CN
xcg
xcp

General control configuration.

Ares-I rigid-body parametric uncertainty

Nominal values (Unit)
2.186 108 (slug-ft2 )
2.361 106 (lb)
0.1465
53.19 (ft)
121.2 (ft)

Relative uncertainties
5%
5%
6%
10%
10%

where Gp (s) is a perturbed plant model, which may be represented by the block diagram in
Figure 3.15. I (s) is any stable transfer function which at each frequency is less than or equal
in magnitude to 1. The subscript I denotes input, but for SISO systems it is not important
that whether the perturbation is considered at the input or output of the plant.
The multiplicative weight are calculated by the Robust Toolbox for 50 samples of the
rigid-body model with parametric uncertainty,
WI (s) =

0.11745(s + 9.741)(s + 0.5113)(s + 0.01471)(s2 + 2.52s + 3.645)

(s + 9.463)(s + 0.2936)(s + 0.04434)(s2 + 2.837s + 3.447)

(3.45)

As seen from the blue solid line and red dashed line in Figure 3.16, the perturbed plant
model Gp (s) can cover a range of 50 samples of the nominal plant Grigid (s) with parametric

51

Figure 3.14

M- structure for robust stability analysis.

Figure 3.15

Plant with multiplcative uncertainty.

uncertainties in Table 3.4. Furthermore, Figure 3.17 shows that the upper and lower bounds
of Gp (s) cover the gain uncertainty of those samples.
3.6.2

Uncertainty Description of Flexible-Body Model

Another significant uncertainty source is the structural flexibility of the launch vehicle.
Based on [47], vibration frequencies should be accurate to within 5% for the first bending
mode and 10% for the second through the fourth or fifth bending modes. One important
requirement of the Ares-I ascent FCS is that the resulting control system is stable with 10%

52

Bode Diagram

Sample
Uncertainty Model

20

Magnitude (dB)

0
20
40
60
80
100
135

Phase (deg)

180
225
270
315
360
4

10

Figure 3.16

10

10

10
Frequency (rad/sec)

10

10

10

Bode plot of parameter uncertainty plant and perturbed plant

samples.

natural frequency uncertainty of first three bending modes throughout the first stage flight.
Uncertainty modeling of the flexible structure is critical to evaluate robustness of a controller design. Usually, norm-bounded additive or multiplicative perturbations of a nominal
model in the frequency domain are used to account for uncertainty in the model frequencies,
damping ratios and mode shape matrix of the model [48, 49, 50]. Such approaches to uncertainty modeling in large flexible launch vehicles do not handle natural frequencies shift very
well. Slight variation in the mode frequencies usually causes the associated dynamic perturbations to be large in the -norm sense. This will make the uncertainty model too conservative
for robust stability analysis. In this section, structured uncertainty is adopted to model real
parameter uncertainty.
The nominal values of flex frequency are given in Table 3.2. The uncertainty model of flex
frequency can be rewritten as
i = i (1 + r i )

(3.46)

where i is the nominal value of ith bending mode natural frequency, r = 10% is the relative

53

Bode Diagram
20
Sample
Lower bound
Upper bound
0

Magnitude (dB)

20

40

60

80

100
4
10

10

10

10
Frequency (rad/sec)

Figure 3.17

10

10

10

Bode plot magnitude.

uncertainty in the parameter, and i is any real scalar satisfying |i | 1.

For the ith bending mode, the perturbed transfer function can be written as
Gpi (s) =

i i T
i i T
= 2
s2 + 2i + i2
s + 2 i (1 + r i ) + 2i (1 + r i )2

(3.47)

where |i | 1 and thus |2i | |i |. We could use a larger uncertainty i to replace |2i |.
After replacing 2i by i in the denominator of Gpi (s), we get
Gpi =

s2

+ 2 i +

2i

i i T
+ i [2 i r s + i (2r + r2 )]

(3.48)

Define the weight function for the ith bending mode as

2 i r
i (2r + r2 )
s+
i i T
i i T

(3.49)

Gi (s)
; |i | 1
1 + Gi (s)Wi (s)i

(3.50)

Wi (s) =
then we obtain
Gpi (s) =

This may be represented by the block diagram in Figure 3.18.

54

Figure 3.18

Block diagram of perturbed transfer function Gpi (s).

The weight functions can be found as

W1 (s) = 0.002111s 2.681

(3.51a)

W2 (s) = 0.003789s + 11.31

(3.51b)

W3 (s) = 0.01149s 65.53

(3.51c)

As seen in Figure 3.19, perturbed models could represent frequency uncertainty of Gf lex (s)
very well. It covers the whole range of the frequency shift.

3.6.3

Robust Stability Analysis

The overall structure of the uncertainty model of Ares-I is shown in Figure 3.20.
The structure of uncertainty perturbation is written as a block-diagonal matrix.

I 0
0
0

0 1 0
0

= diag{I , 1 , 2 , 3 } =

0 2 0
0

0
0
0 3

(3.52)

The M- system in Figure 3.14 is stable for all allowed perturbations with () 1, ,
if and only if
(M (j)) 1,

(3.53)

55

Bode Diagram
40
Sample
Left bound
Right bound

Magnitude (dB)

20
0
20
40
60
80

Phase (deg)

540

360

180

0
0

10

Figure 3.19

10
Frequency (rad/sec)

10

Bode plot samples of Gf lex (s) with frequencies uncertainty and

P
the boundary of perturbed models 3i=1 Gpi (s).

(M ) can be calculated by the MATLAB Robust Control Toolbox. Figure 3.21 clearly
shows that the upper bound of (M ) is smaller than 1. The maximum value of (M )
is 0.8858. It means that the flight control system is stable with respect to the rigid-body
parameter uncertainties in Table 3.4 and 10% bending mode frequency uncertainties of all
three modes.

56

Figure 3.20

Block diagram of perturbed attitude control system.

57

0.9
upper bound
lower bound

0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
3
10

10

Figure 3.21

10

10
10
Frequency (rad/sec)

10

-plot for RS of structural filters design.

10

58

CHAPTER 4.

UNCONTROLLED ROLL DRIFT

4.1

Introduction

The roll motion of Ares-I CLV under nominal flight conditions is actively stabilized by
its RCS equipped with thrusters. However, in this chapter, we examine the feasibility of
maintaining the pitch/yaw attitude stability as well as the ascent flight performance of Ares-I
CLV during its ascent phase but in the event of disabled or failed roll control. This situation
can occur when the roll-axis disturbance torque unexpectedly exceeds the control authority of
the RCS of a slender launch vehicle.
A simple pitch/yaw control logic will be proposed for such a technically challenging problem by exploiting the inherent versatility of a quaternion-based attitude control system. The
proposed pitch/yaw control logic only requires the desired inertial attitude quaternion to be recomputed using the actual uncontrolled roll angle information. This simple approach achieves
an ascent flight trajectory identical to the nominal flight case with active roll control. Another approach that utilizes a simple adjustment of the proportional-derivative gains of the
quaternion-based flight control system without active roll control is also presented in this chapter. This approach doesnt require the re-computation of desired inertial attitude quaternion.
Linear stability criterion is developed for proper adjustments of attitude and rate gains. The
linear stability analysis results are validated by nonlinear simulations of the ascent flight phase.
However, the first approach, requiring a simple modification of the desired attitude quaternion,
is recommended for the Ares-I as well as other launch vehicles in the event of no active roll
control.

59

4.2

Pitch/Yaw Closed-Loop Instability Caused by Uncontrolled Roll Drift

Simulation results of a reference Ares-I CLV with a baseline ascent flight control system
but in the event of uncontrolled roll drift are provided in Figures 4.1, 4.2, and 4.3. An M-file
based nonlinear 6-DOF simulation program is used for simulation of this nominal case. As can
be seen from these figures, the pitch/yaw flight control system becomes unstable slightly after
t = 60 sec although it maintains closed-loop stability during the early ascent phase.
Actual output
Command

q1

1
0
1

20

40

60

80

100

120

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

q2

1
0
1

q3

1
0
1

Figure 4.1

Attitude quaternion for an unstable closed-loop system caused

by uncontrolled roll drift.

60

Actual output
Command

1 (deg)

200
0
200

20

40

60

80

100

120

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

2 (deg)

100
0
100

3 (deg)

200
0
200

Figure 4.2

Euler angles for an unstable closed-loop system caused by uncontrolled roll drift.

10

y (deg)

5
0
5
10

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

10

z (deg)

5
0
5
10

Figure 4.3

Gimbal angles for an unstable closed-loop system caused by

uncontrolled roll drift.

61

4.3

Stability Analysis

In this section we briefly describe the rotational equations of motion of the Ares-I CLV
for its ascent flight control analysis and design. Details of the six-degrees-of-freedom (6-DOF)
equations of motion of the Ares-I CLV can be found in Chapter 2. Also, detailed discussions
of ascent flight control analysis and design for the Ares-I under nominal flight conditions can
be found Chapter 2 and Chapter 3.
The angular velocity vector
~ of the vehicle is expressed as

~ = p~i + q~k + r~k

Rotational equations of

Ixx Ixy Ixz

xy Iyy Iyz

Ixz Iyz Izz

motion of a reference Ares-I CLV are described by

p
0 r q Ixx Ixy Ixz p

q
0 p
= r
Ixy Iyy Iyz q

r
q p
0
Ixz Iyz Izz
r

Taero.x Trkt.x Trcs

+ T
+ 0
+
T
aero.y rkt.y

Taero.z
Trkt.z
0

(4.1)

(4.2)

The rocket thrust is simply modeled as Equation. (2.41) and Equation. (2.42). The bodyaxis components of the rocket thrust-generated torque are expressed by Equation. (2.44). The
commanded quaternion (q1c , q2c , q3c , q4c ) from an ascent guidance system, are used to generate
the attitude-error quaternion, are expressed by Equation. (2.46)
The classical proportional and derivative (PD) control laws of the ascent flight control
system utilizing the quaternion-error feedback concept are described by
Trcs = 2Kpx q1e Kdx p

(4.3a)

y = 2Kpy q2e Kdy q

(4.3b)

z = 2Kpz q3e Kdz r

(4.3c)

A simplified block diagram representation of the quaternion-based ascent flight control system of Ares-I CLV is provided in Figure 4.4. Detailed discussions of the advantages of the

62

Table 4.1

Reference Ares-I CLV parameters at t = 60 sec.

Parameters
Ixx
Iyy
Izz
T
`
Kpy
Kpz
Kdy
Kdz

Initial values
9.2356105
2.1860108
2.1860108
2.3608106
53
1.3484
1.3484
1.5023
1.5023

Units
slug-ft2
slug-ft2
slug-ft2
lb
ft
rad
rad
rad-sec
rad-sec

quaternion-feedback control system, especially with its large-angle control capability, can be
found in [16].

Figure 4.4

A simplified block diagram representation of the quaternion

based ascent flight control system of Ares-I CLV.

A summary of the basic parameters of a reference Ares-I CLV is provided in Table 4.1.

4.3.1

Simplified Nonlinear Closed-Loop Pitch/Yaw Dynamics

Assuming uncontrolled, but slow, roll motion and controlled, fast pitch/yaw attitude dynamics of the Ares-I CLV, we consider the pitch/yaw attitude dynamics simply described by

63

T`
y =
Iyy
T`
r
z =
Izz

T`
(2Kpy q2e Kdy q)
Iyy
T`
(2Kpz q3e Kdz r)
Izz

(4.4a)
(4.4b)

where T is the total thrust force and ` = Xg + cx is the thrust force arm. Furthermore, we
also consider the quaternion-error differential equations given by

q1e

q2e

q3e

q4e

q p

q1e

1 r 0

p q

q2e
=

q p 0 r q3e

p q r 0
q4e

(4.5)

Derivation of Equation. (4.5) is provided in Appendix D.

4.3.2

Linear Stability Analysis

For the Ares-I with Iyy = Izz = I, Kdy = Kdz = Kd , Kpy = Kpz = Kp (see Table 4.1), and
with controlled pitch/yaw motions, we have

TI` Kd

r
0

q2e q4e /2

q3e
q1e /2

2 TI` Kp

TI` Kd

2 TI` Kp

q1e /2

q4e /2

q2e

q3e

(4.6)

Note that for controlled pitch/yaw motions, q2e and q3e are small. Furthermore, q1e and q4e
2 + q 2 1 for small
can be assumed to be slowly time varying if p is small. Also note that q1e
4e

q2e and q3e .

The characteristic equation of the linear system described by Equation. (4.6), which is
assumed to have constant coefficients, can be found as
s4 + (2KKd )s3 + (2q4e KKp + K 2 Kd2 )s2 + (2q4e K 2 Kp Kd )s + K 2 Kp2 = 0
where K = T `/I.

(4.7)

64

Table 4.2
s4
s3
s2
s1
s0

Routh arrays.
2q4e KKp + K 2 Kd2
2q4e K 2 Kp Kd
K 2 Kp2

1
2KKd
q4e KKp + K 2 Kd2

2K 3 Kd Kp2 +2q4e K 3 Kd Kp2 +2q4e K 4 Kd3 Kp

q4e KKp +K 2 Kd2
2
2
K Kp

K 2 Kp2
0

According to the Routh stability criterion [51] as illustrated in Table 4.2, if q4e KKp +
K 2 Kd2 > 0 and also if
2K 3 Kd Kp2 + 2q4e K 3 Kd Kp2 + 2q4e K 4 Kd3 Kp
>0
q4e KKp + K 2 Kd2

(4.8)

then we have a stable closed-loop system. Equation. (4.8) can be rewritten as

2K 3 Kd Kp
2
(Kp + q4e
Kp + q4e KKd2 ) > 0
q4e KKp + K 2 Kd2

(4.9)

By defining B = KKd2 /Kp , we obtain the linear stability criterion as

2
q4e
+ Bq4e 1 > 0

which becomes
q4e

B
>
+
2

q4e

B
<

or

(4.10)

B2 + 4
2

(4.11)

B2 + 4
2

(4.12)

This result confirms that the pitch/yaw closed-loop system can become unstable for small
q4e (i.e., for large error q1e ), which can occur without active roll control. Equation. (4.10)
shows that the critical parameter is B, which is determined by physical parameters of the
Ares-I CLV and its control gains. Assuming that q4e is positive, therefore the linear stability
criterion becomes
B>
The plot of function B =

2
1q4e
q4e

2
1 q4e
q4e

(4.13)

is provided in blue line (Figure 4.5). When the data point

(q4e , B) in the region above the blue line, the attitude control system is stable, otherwise it

65
is unstable. For example, at t = 60 sec, B = 0.9595, the corresponding value of q4e is 0.6294,
illustrated by the red dash line in Figure 4.5. In terms of Euler angle it is about 102 deg. It
means that when the roll error is smaller than 102 deg (q4e > 0.6294), the attitude control
system is stable. If the roll error is larger than 102 deg (q4e < 0.6294), the attitude control
system becomes unstable.

2
1.8
1.6

Stable

1.4

Unstable

1.2
1
0.8
0.6
0.4
0.2
0

0.2

0.4

0.6

0.8

q4e

Figure 4.5 Plot of the function B =

2
1q4e
q4e .

In order to illustrate how the specific values of q1e and q4e affect closed-loop system stability,
we consider three cases as described in Table 4.3. Figures 4.6 and 4.7 clearly show that closedloop stability is affected by a large value of q1e . As can be seen in Figure 4.8, the aerodynamic
disturbance makes the case worse. The linear model of uncontrolled roll drift with aerodynamic
disturbance can be found in Appendix E.
Table 4.3
Case numbers
1
2
3

Three cases for root locus stability analysis.

(1e , 2e , 3e ) deg
(30, 0, 0)
(80, 0, 0)
(80, 0, 0)

q1e
0.2588
0.6428
0.6428

q4e
0.9659
0.7660
0.7660

Aerodynamic disturbance
No
No
Yes

66

Root Locus
1
0.8
0.6

Imaginary Axis

0.4
0.2
0
0.2
0.4
0.6
0.8
1
3.5

2.5

1.5
Real Axis

0.5

Figure 4.6 Root locus plot for Case 1.

0.5

67

Root Locus
1.5

Imaginary Axis

0.5

0.5

1.5
3

2.5

1.5
1
Real Axis

0.5

0.5

0.5

Figure 4.7 Root locus plot for Case 2.

Root Locus
1
0.8
0.6

Imaginary Axis

0.4
0.2
0
0.2
0.4
0.6
0.8
1
2.5

Figure 4.8

1.5

1
0.5
Real Axis

Root locus plot for Case 3, showing closed-loop instability with

a nominal loop gain.

68
4.3.3

Nonlinear Stability Analysis

In general, when RCS fails, it becomes an underactuated control problem of an axisymmetric rigid body. Equation. (4.5) can help us to simplify the analysis of this problem by checking
the dynamics of the attitude-error quaternion, which can be described as

Iy Iz
Trcs
p Ix qr 0 Ix

q = Iz Ix rp + T ` + 0
Iy y

Iy

Ix Iy
T`
pq

0
r
Iz
Iz z

q1e

q2e

q3e

q4e

q p

q1e

q2e
1 r 0
p
q

2
q
p
0
r

q3e

p q r 0
q4e

(4.14)

Trcs = 0

(4.15)

(4.16a)

y = 2Kpy q2e Kdy q

(4.16b)

z = 2Kpz q3e Kdz r

(4.16c)

The attitude-error quaternion feedback control law always tries to drive the attitudeerror quaternion from any initial values to (0, 0, 0, 1). In order to simplify the notation,
(q1 , q2 , q3 , q4 ) will be used to replace (q1e , q2e , q3e , q4e ) in Equation.(4.15).
Since the body is axisymmetric, it is assumed Iy = Iz , and p = 0. The whole dynamical
system becomes an autonomous system, x = f (x), where x = (q, r, q1 , q2 , q3 , q4 )T

q
r
q1
q2
q3
q4

p q2 K
dq
K

K
p q3 K
dr


r
2 q2 2q q3
=

r q1 + q q4
2
2


qq + rq
2 1 2 4

2q q2 2r q3

(4.17)

69
p =
where K

T`
Iy Kp

d =
and K

T`
Iy Kd .

Note that the notation of attitude-error quaternion is

changed from (q1e , q2e , q3e , q4e ) to (q1 , q2 , q3 , q4 ), in order to keep expressions simple.
A Lyapunov function candidate can be taken as the energy-like function
V (x) =

1 2
1 2
q +
r + q12 + q22 + q32 + (1 q4 )2

2Kp
2K p

(4.18)

and V (x) = 0 when x = (0, 0, 0, 0, 0, 1)T , otherwise V (x) > 0.

Its derivative V (x) along any trajectory is
d
K
V (x) =
(q 2 + r2 ) 0

Kp

(4.19)

M = {(q, r, q1 , q2 , q3 , q4 ) : q = r = 0, q2 = q3 = 0}

(4.20)

which is negative semidefinite.

Define a set M ,

The set M is a positive invariant set, since

x(0) M x(t) M, t 0

(4.21)

By LaSalles theorem (Invariance Principle) [52], all trajectories approach M as t .

In order to visualize the attitude quaternion a new variable z is defined as z 2 = q22 + q32 =
1q12 q42 and z is a nonnegative real number. Therefore the time history of attitude quaternion
is a trajectory on the spherical surface q12 + q42 + z 2 = 1 (Figure 4.9) or on the surface of the
p
cone z = q22 + q32 . The set M is the circle in the q1 q4 plane (Figure 4.10).
Simulation results are given below to verify both the linear and nonlinear stability analysis.
The initial value of the simulation case can be found in Table 4.4. Note that by linear stability
analysis, if q4 < 0.6294 (q1 > 0.7771), the linear system is unstable. The stable and unstable
regions are shown in Figure 4.11.
For an axisymmetric rigid-body, if the uncontrolled principal axis is an axis of symmetry
and the other two axes are controlled by an attitude quaternion feedback law, the dynamical
system, Equation. (4.17), will converge to an invariant set or a subsystem, Equations. (4.20),
globally. On the other hand, by the Hartman-Grobman theorem [53], the local behavior near

70

1
0.8

0.6
0.4
0.2
0

0.5
0

1
0.5

0.5

0
0.5

q4

Figure 4.9

Table 4.4

Trajectory on the spherical surface q12 + q42 + z 2 = 1.

Three simulation cases for nonlinear stability analysis.

Case numbers
1
2
3

Initial values of quaternion (q1 , q2 , q3 , q4 )

(0, 0.01, 0.01, 0.9999)
(0.5, 0.01, 0.01, 0.8659)
(0.85, 0.01, 0.01, 0.5266)

the invariant set M is governed by the linear system Equation. (4.6). This is the reason why an
oscillation phenomenon can be observed when q1e is very large from Figure 4.20 to Figure 4.24.

71

0.5

0
1
1

0.5
0.5

0
0
0.5

0.5

Figure 4.10

The positive invariant set M .

Stable region
Unstable region

1
0.8
0.6
0.4

q4

0.2
0
0.2
0.4
0.6
0.8
1
1

Figure 4.11

0.5

0
q1

0.5

The stable and unstable regions in M .

72

p (rad/sec)

1
0
1

10

15

10

15

10

15

q (rad/sec)

0.01
0
0.01

r (rad/sec)

0.01
0
0.01

Time (sec)

Figure 4.12

Angular velocity for Case 1.

q1

1
0
1

10

15

10

15

10

15

q2

0.01
0
0.01

q3

0.01
0
0.01

Time (sec)

Figure 4.13

Attitude quaternion for Case 1.

73

10

x 10

q3

2
2

Figure 4.14

4
q2

10
3

x 10

Phase portrait of q2 and q3 for Case 1.

0.01

y (deg)

0
0.01
0.02
0.03

10

15

10

15

0.01

z (deg)

0
0.01
0.02
0.03

Time (sec)

Figure 4.15

Control inputs for Case 1.

74

p (rad/sec)

1
0
1

10

15

20

25

30

35

40

10

15

20

25

30

35

40

10

15

20
Time (sec)

25

30

35

40

q (rad/sec)

0.01
0
0.01
0.02

r (rad/sec)

0.01
0
0.01

Figure 4.16

Angular velocity for Case 2.

q1

0.5
0.4999
0.4998

10

15

20

25

30

35

40

10

15

20

25

30

35

40

10

15

20
Time (sec)

25

30

35

40

q2

0.01
0
0.01

q3

0.01
0
0.01

Figure 4.17

Attitude quaternion for Case 2.

75

10

x 10

q3

5
4

q2

Figure 4.18

10
3

x 10

Phase portrait of q2 and q3 for Case 2.

y (deg)

0.02

0.02

0.04

10

15

20

25

30

35

40

10

15

20
Time (sec)

25

30

35

40

z (deg)

0.02

0.01

0.01

Figure 4.19

Control inputs for Case 2.

76

p (rad/sec)

1
0
1

20

40

60

80

100

120

140

160

20

40

60

80

100

120

140

160

20

40

60

80
100
Time (sec)

120

140

160

q (rad/sec)

0.2
0
0.2

r (rad/sec)

0.2
0
0.2

Figure 4.20

Angular velocity for Case 3.

q1

0.9
0.8
0.7
0

20

40

60

80

100

120

140

160

20

40

60

80

100

120

140

160

20

40

60

80
100
Time (sec)

120

140

160

q2

0.1
0
0.1

q3

0.1
0
0.1

Figure 4.21

Attitude quaternion for Case 3.

77

0.1
0.08
0.06

Start Point

0.04

q3

0.02
0
0.02
0.04
0.06
0.08
0.1
0.1

0.05

Figure 4.22

q3

0
q2

0.05

0.1

Phase portrait of q2 and q3 for Case 3.

End Point

0.1

Start Point

0.05

q1

q2
0.05
0.1
0.05
0.1
0.6

0
0.7

0.8

Figure 4.23

0.05
0.9

0.1

q1 -q2 -q3 plot for Case 3.

78

0.4

y (deg)

0.2
0
0.2
0.4

20

40

60

20

40

60

80

100

120

140

160

80
100
Time (sec)

120

140

160

0.4

z (deg)

0.2
0
0.2
0.4

Figure 4.24

Control inputs for Case 3.

0.02
0.015
0.01
0.005

q2

0.005

0.01
0

Figure 4.25

0.01
0
0.2

0.4

0.01
0.6

0.8

0.02

q2 -q3 plot, from left to right q1 = 0,q1 = 0.2,q1 = 0.4,q1 = 0.6.

79

4.4

New Pitch/Yaw Control Logic with Modified Commanded

Quaternions

In this section, we present an approach for maintaining the pitch/yaw closed-loop stability
even in the event of uncontrolled roll drift. This approach utilizes the inherent versatility
of the quaternion-based attitude control system [16]. This approach simply requires an onboard computation of the desired attitude quaternion using the actual uncontrolled roll angle
information as illustrated in Figure 4.26.

Figure 4.26

A block diagram representation of a proposed method for computing a new set of commanded attitude quaternion.

The attitude quaternion (q1 , q2 , q3 , q4 ) are computed by numerically integrating the

kinematic differential equations Equation. (2.12).
The attitude quaternion (q1 , q2 , q3 , q4 ) for a rotational sequence of C1 (1 ) C2 (2 )
C3 (3 ) are related to Euler angles as follows [16]:
q1 = sin(1 /2) cos(2 /2) cos(3 /2) cos(1 /2) sin(2 /2) sin(3 /2)
q2 = cos(1 /2) sin(2 /2) cos(3 /2) + sin(1 /2) cos(2 /2) sin(3 /2)
q3 = cos(1 /2) cos(2 /2) sin(3 /2) sin(1 /2) sin(2 /2) cos(3 /2)
q4 = cos(1 /2) cos(2 /2) cos(3 /2) + sin(1 /2) sin(2 /2) sin(3 /2)

(4.22)

80
Also, we have the following direction cosine matrix relationship for the rotational sequence
of C1 (1 ) C2 (2 ) C3 (3 )

cos 2 cos 3
cos 2 sin 3
sin 2

sin sin cos cos sin sin sin sin + cos cos sin cos
1
2
3
1
3
1
2
3
3
1

cos 1 sin 2 cos 3 + sin 1 sin 3 cos 1 sin 2 sin 3 sin cos 3 cos 1 cos

2
2
C11 C12 C13 1 2(q2 + q3 ) 2(q1 q2 + q3 q4 ) 2(q1 q3 q2 q4 )

= 2(q q q q ) 1 2(q 2 + q 2 ) 2(q q + q q )

=
C
C
C
1 2
3 4
2 3
1 4
22
23

21
1
3

2
2
2(q1 q3 + q2 q4 ) 2(q2 q3 q1 q4 ) 1 2(q1 + q2 )
C31 C32 C33

(4.23)

For this particular rotational sequence of Euler angles, the Euler angles (2 , 3 ) describe
the inertial orientation of the longitudinal axis of the Ares-I CLV.
The actual Euler angle 1 of the Ares-I CLV can then be determined from the attitude
quaternion (q1 , q2 , q3 , q4 ) as follows:
1 = sgn{C23 } cos

C33
cos 2


= sgn{2(q2 q3 + q1 q4 )} cos

1 2(q12 + q22 )

1 4(q1 q3 q2 q4 )2

)
(4.24)

when |2 | =
6 /2.
Similarly, the commanded angles (2c , 3c ) of the Ares-I CLV can be determined from the
desired attitude quaternion (q1c , q2c , q3c , q4c ) commanded from an ascent guidance system as
follows:

3c

2c = sin1 {2(q1c q3c q2c q4c )}

(
)
2 + q2 )
1 2(q2c
1
3c
p
= sgn{2(q1c q2c + q3c q4c )} cos
1 4(q1c q3c q2c q4c )2

(4.25)
(4.26)

By using the actual Euler angle 1 and the commanded angles (2c , 3c ), we can obtain a
modified set of desired attitude quaternion as follows:
q1c = sin(1 /2) cos(2c /2) cos(3c /2) cos(1 /2) sin(2c /2) sin(3c /2)
q2c = cos(1 /2) sin(2c /2) cos(3c /2) + sin(1 /2) cos(2c /2) sin(3c /2)
q3c = cos(1 /2) cos(2c /2) sin(3c /2) sin(1 /2) sin(2c /2) cos(3c /2)
q4c = cos(1 /2) cos(2c /2) cos(3c /2) + sin(1 /2) sin(2c /2) sin(3c /2)

(4.27)

81

Original command
Actual output

1c (deg)

200
100
0
100

20

40

60

80

100

120

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

2c (deg)

20
40
60

3c (deg)

0
50
100

Figure 4.27

Comparison of new and original attitude Euler angles command.

These new commanded quaternion (

q1c , q2c , q3c , q4c ) are then used to determine the
attitude-error quaternion (q1e , q2e , q3e , q4e ) as follows:

q
1e

q2e

q3e

q4e

q
q3c
q2c
q1c
4c


q1c
q2c
q3c q4c
=

q1c q4c
q3c
q2c

q1c
q2c
q3c
q4c

q
1

q2

q3

q4

(4.28)

The pitch and yaw gimbal control laws are simply the same as the original ascent flight control
laws described by
y = 2Kpy q2e Kdy q

(4.29a)

z = 2Kpz q3e Kdz r

(4.29b)

No adjustment of the control gains of the original ascent flight control system is required for
controlling the pitch and yaw motions without active roll control.

82

Original command
New command

q1c

1
0.5
0
0.5

20

40

60

80

100

120

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

q2c

0
0.5
1

q3c

0.5
0
0.5

Figure 4.28

Comparison of new and original attitude quaternion command.

Stable closed-loop responses of the proposed approach with a modified set of desired quaternion can be seen from Figure 4.29 to Figure 4.32 in the event of uncontrolled roll drift. Furthermore, as can be seen in Figure 4.32, the proposed control approach achieves an identical
ascent flight trajectory as the nominal ascent flight control system with active roll control.

83

Actual output
Command

q1

0.5
0
0.5

20

40

60

80

100

120

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

q2

0
0.5
1

q3

0.5
0
0.5

Figure 4.29

Quaternion for a closed-loop system stabilized by the proposed

control logic employing modified commanded quaternion.
Actual output
Command

1 (deg)

100
0
100

20

40

60

80

100

120

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

2 (deg)

20
40
60

3 (deg)

0
50
100

Figure 4.30

Euler angles for a closed-loop system stabilized by the proposed control logic employing modified commanded quaternion.

84

y (deg)

20

40

60

80

100

120

20

40

60
Time (s)

80

100

120

z (deg)

Figure 4.31

Gimbal angles for a closed-loop system stabilized by the proposed control logic employing modified commanded quaternion.

x 10

with RCS
without RCS

12

Z Position (ft)

10
8
6
4
2
0
10
4

x 10

0
Y Position (ft)

Figure 4.32

10
4

X Position (ft)

x 10

Comparison of ascent trajectories: with roll control system vs.

without roll control system.

85

4.5

Simple Adjustment of Control Gains

The approach presented in the preceding section provides the desired ascent flight performance despite the uncontrolled roll drift. However, it requires on-board computation of
(1 , 2c , 3c ) to generate modified command quaternion (
q1c , q2c , q3c , q4c ).
In this section, we examine the feasibility of achieving pitch/yaw closed-loop stability by
simply adjusting the PD control gains without such on-board computation of (1 , 2c , 3c ).
We introduce a new derivative gain as
d = Kd
K

(4.30)

where is a scale factor to be properly chosen and Kd is the original derivative gain selected
for the nominal flight conditions. And a simple PD control laws in the event of uncontrolled
roll drift are proposed as
dq
y = 2Kp sgn(q4e )q2e K

(4.31a)

dr
z = 2Kp sgn(q4e )q3e K

(4.31b)

As discussed in [16], the term sgn(q4e ) is necessary for a quaternion-based feedback control logic
for accommodating a short angular path. In particular, such a sign change term is necessary
to avoid an undesirable 360 deg flip-over of the CLV. Some detailed analysis of this approach
is presented in this section.
From Equation. (4.11), we notice that a larger value of B = KKd2 /Kp is necessary for
stability when q4e is small. The parameter K = T `/I is determined by the physical parameters
of the rocket. If Kp is decreased, then the overall loop gain of the control system is decreased,
which is not desirable. A simple way to increase the parameter B is to increase the derivative
gain Kd .
From Equation. (4.10), we have
B>

2
1 q4e
q4e

2
K 2 Kd2
1 q4e
>
Kp
q4e

(4.32)
(4.33)

86
or
1
>
Kd

2 )K
(1 q4e
p
q4e K

(4.34)

A root locus plot of case 3 with a new derivative gain Kd = 4Kd only in the pitch channel is
shown in Figure 4.33.
Root Locus
0.8

0.6

Imaginary Axis

0.4

0.2

0.2

0.4

0.6

0.8
2

Figure 4.33

4.5.1

1.5

0.5
Real Axis

0.5

Root locus plot for Case 3 but with a new derivative gain with
= 4 in the pitch channel.

Rigid Body 6-DOF Nonlinear Simulation Results

The Matlab-based program, employing a complete set of 6-DOF nonlinear models of Ares-I
was used to validate the linear stability analysis result. The simulation results are shown in
Figure 4.34-Figure 4.37. A dispersed, but stable, ascent trajectory can be seen in Figure 4.37
for the case with a simple gain adjustment but without active roll control.
Note that after changing the control gain, structural filters also need to be adjusted. The design methodology and design tool are mentioned in Chapter 3. A design example is given here.
Flexible-body 6-DOF Nonlinear Simulation Results are similar to Figure 4.34- Figure 4.37.

87

Command
Actual Output

q1

1
0
1

20

40

60

80

100

120

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

q2

1
0
1

q3

1
0
1

Figure 4.34

Attitude quaternion of 6-DOF nonlinear simulation with

Kd = 4Kd .

Actual output
Command

1 (deg)

200
100
0
100

20

40

60

80

100

120

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

2 (deg)

20
40
60

3 (deg)

0
50
100

Figure 4.35

Euler angles of 6-DOF nonlinear simulation with Kd = 4Kd .

88

y (deg)

1
0
1
2

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

z (deg)

2
1
0
1

Gimbal angles of 6-DOF nonlinear simulation with Kd = 4Kd .

Figure 4.36

x 10

with RCS
without RCS

12

10

Z Position (ft)

0
10
5

x 10

0
Y Position (ft)

10
4

x 10
X Position (ft)

Figure 4.37

Comparison of ascent trajectories.

89

CHAPTER 5.

UNDERACTUATED CONTROL PROBLEM OF AN

AXISYMMETRIC RIGID BODY

5.1

Introduction

The problem of ascent flight control in the event of uncontrolled roll drift can be generalized
as an underactuated control problem. Specifically, it is the problem of attitude stabilization
with less than three independent control torques. System equations for a rigid body rotation
can be written as follows:

p

q =


r

q
1

q2 1

2
q3

q4

Iy Iz
Ix qr
Iz Ix
Iy rp
Ix Iy
Iz pq

q p

u1
Ix
u2
Iy
u3
Iz

(5.1)

q
1

r 0
p q
q2

q p 0 r q3

p q r 0
q4

(5.2)

Note that the attitude kinematic differential equation can also be written in terms of Euler
angles as follows:

= 1
2 cos 2

cos 1 sin 2
cos 2 sin 1 sin 2

0
cos 1 cos 2 sin 1 cos 2

0
sin 1
cos 1

p


q


r

(5.3)

90
We have the following relationship for the rotational sequence of C1 (1 ) C2 (2 )
C3 (3 )

(
1 = sgn{2(q2 q3 + q1 q4 )} cos

1 2(q12 + q22 )

)
(5.4a)

p
1 4(q1 q3 q2 q4 )2

2 = sin1 {2(q1 q3 q2 q4 )}

(5.4b)
(

3 = sgn{2(q1 q2 + q3 q4 )} cos1

1 2(q22 + q32 )
p
1 4(q1 q3 q2 q4 )2

)
(5.4c)

The underactuated control problem has been dealt with [23, 24, 25, 26, 27]. Necessary and
sufficient conditions for controllability of the systems (Equation. (5.1) and Equation. (5.2) ) in
the case that the gas jet actuators yield one, two, or three independent torques are given in
[23]. Particularly, the problem of attitude stabilization of an axisymmetric (Iy = Iz ) spacecraft
using two pairs of gas jet actuators is considered in [25, 26]. A new kinematic formulation is
used to derive the feedback control law.
Without loss of generality, we consider the commanded quaternion to be (0, 0, 0, 1). Then
the attitude quaternion (q1 , q2 , q3 , q4 ) becomes the attitude error quaternion (q1e , q2e , q3e , q4e ).
Therefore
u2 = Kpy q2 Kdy q

(5.5a)

u3 = Kpz q3 Kdz r

(5.5b)

Since the rigid body is assumed to be axisymmetric (Iy = Iz ), we have Kpy = Kpz = Kp and
Kdy = Kdz = Kd .
Based on the analysis of Ares-I ascent flight control system in chapters 3 and 4, a modified
attitude quaternion feedback control law is derived in the next section. It can stabilize an
axisymmetric rigid body to the subsystem M (Equation. (4.20)). Moreover, a new kinematic
formulation is not needed.

91

5.2

Steady-State Oscillations

The equations of motion of an axisymmetric rigid body with a PD-type attitude quaternion
feedback control law can be expressed as

q1

q2

q
3

q4
where a =

IIx
I ,

p q2 K
dq
arp K

p q3 K
dr
apq K

q
p
r
q

q
+
q

2 2
2 3
2 4

p
q
r
2 q1 + 2 q3 + 2 q4

p
q
r
q

q
+
q

2 1
2 2
2 4

q
p
r
2 q1 2 q2 2 q3

(5.6)

p = Kp /I and K
d = Kd /I.
I = Iy = Iz , K

For this autonomous system x = f (x), where x = (q, r, q1 , q2 , q3 , q4 )T , there is one

steady-state oscillation which has the form
M1 = {(q, r, q1 , q2 ,

q1

q4

q3 q 4 ) : q = r

p
2 q4
=

p2 q1

= 0, q2 = q3 = 0}

(5.7)

(5.8)

In this oscillation, the rigid body is rotating around its symmetry axis with the angular velocity p. At the same time, the symmetry axis has the orientation by quaternion (q1 , 0, 0, q4 ),
or by Euler angles (1 , 0, 0). Note that it presents the orientation we want to achieve by a
feedback control law when there are only two independent control inputs u2 and u3 .
The autonomous system Equation. (5.6) may have another steady-state oscillation, when
p is not zero. It has the form
M2 = {(q, r, q1 , q2 , q3 q4 ) :

q 0
=
r
0

q1 = c1 , q4 = c4 }

q

r

(5.9)

(5.10)

q2 0 q2

q3
0
q3

(5.11)

92

q
cos sin q2

r
sin cos
q3

(5.12)

where c1 , c4 , , and are constants as follows:

c1 =

d (q 2 + r2 )
K
pp
K

(5.13a)

c4 =

[(2a 1)p2 q 2 r2 ](q 2 + r2 )

p p2
2K

(5.13b)

p2 + q 2 + r 2
2p

(5.13c)

=q

p
K
2

( (2a1)p 2p(q

2 +r 2 )

(5.13d)
d )2
)2 + (K

(2a 1)p2 q 2 r2
(5.13e)
d
2pK
p
p

The magnitude of vectors (q2 , q3 )T and (q, r)T are q22 + q32 = R and q 2 + r2 = R,
= tan1

= | |R. Note that R

is a positive real zero of a 4th-order polynomial
respectively, where R
2 ]p2 x2 +[(2a1)2 p2 +4K
2 ]p4 x4p4 K
2 (5.14)
f (x) = x4 +(34a)p2 x3 +[(2a1)(2a3)p2 +4K
p
d
d
p2 , f (x) as x and f (x) is continuous in x, f (x) has at
Since f (0) = 4p4 K
least one positive real zero. Thus M2 always exists when p is not zero. The derivation of the
steady-state oscillation M2 can be found in Appendix F.
A numerical example of those oscillations is provided below, using data for Ares-I CLV
at t = 60 sec (see Table 5.1). The constants in Equation. (5.13) are listed in Table 5.2 with
p = 0.005 rad/sec 0.2865 deg/sec.
Table 5.1

Ares-I reference parameters at t = 60 sec

Parameters
a
p

Kp
d
K

Values
0.9958
0.005
1.5441
0.8607

Unit
rad/s
s2
s1

The steady-state oscillations M1 and M2 can be visualized in Figures 5.1 and 5.2. M2 is
illustrated from Figure 5.1 to 5.9.

93

Table 5.2

Parameters of steady-state oscillation M2

Parameters
c1
c4

Values
0.7768
0.6242
0.6994
1.3963
36.90
0.05979
0.08348

Unit
rad
rad
rad/s
s1
deg
rad
rad/s

1
0.8

0.6
0.4

M1

0.2

M2

0
1
0.5

0
0.5

0.5
1
q

q1

Figure 5.1

Steady-state oscillations M1 and M2 on the spherical surface

q12 + q42 + z 2 = 1.

94

1
0.9
0.8
0.7

0.6
0.5
0.4

M2

0.3
0.2
0.1

M1

0.5

0.5

0
1

q3

Figure 5.2

q2

Steady-state
oscillations M1 and M2 on the surface of cone
p
2
2
z = q2 + q3 .

p (rad/sec)

0.01
0
0.01

10

20

30

40

50

10

20

30

40

50

10

20

30

40

50

q (rad/sec)

0.1
0
0.1

r (rad/sec)

0.1
0
0.1

Time (sec)

Figure 5.3

Angular velocity of steady-state oscillation M2 .

95

q1

0.7769
0.7769
0.7769

10

20

30

40

50

10

20

30

40

50

10

20

30

40

50

q2

0.1
0
0.1

q3

0.1
0
0.1

Time (sec)

Figure 5.4

Attitude quaternion of steady-state oscillation M2 .

(deg)

102.6
102.4
102.2
102

10

20

30

40

50

10

20

30

40

50

10

20

30

40

50

2 (deg)

10
0
10

3 (deg)

10
0
10

Time (sec)

Figure 5.5

Euler angles of steady-state oscillation M2 .

96

0.08
0.06
0.04

0.02
0
0.02
0.04
0.06
0.08
0.1

Figure 5.6

0.05

0
q

0.05

0.1

Phase portrait of q and r of steady-state oscillation M2 .

0.05
0.04
0.03
0.02

q3

0.01
0
0.01
0.02
0.03
0.04
0.05
0.06

Figure 5.7

0.04

0.02

0
q2

0.02

0.04

0.06

Phase portrait of q2 and q3 of steady-state oscillation M2 .

97

0.08

q and r
q2 and q3

0.06
0.04
0.02

R
y

0
0.02

||R
0.04
0.06
0.08
0.1

0.05

0
x

0.05

0.1

The relation between vectors (q, r)T and (q2 , q3 )T of steady-state oscillation M2 .

Figure 5.8

10

y (deg)

5
0
5
10

10

20

10

20

30

40

50

30

40

50

10

z (deg)

5
0
5
10

Time (sec)

Figure 5.9

Gimbal angles of steady-state oscillation M2 .

98

5.3

Modified Attitude Quaternion Feedback Control Law

In the last section, it was shown that there are at least two steady-state oscillations of the
autonomous system Equation. (5.6). Thus the original attitude quaternion feedback law can
not always drive Equations. (5.1) and (5.2) to M1 . A new feedback control law is needed.
Define a new state variable , such that = p. A modified attitude quaternion feedback
control law has the form
u2 = Kp [cos(/2)q2 + sin(/2)q3 ] Kd q

(5.15a)

u3 = Kp [ sin(/2)q2 + cos(/2)q3 ] Kd r

(5.15b)

where Kp and Kd are control gains designed to stabilize a rigid body by attitude quaternion
feedback law [16]. is used to catch the angular velocity p.
We may obtain a new autonomous system, x = f (x) as follows:

q arp Kp [cos(/2)q2 + sin(/2)q3 ] Kd q

p [ sin(/2)q2 + cos(/2)q3 ] K
dr
r apq K

q
p
r
q1
2 q2 2 q3 + 2 q4

q =
2r q1 + p2 q3 + 2q q4
2

p
q
r
q3
2 q1 2 q2 + 2 q4

q4
p2 q1 2q q2 2r q3

where I = Iy = Iz , a =

IIx
I ,

(5.16)

p = Kp /I and K
d = Kd /I.
(0) = 1 , K

5.4

Nonlinear Stability Analysis

A Lyapunov function candidate can be taken as the energy-like function

V (x) =

1 2
1 2
q +
r + [q1 sin(/2)]2 + q22 + q32 + [q4 cos(/2)]2
p
p
2K
2K

(5.17)

and V (x) = 0 when x = (0, 0, sin(/2), 0, 0, cos(/2))T , otherwise V (x) > 0.

The derivative of V (x) can be found as
d
K

V (x) =
(q 2 + r2 ) + [q1 cos(/2) q4 sin(/2)](p )
p
K

(5.18)

99
Since = p, we have
Kd 2
V (x) =
(q + r2 ) 0
Kp

(5.19)

which is negative semidefinite. Detailed derivation can be found in Appendix G.

The set M1 is a positive invariant set, since
x(0) M1 x(t) M1 , t 0

(5.20)

By LaSalles theorem (Invariance Principle) [52], all trajectories approach M1 as t .

In addition, we have
V (x) as kxk

(5.21)

The Lyapunov function V (x) is radially unbounded. Thus the modified quaternion feedback
control law Equation. (5.15) can reorient the symmetry axis to the desired direction from an
arbitrary initial orientation. Moreover, the new control law has no restriction on the spinning
rate p, which could be an arbitrary value. Even if p keeps changing due to some disturbance
torque, the new control law still works.

5.5

Simulation Results

Two simulation cases are performed to compare the effect of original and modified quaternion feedback laws. Assuming that the initial spinning rate is p = 0.005 rad/sec.
Table 5.3
Case numbers
1
2

Simulation cases with p = 0.005 rad/sec

Initial values (q, r, q1 , q2 , q3 , q4 )

(0, 0, 0.7769, 0.0598, 0, 0.6268, )
(0, 0, 0.7769, 0.0598, 0, 0.6268, )

Control law
Original attitude quaternion feedback
Modified attitude quaternion feedback

The effect of M2 can be seen from Figure. 5.10 to Figure. 5.18. The trajectory converges to
M2 in a oscillation behavior. Furthermore, the modified quaternion feedback control law eliminates this effect. The trajectory is driven to M1 , which represents the commanded orientation
of the axis of symmetry.

100

1
0.8

0.6
0.4
1
0.2

0.5

0
0.5

0.5
0.5

q1

q4

Figure 5.10

Trajectory on the spherical surface q12 + q42 + z 2 = 1 for Case

1.

Start Point

Figure 5.11

Detailed trajectory on the spherical surface q12 + q42 + z 2 = 1

for Case 1.

101

q (rad/sec)

0.2
0.1
0
0.1
0.2

50

100

150

200

250

300

50

100

150
Time (sec)

200

250

300

r (rad/sec)

0.2
0.1
0
0.1
0.2

Figure 5.12

Angular velocity components q and r for Case 1.

q1

0.8
0.75
0.7

50

100

150

200

250

300

50

100

150

200

250

300

50

100

150
Time (sec)

200

250

300

q2

0.1
0
0.1

q3

0.1
0
0.1

Figure 5.13

Quaternion q1 , q2 and q3 for Case 1.

102

1 (deg)

110
105
100
95

50

100

150

200

250

300

50

100

150

200

250

300

50

100

150
Time (sec)

200

250

300

2 (deg)

10
0
10

3 (deg)

10
0
10

Figure 5.14

Euler angles for Case 1.

10

y (deg)

5
0
5
10

50

100

150

200

250

300

50

100

150
Time (sec)

200

250

300

10

z (deg)

5
0
5
10

Figure 5.15

Gimbal angles for Case 1.

103

0.1
0.08
0.06
0.04

0.02
0
0.02
0.04
0.06
0.08
0.1
0.1

0.05

Figure 5.16

0
q

0.05

0.1

Phase portrait of q and r for Case 1.

0.08
0.06
0.04

q3

0.02
0
0.02
0.04
0.06
0.08
0.1

Figure 5.17

0.05

0
q2

0.05

Phase portrait of q2 and q3 for Case 1.

0.1

104

0.65
0.645
0.64
0.635

q4

0.63
0.625
0.62
0.615
0.61
0.605
0.6
0.75

0.76

0.77

0.78

0.79

0.8

0.81

q1

Figure 5.18

Phase portrait of q1 and q4 for Case 1.

1
0.8

0.6
0.4
0.2

0.5
0

0.5
0

0.5

0.5
1

q4

Figure 5.19

Trajectory on the spherical surface q12 + q42 + z 2 = 1 for Case

2.

105

Start Point

M1

Figure 5.20

Detailed trajectory on the spherical surface q12 + q42 + z 2 = 1

for Case 2.

0.01

q (rad/sec)

0
0.01
0.02
0.03
0.04

10

15

20

25

30

10

15
Time (sec)

20

25

30

r (rad/sec)

0.06
0.04
0.02
0
0.02

Figure 5.21

Angular velocity components q and r for Case 2.

106

q1

0.85
0.8
0.75

10

15

20

25

30

10

15

20

25

30

10

15
Time (sec)

20

25

30

q2

0.1
0
0.1

0
4

q3

x 10

0
2

Figure 5.22

Quaternion q1 , q2 and q3 for Case 2.

1 (deg)

115
110
1

105
100

10

15

20

25

30

10

15

20

25

30

10

15
Time (sec)

20

25

30

2 (deg)

5
0
5

3 (deg)

10
5
0
5

Figure 5.23

Euler angles for Case 2.

107

y (deg)

0
2
4
6

10

15

20

25

30

10

15
Time (sec)

20

25

30

z (deg)

10

Figure 5.24

Gimbal angles for Case 2.

108

5.6

A Special Case

When p = 0, = 1 (0) is a constant. The modified attitude quaternion feedback control

law is equivalent to a quaternion command adjustment.
For a typical attitude quaternion feedback law, we have

q
q
q3c q2c q1c
1e 4c

q2e q3c q4c
q1c q2c

q3e q2c q1c q4c q3c

q4e
q1c
q2c
q3c
q4c

q
1

q2

q3

q4

(5.22)

The command Euler angles are (1c , 2c , 3c ) = (1 , 0, 0), where

q1c = sin(1c /2) cos(2c /2) cos(3c /2) cos(1c /2) sin(2c /2) sin(3c /2)
q2c = cos(1c /2) sin(2 c/2) cos(3c /2) + sin(1c /2) cos(2c /2) sin(3c /2)

(5.23)

q3c = cos(1c /2) cos(2c /2) sin(3c /2) sin(1c /2) sin(2c /2) cos(3c /2)
q4c = cos(1c /2) cos(2c /2) cos(3c /2) + sin(1c /2) sin(2c /2) sin(3c /2)
The new command quaternion is
q1c = sin(1 /2)
q2c = 0

(5.24)

q3c = 0
q4c = cos(1 /2)
The attitude-error quaternion becomes

cos(1 /2)
0
0
sin(1 /2)
q1e

q2e
0
cos(1 /2) sin(1 /2)
0

0
sin(1 /2) cos(1 /2)
0
q3e

sin(1 /2)
0
0
cos(1 /2)
q4e

q1

q2

q3

q4

(5.25)

By the attitude quaternion feedback law, we have

u2 = Kp q2e Kd q = Kp [cos(1 /2)q2 + sin(1 /2)q3 ] Kd q

(5.26a)

u3 = Kp q3e Kd r = Kp [ sin(1 /2)q2 + cos(1 /2)q3 ] Kd r

(5.26b)

109
The quaternion command adjustment is a special case of the modified attitude quaternion
feedback control law, when p = 0.
q1e = sin(2 /2) sin(3 /2)
q2e = cos(1 ) sin(2 /2) cos(3 /2) + sin(1 ) cos(2 /2) sin(3 /2)

(5.27)

q3e = cos(1 ) cos(2 /2) sin(3 /2) sin(1 ) sin(2 /2) cos(3 /2)
q4e = cos(2 /2) cos(3 /2)
When 2 0, 3 0, we have
q1e 0
q2e cos(1 )2 /2 + sin(1 )3 /2
q3e sin(1 )2 /2 + cos(1 )3 /2
q4e 1

(5.28)

110

CHAPTER 6.

CONCLUSIONS

A set of dynamic models of the Ares-I Crew Launch Vehicle, incorporating its propulsion,
aerodynamics, guidance and control, and structural flexibility, has been described in this dissertation. The results of developing a Matlab-based simulation and linearization program by
utilizing NASAs SAVANT Simulink-based program have been discussed. The purpose of the
study was to develop an independent validation tool for the performance and stability analysis
of the ascent flight control system of the Ares-I. A linearized model of the Ares-I was obtained
as a test case of an independent validation of the ascent flight control design and analysis of
the Ares-I.
The fundamental principles of flight control analysis and design for flexible launch vehicles
have also been examined. In particular, the classical drift-minimum and load-minimum
control principles were re-examined, and the performance and stability of launch vehicle ascent
flight control with an additional feedback of angle-of-attack was demonstrated. For a typical
non-collocated actuator and sensor control problem of large flexible launch vehicles, nonminimum-phase filtering of unstably interacting bending modes was shown to be effective
and robust.
Two distinct approaches to the ascent flight control of Ares-I in the event of uncontrolled roll
drift have been investigated. The first approach exploits the inherent versatility of a quaternionbased attitude control system, and it only requires the desired inertial attitude quaternion to
be re-computed using the actual uncontrolled roll angle information. This approach achieved
an ascent flight trajectory identical as the nominal flight case with active roll control. The
second approach only requires a simple adjustment of the proportional-derivative gains of the
quaternion-based flight control system. The first approach is recommended for the Ares-I as

111
well as other launch vehicles in the event of no active roll control.
Finally, an undesired steady-state oscillation is found when the spinning rate is a constant.
Inspired by the method derived to stabilize a large flexible launch vehicle in the event of uncontrolled roll drift, a modified attitude quaternion feedback law is presented in this dissertation.
It is used to stabilize an axisymmetric rigid body by two independent control torques. By
Lyapunovs stability analysis, it is proved that the new control law can achieve an arbitrary
orientation of the symmetry axis with arbitrary spinning rate.

112

APPENDIX A.

A SUMMARY OF THE 6-DOF EQUATIONS OF

MOTION

Total force expressed in

Ftotal.xb

F
total.yb

Ftotal.zb

the body frame:

Faero.xb

= F
aero.yb

Faero.zb

Frkt.xb

+ F
rkt.yb

Frkt.zb

inertial frame:

Ftotal.xi

1
=

m Ftotal.yi

Ftotal.zi

Frcs.xb

+ F
rcs.yb

Frcs.zb

Total force expressed in the inertial frame:

F
Ftotal.xb
total.xi

I/B

F
Ftotal.yb
total.yi = C

Ftotal.zi
Ftotal.zb
Translational equation in the

(A.1)

(A.2)

gx

+ g
y

gz

Rotational equation in the body frame:

0 r q Ixx
Ixx Ixy Ixz p

I

0 p
Ixy
xy Iyy Iyz q = r

Ixz
q p
0
r
Ixz Iyz Izz

Taero.xb Trkt.xb

+
T
aero.yb + Trkt.yb

Taero.zb
Trkt.zb

(A.3)

Ixy Ixz p

Iyy Iyz
q

r
Iyz Izz

Trcs.xb

(A.4)

113

q1

q p

q1

q2 1 r 0
p
q
q
2

2
q
p
0
r
q
q

3
3

p q r 0
q4
q4

(A.5)

114

APPENDIX B.

ADDITIONAL FIGURES FROM 6-DOF SIMULATION

B.1

Atmospheric Model

x 10

3.5
3

Altitude (ft)

2.5
2
1.5
1
0.5
0
900

950

1000
1050
1100
Speed of Sound (ft/sec)

Figure B.1

Speed of sound

1150

1200

115

x 10

1.8
1.6

Altitude (ft)

1.4
1.2
1
0.8
0.6
0.4
0.2
0

0.5

1.5

2.5

Density of Air (slug/ft )

Figure B.2

3
3

x 10

Density of air

Vw.xb (ft/sec)

20
10
0
10

20

40

60

80

100

120

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

Vw.yb (ft/sec)

50
0
50

Vw.zb (ft/sec)

50
0
50
100

Figure B.3

Wind profile

116

8000
6000
4000

Base Force (lb)

2000
0
2000
4000
6000
8000
10000
12000

4
5
Altitude (ft)

Figure B.4

Base force

9
5

x 10

117

B.2

Aerodynamic Coefficient

2.6
2.4
2.2
2

CA

1.8
1.6
1.4
1.2
1
0.8
0.6
0

2
3
Mach Number

Figure B.5 CA

118

0.1
0.12
0.14

CY

0.16
0.18
0.2
0.22
0.24
0.26

2
3
Mach Number

Figure B.6 CY

0.26
0.24
0.22

CN

0.2
0.18
0.16
0.14
0.12
0.1

2
3
Mach Number

Figure B.7 CN

119

5.5
5
4.5

CMp

4
3.5
3
2.5
2
1.5

2
3
Mach Number

Figure B.8 CM p

1.5
2
2.5

CMy

3
3.5
4
4.5
5
5.5

2
3
Mach Number

Figure B.9 CM y

120

Rocket Parameters
6

2.2

x 10

Rocket Weight (lb)

1.8
1.6
1.4
1.2
1
0.8
0.6
0.4

20

40

60
Time (sec)

80

Figure B.10 Rocket weight.

100

120

121

3.5

x 10

Thrust (lb)

2.5

1.5

0.5

20

40

60
Time (sec)

80

100

120

Figure B.11 Rocket thrust.

Ixx (slugft2)

x 10

1
0

20

40

60

80

100

120

20

40

60

80

100

120

20

40

60
Time (sec)

80

100

120

Iyy (slugft2)

x 10

2
1

Izz (slugft2)

x 10

2
1

Figure B.12

Moments of inertia.

122

APPENDIX C.

C.1

Translational equations in

u
0 r q

v
0 p
= r

w
q p
0

where

B/I

LINEARIZATION RESULTS

Nonlinear 6-DOF Equations

the body frame:

u
gx
Faero.xb

1
v + C B/I g + F

y m aero.yb

w
Faero.zb
gz

2(q22

q32 )

+
2(q1 q2 + q3 q4 ) 2(q1 q3 q2 q4 )
1

2
2
=
2(q1 q2 q3 q4 ) 1 2(q1 + q3 ) 2(q2 q3 + q1 q4 )

2(q1 q3 + q2 q4 ) 2(q2 q3 q1 q4 ) 1 2(q12 + q22 )

Rotational equation in the body frame:

Ixx Ixy Ixz p

0 r q Ixx

0 p
xy Iyy Iyz q = r
Ixy

Ixz Iyz Izz

r
q p
0
Ixz

Taero.xb Trkt.xb

+
T
aero.yb + Trkt.yb

Trkt.zb
Taero.zb

T
1

+ T
z
m

T y
(C.1)

Ixy Ixz p

Iyy Iyz
q

Iyz Izz
r

Trcs.xb

+
0

(C.2)

(C.3)

123

C.2

Linear Rigid-Body Model

u = r0 v q0 w w0 q + v0 r + (2gy q2c + 2gz q3c )q1 + (4gx q2c + 2gy q1c 2gz q4c )q2
+ (4gx q3c + 2gy q4c + 2gz q1c )q3 + (2gy q3c 2gz q2c )q4 +

1
Faero.xb
m
(C.4)

v = r0 u + p0 w + w0 p u0 r + (2gx q2c 4gy q1c + 2gz q4c )q1

+ (2gx q1c + 2gz q3c )q2 + (2gx q4c 4gy q3c + 2gz q2c )q3 + (2gx q3c + 2gz q1c )q4
+

T
1
Faero.yb + ( )z
m
m
(C.5)

w = q0 u p0 v v0 p + u0 q + (2gx q3c 2gy q4c 4gz q1c )q1

+ (2gx q4c + 2gy q3c 4gz q2c )q2 + (2gx q1c + 2gy q2c )q3 + (2gx q2c 2gy q1c )q4 (C.6)
+

1
T
Faero.zb + ( )y
m
m

0
Ixx 0

0 I
0
yy

0
0 Izz

q = b

r
b3

(C.7)

where the relatively small products of inertia are ignored, and

b1 = (r0 Iyy Izz r0 )q + (Iyy q0 q0 Izz )r + Taero.xb
(C.8a)
+ (cy T )y + (cz T )z + Trcs
b2 = (r0 Iyy + Izz r0 ))p + ((Ixx p0 + p0 Izz )r + Taero.yb + (cx Xg )T y

(C.8b)

b3 = (q0 Ixx Iyy q0 )p + (Ixx p0 p0 Iyy )q + Taero.zb + (Xg cx )(T )z

(C.8c)

124

1
q1 = (q4c p q3c q + q2c r + r0 q2 q0 q3 + p0 q4 )
2
1
q2 = (q3c p + q4c q q1c r r0 q1 + p0 q3 + q0 q4 )
2
1
q3 = (q2c p + q1c q + q4c r + q0 q1 p0 q2 + r0 q4 )
2
1
q4 = (q1c p q2c q q3c r p0 q1 q0 q2 r0 q3 )
2
Linearization of the aerodynamic forces and moments:

0
0
Faero.xb 0

F

0
aero.yb = 0 CY QS/Vm

Faero.zb
0
0
CN QS/Vm.xb

Taero.xb

T
aero.yb

Taero.zb

(C.9a)
(C.9b)
(C.9c)
(C.9d)

(C.10)

cz
cy
0
Faero.xb

= c

0
Xa + cx
z

Faero.yb

Faero.zb
cy Xa cx
0

0
0
0
u

+
0
CM p QSb/Vm.xb
0
v

0 CM y QSb/Vm
0
w

(C.11)

Linearization of and :
=

1
Vm.xb

(C.12)

1
v
Vm

Linearization of the thrust force and moment:

0
F
rkt.xb

= T
F
z
rkt.yb

Frkt.zb
T y

T
0
c
c
z
y
rkt.xb

Frkt.xb

0
Xg + cx
rkt.yb = cz

Frkt.yb

Trkt.zb
cy Xg cx
0
Frkt.zb

(C.13)

(C.14)

(C.15)

125
Quaternion errors:

q1e

q2e

q3e

q4e

q4c

q3c

q3c q4c

=

q2c q1c

q1c
q2c

C.3

q2c q1c
q1c

q2c

q4c

q3c

q3c

q4c

q1

q2

q4

(C.16)

Linear State-Space Equations

A linearized state-space model of a rigid vehicle is described by

x = Ax + Bu

(C.17)

y = Cx
where x = (u, v, w, p, q, r, q1 , q2 , q3 , q4 )T , u = (Trcs , y , z )T ,
y = (2q1e , 2q2e , 2q3e , p, q, r)T ,

A11 A12 A13

A=
0
A21 A22

0
A32 A33

A11

r0
q0
0

=
p0
r0 CY QS/(mVm )

q0
p0
CN QS/(mVm.xb )

w0 v0
0

A12 =
0
u0
w0

v0 u0
0

(C.18)

(C.19)

(C.20)

A13

(2gy q2c + 2gz q3c )

(4gx q2c + 2gy q1c 2gz q4c )

=
(2gx q1c + 2gz q3c )
(2gx q2c 4gy q1c + 2gz q4c )

(2gx q3c 2gy q4c 4gz q1c ) (2gx q4c + 2gy q3c 4gz q2c )

(4gx q3c + 2gy q4c + 2gz q1c ) (2gy q3c 2gz q2c )

(2gx q4c 4gy q3c + 2gz q2c ) (2gx q3c + 2gz q1c )

(2gx q1c + 2gy q2c )

(2gx q2c 2gy q1c )

(C.21)

126

A21

Ixx 0

=
0 Iyy

0
0

0
1

Izz

h
0

x )CN QS
(Xa +c
Vm.xb

+
h

(Xa +cx )CY QS

Vm

0
Ixx 0

=
0 Iyy 0

0
0 Izz

CM y QSb
Vm

CM p QSb
Vm.xb

A22

r0 Iyy + Izz r0

Iyy q0 q0 Izz

r I + I r
0
Ixx p0 + p0 Izz
0 yy
zz 0

q0 Ixx Iyy q0 Ixx p0 p0 Iyy

0

q4c

q3c

q2c

0
Ixx 0

B2 =
0 Iyy 0

0
0 Izz

cy T
1

0 (c X )T
x
g

0
0

(C.22)

(C.23)

q4c q1c
1 q3c

A32 =

2
q4c
q2c q1c

q1c q2c q3c

0
r0 q0 p0

0
p0 q 0
1 r0

A33 =

2
0
r0
q0 p0

p0 q0 r0 0

B1

B=
B2

0
0
0

B1 =
0
T /m
0

0 T /m
0

cz T + cy T

(cx Xg )T

(C.24)

(C.25)

(C.26)

(C.27)

(C.28)

127

C=

0 0 0 0 0 0

2q4c

2q3c

0 0 0 0 0 0 2q3c

2q2c 2q1c

2q4c

2q1c

2q2c

0 0 0 0 0 0

2q2c

2q1c

2q4c

2q3c

0 0 0 1 0 0

0 0 0 0 1 0

0 0 0 0 0 1

(C.29)

Assuming that p0 = q0 = r0 = 0, v0 = w0 = 0, 2q2e and Vm is a constant, then we get

the rigid-body dynamic model of Ares-I CLV in the pitch plane as follows:
gx
CN QS
T
= +
+
+
y
Vm
mVm
mu0
(Xa + cx )CN QS + CM p QSb
(cx Xg )T
=
+
y
Iyy
Iyy
where =

(C.30a)
(C.30b)

w
Vm .

C.4

Linear Flexible-Body Model

The linear state-space equation of the Ares-I including the flexible-body modes is described
by
+ Bu

x = Ax

(C.31)

y = Cx
where
x = (u, v, w, p, q, r, q1 , q2 , q3 , q4 , 1 , 2 , 3 , 4 , 5 , 6 , 1 , 2 , 3 , 4 , 5 , 6 )T
(C.32)
and

=
A

0
0

=
B

2 2

T mB1

(C.33)

(C.34)

128

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

C=

0 0 0

0 0 0

0 0 0

2q4c

2q3c

2q2c 2q1c

2q3c

2q4c

2q1c

2q2c

2q2c

2q1c

2q4c

2q3c

1 0 0 0 0 0 0
0 1 0 0 0 0 0
0 0 1 0 0 0 0

(C.35)

129

APPENDIX D.

ATTITUDE ERROR QUATERNION KINEMATIC

DIFFERENTIAL EQUATIONS

Equations. (2.12) and (2.46) can be written as

1
q = q
2

(D.1)

qe = Qc q

(D.2)

where q = (q1 , q2 , q3 , q4 )T , qe = (q1e , q2e , q3e , q4e )T , and

q p

r 0
p
q

q
p
0
r

p q r 0

q4c

q3c

q2c q1c

q3c q4c

Qc =

q2c q1c

q1c
q2c

q1c

q2c

q4c

q3c

q3c

q4c

(D.3)

(D.4)

1
T
Note that Qc is an orthonormal matrix; that is, Qc QT
c = I and Qc = Qc . Differentiating

Equation. (D.2) with constant commanded quaternions, we obtain

q e = Qc q

(D.5)

Substituting Equation. (D.1) into Equation. (D.5), we obtain

1
q e = Qc Q1
c qe
2

(D.6)

130
1
By substituting Equations. (D.3) and (D.4) into Qc Q1
c , we can show that Qc Qc = .

An indirect approach to obtaining this relationship is provided as follows.

Let
Qc = q4c I + Q

(D.7)

where

q3c

q3c
0

Q=

q2c q1c

q1c
q2c

q2c q1c
q1c

q2c

q3c

q3c

(D.8)

Note that Q and are skew-symmetric matrices; that is, Q = QT and = T .

Consequently, Q is a symmetric matrix; that is, Q = (Q)T . Then we have
Q = (Q)T = T QT = QT = Q

(D.9)

By using these properties, we can rewrite Qc Q1

c as
1
1
Qc Q1
c = Qc Qc = (q4c I + Q)Qc

(D.10)

which becomes
1
1
1
(q4c I + Q)Q1
c = (q4c I + Q)Qc = (q4c I + Q)Qc = Qc Qc

(D.11)

Finally, we have
Qc Q1
c =

(D.12)

and

q
1e

q2e

q3e

q4e

0
r q

1 r 0
p

=
2

q p 0

p q r

q
1e

q
q2e

r q3e

q4e
0

Note that the constant commanded quaternions are assumed here.

(D.13)

131

APPENDIX E.

LINEAR MODEL OF UNCONTROLLED ROLL DRIFT

WITH AERODYNAMIC DISTURBANCE

CY QS
T
z
m
m

(E.1)

CN QS
T
+ y
m
m

(E.2)

v = u0 r gx +

w = u0 q gx

where =

w
u0 ,

q = M + M y

(E.3)

r = M + M z

(E.4)

q2e =

q1e
q4e
q
r
2
2

(E.5)

q3e =

q1e
q4e
q+
r
2
2

(E.6)

v
u0 .

When attitude error quaternion are small, we have 2q2e and 2q3e .

132

CY QS
mu0

QS

w
0
CNmu

0


q
0
M /u0
=

r
0
M /u0

q2e
0
0


q3e
0
0

u0

2gx

u0

2gx

q4e /2 q1e /2

q1e /2

q4e /2

w
T /m


q M
+

r
0

q2e
0

0
q3e

T /m
0
0
M
0

0
(E.7)

133

APPENDIX F.

F.1

DERIVATION OF A STEADY-STATE OSCILLATION

A Steady-State Oscillation of the Autonomous System

For this autonomous system x = f (x), we have

p q2 K
dq
q
arp K

r apq K
p q3 K
dr

q1 2r q2 2q q3 + p2 q4

q2 r q1 + p q3 + q q4
2
2
2

q q q p q + r q
3 2 1 2 2 2 4

q4
p2 q1 2q q2 2r q3

(F.1)

Assume that there is a steady-state oscillation. q1 and q4 are constants and it has the
following relation between two vectors (q, r)T and (q2 , q3 )T is

q
q2
=X

r
q3

cos sin
X=

sin cos

(F.2)

(F.3)

where | | is the magnitude ratio and is the phase shift between those two vectors. Note that
the matrix X is invertible. Therefore

q 1 cos sin q
q2
1
=X =

sin cos
r
q3
r
Since and are constants, the relation between (q,
r)
T and (q2 , q3 )T is

q
q2
=X

r
q3

(F.4)

(F.5)

134
Equation. (F.1) can be expressed in terms of (q, r)T , (q2 , q3 )T and (q1 , q4 )T as

q2
q Kd ap q
p)

=
+ (K

d
r
ap K
r
q3

q2 1 q4 q1 q 1 0 p q2
=
+

2
2
q3
r
q1 q4
p 0
q3

q1 1 0 p q1 1 r q q2
=

2
2
q4
p 0
q4
q r
q3
Substituting Equation. (F.4) into Equation. (F.6a), we get

q
q Kd ap
q
p )X 1
= M
=
+ (K
d
r
ap K
r
r
Similarly, substituting Equation. (F.2) into Equation. (F.6b), we get

1 0 p q2
q2
q2 1 q4 q1
X +

=N

2
2
q1 q4
p 0
q3
q3
q3
According to Equations. (F.5), (F.7) and (F.8), we obtain

q
q2
q2
q
q2
=X
= XN
= M = MX

r
q3
q3
r
q3

(F.6a)

(F.6b)

(F.6c)

(F.7)

(F.8)

(F.9)

Therefore, the relation between M , N and X must be

M X = XN
In the matrix form, we have

1 0 p
1 q4 q1
Kd ap
p )X 1
X +

+ (K
X = X

2
2
d
q1 q4
p 0
ap K

(F.10)

(F.11)

Note that the preceding equation is a quadratic matrix equation as follows:

XAX + BX + XC = D

(F.12)

135
In the matrix form, the preceding equation can be written as

d ap
1 q4 q1
1 0 p K

p )I
X
X + X
+
X = (K
2
2

p 0
ap Kd
q1 q4

(F.13)

We are going to find out expressions of q1 and q4 . According to the assumption that q1
and q4 are constant, we have

q1 1 0 p q1 1 r q q2 0
+
=
=

2
2
0
p 0
q4
q r
q3
q4

(F.14)

therefore,

1 r q q2
1 0 p q1

2
2
p 0
q4
q r
q3
Substituting Equation. (F.4) into the preceding equation, we obtain

1 0 p q1
1 r q 1 cos sin q

=
2
2

p 0
q r
sin cos
r
q4

(F.15)

(F.16)

The right side of the preceding equation can be simplified

2
2
1 qr cos q sin r sin qr cos
1 r cos q sin r sin q cos q

2
2
q cos r sin q sin r cos
r
q 2 cos rq sin + qr sin r2 cos

q 2 + r2 sin
=

2
cos
(F.17)
Hence,

q1

=
q4
where =

q2

+
p

r2

cos
cos
=

sin
sin

(F.18)

q 2 +r 2
p

F.2

Solution of the Quadratic Matrix Equation

The matrix X is the solution of the Equation. (F.13)

1 q4 q1
1 0 p Kd ap
p )I
X
X + X
+
X = (K
2
2
d
q1 q4
p 0
ap K

(F.19)

136
where

cos
q1
=

sin
q4

(F.20)

The first term on the left side of Equation. (F.13) can be written as

2
q
q
cos

sin

sin

cos

cos

sin

1

1 4

X
X =

2
2
q1 q4
sin cos
cos sin
sin cos

2
0 1 cos sin
=

2
1 0
sin cos

2 sin cos
=

2
cos sin
(F.21)
The second term on the left side of Equation. (F.13) can be written as

1 0 p cos sin 0 p p sin cos

X
=

=
2
2
2
p 0
sin cos
cos sin
p 0
The third term on the left side of Equation. (F.13) can be written as

Kd ap cos sin
Kd ap
X =

ap Kd
sin cos
ap Kd

cos sin
sin cos
d
= K

+ ap

sin cos
cos sin

(F.22)

(F.23)

Substituting Equations. (F.21), (F.22) and (F.23) into Equation. (F.13), the quadratic
matrix equation becomes

2
cos sin
p sin cos
sin cos
d
p )I
(
+ )
+ K

= (K
+ ap
2
2
sin cos
cos sin
cos sin
(F.24)
or

cos sin
(2a 1)p
sin cos
d
p )I

)
+ K

= (K
2
2
cos sin
sin cos
2

(F.25)

137
Therefore, and must satisfy the following equation

(2a1)p 2
(2a1)p 2

sin + Kd cos
cos + Kd sin

2
2
p )I

= (K
(2a1)p 2
(2a1)p 2

cos Kd sin
sin + Kd cos
2
2

(F.26)

Note that the preceding equation can be reduced to only two equations
(2a 1)p 2
d cos = K
p
sin + K
2

(F.27)

(2a 1)p 2
d sin = 0
cos K
2

(F.28)

and

In the matrix form, we get

Since =

q 2 +r 2
p ,

or

(2a1)p 2
2

d
K

d
K

(2a1)p 2
2

sin Kp

cos
0

(F.29)

we find

(2a1)p2 (q 2 +r 2 )
2p

d
K

d
K

(2a1)p2 (q 2 +r 2 )
2p

sin Kp

cos
0

p
K

sin

= (2a1)p2 (q2 +r2 )

2
2

(
)
+
(
K
)
d
cos
2p

(2a1)p2 (q 2 +r 2 )
2p

d
K

(F.30)

(F.31)

If is chosen as
p
K
=q
2
2
2
d )2
( (2a1)p 2p(q +r ) )2 + (K
Finally, solving Equation. (F.13) yields

1
sin

= q

2
2
2
d )2
( (2a1)p 2p(q +r ) )2 + (K
cos

F.3

(F.32)

(2a1)p2 (q 2 +r2 )
2p

d
K

(F.33)

State Equations of the Steady-State Oscillation

Substituting Equations. (F.18) and (F.33) into Equation. (F.8) yields

2
2
2
q2
q2 p + q + r 0 1 q2

=N
=

2p
q3
q3
1 0
q3

(F.34)

138
From Equation. (F.7) we have

Kd ap
p )X 1
M =
+ (K
d
ap K

(F.35)

According to Equations. (F.18) and (F.32), we have

cos sin
(2a 1)p2 (q 2 + r2 ) 2
Kd ap
d )2
M =
) + (K
+ (

2p
d
ap K
sin cos

q
q
2 (q 2 +r 2 )
2 (q 2 +r 2 )
(2a1)p
(2a1)p

d )2 cos
d )2 sin
)2 + (K
ap (
)2 + (K

Kd + (
2p
2p
=
q
q

2
2
2
2
2
2
d )2 sin K
d + ( (2a1)p (q +r ) )2 + (K
d )2 cos
ap + ( (2a1)p 2p(q +r ) )2 + (K
2p
(F.36)
Substituting Equation. (F.33) into the preceding equation yields

(2a1)p2 (q 2 +r2 )
0
ap
(p2 + q 2 + r2 ) 0 1

2p
M =
=

2
2
2
2p
ap + (2a1)p 2p(q +r )
0
1 0
(F.37)
Finally, state equations of the steady-state oscillation are listed here

q2 p2 + q 2 + r2 0 1 q2

2p
q3
1 0
q3

q p2 + q 2 + r2 0 1 q
=

2p
1 0
r
r
and

(F.38)

(F.39)

q
cos sin q2
=

r
sin cos
q3

(F.40)

p
K
=q
2
2
2
d )2
( (2a1)p 2p(q +r ) )2 + (K

(F.41)

where

q1 q 2 + r2 cos

p
q4
sin

(F.42)

139

sin

= q
2
2
2
d )2
( (2a1)p 2p(q +r ) )2 + (K
cos
= tan1

(2a1)p2 (q 2 +r2 )
2p

d
K

(2a 1)p2 q 2 r2
d
2pK

(F.43)

(F.44)

Since there is another constraint

q12 + q22 + q32 + q42 = 1

(F.45)

q 2 + r2 2
)
p

(F.46)

q 2 + r2 = 2 (q22 + q32 )

(F.47)

From Equation. (F.18) we have

q12 + q42 = (
From Equation. (F.2) we get

Define the magnitude square of the vector (q, r)T as x

q 2 + r2 = x

(F.48)

x
2

(F.49)

x 2
x
) + 2 =1
p

(F.50)

and
q22 + q32 =
Therefore, x must satisfy the constraint
q12 + q22 + q32 + q42 = (

Substituting Equation. (F.32) into the preceding equation yields

2 ]p2 x2 + [(2a 1)2 p2 + 4K
2 ]p4 x 4p4 K
2 = 0 (F.51)
x4 + (3 4a)p2 x3 + [(2a 1)(2a 3)p2 + 4K
p
d
d
Define a function f (x) as the left side of equal sign of the preceding equation
2 ]p2 x2 +[(2a1)2 p2 +4K
2 ]p4 x4p4 K
p2 (F.52)
f (x) = x4 +(34a)p2 x3 +[(2a1)(2a3)p2 +4K
d
d
Since x is the magnitude square of the vector (q, r)T , all real positive zeros of the function
f (x) could be x.

140

APPENDIX G.

DERIVATION OF THE DERIVATIVE OF A

LYAPUNOV FUNCTION CANDIDATE

A Lyapunov function candidate is given in Equation. (5.17),

V (x) =

1 2
1 2
q +
r + [q1 sin(/2)]2 + q22 + q32 + [q4 cos(/2)]2
p
p
2K
2K

(G.1)

Expanding the square terms of quaternion, we get

V (x) =

1 2
1 2
q +
r + q12 2q1 sin(/2) + sin2 (/2)

2K p
2K p
+ q22 + q32 + q42 2q4 cos(/2) + cos2 (/2) (G.2)

Therefore, the Lyapunov function can be rewritten as

V (x) =

1
(q 2 + r2 ) + 2 2q1 sin(/2) 2q4 cos(/2)
p
2K

(G.3)

The derivative of V (x) is

1
V (x) =
(q q + rr)
2q1 sin(/2) q1 cos(/2) 2q4 cos(/2) + q4 sin(/2)

Kp

(G.4)

Substituting Equation. (5.16) into the preceding equation yields the derivative of V (x)
along the trajectories of the system, then
q
d q}
p [cos(/2)q2 + sin(/2)q3 ] K
{arp K
V (x) =
p
K
+

r
p [ sin(/2)q2 + cos(/2)q3 ] K
d r}
{apq K
p
K

(rq2 qq3 + pq4 ) sin(/2) (pq1 qq2 rq3 ) cos(/2)

q1 cos(/2) + q4 sin(/2)

(G.5)

141
Expanding the preceding equation yields
d
d
apqr
K
apqr
K
V (x) =
qq2 cos(/2) qq3 sin(/2)
q2
+ rq2 sin(/2) rq3 cos(/2)
r2
p
p
p
p
K
K
K
K
rq2 sin(/2) + qq3 sin(/2) pq4 sin(/2) + pq1 cos(/2) + qq2 cos(/2) + rq3 cos(/2)
[q1 cos(/2) q4 sin(/2)]
(G.6)
Hence,
d
d
K
K
q2
r2 + [q1 cos(/2) q4 sin(/2)]p [q1 cos(/2) q4 sin(/2)]
V (x) =
p
p
K
K

(G.7)

Finally, we obtain a compact form the the derivative of V (x)

d
K

V (x) =
(q 2 + r2 ) + [q1 cos(/2) q4 sin(/2)](p )
p
K

(G.8)

Since = p, we have
d
K
(q 2 + r2 )
V (x) =
p
K

(G.9)

142

BIBLIOGRAPHY

[1] Cook, S., Ares Project Status, Presented at 2nd AIAA Space Exploration Conference,
December 4-6, 2006.
[2] Sumrall, P., The Ares Projects: Progress Toward Exploration, Presented at AIAA Space
2008 Conference, September 10, 2008.
[3] Butt, A., Popp, C. G., Pitts, H. M. and Sharp, D. J., NASA Ares I Launch Vehicle Roll
and Reaction Control Systems Design Status, NASA Technical Reports Server, August
2009.
[4] Norris, L., Tao, Y. C., Hall, R., Chuang, J., and Whorton, M., Analysis of Ares-I Ascent Navigation Options, AIAA 2008-6290, AIAA Guidance, Navigation, and Control
Conference, Honolulu, Hawaii, August 18-21, 2008.
[5] Pinson, R. M., Schmitt, T. L., and Hanson, J. M., Development of a Smooth Trajectory
Maneuver Method to Accommodate the Ares I Flight Control Constrains, AIAA 20086292, AIAA Guidance, Navigation, and Control Conference, Honolulu, Hawaii, August
18-21, 2008.
[6] Pitman, G. R., Inertial Guidance, New York, John Wiley and Sons, 1962, pp. 260-261.
[7] Dukeman, G. A., and Hill, A. D., Rapid Trajectory Optimization for the Ares-I Launch
Vehicle, AIAA 2008-6288, AIAA Guidance, Navigation, and Control Conference, Honolulu, Hawaii, August 18-21, 2008.
[8] Seifert, H., Space Technology, New York, John Wiley and Sons, 1959, Chapter 4.

143
[9] Leondes, C. T., Guidance and Control of Aerospace Vehicles, New York, McGraw-Hill
Book Company, 1963, pp. 191-249.
[10] Greensite, A. L., Analysis and Design of Space Vehicle Flight Control Systems, Spartan
Books, New York, 1970, pp. 44-108, pp. 194-372.
[11] Whorton, M., Hall, C., and Cook, S., Ascent Flight Control and Structural Interaction
for the Ares-I Crew Launch Vehicle, AIAA 2007-1780, April 2007.
[12] Haeussermann, W., Description and Performance of the Saturn Launch Vehicles Navigation, Guidance, and Control System, NASA Technical Reports Server, NASA TN
D-5869, July 1970.
[13] Frosch, J. A. and Valley, D. P., Saturn AS-501/S-IC Flight Control System Design,
Journal of Spacecraft, Vol. 4, No. 8, 1967, pp. 1003-1009.
[14] Blakelock, J., Automatic Control of Aircraft and Missiles, John Wiley & Sons, New York,
1991. pp. 429-437.
[15] Blackburn, T. R. and Vaughan, D. R., Application of Linear Optimal Control and Filtering Theory to the Saturn V Launch Vehicle, IEEE Transactions on Automatic Control,
Vol. AC-16, No. 6, December 1971, pp. 799-806.
[16] Wie, B., Space Vehicle Dynamics and Control, AIAA Education Series, 1998, pp. 129-145,
pp. 381-418.
[17] Wie, B., and Barba, P. M., Quaternion Feedback for Spacecraft Large Angle Maneuvers,
Journal of Guidance, Control, and Dynamics, Vol. 8, No. 3, 1985, pp. 360-365.
[18] Wie, B., Weiss, H., and Arapostathis, A., Quaternion Feedback Regulator for Spacecraft
Eigenaxis Rotations, Journal of Guidance, Control, and Dynamics, Vol. 12, No. 3, 1989,
pp. 375-380.

144
[19] Weiss, H., Quaternion-Based Rate/Attitude racking System with Application to Gimbal
Attitude Control, Journal of Guidance, Control, and Dynamics, Vol. 16, No. 4, 1993, pp.
609-616.
[20] Wie, B. and Byun, K. W., New Generalized Structural Filtering Concept for Active
Vibration Control Synthesis, Journal of Guidance, Control, and Dynamics, Vol. 12, No.
2, 1989, pp. 147-154.
[21] Byun, K. W., Wie, B. and Sunkel, J., Robust Non-Minimum-Phase Compensation for
a Class of Uncertain Dynamical Systems, Journal of Guidance, Control, and Dynamics,
Vol. 14, No. 6, 1991, pp. 1191-1199.
[22] Davis, S., Ares I-X Flight Test-The Future Begins Here, AIAA 2008-7806, AIAA space
2008 Conference and Exposition, San Diego, California, September 9-11, 2008.
[23] Crouch, P. E., Spacecraft Attitude Control and Stabilization: Applications of Geometric
Control Theory to Rigid Body Models, IEEE Transactions on Automatic Control, Vol.
29, No. 4, 1984, pp. 321-331.
[24] Krishnan, H., McClamroch, N. H., and Reyhanoglu, M., Attitude Stabilization of a Rigid
Spacecraft Using Gas Jet Actuators Operating in a Failure Mode, IEEE Conference on
Decision and Control, Tucson, Arizona, pp. 1612-1617.
[25] Tsiotras, P., and Longuski, J. M., Spin-axis stabilization of symmetric spacecraft with
two control torques, Systems & Control Letters, Vol. 23, 1994, pp. 395-402.
[26] Tsiotras, P., Corless, M., and Longuski, J. M., A Novel Approach to the Attitude Control
of Axisymmetric Spacecraft, Automatica, Vol. 31, No. 6, 1995, pp. 1099-1112.
[27] Bymes, C.I., and Isidori. A., On the attitude stabilization of rigid spacecraft, Automatica, Vol. 27, No. 1, 1991, pp. 87-95.

145
[28] Betts, K. M., Rutherford, R. C., McDuffie, J., Johnson, M. D., Jackson, M., and Hall, C.,
Time Domain Simulation of the NASA Crew Launch Vehicle, AIAA 2007-6621, August
2007.
[29] Betts, K. M., Rutherford, R. C., McDuffie, J., Johnson, M. D., Jackson, M., and Hall,
C., Stability Analysis of the NASA Ares-I Crew Launch Vehicle Control System, AIAA
2007-6776, August 2007.
[30] James, R. L., A Three-Dimensional Trajectory Simulation Using Six Degrees of Freedom
with Arbitrary Wind, NASA TN D-641, 1961.
[31] Harris, R. J., Trajectory Simulation Applicable to Stability and Control Studies of Large
Multi-Engine Vehicles, NASA TN D-1838, 1963.
[32] Zipfel, P. H., Modeling and Simulation of Aerospace Vehicle Dynamics, AIAA Education
Series, 1998, pp. 367-427.
[33] Stuelpnagel, J., On the parameterization of the three-dimensional rotation group, SLAM
Review, Vol. 6, No. 4, October 1964.
[34] Brandon, J. M., Derry, S. D., Heim, E. H., Hueschen, R. M., and Bacon, B. J., Ares-I-X
Stability and Control Flight Test: Analysis and Plans, NASA Technical Reports Server,
September 2008.
[35] Hall, C., Lee, C., Jackson, M., Whorton, M., West, M., Brandon, J., Hall, R. A., Jang,
J., Bedrossian, N., Compton, J., and Rutherford, C., Ares-I Flight Control System
Overview, AIAA 2008-6287, AIAA Guidance, Navigation, and Control Conference, Honolulu, Hawaii, August 18-21, 2008.
[36] Chopra, A. K., Dynamics of Structures, Prentice Hall, New Jersey, 1995, pp. 585-600.
[37] Hoelkner, R. F., The Principle of Artificial Stabilization of Aerodynamically Unstable
Missiles, ABMA DA-TR-64-59, September 25, 1959.

146
[38] Hoelkner, R. F., Theory of Artificial Stabilization of Missiles and Space Vehicles with
Exposition of Four Control Principles, NASA Technical Reports Server, NASA TN D555, June 1961.
[39] Harvey, C. A., An Alternate Derivation and Interpretation of the Drift-Minimum Principle, NASA Contract NASw-563, MH MPG Report 1541-TR 15, Minneapolis-Honeywell,
November 22, 1963.
[40] Garner, D., Control Theory Handbook, NASA TM X-53036, April 22, 1964.
[41] Rheinfurth, M. H., The Alleviation of Aerodynamic Loads on Rigid Space Vehicles,
NASA TM X-53397, February 21, 1966.
[42] Martin, D. T., Sievers , R. F., OBrien, R. M., and Rice, A. F., Saturn V Guidance,
Navigation, and Targeting, J. Spacecraft, Vol. 4, No. 7, 1967, pp. 891-898.
[43] Wie, B., and Du, W., Analysis and Design of Launch Vehicle Flight Control Systems,
AIAA 2008-6291, AIAA Guidance, Navigation, and Control Conference, August 2008.
[44] Doyle, J. C., Francis, B. A. and Tannenbaum A. R., Feedback Control Theory, Dover
Publication, New York, 1992, pp. 45-62.
[45] Skogestad, S. and Postlethwaite, I., Multivariable Feedback Control: Analysis and Design,
John Wiley & Sons, 2nd ed., 2005, pp. 259-338.
[46] Zhou, K., Doyle, J. C. and Glover. K., Robust and Optimal Control, Prentice Hall, New
Jersey, 1995, pp. 213-300.
[47] Effects of Structural Flexibility on Launch Vehicle Control Systems, NASA Technical
Reports Server, NASA SP-8036, February 1970.
[48] Balas, G. J., and Doyle, J. C., Robustness and Performance Tradeoffs in Control Design
for Flexible Structures, IEEE Transactions on Control Systems Technology, Vol. 2, No.
4, 1994, pp. 352-361.

147
[49] Balas, G. J., and Doyle, J. C., Control of Lightly Damped, Flexible Modes in the Controller Crossover Region, Journal of Guidance, Control, and Dynamics, Vol. 17, No. 2,
1994, pp. 370-377.
[50] Smith, R. S., Chu, C. C., and Fanson, J. L., The Design of H Controllers for an
Experimental Non-collocated Flexible Structure Problem, IEEE Transactions on Control
Systems Technology, Vol. 2, No. 2, 1994, pp. 101-109.
[51] Franklin, G. F., Powell, J. D., and Naeini, A. E., Feedback Control of Dynamic Systems,
5th ed., Prentice Hall, New Jersey, 2006, pp. 131-139.
[52] Khalil, H. K., Nonlinear Systems, 3rd ed., Prentice Hall, New Jersey, 2010, pp. 126-129.
[53] Sastry, S. S., Nonlinear Systems: Analysis, Stability, and Control, 1st ed., Springer, New
York, 1999, pp. 287-288.