Nonlinear Control of the Reaction Wheel Pendulum Using Passivity-

based control And Backstepping Control

Gajje Phani Teja, Dr. Ashwin Dhabale, Tushar Waghmare,
M.Tech, Assistant professor, Research Scholar,
Department of Electrical Engineering, Department of Electrical Engineering, Department of Electrical Engineering,
Visvesvaraya National Institute of Visvesvaraya National Institute of Visvesvaraya National Institute of
Technology, Technology, Technology,
Nagpur, India. Nagpur, India. Nagpur, India.
[email protected] [email protected]

Abstract—This paper presents a control logic for the control the inverted pendulum [4]. In this paper, control of
inversion and stabilization of the reaction wheel pendulum. the inverted pendulum is achieved using passivity-based
The inverted pendulum control has the applications in the control (PBC) and backstepping control. PBC and
fields of robotics, rocket propulsion control and self-balancing backstepping control are two well-known methods for non-
vehicles. The control logic in this paper was designed using the linear system control design. These methods use the concept
nonlinear control techniques. Control logic for inverting the of Lyapunov stability for the control design. The passivity-
pendulum was designed using passivity-based control and the based control is one of the widely used control techniques
control logic for balancing the pendulum in the inverted
in the applications like walking robots, pendulum systems,
position was designed using backstepping control. The final
switching, synchronization, electrical systems,
control logic was designed by incorporating proper logic for
switching the control between the two control logics. The
electromechanical systems etc.,[5]. The backstepping
simulation results are presented and explained substantiating control also holds good applications like pumping stations,
the designed control logic. ship control and flight control.
The paper is organized such that after the brief
Keywords—reaction wheel pendulum, inverted pendulum, introduction in Section I, some important concepts and
Lagrange’s equations, Rayleigh dissipation function, passivity-
theorems that are useful to understand this paper are
based control, backstepping control.
presented in Section II. The system and its parameters are
I. INTRODUCTION described in Section III. The system dynamics are derived
th
in Section IV. The design of the control is explained in
Since 17 century, pendulum was used for observation, Section V followed by the simulation results and their
measurement and experimentation by many great scientists explanation in Section VI. Finally, the paper is concluded in
[1]. Pendulum exhibits most of the commonly known Section VII.
nonlinearity properties like multiple equilibrium points,
limit cycles etc., making it an interesting system to study the II. BASIC CONCEPTS AND THEOREMS
control design for the nonlinear systems. A pendulum In this section, some important concepts and theorems that
system has infinite stable and unstable equilibrium points. are useful in understanding the paper are given and are used
The inverted position of the pendulum is an unstable in further sections for derivation of dynamics and control
equilibrium point. The challenge is to design a suitable design.
control logic to stabilize the system in inverted position. The
pendulum thus inverted and stabilized in inverted position A. Lagrange’s equations of motion
is called as the inverted pendulum. The inverted pendulum Lagrange’s equations is an alternate way of deriving the
control has got the applications in the areas like walking equations of motion of a system using only scalar energies
humanoid robots, rocket propulsion, self-balancing vehicles in the system.
[2]. The inverted pendulum is achieved using the systems The Lagrange’s equations of motion are given by
like cart-pendulum, reaction wheel pendulum, pendubot, formula
rotational inverted pendulum [3]. The cart-pendulum − =T
̇
problem was used for demonstration of linear feedback
control problem since 1950’s. Later it was used for Where, L is the Lagrangian function,
demonstration of the nonlinear control problem also. In cart- Text is the external torque on the system,
pendulum problem, stabilization of the inverted pendulum xi is the ith generalized coordinate.
is done by moving the entire system to which the pendulum Lagrangian function (L) = T - V.
is hinged. That makes this system model not suitable for
many practical applications. So, the reaction wheel Where, T is the kinetic energy of the system,
pendulum was designed in which the stabilization of the V is the potential energy of the system [6].
inverted pendulum is done by rotation of the reaction wheel
only. The reaction wheel is accelerated or decelerated in B. Lyapunov stability theorem
accordance with the current position and required position Let 𝐱 = 𝟎 be an equilibrium point for equation 𝐱̇ =
of the pendulum which is a little more practical as the whole 𝐟(𝐱) , 𝐟(𝟎) = 𝟎 and D ⊂ R be a domain containing 𝐱 = 𝟎.
system to which the pendulum is hinged do not move away Let V: D → R be a continuously differentiable function such
either to invert or stabilize the pendulum. that
Different control techniques like Fuzzy logic control, V(0) = 0 and V(x) > 0 in D – {0},
Neural network control, PID adaptive control, Energy based
V̇(x) ≤ 0 in D Then 𝐱 = 𝟎 is stable.
control, Hybrid control, sliding mode control are used to

978-1-7281-8873-7/20/$31.00 ©2020 IEEE

Authorized licensed use limited to: ULAKBIM UASL - GAZI UNIV. Downloaded on June 05,2024 at 12:22:45 UTC from IEEE Xplore. Restrictions apply.
Moreover, if φ : Angle of some fixed point on reaction wheel
from the extended line drawn through hinged
V̇(x) < 0 in D – {0} Then x = 0 asymptotically is stable
point and the center of reaction wheel in the
[7].
direction shown in Fig.1.
C. Theorem 1
If a p-input-p-output system ẋ = f(x, u) , y = h(x) is I =I +m +m l ,
1. Passive with a radially unbounded positive definite
storage function and ML = +m l .
2. Zero-state observable,
IV. SYSTEM DYNAMICS
Then the origin x=0 can be globally stabilized by u =
−ϕ(y), where ϕ is any local Lipschitz function such that In this section, the system dynamics by both considering
and without considering the friction are derived.
ϕ(0) = 0 and y ϕ(y) > 0 for all y ≠ 0. [7] The system dynamics are derived using the Lagrange’s
equations of motion which are stated in Section II. For that,
III. SYSTEM DESCRIPTION the Lagrangian function (L) is formed as stated in Section II
In this section, the description of the system and the
experiment are given. Also, parameter notations used in KE = I +m θ̇
this paper are explained.
Let us assume a pendulum consisting of a rod hinged at one
end and the reaction wheel along with motor attached on KE = m l θ̇ + I θ̇ − φ̇
the other end. When freely released from any initial point
T = KE + (KE )
with any initial torque applied, the system always reaches
its stable equilibrium point at the bottom. So, a control m l
system is needed to invert the system and stabilize it at that V= +m l g (1 − cos θ)
2
unstable equilibrium point.
L= I +m θ̇ + m l θ̇ + I θ̇ −

φ̇ − +m l g (1 − cos θ)
(1)
By using lagrange equations for system dynamics,
We get the equations describing the system dynamics as
(I + I )θ̈ − (I )φ̈ + {MLg sin θ} = 0 (2)
I θ̈ − φ̈ =T (3)
By substituting equation (3) in (2), we get
I θ̈ + {MLg sin θ} + T = 0 (4)
In (2)-(4), the friction is neglected. Hence to make the
description more practical, we have to include friction in the
FIG.1 REACTION WHEEL PENDULUM
above equations. This is done using Rayleigh’s dissipation
function along with Lagrange’s equations.
The system parameters used in this paper are explained
below. According to definition, Rayleigh’s dissipation function (D)
mp : Mass of the pendulum body, D = B θ̇ + (B + B )φ̇
mr : Mass of the reaction wheel,
Ip : Moment of inertia of pendulum body about its The lagrange equations with dissipation term are
center of mass,
Ir : Moment of inertia of reaction wheel about its ̇
− + ̇
=T [6]
center of mass,
We get complete system dynamics as
Bp : Frictional coefficient between pendulum and
hinged point, (I + I )θ̈ − (I )φ̈ + {MLg sin θ} + B θ̇ = 0 (5)
Br : Frictional coefficient between pendulum and I θ̈ − φ̈ + (B + B )φ̇ = T (6)
reaction wheel,
Bm : Frictional coefficient of motor, By substituting equation (6) in (5), we get
lp : Length of pendulum body, I θ̈ + {MLg sin θ} + T + B θ̇ − (B + B )φ̇ = 0 (7)
Rr : Radius of reaction wheel,
θ : Angle of the pendulum body from its initial The dynamics of the motor used to drive the reaction wheel
equilibrium position with no external force on are given by following equations
the system, V =I R +L + k φ̇ (8)

Authorized licensed use limited to: ULAKBIM UASL - GAZI UNIV. Downloaded on June 05,2024 at 12:22:45 UTC from IEEE Xplore. Restrictions apply.
T =kI . (9) Ṡ = k S − PE (−T )θ̇ + k (S ) (T ) θ̇ − φ̇
Where, Vm is voltage applied across DC motor terminals, Ṡ = T −k S − PE θ̇ + k (S ) θ̇ − φ̇ =u y
Im is the current flowing inside the DC motor, (18)
Ra is the resistance of DC motor, at the required theta, S should become zero. So, Ṡ ≤ 0
assuming, u = −ky, we can make above condition
La is the inductance of DC motor,
possible.
Imr is the inertia of the DC motor,
(T ) =− −k S − PE θ̇ + k (S ) θ̇ − φ̇
Bm is the friction coefficient of the DC motor,
(19)
kv is the back emf constant, By using above derived control torque input, we can make
kt is the torque constant. system to move on a surface where S1 is around PEreq.
The initial angular velocity of the pendulum and the
V. CONTROL DESIGN initial angular velocity of the reaction wheel are zero. So, Ṡ
This section deals with the design of the control logics for and the control torque are zero at the initial position. Hence
the swing up and the stabilization using passivity-based the system remains stable at that initial state. So, initially
control and backstepping control respectively. Also, the system was disturbed by applying some random voltage
logic for switching the control between them is explained. and then the above derived control torque input is applied
The control design is divided into two parts due to the to the system.
limitations of the methods we use. The control to make the
pendulum swing up to inverted position is designed using B. Stabilization
passivity-based control and the stabilization around the Backstepping control is used for stabilization of the
inverted position is acquired by backstepping control. The pendulum around inverted position. For this method, we
base for above two controls is given by Lyapunov stability consider the practical dynamics of the system i.e.,
theorem which was stated in Section II.
equations (5), (6) and (7).
A. Swing up In backstepping control, the error of a system state
Passivity-based control is used for swing up of the value with respect to the required state value is traced back
to different states error estimations and finally derive the
pendulum. The passivity-based control is designed using
control torque required to make system states reach the
Theorem 1 in Section II. For this method, we consider the required values.
system without friction i.e., (2)-(4). This control can be Backstepping was designed using the concept of
only used for the passive systems. Lyapunov stability in Section II.
A system is said to be passive if and only if there exists To apply the backstepping control, the system state
a positive storage function (S) such that Ṡ ≤ u y where u equations are to be in strict feedback form.
is the input of the system and y is the output of the system. Let us assume T = T − (B + B )φ̇ (20)
So, the storage functions are assumed such that their State space representation for (7) is,
derivative should be multiple of the input variable i.e., Let us assume x = θ, x = θ̇
torque. So, the below storage functions are assumed such ẋ = x (21)
that one of their derivatives contain torque equivalent from ẋ = − g sin x − ( −( x (22)
) )
(4) and the other from (3). As the above state equations are in strict feedback form,
Let us assume some storage functions S1, S2 we can apply backstepping control.
S = MLg (1 − cos θ) + I θ̇ (10) Let us assume some control u1 exists to make (21)
stable. Then by the definition of the Lyapunov stability
S = I θ̇ − φ̇ (11)
theorem in Section II, we can say that there exists a positive
By differentiating (8) and (9), function V1 such that 𝑉̇ ≤ 0.
Ṡ = MLg (sin θ) θ̇ + I θ̇θ̈ (12) Therefore, if ẋ = u is stable system,
Substituting (4) in (12),
then V̇ = ẋ = u ≤ 0. (23)
Ṡ = (−T )θ̇ (13)
To satisfy above condition, u = − (24)
Ṡ = I θ̇ − φ̇ θ̈ − φ̈ (14)
But ẋ = x in our system
Substituting (3) in (14),
Let e = x − u (25)
Ṡ = (T ) θ̈ − φ̈ (15) Modifying equation (10), (11)
Let us assume system storage function S as ẋ = x − u + u = e + u (26)
S= S − PE + (S ) ≥ 0 (16) ML T B
ė = ẋ − u̇ = − g sin x − − x − u̇
Where PEreq is the required potential energy calculated I (I ) (I )
using required theta. =u (27)
Ṡ = k S − PE Ṡ + k (S ) Ṡ (17) Assuming control input u2 makes our system stable, the
Substituting (10), (11) in (17), Lyapunov function V is taken as

Authorized licensed use limited to: ULAKBIM UASL - GAZI UNIV. Downloaded on June 05,2024 at 12:22:45 UTC from IEEE Xplore. Restrictions apply.
V=V + e ≥0 (28) By using above control torque, we reach x = θ =
∂V dV 180 , x = θ̇ = 0, φ̇ = 0
V̇ = V̇ + eė = ẋ + eu = (e + u ) + eu and substituting above two in (40)
∂x dx
T =0
= u +e +u (29)
Substituting above 4 values in (5), (6),
From equation (23), we can say that first term of V̇ is less θ̈ = 0, φ̈ = 0.
than or equal to zero. Now we have to make the second Hence equilibrium is achieved.
term less than or equal to zero.
C. Switching
e +u ≤0 (30)
Now the most important thing is to know when to
to make above condition true, switch the control between these two logics.
+u = −K e , where K1 > 0 (31) In swing up, the final energies of the system change
such that the value of S1 is around PEreq. Since, S1 is sum
u = −K e − (32) of system potential energy and a part of total kinetic energy,
both energy values vary such that S − PE is around 0.
u = −K (x − u ) − = −K x + − (33) The inverted position is one of the solutions of the equation
By comparing equation (22) with (33), S − PE = 0 i.e., PE equals to PEreq and the kinetic
energy part of S1 equals to zero. To stabilize the pendulum
− g sin x − ( )
−( )
x − u̇ = −K x + in inverted position, Backstepping control is used.
So, the control is switched to Backstepping control
−
from passivity-based control when S1 ≥PEreq to stabilize
=− g sin x − ( x − u̇ + K x + + the system at the inverted position and again switched back
( ) ) to passivity-based control when S1 ≤PEreq.
VI. SIMULATION RESULTS
(34) In this section, the values of the fixed parameters used in
u̇ = (ẋ ) = (x ) = − (x ) (35) simulation are given. The variable parameter plots for the
simulation of the inverted pendulum with the designed
Substituting (35) in (24) control logic are provided and used to explain the working
=− g sin x − ( x + (x ) + of control logic and the system stability and robustness at
( ) )
the inverted position.
K x + + The system is simulated in MATLAB using the
parameters listed in Table1.
Rearranging, we get
Table.1 FIXED SYSTEM PARAMETERS
( )
=− g sin x − ( )
x + K + −
parameter value
(x ) + (1 + K ) Ra 4.1 ohm
( )
(36) La 0.14 mH
kv (or) kt 0.004
Considering V = x −θ , where K > 0 (37)
Bm (1/758) mN-m/rpm
=K x −θ , =K (38) Imr 1.12 g-cm2
mp 0.02 Kg
Substituting (38) in (36),
Lp 12.5 cm
T =I − g sin x − ( )
x + K +K − mr 0.3 Kg
(x ) + (1 + K )K x − θ Rr 2 cm
( ) Bp 0.0001 N-m-s
Let 𝑎 = (1 + K )K , 𝑎 = K + K − ( )
Br 0.0001 N-m-s
ML g 9.8 m/s2
T =I − g sin x + 𝑎 x − θ + 𝑎 (x ) After an initial random voltage disturbance,
I
passivity-based control takes the control and the variable S1
(39)
is made to reach PEreq. From this point, the system was
Substituting (20) in (39), switched to backstepping control which tries to make the
T =I − g sin x + 𝑎 x − θ + 𝑎 (x ) + control torque zero at inverted position by reducing errors
in required angle and angular velocity of the pendulum to
(B + B )φ̇ zero. Hence making angular accelerations zero. Since the
(40) angular acceleration of reaction wheel is zero at that point,
By using above equation, the pendulum is stabilized around the angular velocity of reaction wheel stays at its value at
the equilibrium point. that stabilizing instant. Hence no system state or variable
are unreal and every value stays with in their finite limits.

Authorized licensed use limited to: ULAKBIM UASL - GAZI UNIV. Downloaded on June 05,2024 at 12:22:45 UTC from IEEE Xplore. Restrictions apply.
Hence, the system can be said to be stable in inverted
position.
The simulation of pendulum with derived control logic was
run for 30 sec without any external disturbance and the
results are plotted as shown in Fig.2.

(f)

FIG. 2 VARIABLE PARAMETERS OF REACTION WHEEL PENDULUM

WITH TOTAL CONTROL AND K1=1, K2=0.0000002, K0=8, K1=12

From the results in Fig.2, we can observe that the

(a) pendulum had reached the stable inverted position i.e., the
point where the angle of pendulum is 180ᵒ, the angular
velocity of the pendulum is zero, the angular acceleration
of both pendulum and the reaction wheel are zero. Also, it
is observed that it is stabilized there since all those values
do not vary. Hence, the pendulum was inverted and
stabilized at inverted position.
Now the simulation was run for 180 sec with
external disturbance torque of -3mN-m at 42sec and 5mN-
m at 62sec are given to the system as shown in Fig. 2 to
verify robustness of the control and the results are plotted.
(b)

Fig.3 EXTERNAL DISTURBANCE TORQUE APPLIED TO REACTION

WHEEL PENDULUM
(c)

(a)
(d)

(b)
(e)

Authorized licensed use limited to: ULAKBIM UASL - GAZI UNIV. Downloaded on June 05,2024 at 12:22:45 UTC from IEEE Xplore. Restrictions apply.
 VII. CONCLUSION
In this paper, the reaction wheel pendulum was
described and its dynamics are derived. The control logics
for the inversion and stabilization of the reaction wheel
pendulum were designed using passivity-based control and
backstepping control respectively. The logic for switching
control between the two controls is presented. The designed
control logic was then substantiated using the simulated
results. The robustness of the designed control was verified
using the simulation results obtained by giving the
(c) disturbance torque of 5mN-m and -3mN-m externally to the
inverted pendulum.
REFERENCES

[1] Matthews M.R., Gauld C. & Stinner A. The Pendulum:

Its Place in Science, Culture and Pedagogy. Science &
Education 13, 261–277 (2004).
https://doi.org/10.1023/B:SCED.0000041867.60452.1
8
[2] The Reaction Wheel Pendulum, Block.D.J.,
(d) Åström.K.J., Spong.M.W., 9781598291940,
https://books.google.co.in/books?id=4VcZHPJgafoC,
2007, Morgan & Claypool Publishers.
[3] Fantoni I. and Lozano R., Non-linear Control for
Underactuated Mechanical Systems, Communications
and Control Engineering, Sontag E. and Thoma M.,
Eds., London: Springer-Verlag, 2002.
[4] O. Boubaker, "The inverted pendulum: A fundamental
benchmark in control theory and robotics,"
International Conference on Education and e-Learning
Innovations, Sousse, 2012, pp. 1-6. doi:
10.1109/ICEELI.2012.6360606
[5] Hill D., Zhao J., Gregg R., and Ortega R., “20 Years of
(e) Passivity- Based Control (PBC): Theory and
Applications,” CDC/CCC 2009. Proceedings of the
48th IEEE Conference on Decision and Control,
Workshop on 20 Years of Passivity-Based Control:
Theory and Applications, IEEE Publ., Piscataway, NJ,
Dec. 2009,
http://webpages.lss.supelec.fr/perso/ortega/sli_ro_wks
hp_cdc091.pdf.
[6] Goldstein herbert (1980), classical mechanics (2nd
ed.), addison-wesley. ISBN 978-0-201-02918-5.
[7] Khalil H.K. (2002). Nonlinear Systems (3rd ed.).
Upper Saddle River, NJ: Prentice Hall. ISBN 978-0-
(f)
13-067389-3.
FIG. 4 VARIABLE PARAMETERS OF REACTION WHEEL PENDULUM [8] Sheng Zhang, Wei-qi Qian “Dynamic backstepping
WITH K VALUES AS FIG 2 FOR THE DISTURBANCES SHOWN IN FIG 3 control for pure-feedback nonlinear systems” 27June
When the disturbance was applied to the system, the 2017 arXiv:1706.08641 .
system moves away from the inverted position. If the [9] P.V. Kokotovic, "The joy of feedback: nonlinear and
system energy gets decreased, the passivity-based control adaptive," in IEEE Control Systems Magazine, vol. 12,
makes it reach the near stable state and the pendulum is no. 3, pp. 7-17, June 1992.
again stabilized there by backstepping control else the [10] Mark W. Spong, Peter Corke, and Rogelio Lozano.
backstepping control itself will stabilize the pendulum. 2001. Brief Nonlinear control of the Reaction Wheel
From the results in Fig.4, the above control system can Pendulum. Automatica 37, 11 (November 2001),
be said to be robust for disturbance torques in between 5mN 1845-1851. DOI=http://dx.doi.org/10.1016/S0005-
and -3mN. 1098(01)00145-5.

Authorized licensed use limited to: ULAKBIM UASL - GAZI UNIV. Downloaded on June 05,2024 at 12:22:45 UTC from IEEE Xplore. Restrictions apply.