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

net/publication/330492207

Study on Inertia Wheel Pendulum Applied to Self-Balancing Electric Motorcycle

Conference Paper · November 2018

DOI: 10.1109/GTSD.2018.8595698

CITATIONS READS
3 750

2 authors, including:

Bang Cong Pham

Ho Chi Minh City University of Technology (HCMUT)
24 PUBLICATIONS 613 CITATIONS

SEE PROFILE

All content following this page was uploaded by Bang Cong Pham on 01 September 2021.

The user has requested enhancement of the downloaded file.

 Study on Inertia Wheel Pendulum Applied to Self-Balancing
Electric Motorcycle

Quang Khanh Ho and Cong Bang Pham

Abstract - Self-balancing electric motorcycles have been recently wheeled balancing vehicle is only suitable in small areas such
developed by high-tech corporations to reaffirm their vision of as park, workshop, etc.
“enjoying the freedom of mobility”, i.e. riders will not have to put
The limitations of the two-wheeled balancing vehicle was
their feet on the ground when they stop. A stabilization system
based on Inertia Wheel Pendulum (IWP) is a potential approach
quickly resolved with self-balancing electric motorcycle in
to this new concept motorcycle. The paper focuses on exploring which riders do not have to put their feet on the ground when
effects of design parameters on dynamic performance of IWP. A they stop. In recent years, automotive corporations have
prototype of IWP is implemented. The understanding of IWP is introduced many prototypes of self-balancing electric
validated through simulation results as well as experimental motorcycle, typically from Honda, Lit Motors as demonstrated
results. The findings have shown that the prototype has been in Fig. 2.
capable of self-balancing under a PID control algorithm.

Keywords: Inertia Wheel Pendulum; Reaction Wheel Inverted

Pendulum; Self-Balancing Electric Motorcycle

I. INTRODUCTION
a) In 2016 by Lit Motors [2] b) In 2017 by Honda [3]
Nowadays, together with cutting-edge technologies, small
Figure 2. Self-balancing electric motorcycles
electric portable transport devices have been developed.
Figure 1 illustrates I2 SE invented by Segway. This device has
a weight of 48kg, be able to speed up 20km/h, and transports For the self-balancing electric motorcycles, there are three
up to a distance of 34km. The dynamics of the Segway are approaches using steering control, gyroscopic stabilization, or
similar to a classic control problem, the inverted pendulum. It reaction wheel to create the balance required. In these
uses brushless DC electric motors in each wheel powered by methods, a self-balancing bicycle robot with a reaction wheel
lithium-ion batteries with balance achieved using tilt sensors, has the same concept as an inverted pendulum that is also
and gyroscopic sensors. The wheels are driven forward or known as Inertia Wheel Pendulum (IWP). In comparison with
backward as needed to return its pitch to upright. the first two approaches of stabilization, IWP has some
advantages of simple modelling and structure that have
attracted interests from researchers.

Figure 3. Prototype of Self-Balancing Electric Motorcycle [4]

Figure 1. Self-balancing scooter I2 SE by Segway [1]
Figure 3 illustrates its prototype developed by Mai [4].
With the advantage of compact, flexible, two-wheeled Most researchers mainly focused on control aspects for the
balancing vehicle has been widely used in life. However, it still IWP, e.g. using LQR and MPC controller by
faces some backlashes of low speed, short distance, potential Kanjanawanishkul [5], studying a two-loop control
dangers posting to inexperienced users. Overall, the two- architecture by Huynh [6]. However, effects of design
parameters on dynamic performance of IWP have not been

* Research supported by R&D funding from Ho Chi Minh City University C. B. Pham is a senior lecturer in Mechatronic engineering at the
of Technology. University of Technology, Vietnam National University Ho Chi Minh City.
Q. K. Ho is a senior student in Mechatronic engineering at the University His research interests are in mechanism design, flexible manufacturing
of Technology, Vietnam National University Ho Chi Minh City. His research systems, robotic systems, and rapid prototyping techniques. (e-mail:
interests are in dynamic and control, programing and mechanism design. (e- [email protected]).
mail: [email protected]).
much explored. In this paper, we mainly focus on studying the 𝑑𝑖𝑎
𝑉𝑚 = 𝑖𝑎 𝑅𝑎 + 𝐿𝑎 + 𝐸𝑏 (1)
effects of pendulum length, wheel mass as well as wheel 𝑑𝑡

inertial moment on responses of IWP. This facilitates the

The back electromotive force (emf) is related to the angular
determination of proper design parameters when applying the
velocity 𝜑̇ by the back-emf constant 𝐾𝑏 as:
IWP to given electric motorcycles. The organization of this
paper is as follows: Section II formulates mathematical models 𝐸𝑏 = 𝐾𝑏 𝜑̇ (2)
of the motor and the IWP. This is followed by studying
dynamic performance through responses of inertia wheel Suppose that the effect of inductance is very small (𝐿𝑎 ≪
motion and pendulum angle in Section III. A simple control 𝑅𝑎 ), Eqn. (1) leads to
approach of PID algorithm is applied to the IWP in Section IV 𝑉𝑚−𝐾𝑏 𝜑̇
that is verified through simulation results in Matlab Simulink 𝑖𝑎 = (3)
𝑅𝑎
and validated through experimental results on an IWP
prototype is presented. The paper is summarized in Section V. The motor torque 𝜏 is related to armature current 𝑖𝑎 by a
motor torque constant 𝐾𝑡 as:
𝑉𝑚 −𝐾𝑏 𝜑̇
II. SYSTEM MODELING 𝜏 = 𝐾𝑡 𝑖𝑎 = 𝐾𝑡 (4)
𝑅𝑎
A self-balancing electric motorcycle based on the concept
of an inverted pendulum as illustrated in Fig. 4, is an unstable
and nonlinear system. To stabilize the system, the following B. Modeling Inertia Wheel Pendulum
three main components are required, i.e. (1) an IMU sensor The IWP system consists of a pendulum and an inertia
that detects the tilt angle of the frame, (2) a controller that is wheel as shown in Fig. 6. The pendulum can rotate freely
used to control motion of an inertia wheel through an electric about the Oz axis perpendicular to the Oxy plane. The inertia
motor, and (3) an inertia wheel that is employed to produce wheel is driven by the DC motor to rotate about an axis in
reactionary torque to balance. parallel to the Oz axis. The pendulum can balance by reaction
force generated by the inertia wheel. The parameter notations
1
2 1. Inertia wheel of the IWP are presented in Table 1.
2. Motor y τ
3. Frame
3
4 4. Wheel
.
φ

O
m2, J2
l2
Figure 4. Structure of a self-balancing electric motorcycle O1
l1 m1, J1
In order to study its dynamic response as well as to θ
implement proper control algorithm to this self-balancing
electric motorcycle, mathematical models of the motor and the
inverted pendulum should be derived.
O x
A. Modeling DC motor
Figure 6. Inertial Wheel Pendulum model
Ra La
+ τ,φ TABLE 1. PARAMETER NOTATIONS OF THE IWP
Symbol Unit Definition
θ rad Angle of pendulum
Vm Eb Motor φ rad Angle of inertia wheel
ia
Inertia moment of pendulum including
J1 kg.m2
_ motor stator
Load Inertia moment of wheel including motor
J2 kg.m2
rotor
Figure 5. DC motor equivalent model
c1 N.m.s/rad Friction factor of pendulum
The equivalent electric circuit of the DC motor and the c2 N.m.s/rad Friction factor of wheel
free-body diagram of the load are shown in Fig. 5. Using m1 kg Mass of pendulum and stator
Kirchhoff’s second law leads to: m2 kg Mass of wheel and rotor
 l1 m Length from origin to COG of pendulum III. EVALUATION OF SYSTEM PERFORMANACE
l2 m Length from origin to COG of wheel In this section, the performance of the IWP is explored by
Kb V/(rad/s) Back-emf constant observing its responses when applying step inputs and
changing the value of certain parameters, e.g. l2, J2 and m2.
Kt N.m/A Motor torque constant
The values of the other parameters are kept unchanged as
Ra Ohm Armature winding resistance shown in Table 2.

Mathematical model of IWP is established by Lagrange TABLE 2. COEFFICIENT VALUES IN THE IWP
method. Energy of system consists of kinetic and potential Symbol Value Symbol Value
energy.
J1 0.01186 (kg.m2) l1 0.1053 (m)
 Translational kinetic:
1 1 J2 0.0005711 (kg.m2) l2 0.14 (m)
𝑇1 = 𝑚1 (𝑙1 𝜃̇ )2 + 𝑚2 (𝑙2 𝜃̇)2 (6)
2 2 c1 0.04 (N.m.s/rad) Kb 0.0987 (V/(rad/s))
 Rotational kinetic:
1 1 c2 0.0001 (N.m.s/rad) Kt 0.0987 (N.m/A)
𝑇2 = 𝐽1 𝜃̇ 2 + 𝐽2 (𝜃̇ + 𝜑̇ )2 (7)
2 2 m1 0.826 (kg) Ra 1.5562 ()
Kinetic energy of the system: m2 0.583 (kg)
𝑇 = 𝑇1 + 𝑇2 (8)
1 1
2 2 ̇ 2 ̇
= (𝑚1 𝑙1 + 𝑚2 𝑙2 + 𝐽1 + 𝐽2 )𝜃 + 𝐽2 𝜃𝜑̇ + 𝐽2 𝜑̇ 2 The nonlinear model of IWP in Eqn. (17) is implemented
2 2
into a Simulink model as illustrated in Fig .7 in which:
Potential energy origin at O. So that, potential energy:  The voltage input Vm is the control variable
𝑉 = (𝑚1 𝑙1 + 𝑚2 𝑙2 )𝑔cos(𝜃) (9)  The four state variables are 𝜃, 𝜃̇, 𝜑, and 𝜑̇
 The output torque  is calculated from Eqn. (4)
From equations (8) and (9), Lagrange equation of IWP:
1
𝐿 = 𝑇 − 𝑉 = (𝑚1 𝑙12 + 𝑚2 𝑙22 + 𝐽1 + 𝐽2 )𝜃̇ 2 ..
θ
.
θ θ
2
1
+ 𝐽2 𝜃̇𝜑̇ + 𝐽2 𝜑̇ 2 − (𝑚1 𝑙1 + 𝑚2 𝑙2 )𝑔cos(𝜃) (10) f1(u)
2

Vm 
Friction energy:
.. .
1 1 φ φ φ
𝑅 = 𝑐1 𝜃̇ 2 + 𝑐2 𝜑̇ 2 (11) f2(u)
2 2

Moment at rotate axis of 𝜃:

𝑑 𝜕𝐿 𝜕𝑅 𝜕𝐿
( ) + − =0 (12)
𝑑𝑡 𝜕𝜃̇ 𝜕𝜃̇ 𝜕𝜃
Figure 7. Simulink model of IWP
⇔ (𝑚1 𝑙12 + 𝑚2 𝑙22 + 𝐽1 + 𝐽2 )𝜃̈ + 𝐽2 𝜑̈ + 𝑐1 𝜃̇ (13)
−(𝑚1 𝑙1 + 𝑚2 𝑙2 )𝑔sin(𝜃) = 0 A. Responses under the step input
From the Simulink model, the voltage input Vm is applied
Moment at rotate axis of  : in step of 10V, 15V, and 20V. The voltage input results in
𝑑 𝜕𝐿 𝜕𝑅 𝜕𝐿
motions of the wheel as well as the pendulum shown in Fig. 8.
( )+ − =0 (14) The simulation results show that:
𝑑𝑡 𝜕𝜑̇ 𝜕𝜑̇ 𝜕𝜑
⇔ 𝐽2 (𝜃̈ + 𝜑̈ ) + 𝑐2 𝜑̇ = 𝜏 (15)  In Fig. 8.a, velocity of the wheel gets the steady state
at a same time about 0,5s. In this time, DC motor
𝜏 is the torque motor. From (4) and (15), it has: generates a moment on wheel as in Fig. 8.b and inertia
𝑉 −𝐾 𝜑̇
𝐽2 (𝜃̈ + 𝜑̈ ) + 𝑐2 𝜑̇ = 𝐾𝑡 𝑚 𝑏 (16) wheel rotates with an acceleration as in Fig. 8.c.
𝑅𝑎
Pendulum is pushed by a force and oscillates around
equilibrium position. As we can see from Fig. 8.d, Vm
From (13) and (16), the nonlinear model of IWP is:
higher makes amplitude of pendulum angle higher
𝑐1 0
𝑚1 𝑙12 + 𝑚2 𝑙22 + 𝐽1 + 𝐽2 𝐽2 𝜃̈ 𝐾 𝐾 𝜃̇ which demonstrates higher force generated on
[ ][ ]+ [ 𝑡 𝑏 ][ ] pendulum. However, frequency oscillation of
𝐽2 𝐽2 𝜑̈ 0 + 𝑐2 𝜑̇
𝑅𝑎 pendulum is the same for all situation, it means Vm
−(𝑚1 𝑙1 + 𝑚2 𝑙2 )𝑔𝑠𝑖𝑛(𝜃) 0 does not affect to frequency oscillation of pendulum.
+[ ] = [ 𝐾𝑡 ] 𝑉𝑚 (17)
0 𝑅𝑎
 When velocity of wheel enters steady state, motor
cannot maintain rotor torque for a long time as in Fig.
8.b. In this state, this torque will equal to friction
 torque, acceleration of wheel is zero, pendulum is Length from origin to COG of wheel l2 is one of important
affected only by gravity force and oscillates gradually parameters to design IWP. To examine the response of system,
back to equilibrium position. a signal Vm = 20V is supplied to the DC motor, the value of l2
is changed into 0.14m, 0.5m, and 1m. The other coefficient
values are fixed.
 In Fig. 9.a, 9.b, 9.c, all situations get the same response
which means l2 does not affect on responses of inertia
wheel motion. However, pendulum angle is different
when l2 changed. With the same voltge to motor, l2
higher get amplitude of pendulum angle smaller which
means smaller sum of force on pendulum. In addition,
frequency oscillate is smaller and stable longer.
 Length of pendulum affects to force on pendulum. It is
useful to choose motor fit with l2. Large value of l2
makes negative effect on pendulum angle response.
a) Wheel velocity b) Motor torque

C. Effect of the inertia moment of wheel (J2)

Another important parameter of IWP is inertia moment of
wheel J2. The same as situation with l2, a step input Vm = 20V
will be supplied to DC motor. Value of J2 is changed between
0.0005kg.m2, 0.002kg.m2, and 0.005kg.m2. The other remain
parameters will be unchanged from Table 2.

c) Wheel acceleration d) Pendulum angle

Figure 8. Responses of IWP under the step input Vm

B. Effect of the length from origin to COG of wheel (l2)

a) Wheel velocity b) Motor torque

a) Wheel velocity b) Motor torque

c) Wheel acceleration d) Pendulum angle

Figure 10. Responses of IWP in case changing J2

The simulation results show that:

 In Fig. 10.a, settling time of wheel velocity is longer
when J2 higher and steady state value is the same for
all cases. In short time condition, torque motor is equal
at 0s then extends longer in case J2 higher as in Fig.
c) Wheel acceleration d) Pendulum angle
10.b. In Fig. 10.c, acceleration of wheel is different at
Figure 9. Responses of IWP in case changing l2
0s: inertia moment of wheel smaller has acceleration
 at start moment higher but more quickly converging to designed as in Fig. 12. In the first loop, the pendulum angle is
zero than other. controlled by a PD controller. In the second loop, the wheel
 As a result, in Fig. 10.d, amplitude of pendulum angle velocity is controlled by a P controller. Sum of the two
is higher in case J2 higher which ease to understand controller outputs is Vm voltage supplied to DC motor.
that because of torque motor affect more longer.
 J2 affects positively to accelerated process of motor so θ=0 -
J2 higher is better to remain moment on pendulum but + PD Controller θ
+ Self-Balancing
mass of system increases.
. +
Electric Motorcycle .
φ=0 φ
+
- P Controller
D. Effect of the mass of wheel (m2)
Figure 12. Control strategy for balancing IWP

A. Simulation Results

a) Wheel velocity b) Motor torque

Figure 13. 3D Model of Self-Balancing Electric Motorcycle

A 3D model of self-balancing motorcycle is designed in

Solidworks as illustrated in Fig. 13. Then, the 3D model is
embedded into Matlab environment through SimMechanics
Link. That results in a Simulink block diagram as shown in
Fig. 14, including:
 “Environment” block, which is under the gravity
 “Ground” block representing the road
 “Body Axis” block representing wheel contact line
with the road
c) Wheel acceleration d) Pendulum angle
 “Body” block representing the body of motorcycle
Figure 11. Responses of IWP in case changing m2
 “Wheel Axis” block representing the motor axis
 “Wheel” block representing the inertia wheel
The last parameter to simulate is the mass of wheel m2.
Like l2, J2, a step input Vm = 20V is supplied to DC motor and
value of m2 is changed between 0.5kg, 0.8kg, and 1kg. The
other parameters are unchanged.
The simulation results show that: Figure 14. Simulink block diagram of self-balancing motorcycle
 From Fig. 11.a, 11.b, 11.c, velocity, acceleration,
torque motor on inertia wheel is the same for all cases
of m2. Therefore, mass of wheel does not affect on
responses of above physical parameters.
 From Fig. 11.d, m2 higher gets amplitude of pendulum
angle smaller because of effect of gravity.
 m2 affects negatively to performance of system.

IV. CONTROL IMPLEMENTATION

In normal state, IWP always stable on bottom so it needs a
controller to control torque on origin to balance with torque
produced by gravity. Steady state of system is pendulum angle Figure 15. Simulation results
is zero, velocity of wheel is zero in order to speed up and get a (𝐾𝑃,𝜃 = 900, 𝐾𝐷,𝜃 = 150, 𝐾𝑃,𝜑̇ = 1.7)
maximum torque as well. Therefore, a simple PD controller is
 Simulation result of PD controller is illustrated as Fig. 15.
Pendulum angle θ balances at 00 from the angle deviation 100.
Wheel velocity increases from zero to 130rad/s then enters
steady state zero.
B. Experimental Results
Figure 16 is a prototype of IWP. System main parts consist
of base, encoder to measure angle of pendulum, wheel connect
with DC motor. The electric part to control IWP consists of
Arduino Uno, Driver BTS7960, adapter 24V-10A and a
switch.
a) Responses with initial deviation of 100

a) Physical model b) Electric components

Figure 16. Prototype of IWP

In normal state, pendulum stable at bottom so to move to b) Responses with initial deviation of -100
upper position, swing up process is controlled by a ON-OFF Figure 19: Dynamic performance of IWP to maintain its upward balance
controller with Vs = 15V as in Fig. 17. Results of swing up and
balance controller are shown in Fig. 18. V. CONCLUSION
Vm (V) This paper has explored the principal to balance the IWP.
It has been found that the pendulum is only activated under the
+Vs acceleration of the wheel. In addition, the effects of pendulum
length, wheel mass as well as wheel inertial moment on
responses of IWP have also been studied. The understanding
. of IWP responses are very helpful to determine the stable
θ (rad/s)
region as well as the motor torque needed to achieve the
-Vs
desired balance. Simple control of PID algorithm has been
successfully implemented to IWP. Future work will be more
Figure 17. Control strategy for swing up process focused on control when the IWP is implemented on self-
balancing electric motorcycles.

ACKNOWLEDGMENT
The authors would like to thank the Faculty of Mechanical
Engineering at Ho Chi Minh City University of Technology
for sponsoring this research program.

REFERENCES
[1] http://www.segway.com/products/professional/segway-i2-se.
[2] http://litmotors.com/
[3] http://www.honda.com/mobility/riding-assist
[4] Mai Thị Thu Hà, Thiết kế, chế tạo và thử nghiệm mô hình xe hai bánh
a) Pendulum angle b) Wheel velocity tự cân bằng, Luận văn thạc sĩ, Thai Nguyen University, 2014.
Figure 18. Real hardware performance for swing up and balance [5] Kiattisin Kanjanawanishkul: LQR and MPC controller design and
comparision for a stationary self-balancing bicycle robot with a reaction
wheel. Kbernetika – volume 51(1), pages 173-191, 2015.
The IWP is able to balance from 100 and enters the steady [6] T.S. Nguyen and T.H. Huynh, Study on a Two-loop Control
state as shown in Fig. 19. Architecture to Balance an Inertia Wheel Pendulum. 3rd National
Foundation for Science and Technology Development Conference on
Information and Computer Science, pages 29-33, 2016.

View publication stats