VOL. 6, NO.

3, MARCH 2011 ISSN 1819-6608

ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

DYNAMIC MODELING AND CONTROL OF A TWO WHEELED

ROBOTIC VEHICLE WITH A VIRTUAL PAYLOAD
K. M. Goher1, M. O. Tokhi2 and N. H. Siddique3
1
Department of Mechanical and Industrial Engineering, College of Engineering, Sultan Qaboos University, Oman
2
Department of Automatic Control and Systems Engineering, The University of Sheffield, United Kingdom
3
School of Computing and Intelligent Systems, University of Ulster, United Kingdom
E-Mail: [email protected]

ABSTRACT
One of the challenging issues to consider in balancing a two-wheeled robotic machine (TWRM) is when the load
carried by the machine is changing position along the vehicle intermediate body (IB). An issue of interest in this case is the
resulting impact on the system behaviour due to changing position of the load. Further complications arise with changing
the size of the load. This work presents investigations into controlling a TWRM with a payload positioned at different
locations along its IB. Two types of control techniques are developed and implemented on the system, the traditional
proportional-derivative (PD) control and fuzzy logic (FL) control. PD and PD-fuzzy logic control techniques are developed
to balance the vehicle with a payload incorporating two different scenarios. Firstly, the payload is positioned at different
locations along its IB. Secondly, it is considered to perform a continuous sliding motion along the IB. The balancing of the
robot has to be achieved during the motion of the vehicle and the payload along the IB. An external disturbance force is
applied to the rod which constitutes the IB in order to test the robustness of the developed controllers. Investigations are
carried out on the effect of changing the level and duration of the disturbance force, and changing the speed of the payload
on the system during the balancing mode. Simulation results of both control algorithms are analyzed on a comparative
basis.

Keywords: wheeled robot, dynamic modeling, FL control, PD control, inverted pendulum.

1. INTRODUCTION recent challenges in such applications is the development

The research on balancing two-wheeled robots of control schemes to help a disabled or an elderly person
has gained momentum over the last decade in a number of on a two wheeled wheelchair [30] to move to further
robotics laboratories around the world. This is due to the levels in shopping centers or to have eye to eye contact
inherent unstable dynamics of such systems. The control with others. Industrial applications of such vehicles will
quality of such robots is characterized by the ability to arise to a great extent in the coming years; for instance,
balance on its two wheels and provide spin on the spot. material handling in narrow paths to different heights etc
This additional maneuverability allows easy navigation on (Goher and Tokhi, 2008).
various terrains, turn at sharp corners and traverse small The idea of balancing a robot on two wheels is
steps or curbs. These capabilities have the potential to based on the concept of inverted pendulum model. This
solve a number of challenges in the industrial and public model has been widely used by researches around the
sectors. For example, a motorized wheelchair utilizing this world in controlling not only wheeled robots but other
technology would give the operator greater types of robot as well such as legged robots. In recent
maneuverability and thus access to places most able- years, researchers have applied the idea of a mobile
bodied people take for granted. Small carts built utilizing inverted pendulum model to various problems like
this technology allow humans to travel short distances in a designing walking gaits for humanoid robots, robotic
small area or factories as opposed to using cars or buggies wheelchairs and personal transport systems.
which are more polluting. The rapid increase of the aged In the work carried out by Shiroma et al., (1996)
population in countries like Japan has prompted on ‘Cooperative behaviour of a wheeled inverted
researchers to develop robotic wheelchairs to assist an pendulum for object transportation’, the interaction of
elderly to move around (Takahashi et al., 2000). forces between objects and the robot has been shown by
Mobile wheeled inverted pendulum models have taking into account the stability effects due to these forces.
evoked a lot of interest recently and at least one Ooi (2003) implemented an indirect Kalman filter
commercial product, Segway [28]. Such vehicles are of configuration combining a piezo rate gyroscope sensor and
interest because they have small footprint and possess high an inclinometer to obtain an accurate estimate of the tilt
navigation capabilities. There are many robots without angle and its derivative. He examined the suitability of
manipulation capabilities that balance around two wheels linear control systems like the linear quadratic regulator
such as JOE robot [29]. The nature of this two-wheeled and pole-placement controller in stabilising the system.
vehicle poses several interesting control questions. For Browning (2004) presented a new domain, called Segway
instance, while a person occupies the vehicle, his/her mass Soccer, for investigating the coordination of dynamically
changes the centre of gravity of the vehicle which in turn formed, mixed human-robot teams within the realm of a
has an impact on the control technique used. One of the team task that requires real-time decision making and

7
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

response. Salerno and Angeles, (2004), presented a The aforementioned researches have concentrated
multivariable control by differentiable state feedback of on developing control algorithms to keep the two-wheeled
semi-autonomous self-balancing two-wheeled quasi- inverted pendulum robot in a balancing state. Other works
holonomic robots. The analysis of robot stability, based on have discussed the dynamics of the system. All these
the Lyapunov linearization method, was provided as well. works have considered a fixed position for the load, which
Pathak (2005) analyzed the dynamic model of a wheeled is mainly the global mass of the robot or the rod attached
inverted pendulum (e.g., Segway, Quasimoro, and Joe) to the axle of the driving wheels. The load is considered to
from a controllability and feedback lineariability point of be concentrated at the centre of mass (COM) of the whole
view. Kim (2005) discussed the enhancement of floor- system. The work presented in this paper considers
driving ability of a two wheeled inverted-pendulum-type challenging control solutions of balancing a two-wheeled
autonomous vehicle based on the dynamics of the robotic robotic machine with changing position and size of the
mechanism using Kane’s dynamic modeling. Randal load.
(2005) applied the idea of non-linear control strategy, A system with two-wheeled vehicle is considered
where two types of semiconductor sensors were used to to be highly non linear and under actuated. Such systems
provide tilt information to the robot to balance. The are considered to be utilized for outdoor environmental
sensors used were gyroscopes and accelerometer. usage where many unexpected conditions could apply to
Most of the previously mentioned works showed the vehicle and affect its performance. These include the
the practical validity of the robot from controllability presence of sources of disturbances such as sudden impact
perspective. However, not much work on the dynamics of forces or obstacles during maneuvering of the vehicle. The
this kind of robot has been reported. One of the most uncertainties due to such effects need to be considered
important contributions discussing the dynamics of two- when designing control strategies for outdoor machines.
wheeled inverted pendulum robot is the work carried by Poorly modeled behaviour of loads carried by the vehicle
Kim (2005). The exact dynamics have been investigated in is another issue which needs to be considered when
order to provide necessary information for further design dealing with such systems. The aim of this work is to
and production. Tsai (2006) developed techniques for develop a simple, effective and robust control approach
system design, modeling and adaptive control of a capable of balancing a TWRM under the effect of a
personal two-wheeled transporter driven by two DC disturbance force varying in magnitude, position and
servomotors. A mechatronic system structure for the duration. The control scheme needs to be able to cope with
vehicle was described and its mathematical modeling changes in the speed and position of an attached payload.
incorporating the friction between the wheels and motion Furthermore, an investigation of the impact of changing
surface was derived as well. Two adaptive two-degrees-of- the length of the intermediate body is considered in this
freedom (DOF) controllers were designed to achieve self- study.
balancing and rotation control. Experimental results The rest of the paper is organized as follows:
revealed that the proposed controllers were capable of Section 2 describes the system dynamics considering a
providing appropriate control actions to steer the vehicle disturbance effect applied to the IB and the frictional
in a desired manner. Hu and Tsai (2006) investigated resistance at the joint connecting the wheel axle and the
robust motion pilot control of an auto-balancing two- IB. Section 3 proposes a control strategy utilizing a
wheeled cart where the feedback control system was conventional PD control scheme to test the vehicle model
solely based on the angle of inclination without using any under certain varying dynamic conditions followed by an
velocity information. H∞ control was employed to design implementation of a PD-like fuzzy control approach for
a robust stabilizing controller. further investigation of a suitable control scheme. A
The type of intelligent robot proposed in this numerical simulation analysis is carried out in section 4
work is a mobile robot with a two-wheeled inverted based on the two suggested control techniques. The paper
pendulum. This design was chosen because its mechanism is concluded in section 5 highlighting the achievements of
has an innately clumsy motion for stabilizing the robot’s the work.
body posture. The robot has a body with two wheels for
moving in a plane and a head similar to a human head for 2. MODELING OF THE TWRM
controlling the motion. Two independent driving wheels The TWRM system considered in this work
are used for position control, for fast motion in a plane comprises a rod on an axle incorporating two wheels as
without casters. Two-wheeled machines have different described in Figure-1. The robot is powered by two DC
applications due to their advantages which arise from their motors driving the vehicle wheels. A payload is attached
special design. For example, a two wheeled vehicle may to the IB of the vehicle. This payload is considered to act
be safer for the occupants while simultaneously being at different positions along the IB in static and dynamic
more agile to navigate narrow city streets. Furthermore, manners. The payload is considered to simulate the weight
the reduced volume and lower mass of this configuration of a person on a wheelchair or an object to be handled to
would increase fuel efficiency and overall functionality. different heights if a manual handling task is given to the
However, because such a vehicle would be inherently robot. Rigid body dynamics are assumed in this study. A
unstable it would require an intelligent control mechanism reference Cartesian coordinate frame designated as OXYZ
to provide dynamic balancing. attached to the axle connecting the wheels with its origin

8
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

located at the vehicle centre point O as shown in Figure-1 wheels of the robot, where H FR and H FL are the frictional
is used for the angular and translational motion of the forces between the wheel and the surface. H TR , H TL ,
vehicle. The Z − axis points vertically upward, the
VTR and VTL are the horizontal and vertical reaction forces
X − axis is parallel and coincides with the axle of the
wheels, and the Y − axis is determined according to the respectively. H R , H L , PR , PL are the horizontal and
right-hand rule in the rectangular coordinate system. The vertical reaction forces between the IB and the wheel axle
IB is considered to be balanced if it coincides with the at the connecting joint.
positive Z − axis. Partial angular deviation from the
Z − axis causes an imbalance for the vehicle with a tilt
angle θ P from the Z − axis. The two main dynamic modes
of the vehicle which will be considered in both the
mathematical modelling and control processes are the
angular deflection of the IB from the Z − axis and the
linear motion of the entire system in the Y − direction.

Upright position
Payload
θP Z Figure-2. Free body diagrams of the vehicle wheels.

Right wheel 2.1 Dynamic model of TWRM

The dynamic model of the system is described in
Y this section by introducing the governing differential
equations describing the dynamics of the two-wheeled
Left wheel vehicle. The model is derived based on the Newton-Euler
equations of motion. The dynamic model of the entire
O system consists of two separate sub-models, namely: the
non-linear model of the inverted pendulum and a linear
model of the permanent magnet direct current motor
X activating the wheels.
The dynamics of vehicle wheels and intermediate
Figure-1. Schematic diagram of the two-wheeled vehicle. body are initially analysed separately and this will lead to
the following two non-linear equations of motion. The tilt
The pendulum and wheel dynamics are initially angle acceleration θ&& and vehicle acceleration Y&& ,
P
analyzed separately, and this will eventually lead to two describing the main two dynamic behaviours of the
equations of motion which completely describe the
TWRM, are obtained as
behaviour of the TWRM. Consider Figure-2, which
represents the free body diagrams of the right and left

1 ⎛ kmke k
θ&&P = ⎜− ( y& L + y& R ) + m (V L + V R ) + M F
(I eq + (M + M p )L g ) ⎝ RR w
2 ⎜ 2
RR w
(1)
⎞
(M + M P )L g g sin θ p + (M + M P )L g Y&& cos θ p + F cos θ P (S − 2 L g )⎟⎟
⎠
and

1 ⎛ kmke & k
Y&& = ⎜⎜ − 2
Y + m (V L + V R )
⎛ ⎛ J ⎞ ⎞ RR w RR w
⎜ 2 ⎜ M w + wy2 ⎟ + (M + M p )⎟ ⎝
⎜ ⎜ R w ⎟⎠ ⎟ (2)
⎝ ⎝ ⎠

(M + M P )Lg θ&&p cos θ p + (M + M P )Lg θ&p2 sin θ p − F ⎞⎟⎟

⎠

Definition of the variables and constants is given VR of the left and right wheels are defined in terms of
in the nomenclature. The applied terminal voltage VL and motor torques TL and TR as follows:

9
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

R ⎛ k k ⎞ Below such limit, it was found that the developed

VL = ⎜⎜ TL + m e y& L ⎟⎟ (3)
controller would not completely keep the IB at a stable
km ⎝ RRw ⎠ position and result in an accumulated error in the
corresponding measurements.
R ⎛ k k ⎞
VR = ⎜⎜ TR + m e y& R ⎟⎟ (4)
km ⎝ RRw ⎠ -
Yd k1
+ Fd
The final torque required for the tilt angle Desired
vehicle Controller

acceleration θ&&P and linear vehicle acceleration Y&& is displacement

d
k2 θP Z
expressed as dt
τ1
Ym
Y
R ⎛ k k ⎞ R ⎛ k k ⎞ (5)
VL + VR = ⎜⎜ TL + m e y& L ⎟⎟ + ⎜⎜ TR + m e y& R ⎟⎟ d τ2
km RRw k4 X O
θm
⎝ ⎠ km ⎝ RRw ⎠ dt

Controller

Detailed descriptions of linear model of the DC θ d k3

motors, non-linear model of the wheels and the IB Desired +
-
Tilt angle
dynamics are provided in the Appendix.

3. CONTROL STRATEGY Figure-3. Implementation of a PD control strategy.

The intermediate body is considered to be
balanced at the upright position if it coincides with the 3.1 Fuzzy control of TWRM
positive Z −axis. Partial angular deviation from the Z −axis Conventional control techniques require accurate
causes an imbalanced mode for the vehicle with a tilt mathematical models to describe the dynamics of the
angle θ P from the Z − axis. The main two dynamic system under study. One of the major limitations of
activities of the vehicle which will be considered in both conventional control systems is their inability to cope with
the mathematical modelling process and implementation changes in the plant dynamics with time and actuator
of the control technique for the angular deflection of the saturation, which add to the nonlinearity of the system
IB from the Z − axis and the linear motion of the entire (Lin et al., 1996). These techniques result in tracking
system in the Y − direction. The IB is considered initially error, when the payload varies fast, and overshoot during
in the upright position; θ P = 0 . With a disturbance force transients. An interesting alternative that could be
investigated is the use of fuzzy logic control (FLC)
F d applied to the rod, the IB will move from the upright methods (Mamdani and Assilian, 1975, Zadeh, 1965,
position with an angle, measured from that position, 1973). In recent years, FLC has attracted considerable
directly proportional to the magnitude and direction of the attention as a tool for novel control approaches because of
disturbance force. the variety of advantages that it offers over classical
The control strategy is based on minimizing the control techniques (Zadeh, 1973). The FLC paradigm has
error between the actual angular position θ m of the IB and been developed as an alternative to conventional model-
the desired (stable) angular position θ d , which is zero. based control systems (Mamdani, 1974). It does not
require a mathematical model of the plant and can be
The two control signals resulting from both feedback applied equally to linear and nonlinear systems.
loops are added to constitute the control signal in terms of It is evident that human knowledge is becoming
the motor terminal voltage required to balance the vehicle more and more important in control systems design due to
at the target position (Figure-A2 in the appendix). the fact that many industrial processes are highly nonlinear
The control algorithm is implemented first using and have un-modeled plant dynamics and uncertainties.
a conventional PD controller, shown in Figure-3, for This experiential perspective in controller design requires
testing the model. For purposes of comparison, the process the acquisition of heuristic and qualitative, rather than
is repeated using a PD-like fuzzy control algorithm, as quantitative, knowledge or expertise from the human
shown in Figure-4. For both control strategies, the operator. During the past several years, fuzzy control has
measured error and its rate of change for both the IB tilt emerged as one of the most active and powerful areas for
angle and the vehicle position are used as control inputs, research in the application of such nonlinear real world
while the controller output is the required motor terminal systems using fuzzy set theory (Zadeh, 1973). Fuzzy
voltage V a . control approach has been applied very successfully in
The PD controller parameters were tuned using a practical problems, especially when conventional control
trial and error method to achieve a desired system techniques do not yield satisfactory performance. The
performance. A disturbance force F d is applied on the IB great advantage of fuzzy controllers is the ability to
after 10 seconds from the start of the simulation and then introduce the knowledge of human experts about proper
at each 90 second interval. This interval was chosen as a and correct control of a plant in the controller (Piegat,
safe limit for repeating such impact forces to the IB. 2001, 2006; von Alrrock, 1995).

10
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Fuzzy-based control has been used extensively in

the field of inverted pendulum-like machines and wheeled The control inputs are the error and the derivative
robotic machines. Shi et al., (2004) implemented a of error for the two measured variables of the system
predictive fuzzy control technique on a two wheeled robot Ym and θ m whereas the control outputs are the terminal
for navigation. Hladek (2007) designed and implemented a
voltages of the DC motors. As the type of motion
multi-agent fuzzy expert system for robotic soccer control.
considered in this study is linear, the two control signals
Astudillo et al., (2006) developed a tracking controller for
from both the feedback loops are identical and hence the
the dynamic model of unicycle mobile robot, based on
motor terminal voltage considered as the system input is a
fuzzy logic theory, by integrating a kinematic controller
summation of both the voltages to the left and right
and a torque controller. Maravall et al., (2005) constructed
wheels. The tilt angle error and its time derivative are
a hybrid fuzzy control system that incorporates PD control
defined as:
into a Takagi-Sugeno fuzzy control structure for
stabilizing an inverted pendulum via a vertical force. In eθ = θ d − θ m (6)
the work presented by Saifizul et al., (2006), a Takagi-
Sugeno fuzzy controller with adaptive neuro-fuzzy d (eθ ) θ (k ) − θ (k − 1) (7)
e&θ = =
inference system (ANFIS) architecture has been used to dt ∆t
affirm the stability condition of a self-erecting single
inverted pendulum. Sun and Er (2004) proposed a hybrid Similarly, displacement error and its derivative
fuzzy controller for robotic systems by combining a fuzzy are defined as:
gain scheduling method and a fuzzy proportional-integral- eY = Y d − Y m (8)
derivative (PID) controller to solve nonlinear control
problems such as pole balancing and multilink robot d (eY ) Y (k ) − Y (k − 1) (9)
manipulation. e&Y = =
dt ∆t
For the required linear displacement of the cart
wheels, sufficient torque needs to be applied at the wheels. Where θ d is the desired tilt angle, θ m is the measured tilt
A fuzzy controller can be designed to produce the torque
angle of IB, Yd is the desired linear vehicle position, and
using the error and change of error of the vehicle
displacement. However, there is still the effect of the Ym is the measured vehicle position.
disruption applied to the IB with the disturbance force. As Five linguistic variables for the error and
the torque produced by the fuzzy controller will not be derivative of error, shown in Table-1, are chosen for each
enough for achieving the upright position of the IB, an inputs and outputs: negative big (NB), negative small
additional torque needs to be produced to bring the rod (NS), zero (Z), positive small (PS), and positive big (PB).
back to the upright position using the tilt angle information Triangular membership functions (MFs) are chosen for
as input to the fuzzy controller. This will result in a multi- each linguistic variable.
input multi-output (MIMO) fuzzy controller, which will
incur a huge time consuming rule-base. Therefore, for Table-1. FLC rule base.
simplicity and reducing the processing time, the fuzzy
controller is split into two PD-like fuzzy controllers, as Change of error
shown in Figure-4, utilizing the error and the derivative of Error NB NS Z PS PB
error for both the measured tilt angle of IB and linear NB NB NB NB NS Z
displacement of the vehicle. This will reduce the rule-base NS NB NB NS Z PS
drastically and the associated processing time. The Z NB NS Z PS PB
controller outputs τ 1 and τ 2 are added into a final
PS NS Z PS PB PB
PB Z PS PB PB PB
output as shown in Figure-4.
Triangular membership functions are chosen for
Desired
- Fd
inputs and outputs. The membership functions for IB angle
Tilt angle
θd e&θ FLC error, derivative of IB angle error, vehicle position error,
d
k1 τ1 θm derivative of vehicle position error and voltage input are
+ dt θP Z
k5 (IB Tilt angle) shown in Figure-5. Normalized universes of discourse are
k2
eθ Y used for signals. The parameters of the two PD-like FL
eY
k3 X controllers k1 , k 2 k 3 and k 4 are selected and tuned to
O
Ym
k6
+ d
k4 τ2 (Vehicle achieve a desired system performance. Scaling factors k 5
yd dt e&Y position)
Desired
FLC and k 6 are selected such that they activate the system to
position
-
generate the desired output. To construct a rule base, the
IB angle error, the derivative of IB angle error, the vehicle
Figure-4. Schematic description of the fuzzy logic position error, the derivative of vehicle position error and
control scheme.

11
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

the voltage input are partitioned into five primary fuzzy sets.
NB NS Z PS PB NB NS Z PS PB
1 1

0.8 0.8
Degree of membership

Degree of membership
0.6 0.6

0.4 0.4

0.2 0.2

0 0

-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1

Tilt angle error Tilt angle derivative of error
(a) (b)
NB NS Z PS PB NB NS Z PB
1 PS
1

0.8 0.8
Degree of membership

Degree of membership

0.6 0.6

0.4 0.4

0.2 0.2

0 0

-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1

Vehicle position error Vehicle position derivative of error

(c) (d)
Figure-5. Membership functions of the fuzzy logic controllers.

4. SIMULATION AND ANALYSIS OF RESULTS

Simulations were carried out on the system using Table-3. Two-wheeled inverted pendulum
the numerical parameters describing the system features as mechanical properties.
listed in Tables 2 and 3.
Parameter Description Units
Table-2. DC motor technical specifications. R w = 0 .1 Wheel radius m
M w = 1 Wheel mass kg
Parameter Description Units
R=3 Nominal terminal resistance Ohms M p = 5 Pendulum mass kg

L=0 Rotor inductance H M = 10 Payload kg

Ke = 0.0061 Back EMF constant Vs / rad M a = 0 .7

Wheels connecting axle mass kg
V e = 0 Back EMF voltage Volts
M a = 0 .5
Ir = 0 Rotor inertia Motor mass kg
kgm 2
M gearbox = 0.35
Reduction gear box mass kg

l = 0 . 25 Length to the pendulum centre m

of mass centre of mass

The simulations were conducted to highlight the

way the vehicle will behave with the effect of changing

12
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

the values of different variables and the control effort

The location of the global centre of mass of the
required to accomplish the required task. Investigation is
IB will be affected as:
carried out with the effect of the following variables:
ƒ The level of the applied disturbance force; force Lg =
(M l + QM )
p (12)
amplitude (M + M )
p

ƒ The duration of the disturbance force; force duration

ƒ The position of disturbance force, 4.1.1 Effect of different disturbance levels
ƒ Position and speed of the payload. The effect of changing the level of the applied
ƒ IB length. disturbance force is considered in this section. Four levels
of disturbance amplitudes including zero are considered to
4.1 Payload static behaviour test the validity of the developed controller in coping with
Let the payload, M be located at a distance Q external disturbances. The simulations are carried out by
from the IB origin O , as shown in Figure-A2 (Appendix), considering the disturbance force applied at different
where Q can be expressed as a factor y multiplied by the positions on the IB of the vehicle. Such assumptions may
be useful in applications such as extended height of crane
rod half length l as: arms if subjected to sudden external forces through their
Q = yl (10) stems especially during motion.
Figures 6 to 11 show the system performance
Where 0 ≤ y ≤ 2 according to the position of the with the disturbance force applied at the upper end, mid-
span and lower end of the IB. It can be noted from Figures
payload M from the origin of the IB. Changing the
6(a), 7(a), 8(a), 9(a), 10(a) and 11(a) that the higher the
payload position will affect both the overall moment of
level of the disturbance force the longer the fluctuation of
inertia, I g and the location of the global centre of mass of the rod angle from the upright position. It can also be
the IB, L g . The overall moment of inertia of the IB is noted that the closer the location of the applied disturbance
force to the IB origin the harder it is to balance the system
affected and modified as: as it takes the control system a long time to bring the rod
M p (2l )
2 to the target position.
+ M p (L g − l ) + M (Q − L g ) (11)
2 2
Ig =
12

Disturbance applying at the pendulum upper end

Disturbance applying at the pendulum upper end
0.3 0.7
Disturbance amplitude = 0 N

0.2 Disturbance amplitude = 30 N 0.6

Disturbance amplitude = 60 N

0.1 Disturbance amplitude = 100 N 0.5

Cart displacement, m
Pendulum angle, deg

0 0.4

-0.1 0.3

0.2 Disturbance amplitude = 0 N

-0.2
Disturbance amplitude = 30 N

-0.3 0.1 Disturbance amplitude = 60 N

Disturbance amplitude = 100 N
-0.4 0
0 50 100 150 200 0 50 100 150 200
time, sec time, sec

(a) IB tilt angle (b) Cart displacement

13
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Disturbance applying at the pendulum upper end

4
Disturbance amplitude = 0 N
Disturbance amplitude = 30 N
3
Disturbance amplitude = 60 N

Motor terminal voltage, volt

Disturbance amplitude = 100 N
2

-1

-2
0 50 100 150 200
time, sec
(c) Motor terminal voltage
Figure-6. System performance and control signal (Upper end disturbance application); y = 2, PD control.

The cart linear displacement is presented in Changing the level of the disturbance force also
Figures 6(b), 7(b), 8(b), 9(b), 10(b) and 11(b). It is noted affected the control effort, as clearly shown in Figures
that increasing the level of the applied disturbance tends to 6(c), 7(c), 8(c), 9(d), 10(d) and 11(d). Higher levels of the
decrease the cart overshoot beyond the specified limit. applied force tended to increase the time the control signal
That is because the amount of the applied force acts as a takes to settle down. The location of the disturbance force
drag force for the cart and slows the cart which in turn also greatly affected the control behaviour. The closer the
increases the rise time of the system response. This position of the force to the rod origin the more the control
phenomenon is clearly repeated when the force is effort fluctuation and the higher the control effort at the
reapplied on the system. The lower the level of the time of applying the force.
disturbance forces the faster the cart in achieving the
desired position.

Disturbance applying at the pendulum mid-span Disturbance applying at the pendulum mid-span
1.5 0.7

1 0.6

0.5
Pendulum angle, deg

0.5
cart displacement, m

0.4
0
0.3
-0.5
Disturbance amplitude = 0 N 0.2 Disturbance amplitude = 0 N
Disturbance amplitude = 30 N
-1 Disturbance amplitude = 30 N
Disturbance amplitude = 60 N 0.1 Disturbance amplitude = 60 N
Disturbance amplitude = 100 N
Disturbance amplitude = 100 N
-1.5
0 50 100 150 200 0
time, sec 0 50 100 150 200
time, sec
(a) IB tilt angle
(b) Cart displacement

14
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Disturbance applying at the pendulum mid-span

4
Disturbance amplitude = 0 N
Disturbance amplitude = 30 N
3
Disturbance amplitude = 60 N

Motor terminal voltage, volt

Disturbance amplitude = 100 N
2

-1

-2
0 50 100 150 200
time, sec

(c) Motor terminal voltage

Figure-7. System performance and control signal (Mid-span disturbance application); y =1, PD control.

Disturbance applying at the pendulum lower end Disturbance applying at the pendulum lower end
3 0.7
Disturbance amplitude = 100 N
Disturbance amplitude = 60 N 0.6
2
Disturbance amplitude = 30 N
Disturbance amplitude = 0 N 0.5
Cart displacement, m
Pendulum angle, deg

1
0.4
0
0.3

-1
0.2 Disturbance amplitude = 0 N
Disturbance amplitude = 30 N
-2 0.1 Disturbance amplitude = 60 N
Disturbance amplitude = 100 N
0
-3 0 50 100 150 200
0 50 100 150 200 time, sec
time, sec

(a) IB tilt angle (b) Cart displacement

Disturbance applying at the pendulum lower end
4
Disturbance amplitude = 0 N
Disturbance amplitude = 30 N
3
Disturbance amplitude = 60 N
Motor terminal voltage, volt

Disturbance amplitude = 100 N

2

-1

-2
0 20 40 60 80 100 120 140 160 180 200
time, sec

(c) Motor terminal voltage

Figure-8. System performance and control signal (Lower end disturbance application); y = 0, PD control.

15
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

0.05 0.7
FLC FLC
0 PD PD
0.6

Cart displacement, m
-0.05 0.5
Tilt angle, deg

-0.1 0.4

-0.15 0.3

-0.2 0.2

-0.25
0.1

-0.3
0 20 40 60 80 100 0
0 20 40 60 80 100
time, sec
time, sec
(a) IB tilt angle (b) Cart displacement
0.035 8
FLC FLC
7 PD
0.03 PD
6
0.025
Control effort, volt

5
Cart velocity, m/s

0.02 4

0.015 3

2
0.01
1
0.005
0
0
-1
-0.005 -2
0 20 40 60 80 100 0 20 40 60 80 100
time, sec time, sec
(c) Cart velocity (d) Control effort
Figure-9. Simulation results (F = 0 N).

16
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

0.6 0.7
FLC FLC
PD 0.6 PD
0.4

Cart displacement, m
0.2 0.5
Tilt angle, deg

0 0.4

-0.2 0.3

-0.4 0.2

-0.6 0.1

-0.8 0
0 20 40 60 80 100 0 20 40 60 80 100
time, sec time, sec
(a) IB tilt angle (b) Cart displacement
0.035 8
FLC FLC
0.03 7 PD
PD
0.025 6
Control effort, volt

0.02 5
Cart velocity, m/s

0.015 4

0.01 3

0.005 2

0 1

-0.005 0
-0.01 -1
-0.015 -2
0 20 40 60 80 100 0 20 40 60 80 100
time, sec time, sec
(c) Cart velocity (d) Control effort
Figure-10. Simulation results (F = 30 N, t = 0.1 sec, Q =50cm).

17
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

2 0.7
FLC FLC
1.5 PD PD
0.6

Cart displacement, m
0.5
Tilt angle, deg

0.5
0.4
0
0.3
-0.5

0.2
-1

-1.5 0.1

-2 0
0 20 40 60 80 100 0 20 40 60 80 100
time, sec time, sec
(a) IB tilt angle (b) Cart displacement
0.04 8
FLC FLC
7
0.03 PD PD
6
0.02
Control effort, volt

5
Cart velocity, m/s

0.01
4
0
3
-0.01
2
-0.02
1
-0.03
0
-0.04 -1

-0.05 -2
0 20 40 60 80 100 0 20 40 60 80 100
time, sec time, sec
(c) Cart velocity (d) Control effort
Figure-11. Simulation results (F = 100 N, t = 0.1 sec, Q =50cm).

4.1.2 Effect of disturbance duration For the cart linear displacement, as described in
The duration of applying such disturbances is Figures 12(b), 13(b) and 14(b) the situation was not as
another variable of interest. Four different durations of the good as for the tilt angle; increasing the duty cycle of the
disturbance force are used to estimate the safe limits force made the system slower in achieving the target and
beyond which the developed controller will not be able to increased the rise time and decreased the cart overshoot.
cope. But for shorter duty cycles, the overshoot tended to be
Figures 12(a), 13(a) and 14(a) show the IB tilt higher which is a characteristic of such impulse
angle, with different duty cycles of disturbance force of disturbance force but accelerated the system behaviour to
100 N and 30 N. It is noted that the overshoot in the IB tilt reach the desired position.
angle increased as the period of the force was increased, The control effort is presented in Figures 12(c),
and the controller still coped well with such higher periods 13(d) and 14(d). It is noted that the longer the disturbance
of time especially for the pendulum tilt angle. The system duty cycle the higher the control effort required and the
could recover faster to the desired position with shorter longer the time for the control signal to settle down.
duty cycles of the force.

18
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Disturbance amplitude = 100 N Disturrbance amplitude = 100 N

0.3 0.7
duty cycle = 1 sec
duty cycle = 0.5 sec 0.6
0.2
duty cycle = 0.25 sec
duty cycle = 0.1 sec

Cart displacement, m
0.1 0.5
Pendulum angle, deg

0 0.4

-0.1 0.3

-0.2 0.2
duty cycle = 1 sec
duty cycle = 0.5 sec
0.1 duty cycle = 0.25 sec
-0.3
duty cycle = 0.1 sec
0
-0.4 0 50 100 150 200
0 50 100 150 200
time, sec
time, sec

(a)IB tilt angle (b) Cart displacement

Disturbance amplitude = 100 N
4
duty cycle = 1 sec
duty cycle = 0.5 sec
3
duty cycle = 0.25 sec
Motor terminal voltage, volt

duty cycle = 0.1 sec

2

-1

-2
0 50 100 150 200
time, sec

(c) Motor terminal voltage

Figure-12. System performance and control signal (Different disturbance duty cycles), PD control.

19
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

0.8
FLC
0.6 PD
0.5
0.4

Cart displacement, m
0.4
Tilt angle, deg

0.2

0 0.3

-0.2
0.2
-0.4

0.1
-0.6 FLC
PD
-0.8 0
0 50 100 150 0 50 100 150
time, sec time, sec
(a)IB tilt angle (b) Cart displacement
0.035 8
FLC FLC
0.03 PD PD
6
0.025
Control effort, volt
Cart velocity, m/s

0.02 4

0.015
2
0.01

0.005 0
0
-2
-0.005

-0.01
0 50 100 150 -4
0 50 100 150
time, sec
time, sec
(c) Cart velocity
(d) Motor terminal voltage
Figure-13. Simulation results (F = 30 N, t = 0.1 sec, Q =15cm).

20
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

1.5 0.7
FLC FLC
PD PD
1 0.6

Cart displacement, m
0.5
0.5
Tilt angle, deg

0.4
0
0.3
-0.5
0.2

-1
0.1

-1.5 0
0 20 40 60 80 100 0 20 40 60 80 100
time, sec time, sec
(a) IB tilt angle (b) Cart displacement
0.04 8
FLC FLC
PD 7 PD
0.03
6
0.02
Control effort, volt

5
Cart velocity, m/s

0.01 4

3
0
2
-0.01 1

0
-0.02
-1
-0.03
0 20 40 60 80 100 -2
0 20 40 60 80 100
time, sec
time, sec
(c) Cart velocity
(d) Motor terminal voltage
Figure-14. Simulation results (F = 30 N, t = 0.2 sec, Q =15cm).

4.1.3 Effect of payload position global centre of mass and the whole moment of inertia of
The third variable of interest is the location of the the pendulum as presented earlier.
payload attached to the IB. The system behaviour Three different levels of the disturbance force are
including the IB tilt angle and the cart linear displacement implemented here with the assumption of the force at the
are presented and the control effort required for three upper end of the rod. Figures 15 to 18 show the system
different locations of the payload. The load is positioned at performance with different load positions and disturbance
upper, mid-span and the lower end of the IB. Changing the forces of 30 N, 100 N and 0 N respectively.
location of the load mainly affects the location of the

21
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Disturbance amplitude = 30 N, Payload = 50 Kg, Z = 2l Disturbance amplitude = 30 N, Payload = 50 kg, Z = 2l

8 0.7
y=0 y=0
6 y=l y=l
0.6
y = 2l y = 2l
4
0.5

Cart displacement, m
Pendulum angle, deg

2
0.4
0
0.3
-2
0.2
-4

-6 0.1

-8 0
0 50 100 150 200 0 50 100 150 200
time, sec time, sec

(a) IB tilt angle (b) Cart displacement

Disturbance amplitude = 30 N, Payload = 50 kg, Z = 2l
5
y=0
y=l
4
y = 2l
Motor terminal voltage, volt

-1

-2
0 50 100 150 200
time, sec

(c) Motor terminal voltage

Figure-15. System performance and control signal (F = 30 N), PD control.

22
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Disturbance amplitude = 100 N, Payload = 50 kg, Z = 2l Disturbance amplitude = 100 N, Payload = 50kg, Z = 2l
25 0.7
y=0 y=0
20
y=l 0.6 y=l
15 y = 2l y = 2l
Pendulum angle, deg

10 0.5

Cart displacement, m
5
0.4
0
0.3
-5

-10 0.2
-15
0.1
-20

-25 0
0 50 100 150 200 0 50 100 150 200
time, sec time, sec

(a) IB tilt angle (b) Cart displacement

Disturbance amplitude = 100 N, Payload = 50 kg, Z= 2l
6
y=0
5
y=l
4 y = 2l
Motor terminal voltage, volt

-1

-2

-3

-4
0 50 100 150 200
time, sec

(c) Motor terminal voltage

Figure-16. System performance and control signal (F = 100 N), PD control.

It is noted in Figures 15(a), 16(a), 17(a) and 18(a) this means that the harder it is to balance the IB when
that the closer the payload to the IB origin the harder the either the force is applied at lower positions of the rod or
balancing in the upright position. More fluctuations of the the payload is closer to the origin. Hence, the worst and
rod occur as the load is closer to the origin. This hardest case of balancing is more likely to happen when
phenomenon is similar to the system behaviour for the the payload and disturbance force are both close to the IB
case of applying the disturbance force at the lower end of origin.
the rod as previously described in Figure-13. Therefore,

23
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

0.3 0.7
FLC FLC
0.2 PD PD
0.6

Cart displacement, m
0.1 0.5
Tilt angle, deg

0 0.4

-0.1 0.3

-0.2 0.2

-0.3 0.1

-0.4
0 20 40 60 80 100 0
0 20 40 60 80 100
time, sec time, sec
(a) IB tilt angle (b) Cart displacement
0.035 8
FLC
FLC
0.03 PD
PD
6
0.025
Cart velocity, m/s

Control effort, volt

0.02 4

0.015
2

0.01
0
0.005

0 -2

-0.005
0 20 40 60 80 100 -4
0 20 40 60 80 100
time, sec
time, sec
(c) Cart velocity (d) Motor terminal voltage
Figure-17. Simulation results (F = 0 N, Q =15cm).

Changing the payload position in the previous value of the disturbance force applied on the rod. The
manner did not affect the cart linear displacement control effort appears to have more fluctuations when the
significantly as noted in Figures 15(b), 16(b), 17(b) and payload was closer to the origin as noted in Figures 15(c),
18(b). This is still not expected to happen for longer 16(c), 17(d) and 18(d).
pendulum rods. However, the effect was more due to the

24
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

0.1 0.7
FLC FLC
0.05 PD 0.6 PD

Cart displacement, m
0.5
-0.05
Tilt angle, deg

0.4
-0.1

-0.15 0.3

-0.2
0.2
-0.25
0.1
-0.3

-0.35 0
0 20 40 60 80 100 0 20 40 60 80 100
time, sec time, sec
(a) IB tilt angle (b) Cart displacement
0.035 8
FLC FLC
0.03 PD 7 PD
6
0.025
Control effort, volt

5
Cart velocity, m/s

0.02
4
0.015 3

0.01 2

1
0.005
0
0
-1
-0.005
0 20 40 60 80 100 -2
0 20 40 60 80 100
time, sec
time, sec
(c) Cart velocity
(d) Motor terminal voltage
Figure-18. Simulation results (F = 0 N, Q =25cm).

4.2. Virtual dynamic payload motion 4.2.1 Motion of the payload and the centre of mass
For dynamic motion of the payload, the payload The linear motion executed by the payload along
M is assumed to move virtually up and down with a the IB is described in Figures 19 and 20. The payload is
simple harmonic motion expressed as: assumed to move up and down along the IB as a simple
harmonic motion.
y = Amp sin ( freq * t + phase) + bias (13) As the payload is moving in this way, the
Amp = 1 location of the COM of the IB will keep changing as noted
Select freq = f in Figures 19(b) and 20(b). The payload is assumed
initially at the mid-span of the IB, then moving up till the
phase = 0 upper end and then down to the lower end of the rod and
bias = 1 continuing up and down until the end of the simulation.
Then the motion of the payload can be described as: As noted in Figure-20(a), the COM of the IB will
never reach the extreme ends of the IB unless the mass of
Q = lsin ( f * t) + 1 (14) the rod is assumed negligible compared to the payload.
Figure-19(b) shows the velocity of the payload along the
Q& = lfc os ( f * t) (15) IB. The payload speed is changing dramatically at the start
of the simulation; such behaviour is not desirable in the
and
system performance as the payload acceleration will
&& = −lf 2 sin ( f * t)
Q (16) increase and that in turn leads to a jerky motion. Such
situation is dominant for higher changes of velocity in a
Where Q , Q& and Q && are the linear displacement, velocity finite time.
and acceleration of the payload

25
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

0.14 0.12

0.11
0.12
0.1
Payload linear displacement, m

0.1 0.09

COM displacement, m
0.08
0.08
0.07
0.06
0.06

0.04 0.05

0.04
0.02
0.03

0 0.02
0 50 100 150 200 0 50 100 150 200
time, sec time, sec

(a) Payload displacement (a) COM displacement

0.015 0.01

0.008
0.01
0.006
Payload linear velocity, m/s

COM linear velocity, m/s

0.004
0.005
0.002

0 0

-0.002
-0.005
-0.004

-0.006
-0.01
-0.008

-0.015 -0.01
0 50 100 150 200 0 50 100 150 200
time, sec time, sec

(b) Payload velocity (b) COM velocity

Figure-19. Payload linear motion (Frequency = 0.1 Hz). Figure-20. COM linear motion (Frequency = 0.1 Hz).

The speed of the COM position, as presented in developed controller such different impact levels. The
Figure-20(b), is around half the value of the payload simulation is carried out by considering the disturbance
speed. This will be different for different values of force applied at the upper end of the IB of the vehicle.
payload and rod mass.

4.2.2 Effect of different disturbance levels

The effect of changing the level of the applied
disturbance force is considered in this section. Four levels
of disturbance amplitudes are considered from 10 N to 150
N in order to test the performance and robustness of the

26
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Disturbance applying at the pendulum upper end Disturbance applying at the pendulum upper end
0.8
Disturbance amplitude = 150 N
0.6 Disturbance amplitude = 100 N
0.5
Disturbance amplitude = 60 N
0.4
Disturbance amplitude = 10 N

Cart displacement, m
Pendulum angle, deg

0.4
0.2

0 0.3

-0.2
0.2
-0.4
Disturbance amplitude = 150 N
0.1 Disturbance amplitude = 100 N
-0.6
Disturbance amplitude = 60 N
Disturbance amplitude = 10 N
-0.8 0
0 50 100 150 200 0 50 100 150 200
time, sec time, sec

(a) IB tilt angle (b) Cart displacement

Disturbance applying at the pendulum upper end
4
Disturbance amplitude = 150 N
Disturbance amplitude = 60 N
3 Disturbance amplitude = 60 N
Disturbance amplitude = 10 N
Motor terminal voltage, volt

-1

-2
0 50 100 150 200
time, sec

(c) Motor terminal voltage

Figure-21. System performance and control signal (Different disturbance levels), PD control.

It can be noted from Figures 21(a), 22(a) and the system response. This phenomenon is clearly repeated
23(a) that the higher the level of the disturbance force the when the force is reapplied on the system. It can also be
higher the fluctuation of the rod angle from the upright noted that the lower the level of the disturbance force, the
position and the longer it takes the IB to balance at the faster the cart in achieving the desired position.
upright position. Changing the level of the disturbance force also
The cart linear displacement is presented in affected the behaviour of the controller, which is clearly
Figures 21(b), 22(b) and 23(b). It is noted that increasing presented in Figures 21(c), 22(d) and 23(d). Higher levels
the level of the applied disturbance tends to decrease the of the applied force tended to increase the time the control
cart overshoot beyond the specified limit. This is because signal takes to settle down and also required a lot of
the amount of the applied force acts as a drag for the cart control effort.
and slows the cart which in turn increases the rise-time of

27
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

0.4 0.7
FLC FLC
0.3 PD PD
0.6
0.2

Cart displacement, m
0.5
0.1
Tilt angle, deg

0 0.4

-0.1 0.3
-0.2
0.2
-0.3
0.1
-0.4

-0.5 0
0 20 40 60 80 100 0 20 40 60 80 100
time, sec time, sec
(a) IB tilt angle (c) Cart displacement
0.04 8
FLC
FLC
PD
0.03 6 PD
Cart velocity, m/s

4
Control effort, volt

0.02

2
0.01

0
0
-2
-0.01
-4
-0.02
0 20 40 60 80 100 -6
time, sec 0 20 40 60 80 100
time, sec
(b) Cart velocity
(d) Motor terminal voltage
Figure-22. Simulation results (F = 30 N, t = 0.1 sec, f = 0.1 Hz).

28
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

1.5
FLC
PD
1 0.5

Cart displacement, m
0.5 0.4
Tilt angle, deg

0 0.3

-0.5 0.2

-1 0.1 FLC
PD
-1.5 0
0 20 40 60 80 100 0 20 40 60 80 100
time, sec time, sec
(a) IB tilt angle (b) Cart displacement
0.04 8
FLC FLC
0.03 PD 6 PD
0.02
4
Control effort, volt

0.01
Cart velocity, m/s

0 2
-0.01
0
-0.02

-0.03 -2
-0.04
-4
-0.05

-0.06 -6
0 20 40 60 80 100 0 20 40 60 80 100
time, sec time, sec
(c) Cart velocity (d) Motor terminal voltage
Figure-23. Simulation results (F = 100 N, t = 0.1 sec, f = 0.1 Hz).

4.2.4 Effect of disturbance duration which the developed controller will not be able to cope
The duration of applying such disturbances is with. Two different levels of disturbance force are used to
another variable of interest. Four different periods of the test the validity of the developed control algorithm at
disturbance force are used to estimate safe limits beyond different disturbance duty cycles.

29
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Disturbance amplitude = 30 N Disturbance amplitude = 30 N

3 0.7
duty cycle = 1 sec
duty cycle = 0.5 sec
duty cycle = 0.25 sec 0.6
2
duty cycle = 0.05 sec
0.5

Cart displacement, m
Pendulum angle, deg

1
0.4
0
0.3
duty cycle = 0.05 sec
-1
0.2 duty cycle = 0.25 sec
duty cycle = 0.5 sec
-2 0.1 duty cycle = 1 sec
duty cycle = 2 sec
0
-3 0 50 100 150 200
0 50 100 150 200
time, sec
time, sec

(a) IB tilt angle (b) Cart displacement

Disturbance amplitude = 30 N
4
duty cycle = 1 sec
duty cycle = 0.5 sec
3 duty cycle = 0.25 sec
duty cycle = 0.05 sec
Motor terminal voltage, volt

-1

-2
0 50 100 150 200
time, sec

(c) Motor terminal voltage

Figure-24. System performance and control signal (Disturbance amplitude = 30 N), PD control.

30
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Disturbance amplitude = 100 N Disturbance amplitude = 100 N

10 0.7
duty cycle = 1 sec
8 duty cycle = 0.5 sec
duty cycle = 0.25 sec 0.6
6 duty cycle = 0.05 sec

4 0.5
Pendulum angle, deg

Cart displacement, m
2
0.4
0

-2 0.3

-4 0.2 duty cycle = 0.05 sec

-6 duty cycle = 0.25 sec
0.1 duty cycle = 0.5 sec
-8
duty cycle = 1 sec
-10 0
0 50 100 150 200 0 50 100 150 200
time, sec time, sec

(a) IB tilt angle (b) Cart displacement

Disturbance amplitude = 100 N
4

3
Motor terminal voltage, volt

-1 duty cycle =1 sec

duty cycle = 0.5 sec
-2 duty cycle = 0.25 sec
duty cycle = 0.05 sec
-3
0 50 100 150 200
time, sec

(c) Motor terminal voltage

Figure-25. System performance and control signal (Disturbance amplitude = 100 N), PD control.

From Figures 24(a), and 25(a), it is noted that the values of time, the overshoot tended to be higher, which is
amount of overshoot in the IB tilt angle increases as the a characteristic of such impulse disturbance force, but
period of the force is increased. Also the system can accelerated the system to reach the desired position. It can
recover faster to the desired position with lower values of also be noted that the effect of repeating the disturbance
the duty cycle of the force. The controller still copes well force is more dominant at higher levels of disturbance
with such higher periods of time especially for the force than the situations at the lower levels. The reason
pendulum tilt angle. The amount of tilt angle overshoot behind that is the accumulation of errors in the cart
and the duration of fluctuations until the system settles position.
down are proportional to the level of the disturbance force The control effort is presented in Figures 24(c),
applied to the system. and 25(c). It is noted that the longer the disturbance duty
For the cart linear displacement, as described in cycle the higher the control effort required and the longer
Figures 24(b), and 25(b), the situation is not as good as for the time for the control signal to settle down.
the tilt angle, as increasing the force duty cycle made the The same conclusions will hold for the results in
system slower in achieving the target, increased the rise- Figures 22 and 26, based on the implementation of PD-
time and decreased the cart overshoot. However, for lower like fuzzy control technique.

31
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

1
FLC
0.8 PD
0.5
0.6

Cart displacement, m
0.4
0.4
Tilt angle, deg

0.2

0 0.3

-0.2
0.2
-0.4

-0.6
0.1 FLC
-0.8
PD
-1 0
0 20 40 60 80 100 0 20 40 60 80 100
time, sec time, sec
(a) IB tilt angle (c) Cart displacement
0.04 8
FLC FLC
0.03 PD PD
6

0.02
4
Control effort, volt
Cart velocity, m/s

0.01
2
0
0
-0.01

-0.02 -2

-0.03 -4

-0.04
0 20 40 60 80 100 -6
0 20 40 60 80 100
time, sec
time, sec
(b) Cart velocity
(d) Motor terminal voltage
Figure-26. Simulation results (F = 30 N, t = 0.2 sec, f = 0.1 Hz).

4.2.5 Effect of payload speed the payload will increase around 6 times based on the
Simulations were carried out with the effect of formula in equation (15).
changing the payload speed along the IB. The speed of the As the payload is moving in this way, the
payload can be changed if the frequency of simple location of the COM of the IB will keep changing as noted
harmonic motion of the payload is changed. Four different in Figures 27(b) and 28(b). The payload is assumed
levels of such frequency are implemented to investigate initially at the mid-span of the IB, then moving upward till
the effect of speed change on the performance of the the upper end and then down to the lower end of the rod
system and the ability of the controller to overcome the and continuing up and down until the end of the
results which may arise in such cases. simulation.
The linear motion executed by the payload along As noted in Figures 27(b) and 28(b), the COM of
the IB is described in Figures 27 (a) and 28(a). The the IB will never reach the extreme ends of the payload
payload is assumed to move up and down along the IB in a unless the mass of the rod is assumed negligible compared
simple harmonic motion. As the frequency of the payload to the payload.
motion increases 10 times from 0.1Hz to 1Hz the speed of

32
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

0.5 0.5

0.45 0.45

0.4 0.4

Payload position, m
0.35 0.35
Load position, m

0.3 0.3

0.25 0.25

0.2 0.2

0.15 0.15

0.1 0.1

0.05 0.05

0 0
0 20 40 60 80 100 0 20 40 60 80 100
time, sec time, sec
(a) (a)
0.45 0.45

0.4 0.4

0.35 0.35
IB COM position, m

COM position, m

0.3 0.3

0.25 0.25

0.2 0.2

0.15 0.15

0.1 0.1

0.05 0.05
0 20 40 60 80 100 0 20 40 60 80 100
time, sec time, sec
(b) (b)
Figure-27. Payload and COM motion (Motion frequency, Figure-28. Payload and COM motion (Motion frequency,
f = 0.1 Hz). f = 1 Hz).

33
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

0.7
Frequency = 0.1 Hz
0.1 Frequency = 0.3 Hz
Frequency = 0.5 Hz 0.6
Frequency = 1 Hz

0.05 0.5
Pendulum angle, deg

Cart displacement, m/s

0.4
0
0.3

-0.05
0.2 Frequency = 0.1 Hz
Frequency = 0.3 Hz
-0.1 0.1 Frequency = 0.5 Hz
Frequency = 1 Hz
0 50 100 150 200 0
0 50 100 150 200
time, sec
time, sec
(a) IB tilt angle (b) Cart displacement
4

Frequency = 0.1 Hz
Frequency = 0.3 Hz
3
Frequency = 0.5 Hz
Frequency = 1 Hz
Motor terminal voltage, volt

Frequency = 1.5 Hz
2

-1

-2
0 20 40 60 80 100 120 140 160 180 200
time, sec

(c) Motor terminal voltage

Figure-29. System performance and control signal (different payload speeds), PD control

The IB tilt angle is shown in Figures 29(a), 30(a) 30(b) and 31(b). The control signal is presented in Figures
and 31(a). It is noted that the angle overshoot increased as 29(c), 30(d) and 31(d).). It is noted that changing the
the speed of the payload increased. The faster the payload payload speed had a very limited effect on the controller
speed the longer before setting and the higher the output signal especially at lower speeds. However, the
fluctuation in the tilt angle. effect was more noticeable at higher speeds.
For such levels of payload speed, there was no
effect on the cart displacement as noted in Figures 29(b),

34
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

0.1 0.7
FLC FLC
0.05 PD 0.6 PD

Cart displacement, m
0 0.5
Tilt angle, deg

-0.05 0.4

-0.1 0.3

-0.15 0.2

-0.2 0.1

-0.25 0
0 20 40 60 80 100 0 20 40 60 80 100
time, sec time, sec
(a) IB tilt angle (b) Cart displacement
0.035 8
FLC FLC
0.03 PD PD
6
0.025
4
Control effort, volt
Cart velocity, m/s

0.02

0.015 2

0.01
0
0.005
-2
0

-0.005 -4

-0.01
0 20 40 60 80 100 -6
0 20 40 60 80 100
time, sec
time, sec
(c) Cart velocity
(d) Motor terminal voltage
Figure-30. Simulation results (F = 0 N, f = 0.1 Hz).

35
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

0.15 0.7
FLC FLC
0.1 PD PD
0.6

0.05

Cart displacement, m
0.5
Tilt angle, deg

0
0.4
-0.05
0.3
-0.1

-0.15 0.2

-0.2 0.1

-0.25
0 20 40 60 80 100 0
0 20 40 60 80 100
time, sec
time, sec
(a) IB tilt angle
(b) Cart displacement
0.035 8
FLC FLC
0.03 PD
6 PD
0.025
4
Control effort, volt
Cart velocity, m/s

0.02

0.015 2
0.01
0
0.005
-2
0

-0.005 -4
-0.01
0 20 40 60 80 100 -6
time, sec 0 20 40 60 80 100
time, sec
(c) Cart velocity
(d) Motor terminal voltage
Figure-31. Simulation results (F = 0 N, f = 1 Hz).

4.2.6 Effect of rod length 5. CONCLUSIONS

The effect of length of the intermediate body is A mathematical model of a TWRM system has
investigated in this section. The system performance been developed in planar coordinate frame. An external
alongside the control signal is presented utilising two disturbance force has been applied to the IB, and a payload
different lengths of the IB. As noted in Figure-32 the has been attached to the rod. Two control techniques; PD
shorter the IB the harder the balance; short rods lead to and PD-like fuzzy control have been developed and
high fluctuations of the rod about the upright balance implemented to balance the TWRM. Simulations have
position and long periods of oscillations until settling been carried out to address the effect of applying a
down. However changing the IB length, with such values, disturbance force on the system with attached payload at
did not affect the linear motion of the vehicle. Changing different locations along the IB. The system dynamic
the IB length corresponds to the position of the rod centre behaviour has been presented alongside the control effort
of mass which becomes closer to the vehicle centre point for achieving the required performance in those different
with shorter rods. cases. Simulations have been carried out for different
cases of changing two variables namely; the level and
duration of a disturbance force and the position and speed
of the payload.

36
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

2 0.7
Rod length = 20 cm Rod length = 20cm
1.5 Rod length = 50 cm Rod length = 50cm
0.6

Cart displacementm, m
0.5
Pendulum angle, deg

0.5
0.4
0
0.3
-0.5

-1
0.2

-1.5 0.1

-2 0
0 50 100 150 200 0 50 100 150 200
time, sec
time, sec
(a) IB tilt angle (b) Cart displacement
5
Rod length = 20cm

4 Rod length = 50cm

Motor terminal voltage, volt

-1

-2
0 50 100 150 200
time, sec

(c) Cart velocity

Figure-32. System performance and control signal (Different IB lengths), PD control.

Several observations have been made from the system Thus, the higher the level at which the force is
dynamic equations and have been proved using applied the easier it is to balance the IB and the longer the
simulations. Those observations can be summarised as time it takes the vehicle to settle down. Increasing the
follows: control efforts VL and VR has the effect of increasing the
A disturbance force directly affects balancing of deviation in the tilt angle of the IB and accelerating the
the IB in the upright position; Increasing the disturbance entire system out of the target. Increasing the inertia of the
force tends to increase the deviation of the IB from the various system components tends to add more damping to
target balance position. Moreover, the disturbance force the system and decrease possible oscillations. However, it
acts as drag on the entire system, specifically if applied will require a lot of control effort to manoeuvre the vehicle
close to the IB lower end, which tends to decelerate the and hence more energy consumption.
vehicle and in turn increase the energy consumption of the In conclusion, multiple conditions can result in
system. increasing the complexity of the control system in
The position of the disturbance has been shown balancing the TWRM: COM close to the vehicle centre
to be a matter of concern. The results can be stated as: point (IB lower end), disturbance force applied close to
ƒ If S > 2 L g (force applied above the location of the that end, shorter IB, higher level of disturbance force,
longer period of disturbance effect or high payload speed.
COM) then the effect will be to increase the amount of
Based on the results achieved, it is concluded that the
deviation from the target location
developed control algorithm coped well in those cases
ƒ If S < 2 Lg (force applied below the location of the
with different limits for such disturbances and objectives
COM) then the effect will be to help the controller to of the study were met to a great extent.
achieve balance by decreasing the deviation in the tilt Further studies will investigate the effect of
angle. changing the direction of applying the disturbance force

37
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

on system behaviour and develop the mechanical Km Ke K

TR = − y& R + m VR (A11)
adjustments for the vehicle for safe operation during such RRw R
cases. The design of control algorithms to deal with
various unexpected external disturbances acting on the Substituting for TR from Equation (A11) into
system is another area of future interest.
Equation (A9) yields:
Appendices Km Ke K
J wxφ&&Ry = − y& R + m VR + (H FR − H TR )Rw (A12)
RRw R
Linear model of the DC motor
Km Ke K J
Motor torque M w &y&R = − y& R + m VR − wx φ&&Ry − H R (A13)
RRw R Rw
τ m = k mi (A1) Km Ke K J
M w &y&L = − y& L + m VL − wx φ&&Ly − H L (A14)
RRw R Rw
Back electro magnetic force (EMF) Ve :
Linear transformations:
Ve = k e ω (A2)
&y&R = Rwφ&&Ry , &y&L = Rwφ&&Ly (A15)
Kirchhoff’s voltage Law of the DC motor:
Linear motion of centre of mass of the right and
di
Va − Ri − L − Ve = 0 left wheels:
dt
(A3) ⎛ J ⎞ K K K
⎜⎜ M w + wx2 ⎟⎟ &y&R = − m e y& R + m VR − H R (A16)
Equation of motion for the motor: ⎝ Rw ⎠ RRw R
dω
∑M = τm − k f −τa = IR
dt
(A4) ⎛ J
⎜⎜ M w + wx2
⎞ K K K
⎟⎟ &y&L = − m e y& L + m VL − H L (A17)
⎝ Rw ⎠ RRw R
di R k V
= i+ eω+ a (A5)
dt L L L Intermediate body dynamics
Consider Figures A1 and A2 which represent the
dω k m k τ free body diagrams of the IB with an external applied
= i− f − a (A6)
dt IR IR IR disturbance force F . Applying the Newton’s second law of
motion in the Y direction yields
Non-Linear Model of the wheels and the IB
∑F Y = (M + M p )Y&&
Wheels Dynamics
Translational equilibrium of the wheels: (H R + H L ) = (M + M p ) (Y&& + Lg θ&&p cos θ p − )
Lg θ& p2 sin θ p + F (A18)

∑F yR = M w &y&R Considering the sum of the applied forces in a

direction perpendicular to the rod gives,
M w &y&R = H TR − H FR − H R (A7) ∑F ⊥ IB = (M + M p ) Y&& cos θ P

∑F zR = M w &z&R (HR + HL ) cosθ p + (PR + PL ) sinθ p − (M + M p ) (g sinθ p + Lgθ&&p ) − F cosθ p (A19)

= (M + M p )Y&&cosθ p
M w &z&R = VTR − VR − M w g (A8)

Rotational equilibrium of the wheel around its

centre point O R :

∑M = J wx φ&&Ry

J wxφ&&Ry = TR + H FR R w − H TR Rw (A9)

From motor dynamics, the torque driving the

right wheel can be expressed as:
Km Ke K (A10)
TL = − y& L + m VL
RRw R

38
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Z [(H L + H R ) cosθP + (PL + PR ) sinθP ]Lg = I gθ&&p − Km Ke ( y& R + y& L ) (A23)

RRw
Payload
+
Km
(VL + VR ) + F cosθ p (Z − Lg ) + M F
F RRw

Y&& = Y&&L + Y&&R (A24)

Nomenclature
θP
(M+MP) H FR and Frictional forces between the wheel and the
H FL ground
H TR , H TL Horizontal reaction forces
Y 2R VTR and Vertical reaction forces
O
VTL
HR , HL Horizontal reaction forces between the IB
and wheel axle
HF HT Vertical reaction forces between the IB and
PR , PL
VT wheel axle
Mw Mass of the vehicle wheel
θ&&P Angular acceleration of the IB around X -
Figure-A1. Schematic diagram of the vehicle. axis
Y&& Linear acceleration of the centre point O of
the vehicle
Z Fz R Resultant force acting on the right wheel in
the Z -direction
FyR Resultant force acting on the right wheel in
F the Y -direction
F External applied disturbance force
θP Y&& Linear acceleration of the centre point O of
the vehicle.
I eq θ&&P MF Frictional moment at the joint connecting
(M+ MP) the IB and the wheel axle
c c and Coulomb and viscous friction coefficients
Q S Lg l MF cv
respectively

V L and VR System input voltage

Y O HL + HR Va Applied voltage
km Proportional constant.
ke Back EMF constant
PL + PR
τm Motor produces a torque
FigureA2. Schematic diagram of the IB. IR Armature inertial load
Va Applied voltage
Finally, equating the sum of the applied moments τa Applied torque
including the frictional moment at the joint connecting the
θ Angular displacement of motor shaft
IB and axle around the global centre of mass yields, ω Angular velocity of motor shaft
∑ Moments O = I eqθ&&p REFERENCES
[(H L + H R )cosθ P + (PL + PR )sinθ P ]Lg − (TR + TL ) − M F (A20)
− F cosθ p ( S − Lg ) = I gθ&&p [1] Browning B., Rybski P. E., Searock J. and Veoso M.
M. 2004. Development of a soccer-playing
M F = ccθ&p + cv sinθ&p (A21) dynamically-balancing mobile robot. Proceedings of
the International Conference on Robotics and
Km Ke Automation, New Orleans, LA.
TL + TR = − ( y& L + y& R ) + K m (VL + VR ) (A22)
RRw R [2] D’Andrea R. and Earl M. G. 2005. Design and
implementation of a minimum time transition for an
inverted pendulum. Proceedings of the Asian

39
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

Conference on Industrial Automation and Robotics, [13] Lee T. H., Leung F. H. F. M., Tam P. K. S. 1999.
Bangkok, Thailand. Position control for wheeled mobile robot using a
fuzzy controller. Proceedings of the 25th annual
[3] Goher K. M. K. and Tokhi M. O. 2008. Modelling, conference of the IEEE Industrial Electronics Society,
simulation and balance control of a two-wheeled San Jose, California, USA.
robotic machine with static variation in load position.
Proceedings of the 22nd European Conference on [14] Mamdani E. H. and Assilian S. 1974. Application of
Modelling and Simulation, Nicosia, Cyprus. fuzzy algorithms for control of simple dynamic plant.
Proceedings of IEE, Part D: 121(12): 1585-1588.
[4] Goher K. M. K. and Tokhi M. O. 2008. Robust
control of a two-wheeled robotic machine. [15] Mamdani E. H., Assilian S. 1975. An experiment in
Proceedings of the 8th Portuguese Conference on linguistic synthesis with a fuzzy logic controller.
Automatic Control, Vila Real, Portugal. International Journal of Man-Machine Studies. 7(1):
1-13.
[5] Goher K. M. K. and Tokhi M. O. 2008. Balance
control of a TWRM with a static payload. Proceedings [16] Maravall D., Zhou C. and Alonso J. 2005. Hybrid
of the 11th International Conference on Climbing and fuzzy control of the inverted pendulum via vertical
Walking Robots and the support Technologies for forces. International Journal of Intelligent Systems.
Mobile Machines (CLAWAR 2008), The University 20: 195-211.
of Coimbra, Coimbra, Portugal.
[17] Marzi H. 2005. Multi input fuzzy control of an
[6] Goher K. M. K. and Tokhi M. O. 2008. Balance inverted pendulum using an armature controlled DC
Control of a TWRM with a Dynamic Payload. Motor, Robotica. 23(6): 785-788.
Proceedings of the 11th International Conference on
Climbing and Walking Robots and the support [18] Pathak K., Franch J. and Agrwal S. K. 2005. Velocity
Technologies for Mobile Machines (CLAWAR 2008), and position Control of a Wheeled Inverted Pendulum
The University of Coimbra, Coimbra, Portugal. by Partial Feedback Linearization. IEEE Transactions
on Robotics. 21(3): 505-513.
[7] Hladek D. 2007. Multi-agent fuzzy control of the
robotic soccer. Proceedings of the 5th Slovakian- [19] Pawlowski S., Dutkiewicz P., Kozlowski P.,
Hungarian Joint Symposium on Applied Machine Wroblewski W. 2001. Fuzzy logic implementation on
Intelligence and Informatics, Popard, Slovakia. mobile robot control. The 2nd Workshop on Robot
Motion and Control. pp. 65-70.
[8] Ibrahim U. F. S. U., Azlan C. A., Abo Osman N. A.
and Saifizul A. A. 2006. Intelligent control of self- [20] Piegat A. 2006. What is not clear in fuzzy control
erecting inverted pendulum via adaptive neuro-fuzzy systems? International Journal Applied Mathematics
inference system. American Journal of Applied and Computer Science. 16(1): 37-49.
Sciences. 3(4): 1795-1802.
[21] Saifizul A. A., Zainon Z., Abu Osman N. A., Azlan C.
[9] Ishikawa S. 1991. A Method of indoor mobile robot A. and Ungko Ibrahim U. F. S. 2006. Intelligent
navigation by fuzzy control. Proceedings of the control for self erecting inverted pendulum via
IEEE/RSJ International Workshop on Intelligent adaptive neuro-fuzzy inference system. American
Robots and Systems. (Intelligence for Mechanical Journal of Applied Sciences. 3(4): 1795-1802.
Systems), Osaka, Japan.
[22] Salerno A., Angles J. 2004. The control of semi-
[10] Kawaji S. and Maeda T. 1991. Fuzzy servo control autonomous self balancing two-wheeled
system for an inverted pendulum. Proceedings of the quasiholonomic mobile robot. Proceedings of the 15th
Internet Fuzzy Engineering Symposium. 2: 812-823. CISM-IFToMM Symposium on Robot Design,
Montreal, Canada.
[11] Kim Y., Kim H. K. and Kwak Y. K. 2005. Dynamic
analysis of a nonholonomic two-wheeled inverted [23] Shi E. X., Huang W. E. and Ling Y. Z. 2004. Fuzzy
pendulum robot. Journal of Intelligent and Robotic predictive control of wheeled mobile robot based on
Systems. 44(1): 25-44. multi-sensors. Proceedings of the 3rd International
Conference on Machine Learning and Cybernetics,
[12] Kim Y., Kim H. K. and Kwak Y. K. 2006. Improving Shanghai, China.
driving ability for a two-wheeled inverted pendulum-
type autonomous vehicle, Part D: Journal of [24] Sun Y. L., Er M J. 2004. Hybrid fuzzy control of
Automobile Engineering. 220(2): 165-175. robotic systems. IEEE Transactions on Fuzzy
Systems. 12(6): 755-765.

40
 VOL. 6, NO. 3, MARCH 2011 ISSN 1819-6608
ARPN Journal of Engineering and Applied Sciences
©2006-2011 Asian Research Publishing Network (ARPN). All rights reserved.

www.arpnjournals.com

[25] Tsai C. C., Lin S. C. and Luo W. L. K. 2006. Adabtive

steering of a self-balancing two-wheeled transporter.
Proceedings of the Automatic Control Conference,
Tamsui, Taiwan.

[26] Tsai M. C. and Hu J. S. 2006. Robust control of auto-

balancing two-wheeled cart motion pilot. Proceedings
of the Automatic Control Conference, Tamsui,
Taiwan.

[27] http://www.segway.com/.

[28] http://leiwww.efpl.ch/joe.

[29] http://www.ibotnow.com/ibot-in-action.html.

41