We are IntechOpen,

the world’s leading publisher of

Open Access books
Built by scientists, for scientists

7,300
Open access books available
191,000
International authors and editors
210M Downloads

Our authors are among the

154
Countries delivered to
TOP 1%
most cited scientists
14%
Contributors from top 500 universities

Selection of our books indexed in the Book Citation Index

in Web of Science™ Core Collection (BKCI)

Interested in publishing with us?

Contact [Link]@[Link]
Numbers displayed above are based on latest data collected.
For more information visit [Link]
 3

Modular Fuzzy Logic Controller for Motion

Control of Two-Wheeled Wheelchair
Salmiah Ahmad1, N. H. Siddique2 and M. O. Tokhi3
1International Islamic University Malaysia,
2Ulster University,
3The University of Sheffield
1Malaysia
2,3United Kingdom

1. Introduction
Most of the wheelchair users are paraplegics, who are not able to move on their own due to
permanent injury in their lower extremities. These wheelchairs are four-wheeled and have
certain limitations due to design and control mechanism. For example, the wheelchairs
cannot move to a higher level, lift the front wheel and stay in an upright position. As a
result, wheelchair users cannot reach certain heights to pick and place things on the shelves,
and cupboards, etc. without any assistant and also cannot have eye-to-eye conversation with
normal people effectively. On the other hand, a two-wheeled wheelchair has a unique
characteristic that may help disabled and elderly people who use the wheelchair as the main
means of transport and can also use the wheelchair for these added advantages. Now the
idea is to transform the standard four-wheeled wheelchair into a two-wheeled upright
wheelchair to facilitate such manuoverability. The front wheels (casters) can be lifted up and
stabilized as an inverted pendulum, thus increasing the level of height achievable while in
the upright position. Similarly, when this upright position is no longer needed it may be
transformed back into its normal four-wheeled position. The schematic diagram of the two-
wheeled wheelchair is shown in Figure 1. The transformation will result in a highly
nonlinear and complex system. Since a human has quite significant mass sitting on the
wheelchair, the two-wheeled wheelchair can be modeled with double links that mimic
double inverted pendulum scenario that need a clever control strategy.
Most of the classical control design methodologies such as Nyquist, Bode, state-space, optimal
control, root locus, H  , and -analysis are based on assumptions that the process is linear and
stationary and hence is represented by a finite dimensional constant coefficient linear model.
These methods do not suit complex systems well because few of those represent uncertainty
and incompleteness in system knowledge or complexity in design. But the fact is the real
world is too complex. As the complexity of a system increases, quantitative analysis and
precision become difficult. The increasing complexity of dynamical systems such as this
coupled with stringent performance criteria, which are sometimes subject to human
satisfaction, necessitates the use of more sophiticated control approaches. However, many
processes that are nonlinear, uncertain, incomplete or non-stationary have subtle and

[Link]
38 Fuzzy Logic – Controls, Concepts, Theories and Applications

interactive exchanges with the operating environment and are controlled by skilled human
operators successfully. Rather than mathematically model the process, the human operator
models the process in a heuristic or experiential manner. It is evident that human knowledge is
becoming more and more important in control systems design. This experiential perspective in
controller design requires the acquisition of heuristic and qualitative, rather than quantitative,
knowledge or expertise from the human operator. During the past several decades, fuzzy
control has emerged as one of the most active and powerful areas for research in the
application of such complex and real world systems using fuzzy set theory (Zadeh, 1965).

δ2
δ1

Link2

2

 R and  L Link1
on each wheel

Fig. 1. Schematic diagram of wheelchair with three under actuated joints

Due to many significant advantages of wheelchair usage, this research presents findings of
the research carried out on the implementation of new architecture of modular intelligent
control strategies on the two-wheeled wheelchair model. The multi-objective control
involves lifting and stabilizing of Link1 and Link2 of double-inverted pendulum like two-
wheeled wheelchair, wheelchair backward and forward motion control as well as position.
It is hoped that the proposed model, mechanisms and control could be of benefit to a
wheelchair user, thus enhancing wheelchair technology for paraplegics and elderly.

2. Intelligent control approach

Intelligent control systems have evolved from existing controllers in a natural way
competing demanding challenges of the time and are not defined in terms of specific
algorithms. They employ techniques that can sense and reason without much a priori
knowledge about the environment and produce control actions in a flexible, adaptive and
robust manner (Harris, 1994). In general, by intelligent control approaches, it is mainly
meant the methodologies of fuzzy logic, neural networks, and genetic algorithms. These
methodologies have shown to be effective in controlling complex nonlinear systems. The
control of complex nonlinear systems has been approached over the last few decades using
fuzzy logic techniques due to the fact that fuzziness itself is easy to implement and can be

[Link]
Modular Fuzzy Logic Controller for Two-Wheeled Wheelchair 39

described by expert knowledge, normally possessed by human. A fuzzy logic controller

(FLC) has the basic configuration illustrated in Figure 2.

Rule-base

Input Inference Output

Fuzzification Defuzzification
Mechanism

Fig. 2. Fuzzy logic control

Generally, a fuzzy logic controller consists of the following components:

i. Fuzzification
ii. Inference mechanism
iii. Rule-base
iv. Defuzzification

Fuzzification is a process of transforming an observed input space to fuzzy sets within a

universe of discourse. This process consists of associating to each fuzzy set a membership
function (MF). These functions can be thought of as maps from the real numbers to the
interval I   0,1 . If there are n fuzzy sets associated with a given quantity x  R , such n
maps Fi : R  I , i  1, , n are defined. They determine to what extent the linguistic label
associated with fuzzy set i characterizes the current value of x. There are different kinds of
MFs used in designing fuzzy controllers. The most common choices are triangular,
trapezoidal, Gaussian and bell shaped MFs. There is no exact method for choosing an MF,
and the designer mainly relies upon an expert knowledge or use heuristic rule.
Inference is used to describe the process of formulating a nonlinear mapping from a given
input space to an output space. The mapping then provides a basis from which decisions
can be taken. The process of fuzzy inference involves the MFs, fuzzy logic operators and
rule-base. Generally there are three types of commonly used fuzzy inference. They differ
mainly in the consequent part of their fuzzy rules, aggregations and defuzzification
procedures. Thus selecting a different fuzzy inference will result in different
computational time. The three common fuzzy inferences are: Mamdani fuzzy inference,
Sugeno fuzzy inference and Tsukamoto fuzzy inference. The choice of a particular
inference mechanism is eventually problem dependent and availability of information
about the system in question.
Mamdani type fuzzy modeling was proposed as the first attempt to control a steam engine
and boiler by a set of linguistic control rules by (Mamdani 1974). In this type of inference,
Max-min is the most common rule of composition used. In this composition rule, the
inferred output of each rule is a fuzzy set chosen from the minimum firing strength. On the

[Link]
40 Fuzzy Logic – Controls, Concepts, Theories and Applications

other hand, in max-product rule of composition the inferred output of each rule is a fuzzy
set scaled down by its firing strength via algebraic product.
A fuzzy system is characterized by a set of linguistic statements based on expert knowledge.
The expert knowledge is usually in the form of if-then rules, which are easily implemented
by fuzzy conditional statements in fuzzy logic. The collection of fuzzy rules that are
expressed as fuzzy conditional statements forms the rule base or the rule set of an FLC. A
rule consists of two parts, antecedent and consequent. For example, typical rule in
Mamdani-type fuzzy model with four-inputs and three-outputs FLC can be expressed by
the following linguistic conditional statement.

If (X1 is Ai) and (X2 is Bj) and (X3 is Ck) and (X4 is Dl)

then (Y1 is Up) and (Y2 is Vq) and (Y3 is Wr)

where {X1, X2, X3, X4} are the inputs with linguistic terms { Ai, Bj, Ck, Dl} and {Y1, Y2, Y3} are
the outputs with linguistic terms {Up, Vq, Wr}.

Defuzzification is basically a mapping from a space of fuzzy control actions defined over an
output universe of discourse into a space of nonfuzzy (crisp) control actions. In a sense this
is the inverse of the fuzzification even though mathematically the maps need not be inverses
of one another. In general, defuzzification can be viewed as a function DF : I n  R ,
mapping a fuzzy vector x F with n fuzzy sets to a real number. There are different methods
of defuzzification. However, simple methods are available to use depending on the
application, among them Centre of Gravity Method (COG), and Weighted Average Method
are widely used in Mamdani-type FLC and Sugeno-type FLC. Each method is problem
dependent, but the experts should know that these methods are available and should try to
see which works best for the application.
The two-wheeled wheelchair model involves lifting and stabilizing the two links (Link1 and
Link2) similar to a double-inverted pendulum and hence is a multi-objective control
problem. Considering the complexity and non-linearity of the wheelchair, the controller has
to be designed in such a way to produce the required torques, namely  R ,  L and  2 , for
acting at three different locations on the wheelchair for lifting the casters/chair and
stabilizing the system. The torque  R and  L represent the input torque to the right and left
wheels respectively.  2 represents the torque between Link1 and Link2 to cater for the
whole weight of the human body. Angular positions of Link1 and Link2,  1 and  2
respectively, are measured using sensors attached to the wheelchair. This characterizes the
system as a highly nonlinear multi-input multi-output (MIMO) system. Fuzzy logic control
is therefore very appropriate to use in this case. To achieve upright position for the two
links, they need to be lifted and stabilized to zero degree (relative to vertical axis) upright
position. This may be realised with a single controller. However, this will lead to a huge
fuzzy rule-base. A conventional fuzzy controller with 4 inputs e 1 , e 1 , e 2 , e 2  and 3
outputs  R , L , 2  (inputs-outputs are shown in Figure 3) has significant drawback in
terms of computational complexity, which increases with the dimension of the system
variables; the number of rules increases exponentially as the number of system variables
increases. A strategy is sought to simplify the development process and reduce the

[Link]
Modular Fuzzy Logic Controller for Two-Wheeled Wheelchair 41

geometric progression in the number of required rules for general purpose tracking and
control situations. Moreover, it should be achieved without compromising the robustness
and capability of the complete system.

Fig. 3. MIMO FLC for lifting and stabilizing the wheelchair

A generic problem with an FLC is that the number of rules grow exponentially with the
number of input-output variables and linguistic terms for each variable. For a complete rule-
 
base with input variables Xi |i  1, , n with linguistic terms Aij | j  1, , mi and output
 
variables Yk | k  1, , l with linguistic terms Bkj | j  1, , pk , the number of rules will be

n
R  mi (1)
i 1

The rules have the form

If (X1 is A11) and ... and (Xn is Anm) Then (Y1 is B11) and ... and (Yl is Blp)
This large number of rules complicates the design of an FLC, because for each of the R
different premises the expert must provide a combination of term sets for the output
variables, which is nearly impossible for a human expert to guess. It is possible to omit a set
of rules if it could be guranteed that a certain combination of input-output variables will
never occur during control of the dynamic system. A modular structure of FLCs with
minimum number of input-output variables can reduce the number of rules R.

3. Modular fuzzy control

For large scale and complex systems, the reduction in computation and design complexity
remains a challenge of intelligent control systems. Hierarchical and modular methodology
have gained wide popularity because of its simplicity in design and robustness. There are
several approaches in decomposing a system into modules such as decentralized approach,
time-scale decomposition, hierarchical system, and workspace decomposition (Siljak, 1991).
For control problems with multiple objectives of different priority, sub-controller with a
subset of input-output variables can be designed for each objective. Furthermore, each
antecedent can be decomposed into single input modules. Each fuzzy module is designed to

[Link]
42 Fuzzy Logic – Controls, Concepts, Theories and Applications

handle one specific input affiliated with one of the decoupled antecedents Xi |i  1, , n
and produces a crisp action Yk |k  1, , l where k  1, , l . Such a generic modular
architecture is shown in Figure 4.

Fig. 4. Modular FLC

A typical fuzzy rule issues an appropriate output action by evaluating the related inputs
from the measurement data. In the conventional IF-THEN fuzzy inference formulation, all
of the system’s input parameters are suggested as antecedents in the fuzzy rule. The total
possible number of fuzzy rules that can be generated for the rule base is Lk where k is the
number of inputs and L is the number of fuzzy linguistic terms or MFs. As compared to the
modular FLC design, each input represents one fuzzy control module. The total number of
rules for each module is determined by the number of MFs L. Thus, the total number of
fuzzy rules for all k modules is kL. This clearly shows a significant reduction in the number
of fuzzy rules from Lk to k as well as savings in computation.
The mathematical model of the two-wheeled wheelchair incorporates three independent
actuators; derived from Figure 1, corresponding to control output to be fed into the system.
The angular position of Link1 and Link2, denoted as δ1 and δ2 respectively, are the
controlled variables that will determine the system performance. The control challenge
relates to the fact that there is more than one mechanism acted upon with the same actuator.
For example, to transform the wheelchair into an upright two-wheeled wheelchair, the
torques determined by fuzzy control are located at both right and left wheels. At the same
time, if linear motion is considered, the same actuator needs to provide enough torque such
that the wheelchair will still move forward or backward while in the upright position.
Lifting and stabilizing consist of two system output parameters to be considered, namely
angular position of Link1, δ1 and angular position of Link2, δ2. Therefore a modular fuzzy
logic control (MFC) is adopted to realize this multi-function two-wheeled wheelchair.
The MIMO system with an objective of achieving zero degree upright position is
decomposed into small and simpler subsystems: Link1-lifting, Link1-stabilizing, Link2-
lifting, and Link2-stabilizing. The structure of the modular FLC for the wheelchair is
illustrated in the block diagram in Figure 5. Accordingly, this type of FLC can deal with, for
example, N subsystems located at different levels, where each subsystem manages its own
control strategy and communicates with the coordinator. The coordinator comprises a pair

[Link]
Modular Fuzzy Logic Controller for Two-Wheeled Wheelchair 43

of switches that gathers information from the subsystems and sends supervisory (threshold
condition) instructions back to the subsystems. The supervisor in this case is the condition,
(if the angular position error of Link1 and Link2, -5° < e < 5°, then Link1-stabilizing and
Link2-stabilizing are activated). In this case, the switches coordinate the condition fulfilment
of all the criteria for the activation of actuator to work accordingly. The reference position
for lifting and stabilizing of both links is 0 degree at the upright position. The parameters ‘a’
and ‘b’ in the figure show the fuzzy input scaling factors (input gain) such that if the
stabilizing subsystem of Link1 or Link2 is activated, the sensitivity of the fuzzy inputs is
increased by giving higher gain (about 10 times) of a and b. The outputs from the system
that are fed back to the controller are the angular position of Link1 (δ1) and the angular
position of Link2 (δ2). The control approach using this modular strategy is believed to work
well with the independently allocated tasks. In the figure, eδ1 shows the angular position
error of Link1, ∆eδ1 represents the change of angular position error of Link1. The effect of
Link2 onto Link1 is taken into account by using the angular position error of Link2, eδ2 as
the fuzzy input for FLC1 and FLC2. In these two controllers, eδ2 represents the angular
position error of Link2, while ∆eδ2 represents the change of angular position error of Link2.
Similarly the effect of Link1 onto Link2 is taken into account by using the angular position
error of Link1, eδ1 as the input for FLC3 and FLC4.

- Lifting Coordinator
+
0 FLC 1

a
Stabilizing
Supervisor

Switch
FLC 2

 R,  L δ1
Lifting

δ2
+
0 FLC 3 2
-

b
Stabilizing
Switch
FLC 4

Subsystems
Fig. 5. Modular FLC for Two-wheeled wheelchair.

4. MFC for two-wheeled wheelchair

The MFC is also known as hierarchical fuzzy control (HFC), and the two terms are used
interchangeably. It is discussed in detail for two-wheeled application in (Ahmad et al. 2011).

[Link]
44 Fuzzy Logic – Controls, Concepts, Theories and Applications

The goal of the controller is to produce the required torques, namely  R ,  L and  2 , for
acting at three different locations on the wheelchair for lifting and stabilizing. The
torques  R and  L represent the input torque to the right and left wheels respectively. On
the other hand,  2 represents the torque between Link1 and Link2 to be used to cater for the
whole weight of the human body. Angular positions of Link1 and Link2,  1 and
 2 respectively, are measured using sensors attached to the wheelchair in Visual Nastran
(VN). To achieve upright position of the two links, they need to be lifted and stabilized at
zero degree upright position. The goal may be treated as a single objective control that is
having Link1 and Link2 at the 0 degree upright position with one controller. This will
increase significantly the computational complexity, which increases with the number of
system variables; the number of rules increases exponentially as the number of system
variables increases.

4.1 Rules reduction strategy for general purpose tracking and control situations
The strategy is sought without compromising the robustness and capability of the system.
Such a strategy relies mainly on three concepts, (Ahmad et al. 2011).
 Independence
 Functional Relationship
 Command Manipulation
To assess the effect of coupling in the fuzzy control, the system is tested with two different
configurations, which mainly differ at the input side of the controller, as shown in Table 1.

With coupling effect Without coupling effect

Link1 - Angular position error of - Angular position error of Link1, eδ1
(Lifting & Link1, eδ1 - Change of angular position error of
Stabilizing) - Change of angular position Link1, ∆eδ1
error of Link1, ∆eδ1
- Angular position error of
Link2, eδ2
Link2 - Angular position error of - Angular position error of Link2, eδ2
(Lifting & Link2, eδ2 - Change of angular position error of
Stabilizing) - Change of angular position Link2, ∆eδ2
error of Link2, ∆eδ2
- Angular position error of
Link1, eδ1
Rules of 5 x 5 x 3 = 75 rules 5 x 5 = 25 rules
each lifting
and
stabilizing

Table 1. Different input configurations of modular fuzzy logic controller

[Link]
Modular Fuzzy Logic Controller for Two-Wheeled Wheelchair 45

The system was tested with both configurations and the performances, with and without
coupling were comparably similar, see Figure 6. Therefore, as seen the second configuration
performed well with fewer fuzzy rules fired, and this configuration is used in implementing
the motion control for two-wheeled wheelchair.

100 10
coupled
of Link1 (degree)

of Link2 (degree)
Angular position

Angular position
decoupled 0
50
-10
0
-20 coupled
decoupled
-50 -30
0 10 20 30 0 10 20 30
Time (s) Time (s)

200 100
Right torque coupled
Link1 and Link2 (Nm)
Wheel Torques (Nm)

Left torque coupled

Torque between

100 Right torque decoupled 0

Left torque decoupled

0 -100 coupled
decoupled

-100 -200
0 10 20 30 0 10 20 30
Time (s) Time (s)

Fig. 6. System performance comparison between coupled fuzzy inputs and decoupled fuzzy
inputs in terms of δ1, δ2, R, L and 2.

The MFC is thus adopted for the two-wheeled wheelchair mechanisms, and the
corresponding research objectives are:
 Lifting and stabilizing control
 Linear motion control (forward or backward)
 Steering motion control
The MFC can be divided into two significant categories, primary and secondary (Bessacini
and Pinkos 1995). The controller is categorized according to different objectives. The control
structure for achieving an upright two-wheeled maneuverable wheelchair is depicted in
Figure 7. The general function of MFC is to minimize the errors in system responses
considered. The primary goal unit caters for the upright control, which consists of lifting
and stabilizing to the upright position and the transformation back to normal four-wheeled
position of Link1 and Link2. These controllers are active most of the time even during
maneuver. The secondary unit is activated by the coordinator (switch), with certain
condition pre-set for output activation. It consists of different unique objectives involving
linear motion control, steering control, additional chair height extension control. Each
objective in the secondary goal unit is discussed in detail in the following sections.

[Link]
46 Fuzzy Logic – Controls, Concepts, Theories and Applications

Fig. 7. Adapted modular intercepts fuzzy logic system (Bessacini and Pinkos 1995)

4.2 Simulation based performance analysis

The overall motion control for two-wheeled wheelchair is represented in Figure 8.
a. FLC for linear motion
The linear motion control generally consists of forward and backward (reverse) motion
control. They are both characterized as secondary systems (Bessacini and Pinkos 1995) since
the system needs to fulfill the primary target to achieve the upright position for both links.
Therefore MFC as discussed in Section 4 is very appropriate to implement.
Similar structure of FLC used for lifting and stabilizing is adopted for linear motion control.
The controls differ in terms of input and output scaling factors due to different reference
points executed. The control strategy designed in Matlab/Simulink was integrated with
wheelchair model, which was developed in VN software environment as a plant. The
motion (forward, backward or steering) takes place after lifting and stabilizing has been
achieved. Results show that the MFC strategy designed works very well and gives good
system performance.
In the current studies of wheelchair mobility, much research has been conducted on
wheelchair mobility in large spaces (outdoor mobility) (Vries et al. 1999; Wong et al. 2007). In
those researches, the distance and angle are considered at the same time to give output torque
of the wheels. On the other hand, note that the two-wheeled wheelchair is designed for use in
confined spaces, such that the linear motion and the steering motion are independent.

[Link]
Modular Fuzzy Logic Controller for Two-Wheeled Wheelchair 47

Fig. 8. Block diagram of two-wheeled wheelchair system motion

This confined space is normally found in the domestic environment (home, office and
library). Within such environment, linear motion is executed alone before steering is done
and vice versa. The block diagram for linear motion control of two-wheeled wheelchair is
shown in Figure 9.

Fig. 9. Block diagram for linear motion control

The FLC for linear motion (FLC3) consists of two inputs and two outputs. The controller
inputs are the position error, e and the change of position error, ∆e, while the controller

[Link]
48 Fuzzy Logic – Controls, Concepts, Theories and Applications

outputs are the torques, τR and τL. The fuzzy inputs are normalized so that they can be
generalized and then processed using the fuzzy rules. Moreover, the input normalization is
done due to the complexity of predetermining the range of change of position error, ∆e.
Gaussian (bell shaped) type membership functions with default parameters given by
Matlab/Simulink are used for all inputs and outputs. The membership levels for each input
and outputs are five in total. These comprise Negative Big (NB), Negative Small (NS), Zero

1NB NS Z PS PB 1NB NS Z PS PB
Degree of membership

Degree of membership
0.5 0.5

0 0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
Error ChangeOfError

1NB NS Z PS PB 1NB NS Z PS PB
Degree of membership

Degree of membership

0.5 0.5

0 0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
Right torque Left torque

Fig. 10. Membership functions for inputs and outputs for FLC3 of linear motion control

∆e
NB NS Z PS PB
e
PB PB PB PS Z
NB
PB PB PB PS Z
PB PB PS Z NS
NS
PB PB PS Z NS
PB PS Z NS NB
Z
PB PS Z NS NB
PS Z NS NB NB
PS
PS Z NS NB NB
Z NS NB NB NB
PB
Z NS NB NB NB

Table 2. Fuzzy rules for linear motion

[Link]
Modular Fuzzy Logic Controller for Two-Wheeled Wheelchair 49

(Z), Positive Small (PS) and Positive Big (PB). The membership function for inputs and
outputs of FLC3 is shown in Figure 10. Table 2 shows the implemented fuzzy rules for FLC3
controller. The two consecutive rows in the output part represent two fuzzy outputs, τR (first
row) and τL (second row). The rules developed are predetermined using expert knowledge
available such that all the errors should be brought back to the reference point immediately.
Forward Motion
The system was commanded to move forward after 4s, where at this time the two links had
been stabilized at the upright position. Figure 11 shows the final position of forward
mechanism execution while Figure 12 to Figure 18 show the results over 15s of simulation
time for forward movement of the two-wheeled wheelchair. The wheelchair was set to move
1.5m forward from its initial position. The results show that the FLC approach worked very
well with the wheelchair system on two wheels. Figure 11 shows the final wheelchair
position when it was set to move forward to 1.5m from the origin. It is noted from Figures
12 and 13 that both links settled after 4s from starting time of linear motion, which can be
considered quite good performance for the initial attempts of parameters setting. Link1
tilted with a positive angle from the 0° upright position. This configuration was
automatically adjusted to initiate the forward motion. The corresponding wheelchair
position is shown in Figure 14. It is noted that as much as 0.1m of the steady state error
appeared when it settled. Figure 15 shows the wheel torques (τR and τL) from the lifting and
stabilizing controller of Link1 (FLC1), and the wheel torques from the linear motion control
is shown in Figure 16. The torques vary from +40Nm to -40Nm during the forward motion
with positive slope during initial phase of travel. The resultant wheel torques contributed by
the lifting and stabilizing control as well as the linear motion control are shown in Figure 17.
The torque between Link1 and Link2 (τ2) given by (FLC2) is shown in Figure 18.

Fig. 11. Final position of 1.5m forward motion

[Link]
50 Fuzzy Logic – Controls, Concepts, Theories and Applications

100
Angular position of Link1 (degree)

80

60

40

20

-20
0 5 10 15
Time (s)

Fig. 12. Angular position of Link1, δ1 (degree)

10
Angular position of Link2 (degree)

-10

-20

-30
0 5 10 15
Time (s)

Fig. 13. Angular position of Link2, δ2 (degree)

2
Wheelchair position (m)

1.5

0.5

0
0 5 10 15
Time (s)

Fig. 14. Wheelchair position, x (m)

[Link]
Modular Fuzzy Logic Controller for Two-Wheeled Wheelchair 51

150
and stabilizing control (Nm)

Right torque
Wheel torques from lifting

100 Left torque

50

-50

-100
0 5 10 15
Time (s)

Fig. 15. Wheel torques, τR and τL due to lifting and stabilizing control, FLC1 (Nm)

100
linear motion control (Nm)

Right torque
Left torque
Wheel torques from

50

-50

-100
0 5 10 15
Time (s)

Fig. 16. Wheel torques, τR and τL due to linear motion control, FLC3 (Nm)

150
Right torque
Resultant torques (Nm)

100 Left torque

50

-50

-100
0 5 10 15
Time (s)

Fig. 17. Resultant wheel torques due to FLC1 and FLC3 (Nm)

[Link]
52 Fuzzy Logic – Controls, Concepts, Theories and Applications

100

50
Link1 and Link2 (Nm)
Torque between

-50

-100

-150

-200
0 5 10 15
Time (s)
Fig. 18. Torque between Link1 and Link2, τ2, FLC2 (Nm)

b. FLC for steering motion

A steering motion is needed when the two-wheeled wheelchair needs to change its
direction. The two-wheeled wheelchair can rotate to the right or to the left depending on
which direction it is commanded. There are two different approaches where steering could
be realized (Tanimoto et al., 2009). Similar direction of wheel rotation with different
magnitudes could lead to steering motion (moving both wheels forward with different
magnitudes). The first approach causes bigger turning radius as compared to the second
approach. The second approach to realize steering motion is by giving different direction of
wheel rotation (moving right wheel forward and left backward). The output torques in this
work given by the FLC used for steering motion covers both approaches according to the
steering error and the change of the steering error. In contrast to normal steering for mobile
robots, steering motion for the two-wheeled wheelchair is executed after the upright
position has been achieved; Link1 and Link2 at the 0° upright position. Therefore the
complexity in this configuration is higher than the steering motion using four wheels, since
other motion controls are active at the same time.
A block diagram for steering motion control of two-wheeled wheelchair is shown in Figure 19.
As discussed earlier, for reasons of simplicity, the torques applied to the two wheels are the
same in magnitudes (one output torque from the controller) so as to move the wheelchair only
forward or backward. Then each right and left wheel torque is made independent to realize
the steering motion. The weight here represents the human body weight, for which an average
70kg human is used. Sensors are attached at the respective reference bodies for control and
measurement. The control signals applied to the wheelchair model comprise the right torque,
τR (Nm), left torque, τL (Nm) and torque between Link1 and Link2, τ2 (Nm). The measured
outputs from the wheelchair system that consist of the angular position of Link1, δ1, (degree),
angular position of Link2, δ2, (degree) and wheelchair rotation angle about the vertical axis, ψ
(degree) are compared with the target references.
The wheelchair system modeled in VN software environment was used as a plant and
controlled with the developed FLC in the Matlab/Simulink environment. The steering
motion introduced takes place after the lifting and stabilizing mechanism has been achieved.

[Link]
Modular Fuzzy Logic Controller for Two-Wheeled Wheelchair 53

This new capacity increased the number of DOF of the two-wheeled wheelchair. Thus, it is
noticeable challenge to control the two-wheeled wheelchair where limited actuators are
available for different functions. Therefore, suitable controllers are needed, and FLC is
adopted. Results show that the FLC strategy works well and gives good system
performance.

Fig. 19. Block diagram for steering motion control

Two inputs and two outputs FLC is developed to control the steering motion. The
membership functions used are shown in Figure 20. The membership levels for each input

1NB NS Z PS PB 1NB NS Z PS PB
Degree of membership

Degree of membership

0.5 0.5

0 0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
WCrotError ChangeofError

1NB NS Z PS PB 1NB NS Z PS PB
Degree of membership

Degree of membership

0.5 0.5

0 0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
Right torque Left torque

Fig. 20. Membership levels for inputs and outputs of steering control

[Link]
54 Fuzzy Logic – Controls, Concepts, Theories and Applications

and output comprise Negative Big (NB), Negative Small (NS), Zero (Z), Positive Small (PS)
and Positive Big (PB). The two inputs used were the wheelchair rotation error (eψ) and the
change of wheelchair rotation error, ∆eψ. The controller outputs are the right and left wheel
torques, τR and τL. All membership functions of input and output parameters are
normalized for ease of control. Table 3 shows the implemented fuzzy rules for steering
motion control (FLC4), where the first row relates to right-wheel torque and the second
(shaded) row relates to the left-wheel torque.

∆eψ
NB NS Z PS PB
eψ
NB NB NB NS Z
NB
PB PB PB PS Z
NB NB NS Z PS
NS
PB PB PS Z NS
NB NS Z PS PB
Z
PB PS Z NS NB
NS Z PS PB PB
PS
PS Z NS NB NB
Z PS PB PB PB
PB
Z NS NB NB NB

Table 3. Fuzzy rules for steering motion

Steering to 30°
The final position of the wheelchair system can be seen in Figure 21. Figures 22 and Figure
23 show the angular positions of Link1, δ1 and Link2, δ2 respectively when the system was
set to steer at 30° causing the two-wheeled wheelchair to rotate to the left from its initial
position. Both links settled with small steady state error after the steering settlement was
achieved. As noted, they settled in less than 4s. As noted in Figure 24, the wheelchair rotated
very near to 30°, with < 0.1° of the steady state error. The output torques from each lifting
and stabilizing control of Link1 as well as the steering control are shown in Figures 25 and
26 respectively. Note that the output torques from FLC1 had the same magnitude and
direction for both right and left wheels. On the other hand, the output torque from FLC4
had the same magnitude but different in direction representing the fuzzy rules output for
steering motion. The torque between Link1 and Link2 can be seen in Figure 27. As noted, it
changed between +30Nm and -20Nm to maintain the upright stability of Link2 with human
payload during the steering motion. The resultant torques for both fuzzy controllers
(FLC1+FLC4) is shown in Figure 28. The system was then tested to rotate at a different angle
(negative angle leading to rotation to the right).

[Link]
Modular Fuzzy Logic Controller for Two-Wheeled Wheelchair 55

Fig. 21. Final steering position for 30° reference point

100
Angular position of Link1 (degree)

80

60

40

20

-20
0 5 10 15
Time (s)

Fig. 22. Angular position of Link1, δ1 (degree)

Angular position of Link2 (degree)

10

-10

-20

-30
0 5 10 15
Time (s)

Fig. 23. Angular position of Link2, δ2 (degree)

[Link]
56 Fuzzy Logic – Controls, Concepts, Theories and Applications

40
Wheelchair rotation (degree)

30

20

10

-10
0 5 10 15
Time (s)
Fig. 24. Wheelchair rotation, ψ (degree)

150
and stabilizing of Link1 (Nm)

Right torque
Wheel torques from lifting

100 Left torque

50

-50

-100
0 5 10 15
Time (s)
Fig. 25. Wheel torques (τR and τL) from FLC1 (Nm)

100
Right torque
steering control (Nm)

Left torque
Wheel torques from

50

-50

-100
0 5 10 15
Time (s)

Fig. 26. Wheel torques (τR and τL) from steering motion control, FLC4 (Nm)

[Link]
Modular Fuzzy Logic Controller for Two-Wheeled Wheelchair 57

50
Link1 and Link2 (Nm)

0
Torque between

-50

-100

-150

-200
0 5 10 15
Time (s)
Fig. 27. Torque between Link1 and Link2, τ2 (Nm)

150
Resultant wheel torques (Nm)

Right torque
100 Left torque

50

-50

-100
0 5 10 15
Time (s)
Fig. 28. Resultant wheel torques (Nm)

5. Conclusion
Fuzzy logic is one of the control techniques that is very close to human feelings and
expressions. It can be easily understood and implemented although the knowledge about
classical or conventional control system is not much identified. Nevertheless the general
knowledge of the system involved must be generally known otherwise it is difficult to
formulate a fuzzy controller for such system. If the system involved is known to be linear,
and simple thus it is more worth to start with conventional Proportional-Integral-
Differential (PID) controller. Otherwise if the system is known to be very complex, nonlinear
and ill-defined type of system, then it is suggested to use one of the computational
approaches such as fuzzy logic. This method was successfully implemented in the two-
wheeled wheelchair system where a modular fuzzy control (MFC) was developed and
implemented for controlling lifting and stabilizing mechanism, linear and steering motion
control. Note that since a wheelchair is a main means of transport for disabled and elderly
people, this two-wheeled wheelchair system would allow the user to achieve a higher level
of height without assistance and hence independence. The wheelchair has been modeled as
a double inverted pendulum. The integrated two-wheeled wheelchair with a human model

[Link]
58 Fuzzy Logic – Controls, Concepts, Theories and Applications

has been imported as the plant into Matlab/Simulink environment for control and
evaluation purposes. Therefore, fuzzy logic techniques have been found suitable for control
of the two-wheeled wheelchair.
A Modular Fuzzy logic Control (MFC) approach has been adopted, where the control tasks are
divided into primary and secondary tasks (subsystems), and FLC modules have been
designed and executed for the various control tasks accordingly. Among the control tasks,
lifting and stabilizing in the upright position are considered as the primary control system
task. Secondary system tasks include linear motion and steering motion. The MFC strategy
developed is based on a hierarchical approach whereby the primary subsystem must be
executed followed by selection of secondary subsystems. Both linear and steering motions
have been successfully controlled independently using a two-input two-output PD-type FLC.
The proposed MFC has been successfully implemented and tested within simulated
exercises for two-wheeled wheelchair application. The results presented proved that the
MFC approach works very well in controlling highly nonlinear systems such as a
wheelchair on two wheels and significantly reduces the number of rules.

6. Acknowledgment
The authors would like to express their appreciation to the International Islamic University
Malaysia (IIUM) for sponsoring the publication of this book chapter.

7. References
Ahmad, S., N. H. Siddique and M. O. Tokhi (2011). A modular fuzzy control approach for
two-wheeled wheelchair. Journal of Intelligent and Robotic Systems Springer Journal,
Vol. 64(3-4), pp. 401-426.
Bessacini, A. F. and R. F. Pinkos (1995). A hierarchical fuzzy controller for intercept
guidance with a forbidden zone. Newport, Naval Undersea Warfare Center
Division.
Engelbrecht, A. P. (2007). Computational intelligence: An introduction. Chichester, England,
John Wiley & Sons.
Mamdani, E. H. (1974). Application of fuzzy algorithms for control of a simple dynamic
process. Proceeding of IEEE 121(12): 1585-1588.
Reznik, L. (1997). Fuzzy Controllers. Oxford, Newnes.
Siljak, D. (1991) Decentralized Control of Complex Systems, Academic Press.
Vries, T. J. A. D., C. V. Heteran and L. Huttenhuis (1999). Modeling and control of a fast
moving, highly maneuverable wheelchair. International Biomechatronics Workshop,
Enschede, Netherlands.
Wong, C.-C., H.-Y. Wang, S.-A. Li and C.-T. Cheng (2007). Fuzzy controller designed by GA
for two-wheeled mobile robots. International Journal of Fuzzy Systems 9(1).
Zadeh, L. A. (1965). Fuzzy sets. Information and Control 8(3): 338 - 353.

[Link]
 Fuzzy Logic - Controls, Concepts, Theories and Applications
Edited by Prof. Elmer Dadios

ISBN 978-953-51-0396-7
Hard cover, 428 pages
Publisher InTech
Published online 28, March, 2012
Published in print edition March, 2012

This book introduces new concepts and theories of Fuzzy Logic Control for the application and development of
robotics and intelligent machines. The book consists of nineteen chapters categorized into 1) Robotics and
Electrical Machines 2) Intelligent Control Systems with various applications, and 3) New Fuzzy Logic Concepts
and Theories. The intended readers of this book are engineers, researchers, and graduate students interested
in fuzzy logic control systems.

How to reference
In order to correctly reference this scholarly work, feel free to copy and paste the following:

Salmiah Ahmad, N. H. Siddique and M. O. Tokhi (2012). Modular Fuzzy Logic Controller for Motion Control of
Two-Wheeled Wheelchair, Fuzzy Logic - Controls, Concepts, Theories and Applications, Prof. Elmer Dadios
(Ed.), ISBN: 978-953-51-0396-7, InTech, Available from: [Link]
controls-concepts-theories-and-applications/modular-fuzzy-logic-control-for-two-wheeled-wheelchair-motion-
control

InTech Europe InTech China

University Campus STeP Ri Unit 405, Office Block, Hotel Equatorial Shanghai
Slavka Krautzeka 83/A No.65, Yan An Road (West), Shanghai, 200040, China
51000 Rijeka, Croatia
Phone: +385 (51) 770 447 Phone: +86-21-62489820
Fax: +385 (51) 686 166 Fax: +86-21-62489821
[Link]
© 2012 The Author(s). Licensee IntechOpen. This is an open access article
distributed under the terms of the Creative Commons Attribution 3.0
License, which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.