WHEELCHAIR OBSTACLE AVOIDANCE BASED ON FUZZY

CONTROLLER AND ULTRASONIC SENSORS

Malek NJAH and Mohamed JALLOULI

Research unit on Control & Energy Management (CEM)

University of Sfax, Sfax Engineering School, BP W,1173-3038 Sfax, Tunisia.
Phone: (216-74) 274.088, Fax. (216-74) 275.595.
e-mail: njahmalek@yahoo.fr and mohamed.jallouli@enis.rnu.tn

ABSTRACT maining at the individual, if the goal is to reduce disability

rather than the disability, the projects seek to restore some
Electric wheelchair is one of the most used for the
of functional capacity lost by relying on aid technical. This
movement of disabled and aged people. This work in-
article discusses a method of navigation for obstacle avoid-
troduces an obstacle avoidance system aimed for provid-
ance using techniques developed in robotics. The peculiar-
ing more autonomous navigation of a electric wheelchair
ity of this electric wheelchair for disabled’s own powers of
(EW) in unknown indoor environments. These technolo-
perception, action and information processing. This gives
gies seek to increase the independence of people with dis-
the possibility of diagnosis, decision making and above
abilities and improve their quality of life by making the
all interaction with the environment by avoiding the ob-
most of each individuals abilities. Furthermore, the in-
stacles facing the courtyard of navigation. The goal is not
tegration of an ultrasonic sensor to avoid obstacle and a
to achieve a self wheelchair but on the contrary, as far as
fuzzy controller to generates velocity for aim to join the
this may foster a cooperation between man and machine
target position.
for the primary purpose of service to that person. Auton-
A prototype of EW has been equipped with control unit
omy of handicapped people relies on a third person, or
based on two micro-controller. The first for manage the
on a manual wheelchair. For people with severe handicap
motors velocity. The second for explore the ultrasonic
the powered wheelchair is usually prescribed. Security
sensors. Two micro-controller exchanges the information
near staircases, obstacle avoidance and door passing re-
with a PC board type PC104, which is used to process data
quire practice and concentration, even with a wide variety
from sensors and encoders. The information is processed
of special interfaces (voice, joystick, touch screen, key-
by a control algorithm based on fuzzy logic. The control
board...). Various and numerous works [4] ,[3],[2] provide
algorithm is optimized using the gradient method to min-
intelligent cooperative solutions between human and on-
imize the path traveled to reach the desired position. The
board automatic functions to drive powered wheelchairs.
practical implementation demonstrates the algorithm va-
lidity for obstacle avoidance and goal achievement with a
minimum path and greater security.
The WAD wheelchair (Wheelchair Attractor Dynam-
Index Terms— Wheelchair, micro-controller, PC104, ics) [14], [6] is another model that arranges a help system
ultrasonic sensor, fuzzy logic, obstacle avoidance. navigation adapted to a large resides public. The com-
mand architecture of this Project is installed currently on
1. INTRODUCTION an electric wheelchair CRUISER model of INVACARE. It
is equipped with a DX system of basis (joystick and power
Several studies have shown that both children and adults module), to which we added the module DX Key, that per-
benefit substantially from access to a means of indepen- mits interfacing via the computer parallel port.The WAD
dent mobility, including power wheelchairs [7]. For young project is aimed to provide limited autonomy to electrical
children, independent mobility serves as the foundation for wheelchairs. The primary focus is to provide a secure ob-
much early learning. For adults,independent mobility is stacle avoidance behavior based on infrared distance sen-
an important aspect of self-esteem. For all individuals, in- sors which generate contributions to the heading direction
dependent mobility increases vocational and educational dynamics that steers the wheelchair away from obstruc-
opportunities. The independent mobility reduces depen- tions. Moreover, the attractor dynamics approach is used
dence on care givers and family members and promotes to integrate the obstacle avoidance behavior to a user de-
feelings of self-reliance [13]. fined target acquisition behavior, in which the direction
Science and technology offer several ways to reduce and the distance to the target are indicated by the user at
the disadvantage caused by disabilities. The focus area is different points in time.
located at of the environment and of the individual. Re-

978-1-4673-5285-7/13/$31.00 ©2013 IEEE

 2. BACKGROUND
𝑑𝑋 𝑉𝑅 + 𝑉𝐿
= cos𝜃 (1)
In this paper, a fuzzy controller is synthesized while us- 𝑑𝑡 2
ing the kinematics model of the electric wheelchair. This 𝑑𝑌 𝑉𝑅 + 𝑉𝐿
= sin𝜃 (2)
controller gives the possibility to reach a desired position 𝑑𝑡 2
and avoid obstacles. Finally this controller implementa- 𝑑𝜃 𝑉𝑅 − 𝑉𝐿
tion has been achieved on a card developed to manage the = (3)
𝑑𝑡 𝐿
data transfer and send the command of the system from a
The discreet form of the kinematic model :
computer. The information delivered by the encoders and
the ultrasonic sensors serve to verify and to compare the 𝑉𝑅𝑘 + 𝑉𝐿𝑘
𝑋𝑘+1 = 𝑋𝑘 + 𝑇 cos𝜃𝑘 (4)
results found in simulation. 2
𝑉𝑅𝑘 + 𝑉𝐿𝑘
𝑌𝑘+1 = 𝑋𝑘 + 𝑇 sin𝜃𝑘 (5)
2
𝑉𝑅𝑘 − 𝑉𝐿𝑘
3. THE KINEMATIC MODEL 𝜃𝑘+1 = 𝜃𝑘 + 𝑇 (6)
𝐿
* T is the sampling period.
Wheelchair contains two independent driving wheels, which
can be oriented and commanded by acting on the speed of
each wheel. 4. SYNTHESIS OF THE FUZZY CONTROLLER
NAVIGATION
The wheelchair is a vehicle based on two independent
coaxial driving wheels located at the rear and two free- Several papers [11] [12] propose solutions of control based
rotating wheel that ensures the static stability of the ve- on a fuzzy theory. The controller proposed in this paper
hicle. It is therefore a unicycle mobile base whith non- permits to the wheelchair to reach the target position (𝑋𝑇
holonomic character. , 𝑌𝑇 ) from a current position (𝑋, 𝑌 ). The controller out-
puts are the velocities 𝑉𝐿 and 𝑉𝑅 which are necessary to
Y apply respectively to the left and the right wheels of the
T arget wheelchair to orient it and to move it so that it reaches the
YT desired position.

The controller inputs the polar coordinates of the target

θT point calculated with respect to a reference mark related to
the wheelchair. The expressions of these two variables are
d
given by the following equations :
X1

√
= (𝑋𝑇 − 𝑋)2 + (𝑌𝑇 − 𝑌 )2
ϕ
Y1
VL 𝑑 (7)
V 𝜑 = 𝜃𝑇 − 𝜃 (8)
VR
θ where:
Y
O1

(𝑌𝑇 − 𝑌 )
𝜃𝑇 = tan−1 (9)
(𝑋𝑇 − 𝑋)
X
0 XT
Figure 2 represent the wheelchair control system. Hys-
X
terisis swich combine two method of control. First the op-
Figure 1. Scheme of the electric wheelchair. timized fuzzy controller to join target [1] and second opti-
mized fuzzy controller to avoid obstacles. The avoidance
of obstacles is necessary for wheelchair. This functional-
Figure 1 depicts a schematic representation of the elec- ity is essential to allow the safe navigation of a wheelchair
tric wheelchair. in a dynamic environment. It is assumed that the EW is
𝑉𝐿 , 𝑉𝑅 : are respectively the linear speeds of the wheels provided with eight sensors.
left and right. Two sensor is placed in front of the wheelchair, four are
𝐿 : is the distance between the two driving wheels. placed on its left and on its right and two placed on its
𝜃 : is the angle of orientation of the wheelchair. rear. These sensors are used to detect the obstacles in four
𝑉 : is the linear speed of the wheelchair. directions.
The position (𝑋, 𝑌 ) and orientation (𝜃) of the wheelchair Two incremental encoders placed on each motor axis gives
can be represented by a vector to three components 𝑃 = the turn speed of wheel.
(𝑋, 𝑌 , 𝜃)𝑇 .
 XT Hysterisis This figure shows the result given by the navigation
YT calculating dr swich
Wheechair d the distance
algorithm not optimize presented by the red curve and the
df
X
position ϕ between ob- blue curve represent the navigation using optimized fuzzy
relative to stacle and the dl
Y the target controller.
wheelchair
θ Consider now, more simulation trajectory such as (𝑋𝑇 ,
𝑌𝑇 ) = (6000mm,0mm) and a departure angle equal to 0
obtained results are given by the following figure 4.
Optimized
Optimized
fuzzy con-
fuzzy controler 2000
troler to avoid
to join target
obstacle

VR VL 1500

Utrasonic sensors

1000

500 Obstacle

y
target
Encoders 0

Figure 2. Wheelchair control systems. −500

Optimized tragectory

Not optimized tragectory

In this paper, the fuzzy controller to avoid obstacles −1000

0 1000 2000 3000 4000 5000 6000 7000
optimized with gradient method. The contribution our study X
is the optimization of the consequence of fuzzy controller
by means of the gradient method. The implementation of Figure 4. Trajectory given by the fuzzy controller to avoid
this fuzzy controller on an electric wheelchair has been re- obstacles (𝑋𝑇 , 𝑌𝑇 ) = (6000mm,0mm).
alized in order to verify and therefore compare the results
obtained in simulation. The figure 5 represents another case of obstacle avoid-
ance. The objective is to achieve a (𝑋𝑇 , 𝑌𝑇 ) = (5000mm,
5. SIMULATION RESULTS 5000mm).

5500
In this section we describe different result obtained by sim-
Target
ulation. We considered the case for (𝑋𝑇 , 𝑌𝑇 ) = (6000 , 5000

6000) in mm and the start angle is Pi/4. the distance be- 4500 Optimized trajectory
tween the chair and the obstacle, is computed by the inter-
4000 Not optimized tragectory
section of the sensor direction different side of the obsta-
cle. The simulation result shown in figure 2. 3500

3000
7000
Y

2500
Target
6000 Obstacle
2000

1500
5000

1000

4000
500
Y

0
3000 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500
Obstacle
X
Optimized tragectory
2000
Not optimized tragectory Figure 5. Trajectory given by the fuzzy controller to avoid
1000
obstacles (𝑋𝑇 , 𝑌𝑇 ) = (5000mm,5000mm).

0
0 1000 2000 3000 4000 5000 6000 7000
The results obtained by the simulation shows the proper
X functioning of the navigation algorithm. The optimization
of obstacle avoidance algorithm gives better results than
Figure 3. Trajectory given by the fuzzy controller to avoid obtained by not optimizing algorithm.
obstacles (𝑋𝑇 , 𝑌𝑇 ) = (6000 , 6000). In the next section we’ll detail the component integrated
into the Electric Wheelchair and give the practice result
obtained following a test on EW.

6. THE OBSTACLE AVOIDANCE SYSTEM OF

THE WHEELCHAIR

the based control system of the wheelchair composed as

the joystick, power module ,two DC motors and two bat-
teries of 24V. Thereafter we have developed a card to man-
age eight ultrasonic sensors. The sensors function is to Figure 7. The front of the EW.
provide information on the environment and the chair (po-
sition, orientation). These measures will be used in ob-
stacle avoidance programs and autonomous navigation al-
gorithms for cooperation between the handicapped person
and the autonomous wheelchair.
The electronic card is developed based on a micro-
controller PIC 16F877 figure 6. This card manages the ul-
trasonic sensors (SRF04, SRF08) and sends the distances
measured by different mode (one each sensor, sensors front,
rear sensors or all sensors ).

Figure 8. The rear of the EW.

7. EXPERIMENTAL RESULTS

We use the first order Takagi-sugeno fuzzy system as a

fuzzy controller. It is programmed by the declaration of
variables,angle distance, speed of right and left wheel. The
control’s objective is to make the wheelchair avoids obsta-
cles and reach any target position. We suppose that (0,0)
is the starting point control results.
figure 9 represents the obstacle position and the three
Figure 6. Electronic card based on PIC16F877. trajectory traveled by the EW in the three following cases:

* 1 The micro-controller PIC16F877.

* 2,3 The circuit MAX232 used to send and receive 5500

information from the computer. 5000

Not optimized Experimental tragectory

* 4,5 stable power 5v. Optimized experimental tragectory

4500
* 7,8,9,10 four DB9 for connecting the SRF04 sensors
Optimized simulation tragectory
and the electronic card. 4000

* 11,12,13,14 four DB9 for connecting the SRF08 sen- 3500

sors and the electronic card.
3000
We used two type of ultrasonic sensors SRF08 and
Y(mm)

SRF04. They are based on measuring the elapsed time 2500

between the echo transmission and the echo return. The 2000
elapsed time used to define the distance between obstacle
1500
and EW. Now we presented the ultrasonic sensor and elec-
tronic card installed on EW. 1000

The position and the orientation of the wheelchair are 500

estimated by the encoder Figure 7, which considers the
measure of rotations of the driving wheels. 0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500
X(mm)
The control system installed on wheelchair can be rep-
resented by Fig. 8. The control system allows the wheelchair
detect and avoid obstacles. For these reasons, it was neces- Figure 9. trajectory traveled by EW.
sary to determine the instantaneous rotation angle and the
speed of two driving wheel to consider the instantaneous * The red curve obtained by the experimental with not
position and speed of the wheelchair. optimized fuzzy controller to avoid obstacles.
 Based on XML Files with Accessibility Information
6000
”, WISP 2009, 6th IEEE International Symposium on
Intelligent Signal Processing, Budapest, Hungary, pp.
Optimized experimental tragectory
5000 26–28.
Optimized simulation tragectory
[5] K. Chabane, “Exploitation of the redundancy for the
4000
order coordinated of a mobile manipulator of aid to
the handicapped people”, Thesis report, University
Y(mm)

3000 d’Evry, 30 november 2006.

[6] J. Pergandi, P. Mallet, D. Mestre : “Evaluation of a

2000
help to the navigation of a ”intelligent” wheelchair”,
Handicap 2006, november 2006.
1000
[7] Richard C. Simpson, “Smart wheelchairs”,Journal of
Rehabilitation Research and Development, Vol. 42,
0
0 1000 2000 3000 4000 5000 6000 pp. 423-436, July 2005.
X(mm)

[8] T. Lu, K. Yuan, H. Zhu and H. Hu,“An Embedded

Figure 10. Optimized trajectory traveled by EW. Control System for IntelligentWheelchair”, Proceed-
ings of 27 Annual Int. Conf. of the IEEE Engineering
in Medicine and Biology Society, Shanghai, China, 1-
* Black curve obtained by the experimental results to 4 September 2005.
join the target (𝑋𝑇 , 𝑌𝑇 ) = (5000mm,5000mm).
* Blue curve obtained by the simulation of the opti- [9] M. Imamura, R. Tomitaka, Y. Miyazaki, K. Kobayashi
mized fuzzy controller to join target and avoid obstacles. and K. Watanabe, “Outdoor waypoint navigation for
an intelligent wheelchair using differential GPS and
8. CONCLUSION INS”, in SICE annual conference, Sapporo, 2004.

[10] D. Ding, Rory , Fellow, S. Guo, Thomas, “Anal-

In this paper we developed a control system to assist the
ysis of Driving Backward in an Electric-Powred
disabled persons who have some problems to drive a stan-
Wheelchair”, IEEE transactions on control systems
dard wheelchair. We use the fuzzy controller. This con-
technology, vol. 12, 6 November 2004.
troller is of the type Takagi-Sugeno of the order 0. Fuzzy
controller results are satisfying since the target position is [11] KUO Chung-Hsien, HUANG Hsu-Lung and LEE
reached and obstacle avoidance. The gradient method used Ming-Yih , “Development of agent based autonomous
to optimize the fuzzy controller developed on first step. robotic wheelchair control systems”, Biomedical En-
Many other command it possible to use in this electric gineering Applications, vol. 15, pp. 223-234 , 2003.
wheelchair. We integrate the embedded computer PC/104
used to analysis different type of command such us voice [12] C. Kuo, H. Yeh, C. Wu, K. Hsiao, “Development
commands, controlling the wheelchair using eye place- of Autonomous Robotic Wheelchair Controller Using
ment. Embedded Systems”, The 33rd Annual Conference
of the IEEE Industrial Electronics Society (IECON),
Nov. 5-8, 2003, Taipei, Taiwan.
9. REFERENCES
[13] Richard C. Simpson, , Daniel Poirot and Fran-
[1] M. Njah, M. Jallouli and N. Derbel “Optimal cie Baxter “The Hephaestus Smart Wheelchair Sys-
Fuzzy Controller for the Navigation of an Electric tem”,IEEE TRANSACTIONS ON NEURAL SYSTEMS
Wheelchair”, TSSD Transactions on Systems,Signals AND REHABILITATION ENGINEERING, Vol. 10,
and Devices, Vol. 6, 2011. NO. 2, JUNE 2002.
[2] F.Leishman, O.Horn ,G.Bourhis “Smartwheelchair- [14] P. Mallet, G. Schner, “WAD Project where Attractor
control through a deictic approach”, J. Robotics and Dynamics aids wheelchair Navigation”, Proceedings
Autonomous Systems 58, pp. 1149–1158, 2010. of the 2002 IEEE/RSJ Intl. Conference on Intelligent
[3] Y. Takahashi and H. Seki, Memeber, IEEE “Fuzzy Robots and Systems, Lausanne, Switzerland, 2002.
Logic Based Regenerative Braking Control System of [15] M. Mazo, J.C. Garcia, F.J. Rodriguez, J. Ureila, J.L.
Electric Wheelchair for Senior Citizen”, 11th IEEE L-0, F. Espinosa, “Experiences in assisted mobility:
Int. Conf. on Rehabilitation Robotics, Japan, pp.23- the SIAMO project ”, The IEEE Conf. on Control
26, June 2009. Applications , vol. 2, pp. 766-771, Sep. 2002, Glas-
[4] F. Paulo, S. Amaral, J. Carlos, G.Garcia, F. Teodiano, gow,Scotland, U.K.
B. Filho, M. Mazo“Ambient Assisted Route Planner