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

net/publication/266630301

Modelling and Simulation of an User-Wheelchair-Environment System

Conference Paper · November 2012

DOI: 10.2316/P.2012.761-022

CITATIONS READS
2 2,618

5 authors, including:

Ignace Tchangou Toudjeu Yskandar Hamam

Tshwane University of Technology Tshwane University of Technology
12 PUBLICATIONS   51 CITATIONS    452 PUBLICATIONS   4,384 CITATIONS   

SEE PROFILE SEE PROFILE

Karim D. Djouani Barend J Van Wyk

Université Paris-Est Créteil Val de Marne - Université Paris 12 Tshwane University of Technology
335 PUBLICATIONS   3,199 CITATIONS    141 PUBLICATIONS   1,445 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Rail Condition Assessment Using 3D Image Acquisition and Modelling View project

Optimising students throughput View project

All content following this page was uploaded by Karim D. Djouani on 05 January 2016.

The user has requested enhancement of the downloaded file.

 Proceedings of the IASTED International Conference
Modelling and Simulation (AfricaMS 2012)
September 3 - 5, 2012 Botswana, Africa

MODELLING AND SIMULATION OF AN USER-WHEELCHAIR-

ENVIRONMENT SYSTEM
Ignace T. Toudjeu1 , Yskandar Hamam2 , Djouani Karim3, Barend J. van Wyk4 , Eric Monacelli5
1,2,3,4
F’SATI and Tshwane University of Technology, 5 LISV and UVSQY
ToudjeuIT@tut.ac.za, HamamA@tut.ac.za, DjouaniK@tut.ac.za, VanwykB@tut.ac.za, eric.monacelli@lisv.uvsq.fr

ABSTRACT A brief summary and further work are mentioned in the

Research on wheelchairs is on its growing phase and concluding section.
assistive aids brought to wheelchair users in terms of
adaptability to electric wheelchair have not yet been
perfected. This paper presents the development of a User-
Wheelchair-Environment (UWE) system in order to ease
driving evaluation of wheelchair user. The system is
modelled using Matlab/Simulink and it provides useful
information such as linear and angular velocity and
displacement in a 3D-virtual environment. The model
consists of three main components (user interface,
wheelchair model and virtual environment) which are
individually modelled and later integrated altogether to
form the UWE system. Simulation results are also
presented. Figure 1. Block diagram of the UWE system

KEY WORDS 2. User-Wheelchair-Environment Modelling

Wheelchair, driver, simulation, virtual reality, modelling
The UWE system comprises a user controlling the
wheelchair via a joystick interface in a virtual
1. Introduction environment. In this section, we elaborate on the user
interface model, dynamic wheelchair model and a virtual
With the increase in the number of both disabled and environment model also referred to as a 3D world.
senior citizens, there is a great need of assistive systems
for mobility enhancement of wheelchair users. Driving a 2.1 User interface
wheelchair today imposes burdens on their users both
physically and emotional [1][2]. Some improvements This part of the overall model allows the user to control
were done by developing dynamic models which can be the wheelchair using the joystick. It also acts as a data
controllable [3][4][5]. Most of these dynamic models acquisition block for our UWE system.
were restricted to movements on flat surfaces ignoring the
-K- 2
effects of gravitational forces experienced by wheelchair
Angular Velocity
users in real life. In [6] a dynamic model considering the Axes
Gain1 Dead Zone1 (w)
movement on an incline surface, which can improve joyinput
safety of the user, was developed. Other methods using Buttons -K- 1
Linear Velocity
virtual reality were also proposed [7][8][9]. However, Gain2 Dead Zone2 (v)
Joystick Input
these methods were still based on a 2D motion. Therefore
the development of a system which integrates both a 3D
dynamic model and a 3D virtual model is presented. This Terminator

system can be used as an evaluation tool to improve

manoeuvring skills of wheelchair users. Critical
conditions can be simulated without endangering the life
of the subject. The UWE system consists of three main
components. See Figure 1 for block diagram. In section 2,
the modeling of the UWE system is presented. Simulation
of the UWE system is discussed in section 3 and
simulation results are also included.
Figure 2. User interface Simulink model
The joystick input in Figure 2 gets a command from the
external joystick connected to the computer via an USB
port. That command obtained for linear and angular
velocity is then amplified and fed to the Dead Zone
Simulink symbol. The Dead Zones in Figure 2 filter out

DOI: 10.2316/P.2012.761-022 198

all slow movements in any direction due to the high Considering the position of the wheelchair in the 3D
sensibility of the joystick and only the clear intention of space ( xyz -coordinate system) indicated by
user is then used to control the wheelchair. T
q   xg , yg , z g ,   with x , y
g g , z g  being the Cartesian
2.2 Wheelchair dynamic model
coordinate the centre of gravity of the wheelchair, and 
Most electric wheelchairs are conceived as nonholonomic the angle of rotation of the wheelchair the Lagrange
mechanical systems. The rear wheels are independently function is expressed on the following form:
driven by a motor. The two front wheels are castor
wheels. The dynamic models used in this paper is M  q  q  C  q, q  q  G  q   E  q   AT  q   (2)
comparable to the one proposed in [6]. This model
considers a third dimension which is the wheelchair on an where
inclined surface. We will briefly discuss the mathematical
 M 0 0 lM cos sin 
model representing the wheelchair in an incline as  
illustrated in Figure 4.  0 M 0 lM cos cos 
M  q 
 0 0 M 0 
2.2.1 Mathematical model of the wheelchair  
 lM cos  sin lM cos cos 0 Iz 
The mathematical model of the wheelchair is based on its
geometry and kinematic equations. The geometry of the 0 0 0 lM cos  cos  
 
wheelchair is dependent on its physical construction in the 0 0 0 lM cos  sin  
plane (see Figure 3) while the kinematic equations C  q, q    ,
0 0 0 0 
characterize its motion in the space.  
0 0 0 0 
y
X
Y
 Mg sin  cos  
l  
 Mg sin  sin  
L G  G q   
yg 0
 
 Mg sin   yg cos   xg sin   
2b 2r
 
R

xg x  cos  cos 

 r r 
 
Figure 3. Wheelchair platform on a flat plane  sin  sin 
 
E  q 4 2  r r  ,    R  ,
   L 
l  0 0 
 b b 
Mgl sin  
 r r 

   cos  sin  cos  cos sin  l 

Figure 4. Wheelchair platform on an incline plane
Lagrange formalism (see Equation 1) was used to derive
differential equations describing the time displacement of
the wheelchair, yet considering the nonholonomic
kinematic constraints.
1 1
L  M  xg2  y g2  zg2   I z 2  lM cos   xg sin   y g cos  
2 2
Mg sin   xg cos  y g sin  
(1)
A  q  24    and 
 sin  sin   sin  cos cos  0 
the vector of Lagrange multipliers. The left and right
torque is designated by  L and  R respectively. When
both are equal, a rectilinear motion occurs and rotational
motion in the contrary case.
Assuming that the wheels of wheelchair roll without
slipping, lateral movements are not feasible, therefore the
matrix associated to the nonholonomic constraints is
perpendicular to the axis of the driving wheels are written
in equation form as A   q  q  0 . Considering

199
S  q nm , a set of smooth and linearly independent The wheelchair dynamic model can be rewritten as a
square-MIMO nonlinear system on the following form
vector field spanning the null space of A  q  such that
2
S T  q  AT  q   0 , we derive the kinetic equation as 
 x  f  x    g j  x u j
j 1
follow: 
 y   h x , h x T
  1   2  
q  S  q  (3)
where x   2 , u   2 and y   2 are state, control input,
 cos  l cos  sin   and output vectors respectively.
   
sin  l cos  cos  
With S  q    and    . The As the goal is to track the trajectory from a given pair of
 0 l sin   
   reference linear velocity r and angular positioning of the
 0 1 
wheelchair  r , the tracking error is then expressed as
velocities  and  are respectively the linear and
angular velocities of the distinguished point G , the centre
e11     r
of gravity of the wheelchair and user combined.  2 , (6)
e1     r
The derivative of Equation 2 with respect to time gives
and should be forced to zero. This is achieved by input-
q  S  q   S  q  (4) output linearization. A system is input-output linearizable
in a field    2 if it has a relative degree vector
  sin  l cos  cos      1 , 2  obtained from successive derivation of
  
  cos  l cos  sin   Equation 6 with respect to time until the control input
With the matrix S  q    .
 0 0  T
   u1 , u2  appears. Hence, after applied Lie derivative to
 0 0  Equation 6, we have

Taking Equation 3 and 4 into Equation 2 and after  1 u

1  1
simplifications, we obtain a more appropriate control
1

e1    r  L f e1  Lg1 e1 Lg2 e1  
1

system representation for the dynamic model of the   u2 
 2  
e1     r    r  e2
2
wheelchair as follow

e 2      L2 e 2  L L e 2 L L e 2  u1 
 
 1 1  f g2 1  
 g sin     2 r f 1 f g1 1
 u2 
      Mr Mr
 R  
     Mg sin  yg cos  xg sin     b b  L 
     
 Ml 2 cos2  Iz 
 r Iz  Ml2 cos2  r Iz  Ml 2 cos2 
 
  Lg1 e11   g sin 
(5) 
where  Mg sin   yg cos   xg sin  
The model as expressed above, show that the wheelchair
2 2
 L f e1    r
 Ml 2 cos 2  I z
in an inclined flat surface is a nonlinear system. Results
found in [6] can also confirm this nonlinearity, thus the
need of a controller. with a relative degree vector equal to   1, 2  and the
decoupling matrix
2.2.2 Wheelchair controller
 1 1 
Many techniques have been used to control nonholonomic  Lg1 e11 Lg2 e11   
Mr Mr
systems. Dabo [10] applied two approaches to control the D  x    
 L f Lg e12 L f Lg2 e12   b b 
wheelchair and positive results were obtained for both  1 
 r I  Ml 2 cos2 r I z  Ml 2 cos 2 
methods. They are NCGPC (Nonlinear Continuous-time 
 z   
Generalized Predictive Control) and feedback (7)
linearization. The latter, also implemented in [6], is used Being that the sum of the relative degrees  1   2  equal
in this paper. The goal being, for a given command, the
to 3 greater than the wheelchair system described in
motion along the trajectory should correspond.

200
Equation 5, it is then convenient to re-write Equation 3 as 2.3 Environment modelling
follow
In order to simulate the wheelchair behaviour one requires
  g sin   
 1 1  a 3D model for a wheelchair and a 3D model for a
Mr Mr 
       preferred environment. In this work, we used Virtual
    Mg sin   y g cos   x g sin     b b   R 
     r I  Ml 2 cos2   
r  I z  Ml 2 cos 2    L  Reality Modelling Language to create the 3D model of the
   Ml 2 cos2  I z
   z  
     
   0 0  wheelchair and also two virtual environments; one
 
(8) representing a Turn-Right path and the other one a Ramp
path of slope 5 degrees. All the 3D models were created
  g sin   according to the ADA standards. Figures 5-7 illustrate our
  3D models. These 3D models were integrated with the
 Mg sin   yg cos   xg sin    abovementioned models (user interface and the
where f  x    ,
 Ml 2 cos 2  I z  wheelchair dynamic model) to form the UWE system.
   The Simulink model of UWE system is presented in
 
Figure 8.
 1 
 Mr 
 
b
g1  x    ,
 r I  Ml 2 cos 2 
  z 
 0 
 
 
 1 
 Mr 
 
 b  , with x   ,  ,  T .
g2  x   
 r I  Ml 2 cos 2   
  z 
 0 
  Figure 5. 3D model of the electric wheelchair
 

The choice of the state feedback law is very important as

it compensates for the nonlinearity occurring in the input-
output behaviour of the wheelchair. The decoupling
matrix (Equation 7) being invertible for its determinant
not equal to zero for I z  Ml 2 cos 2  0 , the state
feedback law linearizing wheelchair system is given by
Figure 6. 3D-model of a Turn-Right Path
1
 u1   Lg1 e1
1
Lg2 e11  1  L f e11 
    ,
 u2   L f Lg1 e1 L f Lg2 e12 
2 2  L2f e12 
 

T
where   1 ,2  is proportional to the tracking error;
1   K1e11 and 2   K 21e12  K 22 e22 . The gain matrix
should be chosen such that the linearized system is stable.
In order words, the gains K1 , K 21 and K 22 are such that
Figure 7. 3D-model of a Ramp Path
all the poles of the linearized system belong to the left
hand side of the s-plane.

201
 Figure 8. User-Wheelchair-Environment simulation system

3. Simulation Results of the UWE System Linear velocity measurement

10
In this section, both simulation setup and results are
presented. Parameters such as the maximum velocities, km/h
0
the combined mass of the wheelchair and the user, the
calculated gains for stability control and the slope of each
-10
surface plane in the virtual environment are displayed in
Table 1. 0 5 10 15 20 25 30

Angular velocity measurement

Table 1. Simulation setup 0.5

Parameters Values
rad/s
0
User interface
Maximum velocity vmax =12Km/h, max =0.4rad/s
Wheelchair -0.5
Kinematics b=0.35m, l=0.25m, r=0.2m 0 5 10 15 20 25 30
time(s)
Dynamics Mass =80kg, Iz=39.733kgm2
Control gains K1=7, K22=10, K21=200
Environment Figure 9. Velocities corresponding to the travelled
Flat surface Phi=00 trajectory on Turn-Right path presented in Figure 10.
Inclined surface Phi=50
3D Trajectory on the Turn Right Path

3.1 User interface results

This part of the overall model allows the user to control 46

the wheelchair using the joystick. The joystick used here
45.5
is a Logitech ATTACK 3; it also includes various buttons
which can be used to customize other applications. The z 45

gains (Gain1 and Gain2) appearing in Figure 3 correspond

44.5
respectively to the maximum angular and linear velocities
shown in Table 1. Resulting graphs of velocities are 44
-250
50
illustrated in Figure 9 and Figure 11. These graphs are 100
-150
-200

represented with respect to the trajectories created by the 150

200
-100
-50
user. 250 0 x
y

Figure 10. 3D representation of the travelled trajectory by

an user on a flat surface (Turn-Right path, phi=00)

202
3.2 Dynamic model results 4. Conclusion
Results proving the stability, controllability and The UWE system was modelled and simulated in this
manoeuvrability of the whole system are represented in paper. The UWE system comprises three key
this subsection. The velocities, both linear and angular, components: User interface, wheelchair and Environment.
obtained from the user interface correspond to the Each component interacts with one another. This
travelled trajectory. Graphs shown in Figure 11 and interaction is beneficial as relevant information such as
Figure 12 describe respectively the 3D representation of velocities (linear and angular), accelerations (linear and
the trajectories in Figure 9 and Figure 10. angular), and displacement of user-wheelchair in the
environment can be measured from the UWE system.
Linear velocity measurement
Besides the fact that the UWE system is a good measuring
tool and can be used to ease driving evaluation of the
10 wheelchair user, it can also be used to develop a
wheelchair user model. This development of wheelchair
km/h
0
user model and its integration in the UWE system will be
explored in future work.
-10
Acknowledgements
0 10 20 30 40 50 60
The authors would like to thank Tshwane University of
Angular velocity measurement
0.5
Technology and F’SATI for the technical support brought
to them during this research.

rad/s References
0

[1] Y. Adachi, K. Goto, Y. Matsumoto & T. Ogasawara,

Development of Control Assistant System for Roboic
-0.5
0 10 20 30 40 50 60
Wheelchair-Estimation of User’s Behaviour based on
time(s) Measurements of the Gaze and the Environment, Proc.
2003 IEEE International Symposium on Computational
Figure 11. Velocities corresponding to the travelled Intelligence in Robotics and Automation, Kobe, Japan,
trajectory on Ramp path presented in Figure 12. July 16-20, 2003.

3D Trajectory on the Ramp Path [2] Y. Kuno, Y. Murakami and N. Shimada, User and
Social Interfaces by Observing Human Faces for
End
Intelligent Wheelchairs, ACM International Conf. Proc.
100
Series archive, Proc. of the 2001 workshop on Perceptive
80 user interfaces , Orlando, Florida, USA, 2001, pp. 1-4.
Slope = 5 o
60
z [3] H. Emam, Y. Hamam, E. Monacelli and I.
40
Mougharbel, Dynanic Model of An Electrical Wheelchair
Start
with Slipping Detection, EUROSIM 2007.
20 Slope = 0 o
80 Slope = 5 o
60
0
100 [4] M. van Gerven, Modeling and control of an
40 -200
-100 electrically driven wheelchair, MSc report, University of
-300
y 20 -400
x
Twente, Netherlands, Oct. 2006, pp. 3.

Figure 12. 3D representation of the travelled trajectory by
an user on an inclined surface (Ramp path ,phi=50).
The driving activities of an ordinary user are clearly
mapped by the resulting velocities and also by the
trajectories travelled. The straight trajectory is a result of
a very small or null angular velocity. The turning
trajectory as shown in Figure 9 and Figure 11, is a result
of a significant angular quantity.
[5] H. Niniss and A. Nadif, Simulation of the behaviour
of a powered wheelchair using virtual reality, Proc. 3rd
Intl. Conf. Disability, Virtual Reality & Assoc. Tech.,
Alghero, Italy, 2000, pp. 9-14.
[6] S. O. Onyango, Y. Hamam, M. Dabo, K. Djouani and
G. Qi, Dynamic Control of an Electrical Wheelchair on an
Incline, Proc. IEEE AFRICON ’09, Nairobi, Kenya, 23-
25 Sept. 2009, pp. 1-6.

203
 [7] A. Fattouh, Y. Dadam and D. T. Pham, MATLAB-
Based 3D Dynamic Model of a Powered Wheelchair,
Innovation production machines and systems: Fourth
I*PROMS Virtual Int.l Conf., 1-14 Jul., 2008, pp 295-300.

[8] H. Niniss, A. Nadif, Simulation System for Powered

Wheelchairs: Evaluation of Driving Skills Using Virtual
Reality, Proc. 7th European Conf. for the Advancement of
Assistive Technology, AAATE, Dublin, Ireland, 1-3 Sept.
2003, pp. 112-116.

[9] L. Chen and H. Fujimoto, Development of Wheelchair

Experience System and Spatial Recognition in Virtual
Reality Environment, Proc. 22nd IEEE IECON, Taipei,
Taiwan, 5-10 Aug. 1996, pp. 744.

[10] M. Dabo, Commande predictive non lineaire a

temps continu de systemes complexes: application a la
chaise roulante electrique et au moteur diesel
turbocompresse, Editions universitaires europeennes, 28
Oct. 2010.

204

View publication stats