Dynamic modelling and simulation of a three-Wheeled

Omnidirectional Mobile Robot: Bond graph approach

Saumya Ranjan Sahoo Shital S. Chiddarwar Veer Alakshendra
Department of Mechanical Engineering Department of Mechanical Engineering Department of Mechanical Engineering
Visvesvaraya National Institute of Visvesvaraya National Institute of Visvesvaraya National Institute of
Technology, Nagpur - Technology, Nagpur - Technology, Nagpur -
440010 440010 440010
Telephone-9438273289, Telephone-9561050130, Telephone-8446985971,
saumyasynergy@gmail.com s.chiddarwar@gmail.com alakshendra.veer@gmail.com

ABSTRACT An omnidirectional mobile platform is more powerful than a

This paper presents the dynamic analysis of a three-wheeled conventional mobile robot with regular wheels as it provides
omnidirectional mobile robot (TWOMR) using bond graph better manoeuvrability, has ability to turn in confine spaces and to
approach. The robot contain subsystem from different physical move in any direction. Due to its various advantages over
domain; the global model is obtained by integrating the model of conventional mobile robots; these robots are being commonly
each sub domain. Simulation is carried out using BG_V20 toolbox used in domestic applications, omni-wheel chairs, shop floor,
in MATLAB Simulink to show the effectiveness of the proposed medical services and so on. Although the omnidirectional mobile
dynamic model. The simulated results are verified with the robot provides a better mobility; it's quite difficult to develop the
Newton-Euler method to illustrate the efficacy of the dynamic dynamic model of the robot due to complicated mechanical
modelling technique. The developed bond graph model can be structure. Therefore, obtaining the dynamic model and motion
used for various control aspects of the robot. control of the omnidirectional mobile platform is sorely needed.
Dynamic modelling of an omni-wheeled mobile platform using
CCS CONCEPTS the Newton - Euler method is available in literature [1]. Williams
• Computer systems organization → Embedded and cyber- [2] presented the dynamics of an Omnidirectional mobile
physical systems→ Robotics → Robotic components manipulator without disturbance and friction using Langragian
approach. Jose[3] demonstrated the role of bond graph in
Keywords modelling and simulation of mechatronics system. Pathak et al.[4]
also developed a robust trajectory control scheme for a free
Mobile robot; dynamic modelling; bond graph; trajectory tracking
floating space robot using bond graph. Moreover, bond graph can
1. INTRODUCTION also be used for simulation of highly nonlinear system [5]. In this
Dynamic modelling plays an important role in operating and method, a complex system can easily be modelled by considering
controlling a physical system. The dynamic interaction among subsystems and later integrating them to obtain the global
various components of a multi-disciplinary physical system structure [6]. Gor et al.[7] presented a dynamic model of a
(mechatronics system) make it difficult to analyse and design the compliant legged quadruped robot based on the bond graph
system. Current research is focusing on design and understanding approach. Recently, railway brake system and hybrid vehicle
of the mechatronics system. Among various methods, bond graph system are also modelled using bond graph [8]. In this paper,
modelling techniques will play a major role in design of dynamic model of a three wheeled omnidirectional robot is
mechatronics systems in the coming years. Bond graph modelling developed by considering various mechatronics components from
is used in industries like automobile, hydraulics, aerospace, different physical domains using bond graph. To increase the
chemical and recently in bio-medical also. Although elementary effectiveness of the model, friction force and slip (between rollers
modelling of industrial robots exists; the bond graph method and ground) is also taken into consideration.
remains unexplored in the mobile robotics domain. The dynamic The outline of the paper is as follows: in section II, the
modelling of mobile robots is more complicated due to the terminology of bond graph is explained. In section III, the
interconnection of elements from different domains. As bond structure of the robot and the word bond graph is illustrated. In
graph modelling has ability of efficiently dealing with multi- section IV, the dynamic modelling of the robot is done using bond
disciplinary modelling, it provides a better platform to model graph technique. Section V, VI and VII describe the simulation
dynamics of such mobile robotic system. Several software like results and conclusions drawn from the study.
SYMBOLS Shakti, 20-Sim, BG_V20 (MATLAB, Simulink) has
been developed which allow to exploit bond graph modelling 2. BOND GRAPH TERMINOLOGY
technique [3]. The behaviour of a dynamic system satisfies the basic principle of
physics i.e. energy conservation, positive entropy production and
Permission to make digital or hard copies of all or part of this work for personal or classroom use
power continuity. Based on this concept, Prof. Henary M. Paynter
is granted without fee provided that copies are not made or distributed for profit or commercial (1965) [5] proposed a method in which power of a system can be
advantage and that copies bear this notice and the full citation on the first page. Copyrights for
components of this work owned by others than the author(s) must be honored. Abstracting with
mapped in a graphical form irrespective of its domain called as
credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, bond graph. Bond graph being a power conservative technique, it
requires prior specific permission and/or a fee. Request permissions from Permissions@acm.org.
AIR '17, June 28-July 2, 2017, New Delhi, India
uses the power of a system from different physical domains. A
© 2017 Copyright is held by the owner/author(s). Publication rights licensed to ACM. bond graph is a unified graphical language used to model multi-
ACM ISBN 978-1-4503-5294-9/17/06...$15.00.
https://doi.org/10.1145/3132446.3134866
physical system. The power transmission between two elements in
a system represented by a line called as bond and a semi headed
arrow is given at one end of the bond which gives the direction of
AIR 2017, June 28-July 02, 2017, New Delhi, India Saumya et al.

power flow as shown in Fig. 1. The important property of bond junction. Fig. 2 illustrates the 1 and the 0 junction in bond graph
graph is the causality concept which determines the causes and terminology.
effects in the system. The causality is represented by a cross
stroke perpendicular to the bond and placed near the element
whose effort parameter is known.

Figure 1: Bond graph representation

The power variables, effort and flow from different physical

domains are shown in Table 1. Figure 2. Illustration of (A)1- and (B)0-juction in bond graph

Table 1: Power variable in bond graph [5] 3. DESCRIPTION OF 3-WHEELED

Energetic domain Effort (e) Flow (f) OMNIDIRECTIONAL ROBOTIC SYSTEM
Electrical Voltage (V) Current (I) The Fig. 3 shows a three wheel omnidirectional mobile robot
(TWOMR) which consists of an omnidirectional mobile platform
Mechanical Force (F) Velocity (v)
with three omnidirectional wheels installed at 120o from each
translation
Mechanical rotation  Angular
other.

)
Torque( )
velocity(

In addition, to power variable two energy variables, the

generalized moment p(t) and generalized displacement q(t) also
defined in the bond graph as shown in (1).


p(t )  e(t )dt
(1)
q(t )   f (t )dt
Figure 3: Three wheel omnidirectional mobile robot
There are different elements of bond graph based on the energetic 3.1 Kinematics of the Robot
behaviour of the component (Energy dissipation or energy The distance of each wheels from the geometrical Centre of the
storage). The description of different elements of bond graph are robot is r0. . All the wheels have same radius ri (i  1,2,3) . P1 , P2 , P3
shown in Table 2.
are the position vectors of wheel 1, 2 and 3 respectively with
Table 2: Elements of bond graph [5]
respect to moving frame Cxy as shown in the Fig. 4.
Component Nome Description Electrical Mechanical
nclatu
re
Source of SE Generate effort Voltage Force
effort source Source
Source of SF Generate flow Current Velocity
flow source source
Resistive R Dissipate energy Resistor Damper
element
Inductive I Effort storage Inductor Mass
element element
Capacitive C Flow storage Capacitor Spring
element element
Transformer TF Transfer energy Transformer Gear
element
Gyrator GY Transfer energy DC motor Gyroscope

Figure 4: Schematic model of TWOMR

To connect the element, there are two kinds of junctions in bond
graph, namely, the 1 and the 0 junction. The 1-junction is a flow
equalization junction and the 0-junction is an effort equalization
AIR 2017, June 28-July 02, 2017, New Delhi, India Saumya et al.

( X , Y ,  ) and ( x, y,  ) represents the posture of the TWOMR in 3.2 Word Bond Graph of the Robot
global and local co-ordinate frame respectively. The position The word bond graph provides a structured graphical
vector of each wheels Pi (i  1,2,3) relative to the fixed frame is representation of the TWOMR by decomposing it into subsystems
with exchanging power among them. The power bond shows the
cos( )  sin( ) power variable from different subsystems. The power variables
determined from the rotational matrix R( )   
 sin( ) cos( )  uses to analyse the system are
. (Voltage,Current) ={(Vi,ii),(Va,ia)}
),(Fy,
(Force,Linear velocity)={(Fx, x y )}
1  2  1 r  1 
P1  r0  , P2  R
0
   3
   P1  0   ,
 0 2  3 (Torque,Angular velocity)={(  ,  ),(FxC,  ) }
 4  1 r  1 
P3  R    P1   0   (2)
 3  0 2  3

The velocity vector of each wheel corresponding to the linear

velocity and angular velocity of the TWOMR is written as

v1  v x  r0 (3)
1 3
v2   v x  v y  r0 (4)
2 2
1 3
v3   vx  v y  r0 (5)
2 2
Where vi (i  1,2,3) are velocities of the wheels.
Eq (3)-(5) can be written in matrix form as
 
 1 0 r0 
 v1    v x 
   1 3  
v 2     2 r0  v y  (6)
2 
v3      
1 3
  r0 
 2 2 
The inverse kinematic model of TWOMR can be obtained as
Figure 5: Word bond graph of TWOMR
v x   v1 
    ; Where, 3x 3 Fig. 5 represents the word bond graph of a TWOMR which
v
 y  J v 2  J is the Jacobian matrix of TWOMR
   v 3  consist of subsystems like actuator, gear transmission system, the
wheel, the parts with linear and rotational motion. Using this
and vi  ri i information, the complete bond graph model of the robot is
developed and same is given in the next section.
 
 
2 1 1  4. BOND GRAPH OF THE ROBOT
1 1 1  Due to complexity of the TWOMR, in this work the system of
Where, J=  0  
3 3 3 TWOMR is divided into different subsystems.The dynamic model
 of each system has been developed using the unified bond graph
1 1 1 
 r 0 r0 r 0  approch. All the subsystems are integrated to obtain the bond
graph of the TWOMR.
The linear velocity in the robot reference frame is v=[vx,vy,0]T 4.1 The Electromechanical Part: DC Motor
and angular velocity is [0,0,ω]T. q =[ X , Y ,  ]T represents There are three DC motors located within the robot to actuate the
velocity in the generalized global reference frame. wheels. The motor (Fig. 6) has both electrical and mechanical
parts.
To simplify the model of the robot, following assumptions are
made
 Motion surface is a horizontal and smooth plane.
 The Centre of gravity and the Centre of mass of the
robot coincides with each other.
 Moment of inertia of the robot remains constant during
the operation process.

Figure 6: DC motor subsystem

AIR 2017, June 28-July 02, 2017, New Delhi, India Saumya et al.

The electrical part consists of a RL-circuit, the input voltage Where  x ,  y ,  w are the coefficient of viscous friction along
source, the back EMF which is linear to the angular velocity of longitudinal, lateral and yaw direction respectively. In bond
the motor. The RL circuit represented by integral causality as graph, the coulomb friction can be modelled as a modulated
shown in Fig. 7 in which the L element imposed its effect on the dissipated source of effort; while the viscous friction represented
1- junction. All the bonds associated with the 1-junction share the by resistive element as shown in Fig. 9. The full arrow represents
same current. The gyrator element describes the power transfer an activated bond for the modulated element. Fig. 9 shows bond
from electrical to mechanical part and the gyrator constant is kt . graph model of the wheel along with friction elements. It shows
The mechanical part of the motor characterized by the motor the bond graph of one wheel. Model for other wheels are also
inertia (Jm) and the viscous friction of the shaft (Bm). The developed in similar fashion.
complete model of the motor is shown in Figure 7.

Figure 9: Bond graph model of the wheel

Figure 7:Bond graph modelling of the electromechanical part (DC 4.4 Motion along Linear Direction
motor)
This model consists of longitudinal and lateral motion part of the
4.2 The Transmission system: Gear robot. The effort is the tract force from the motor shaft. The
The mechanical gear, which links the motor output power to the coulomb friction and slippage act as a dissipative element inside
load from wheel shaft part is represented by a transformer element the system. The friction model can be expressed as shown in Eq
TF. It scales the flow and effort variable of the motor shaft and (9)
transmits it to the wheel shaft. The bond graph representation of
mg
the gear is given in the Fig. 8. Where, N represents the gear ratio. fi  (  w tan 1 (kv w ) R w ( )   r tan 1 ( kvr ) R r ( )) (9)
3
The elements of linear motion along the longitudinal and
transverse direction are connected with the 1-junction. The mass
of the robot imposed its effect on the 1-junction and gives the
Figure 8: Bond graph modelling of gear.
information about linear velocity. As the system has a moving
4.3 The Wheel Part frame and a fixed frame, due to the gyroscopic effect, the flow
The output torque from the gear acts as an input effort to the velocity along the lateral direction and the yaw component is
wheel system that is the mechanical load part. The ground is plane modelled as a modulated gyrator element in linear direction.
and smooth; here longitudinal friction and lateral slippage act at vw MSe fcwx

the wheel roller and the ground interaction. The modelling of Fx1 Se

ground interaction with the roller is complicated as it varies non- I m

linearly with the sliding velocity of the respective wheel. Two F x2 Se 1

R
kind of ground interactions are taken into consideration; first one
Bx

is the friction along the wheel rotation direction and second is the F x3 Se

along the roller rotation direction i.e. along perpendicular to the

MGY Theta_dot
MGY

vr MSe fcrx

wheel rotation direction. It is assumed that the robot weight is

equally distributed on each wheel. The friction force [2] modelled (a)
as Eq (7) vw MSe fcwy

mg 2
fw   w tan 1 ( kv w ) Fy1 Se

3  Im
(7)
mg 2
fr   r tan 1 ( kv r ) Fy2 Se 1
3  R By

Where, f wi and f ri are the friction force along wheel rotation and Fy3 Se

roller rotation direction respectively.  wi ,  ri are coefficient of MGY Theta_dot

MGY

friction and vwi , vri are sliding velocity in wheel and roller vr MSe fcry

direction respectively. The viscous friction acting on the robot is (b)

linearly varying with the velocity of the robot as shown in Eq(8).
Figure 10: Bond graph of (a) longitudinal motion (b)
This acts as a resistive element to the system.
transverse motion
f fvx
  x x , f fvy
  y y , f fvw
  w  (8) 4.5 The Yaw Motion
The actuated force from the motor creates a moment about the
center of gravity of the robot which act as an effort for yaw
AIR 2017, June 28-July 02, 2017, New Delhi, India Saumya et al.

motion of the robot. The inertia of the robot and viscous friction To move the robot in a straight line path, two wheels of the robot
are connected with 1-junction and share the same angular should rotate in opposite sense; the third wheel being stationary.
velocity. The inertia of the robot imposed its effect on the 1- The two wheels are actuated with 1Nm and -1Nm torque
junction and gives the angular velocity information. The bond (opposite direction), the robot follows a straight line path. The
graph for yaw motion is shown in Fig. 11. direction of the robot motion depends on the position of actuated
vw MSe fcw
wheels. The velocity of the two wheels is shown in Fig.13.

Fd1 Se TF r0

TF I I
r0_

Fd2 Se TF 1
ro__
R Bw

TF
r0___

Fd3 Se

Figure 11: Bond graph of yaw motion

4.6 Bond Graph of the TWOMR
All the three motors, the gear, the three wheels and the linear and
yaw motion are integrated together to create the bond graph of the (a) (b)
three wheeled omnidirectional mobile robot (TWOMR). The bond Figure 13: (a) Straight trajectory followed by the robot (b)
graph of TWOMR is shown in Fig. 12. velocity of the wheels

coulomb_f ric tion_longit udinal_w2

MSe
R v is cous_f ric tion_longitudinal

I La I Jm

Se 1 GY 1 TF TF 0 TF 1 I m

Se GY N R cos _90

R Ra R TF
Bm sin_90

MSe
TF L__1 MSe coulomb_f ric tion_longit udinal_w3
coulomb_f ric tion_longit udinal_w1

MGY
MGY MGY
MGY1

c oulomb_f ric tion_Trans verse_w1

MSe

I La1 I Jm1
R vis cous_f rict ion_lateral

Se 1 GY 1 TF TF 0 TF
Se1 GY1 N1 R1 minus_cos _30
1 I m_

R R a1 R
TF
Bm1
minus_sin_30

TF L__2
MSe MSe
c oulomb_f ric tion_Trans verse_w2 coulomb_frict ion_t ransv erse_w3

TF
minus_s in__30
I La2 I Jm2 coulomb_friction_wheelmot ion_w3

MSe

R visc ous _f rict ion_lateral1

Se 1 GY 1 TF TF 0 TF TF L
Se2 GY2 N2 R2 cos_30

TF L__3
1 I I
R Ra2 R
Bm2

TF L_
TF L__

MSe MSe
coulomb_friction_wheelmot ion_w1 c oulomb_f ric tion_wheelmot ion_w2

Figure 12: Bond graph of the TWOMR

For a circular trajectory, two wheels should be actuated with
5. SIMULATION RESULT torque applied in the same sense. For simulation purpose, these
To validate the model, open loop test is conducted in MATLAB
wheels are actuated with 1 Nm torque.
Simulink using the BG_V20 tool box. Simulation is carried out
using the parameters as shown in Table 3.
Table 3: Simulation parameters
Parameter Unit Value
M Kg 9.5
Iz Kg-m2 0.17
r0 m 0.05
w - 0.26

r - 0.09
AIR 2017, June 28-July 02, 2017, New Delhi, India Saumya et al.

Figure 14. (a)Circular trajectory followed by the robot (b) Figure 16.Torque comparison (a) circular trajectory (b)
velocity of the robot. straight path
The velocity of the two actuated wheels (2,3) is shown in the Fig. 7. CONCLUSION
15, but here due to the friction on the stationary wheel, the third The bond graph formalism is a technique devoted to model
wheel (1) also has angular velocity as shown in this Fig.15. engineering systems as combinations of connected elements. In
this paper, a step by step procedure for bond graph modelling of a
three-wheel omnidirectional mobile robot is described. To
increase the reliability of the model, friction (Coulomb as well as
viscous) and slippage (between roller and the surface) are
considered. To validate the model, open loop simulation is done
for the straight path and circular trajectory and compared with
Newton-Euler method. This study shows that bond graph
technology can be effectively implemented for modelling of
complex mechatronics or multi-domain systems. The causal and
structural properties of Bond graph can be exploited for control
Figure 15. Velocity of the three wheel of TWOMR and supervision of the model.
6. VALIDATION USING NEWTON-EULER 8. REFERENCES
FORMULATION [1] Tuan Dinh Viet, Phuc Thinh Doan .2012, Tracking control of a three-
The dynamic modelling of TWOMR is done using Newton’s wheeled omnidirectional mobile manipulator system with disturbance
and friction, Journal of Mechanical Science and Technology, Springer,
second law of motion and Euler’s law from which the linear and Volume 26, Issue 7, pp 2197–2211
angular momentum balance are obtained as equation 10 [1] [2] Robert L. Williams.2002. Dynamic Model with Slip for Wheeled

 (F  f )R( )d  mP
Omnidirectional Robots. IEEE transactions on robotics and automation,
i fi i 0 vol. 18, no. 3
[3] Jose J.Granda.2002.The role of bond graph modelling and simulation in
r  ( F  f )  I
0 i fi
(10) mechatronics systems: An integrated software tool: CAMP-G,
MATLAB–SIMULINK. Mechatronics.
Fi  u i  vi [4] P.M.Pathak.2008.A scheme for robust trajectory control of space robots.
Simulation Modelling Practice and Theory, Volume 16, Issue 9, October
Where, m and I are the mass and inertia of the robot, f fi (i  1,2,3) 2008, Pages 1337–1349
[5] Yang Qian and Ahmed Rahmani.2009. Bond Graph Modelling and
is the friction force exerted at each wheel . u i is the voltage Simulation of a Dual-Arm Mobile Manipulator. In IEEE Int. Conf. on
supplied to the motor and  ,  are the motor constant. The Robotics and Biomimetics, Bangkok, Thailand, pp. 639-644.
[6] Buddhika L. Samarakoon and Lalith B. Gamage.2015.Design evolution
equations are simulated in MATLAB Simulink. The toque profile of engineering systems using bond graphs and genetic programming.
of the motor while the robot follows a circular trajectory and Mechatronics 33 (2016) 71–83.
straight path shown in the Fig. 16. It shows that the toque profile [7] M.M. Gor and P.M. Pathak .2015.Control oriented model-based
simulation and experimental studies on a compliant legged quadruped
obtained by bond graph approach matches with the Newton- robot. Robotics and Autonomous Systems 72 (2015) 217–234.
Euler’s approach. From these observations, it can be found that [8] B.Six, B. Ould-Bouamama, S. Lintz and A.L. Gehin.2016.Railway
both the dynamic modelling tools give comparable results and Direct Braking System Analysis using Hybrid Bond Grap.,CoDIT’16 -
confirmed that the dynamic model generated using bond graph is April 6-8
[9] Veer.Alakshendra,Shital S. Chiddarwar.2016.Trajectory Tracking
capable to produce results as good as by classical technique. Control of Three-Wheeled Omnidirectional Mobile Robot: Adaptive
Sliding Mode Approach, Proceedings of the 28th International
Conference on CARs & FoF 2016
[10] Tzafestas, S. G. (2014). Introduction to mobile robot control. London:
Elsevier.

(a)

(b)