Dynamic Modelling of A Two-Wheeled Vehicle: Jourdain Formalism
Dynamic Modelling of A Two-Wheeled Vehicle: Jourdain Formalism
Dynamic Modelling of A Two-Wheeled Vehicle: Jourdain Formalism
formalism
Lamri Nehaoua, Hichem Arioui, Nicolas Seguy, Saïd Mammar
February 8, 2013
Abstract
This paper presents a motorcycle direct dynamic formulation by the Jourdain’s
principle approach. This vehicle is considered as an assembly of six rigid bod-
ies and the resulting equation of motion allows to simulate 11 Degrees of Free-
dom (DoF). The vehicle geometry is described and a step by step procedure is
introduced to evaluate the kinematics and the generalized efforts of the consid-
ered vehicle. In addition, to simulate the equation of motion, a Lyapunov based
stabilization is developed to assess the vehicle behavior in response to a propul-
sion/braking torque applied on the vehicle’s wheels and a rider torque exerted on
the motorcycle’s handlebar. Simulation results reveal some dynamics features like
load transfer and counter-steering phenomena.
keywords
Motorcycle modeling, motorcycle control, Jourdain’s dynamics principle.
1 Introduction
During last decades, motorcycles have become a popular transportation tool. Indeed,
the number of two-wheeled vehicles is in constant increase, specifically in urban ar-
eas. However, the risk associated with driving such vehicles is important and a road
accident, even trivial, can have fatal consequences on the two-wheeled users.
Motorcyclists are the most vulnerable road users. They represent less than 2% of
the traffic, but 18% of those killed in 2010 in France [1]. There are many causes to
these fatalities: alcohol, non-compliance with the road rules, poor visibility, lack of
assistive devices and especially the overvaluation of the self mastering capabilities of
such a vehicle. Indeed, with the same power, the motorcycle represents more aggres-
sive dynamics compared to those of a car vehicle. Riding a two-wheeled vehicle is
neither simple nor intuitive like driving a car vehicle. Involved accelerations, critical
equilibrium, trajectory anticipation are, among others, many points that make riding a
∗∗ Corresponding author. Email: [email protected]
1
motorcycle a complicated exercise especially for inexperienced riders who get carried
away by the riding sensations and pleasure.
Fortunately, recent preventive as well as suppressive laws have significantly con-
tributed to decrease number of fatalities. It must be said that motorcyclists’ road safety
has attracted the attention of all academic, industrial and political institutions. In
France, several research projects were undertaken: SUMOTORI 1 (2003-2006), SIMA-
COM2 (2006-2009) [2], SAFERIDER 3 (2008-2010), DAMOTO 4 (2008-2011), and
others with a common point: understand the rider behavior.
To deal with this aim, it is necessary to predict the dynamics of the motorcycle
vehicle and model it. In the study of mechanical systems, modeling usually leads to a
set of differential/algebraic equations (DAE) derived from laws of mechanics. In this
field, two approaches are proposed: the analytical approach and the numerical one.
In the later one, many tools were developed such MSC ADAMS [3], simMechanics
[4] and Modelica [5]. The alternative approach is the development of symbolic equa-
tions of motion which is a time consuming method, but it offers more flexibilities in
the system description. In that case, an algebra package like MBsymba [6] is highly
recommended.
Many works have addressed the modeling of the two-wheeled vehicles. Indeed, the
first study on the stability and balance of a bicycle vehicle was performed by Whipple
[7]. This bicycle is represented by two bodies linked via the steering mechanism.
Wheels are modeled as circular thin disks where the gyroscopic effect is considered.
Other research works have been undertaken [8, 9, 10]. In all these studies, the tire-
road interaction has been neglected. In 1971, Sharp integrated tires’ efforts [11]. The
objective of this study is to examine the effect of parameters uncertainties on the vehicle
stability. The vehicle is modeled as a set of two rigid bodies connected by the steering
mechanism, simulating 4 DoF: the lateral displacement and the yaw, the roll and the
steering rotations.
So far, in all these models, the longitudinal velocity was assumed to be constant and
the vehicle suspension motions were ignored. It was interesting to examine how these
factors influenced the frequency characteristics of the different stability modes. These
aspects have been investigated in a further work of Sharp in 1974 [12]. In 1994, Sharp
published a paper, in which, the motorcycle is represented by a set of four rigid bodies
to simulate 8 DoF. The originality of this model lies in the integration of flexibilities
and the rider tilting motion [13].
In the same perspective, a model incorporating the suspension travels and a more
complex representation of the tire-road interaction is presented in [14]. The advances
made by this model have allowed exploring the dynamics of such a vehicle for a large
motion around equilibrium. In that case, the interaction between the in-plane and the
out-of-plane motion is well described. More recently, the FastBike program presents
a nonlinear model of a motorcycle described as a set of 6 bodies and simulating 11
DoF. Each body is identified by its natural coordinates (a vector of six coordinates:
three Cartesian coordinates of its center of mass and three rotations) and the assembly
1 Safety of motorcyclist against risky environment
2 Simulator for learning to drive motorized two-wheeled vehicles
3 Advanced telematics for enhancing the SAFETY and comfort of motorcycle RIDERS
4 Detection of accident situations on motorcycle
2
is done by introducing geometric constraints [15]. Finally, a modeling technique based
on the recursive Newton-Euler approach is adapted to derive a motorcycle dynamics
model intended for a two-wheeled simulator application [16, 17].
Among the vehicle models, the simulation of single-track one turns out to be par-
ticularly difficult. While a car vehicle remains stable, the stabilization of two-wheeled
vehicles is required and a rider model must be included. The literature of two-wheeled
vehicle stabilization is recent [19, 18, 20]. Generally, a two-layer controller is adopted
to track the reference road trajectory and an inner controller stabilizes the longitudi-
nal and lateral dynamics. For this, several strategies are used: the optimal maneuver
method [21, 22], robust H∞ and LPV 1 control techniques [23].
In this paper, a step by step modeling approach based on the Jourdain’s principle
is presented. This method was firstly used in [24] to derive a car dynamics model and
the modeling prerequisites are inspired from this work. It offers an efficient tool for
deriving motion equations in a systematic way and with less theoretical skills compar-
ing to the Lagrangian and the Newton/Euler approaches. The great efforts concern the
computation of Jacobian matrices and the virtual power of the different external ap-
plied forces. Our aim is to derive a mathematical description of a motorcycle vehicle
with middle complexity, which captivates the essential dynamics issues. This model is
intended for control, state observation and security devices development applications.
In section 2, a brief outline to the mathematical modeling of the motorcycle kinematics
and dynamics is introduced. Section 3 focuses on the computation of the generalized
vector efforts. Section 4 is intended to demonstrate the basics concepts of the tire/road
contact. Finally, a virtual rider controller is introduced in order to regulate the longi-
tudinal speed towards a reference profile and to stabilize the lateral dynamics by using
the roll motion.
3
• the steering angle δ of the front body and the front wheel,
• the front suspension travel λf and the swing arm rotation µ,
• and finally, the front and rear wheels spinning, ζf and ζr .
Su
Gf
Gr
B
ǫ
Gl
ǫ P Gs
Rf Rr
Sl
4
O
io •
ψ
v •
• cr
iv ℜo
•
cf ℜv
jv
jo
Figure 2: Configuration of the vehicle system axis ℜv w.r.t the inertial reference frame
ℜo
r oi = rov + r vi (1)
From equation (1), the velocity and the acceleration of a body i are derived as
following:
v oi = v ov + ω ov × r vi + ṙ vi (2)
aoi = aov + ǫov × r vi + ω ov × (ω ov × r vi + 2ṙvi ) + r̈vi
and, the angular velocity ωoi and the angular acceleration ǫoi vectors of body i are
derived in the same way:
Since the kinematic configuration of each body is well defined in its local reference
frame, it remains to define the kinematic configuration of the movable reference frame
ℜv , it means, to find the expressions of aov and ǫov in equations (2,3) w.r.t v̇ ov and
ω̇ov . Suppose that the linear and angular velocity vectors of ℜv w.r.t the inertial one
ℜo are given by v oov = Rvo v vov and ω oov = Rvo ωvov . It goes that the linear and angular
acceleration vectors of ℜv , expressed in ℜv , can be derived as following:
aov = ω ov × v ov + v̇ ov (4)
ǫov = ω̇ ov
5
kµ
kǫ
iǫ µ
ℜǫ
θ+ǫ
•
P
iµ ℜµ
B iǫ P kϕ
• • iϕ kv
δ
ℜδ ϕ
iδ ℜǫ ℜϕ jϕ
jδ jǫ iϕ • • jv
ℜϕ v v ℜv
From equations (2-4), the linear and angular acceleration vectors of each body i are
expressed in the movable reference frame ℜv by the following equations:
At present, all kinematic quantities are defined, then we introduce the principle of
the virtual power called also Jourdain’s principle.
Moreover, from equations (2, 3, 5), the velocity and acceleration vectors can be
written as following:
6
∂voi ∂ωoi ∂v oi ∂ω oi
v oi = ϑ ω oi = ϑ aoi = ϑ̇ + aR ǫoi = ϑ̇ + ǫR (8)
∂ϑ ∂ϑ ∂ϑ ∂ϑ
where:
T
ϑ= vx vy ψ̇ ż ϕ̇ θ̇ δ̇ λ̇f µ̇ ζ̇f ζ̇r (9)
denotes the vector of generalized velocities and the partial derivatives are called
Jacobian matrices of the velocity vector w.r.t the generalized velocities vector ϑ. Vec-
tors aR and ǫR are referred as the residual acceleration terms. Consequently, from
equations (7, 8) and after some algebraic manipulations, the motion dynamics can be
expressed as:
Mϑ̇ = Q (10)
where M is the mass matrix obtained from the direct computation of Jacobian
matrices:
n
( T T )
X ∂v oi ∂v oi ∂ωoi ∂ωoi
M= mi + Ii (11)
i=1
∂ϑ ∂ϑ ∂ϑ ∂ϑ
and the generalized efforts vector Q can be split as following:
n
( T T )
X ∂voi ∂ω oi
Q= F i,a + M i,a − (12)
i=1
∂ϑ ∂ϑ
| {z }
Qa
n
( T )
X ∂v oi ∂ωoi
mi aR + (Ii ǫR + ω oi × Ii ω oi ) (13)
i=1
∂ϑ ∂ϑ
| {z }
QR
The next subsection presents a step-by-step procedure to derive the Jacobian ma-
trices. These matrices allow the calculation of the mass matrix M and the residual
generalized effort vector Q.
7
where the coordinates of vectors r vP and rP Gr in the reference frame ℜv are
computed by:
0 xGr
rvP = Rϕ 0 r P Gr = Rϕ,θ 0 (15)
z + hp zGr
where the coordinates of the unit vector kϕ in ℜv are obtained by kϕ = Rϕ [0, 0, 1]T
and the components of the angular velocity vectors ω ϕ and ωθ are respectively given
by ω ϕ = [ϕ̇, 0, 0]T , ω θ = Rϕ [0, θ̇, 0]T . By using equations (5, 8), the Jacobian
matrices and the residual acceleration vectors can be deduced by simple algebraic ma-
nipulations.
where the coordinates of the vectors rP B and rBGf are given in the reference
frame ℜv as following:
lP B xGf
rP B = Rϕ,θ,ǫ 0 r BGf = Rϕ,θ,ǫ,δ 0 (20)
0 zGf
8
2.4.3 Lower front body
The front lower body has 1 DoF w.r.t Gf representing the suspension travel λf . Its
center of mass position vector is:
xGl
rBGl = Rϕ,θ,ǫ,δ 0 (25)
λf + zGl
where the coordinates of the unit vector kδ in ℜv are kδ = Rϕ,θ,ǫ,δ [0, 0, 1]T .
xGs
r P Gs = Rϕ,µ 0 (29)
zGs
9
2.4.5 Rear-wheel bodies
The rear-wheel body has three independent DoF w.r.t ℜv including roll rotation, ver-
tical displacement and the tire spin rotation around wheel spin axis j ζ,r . It follows
that:
where the components of the angular velocity are ω ζ,r = Rϕ [0, ζ̇r , 0]. On the other
hand, the position of the rear wheel center of mass is:
xRr
r P Rr = Rϕ,µ 0 (34)
zRr
where ωζf = Rϕ,θ,ǫ,δ [0, ζ̇f , 0]T . In the same way as the rear wheel, the position
of the front wheel center of mass is:
xRf
rBRf = Rϕ,θ,ǫ,δ 0 (40)
λf + zRf
10
By differentiating, the relative linear velocity vector is:
!T 0
06
∂ δ̇
Qa,h = 0 = τr − Cδ δ̇ (43)
∂ϑ
τr − Cδ δ̇ 04
where −Cδ δ̇ is the steer damping torque.
In [25], a more detailed rider-motorcycle interaction is described by taking into
account the rider’s upper torso leaning motion and the connection of his/her arms with
the handlebars.
11
3.3 Propulsion and braking torque
For the front-wheel, the braking force acts between the wheel knuckle and the front
lower body which contains the braking disk. For the rear wheel, the braking torque
is applied between the rear-wheel body and the swing arm assembly where the engine
propulsion torque is applied around the rear-wheel spin axis, hence:
!T 0 !T 0
T
0
∂ ζ̇f τB,f + ∂ ζ̇r τD + τB,r + ∂ µ̇ −τB,r
Qa,pb =
∂ϑ ∂ϑ ∂ϑ
0 0 0
(44)
08
−τB,r
=
τB,f
τD + τB,r
!T
07
∂ λ̇f
Qa,sf = Fs,f = Fs,f (45)
∂ϑ
03
In the same way, the contribution of the rear suspension can be computed as fol-
lowing:
!T
∂ λ̇r
Qa,sr = Fs,r (46)
∂ϑ
where λr is the rear suspension travel given by λ2r = rSl Su · rSl Su . By differenti-
ating, the suspension travel rate λ̇r is deduced and written by the following equation:
12
eSl Su
kδ Su
λf ǫ λr
P
Rf Rr
Sl
-Fs,f -Fs,r
where eSl Su is the rear suspension unit vector. Consequently, the contribution of
the rear suspension in the total generalized effort is given by:
05
eSl Su · (j θ × rP Su )
Qa,sr = 02 Fs,r
(48)
−eSl Su · (j µ × r P Sl )
02
where the coordinates of the unit vectors j θ and j µ in ℜv are respectively equal to
Rϕ [0, 1, 0]T .
Equations (42-48) show the contribution in Qa of each external force/moment ap-
plied on the motorcycle vehicle. In the next section, the tire-road interaction is exposed
and its contribution is demonstrated.
13
4.1 Tires kinematics
To describe the tire motion, a new reference frame ℜT (c, iT , j T , kT ) is introduced at
the contact point c of each wheel’s tire. kT is the normal vector to the road surface.
Vector iT is obtained by the cross product j ζ × kT and j T completes the reference
axis.
•
jζ
iT ρ Fz
Fy
α
voc •
c
Fx
jT
sin γ = j ζ · kT (49)
Besides, the position of the rear and front road/tire contact point c w.r.t the movable
frame reference ℜv is given by:
in which the coordinates of rRr cr and rRf cf in ℜv are respectively given by:
0 −ρf sin ǫ
r Rr c r = Rϕ 0 r Rf c f = Rϕ,θ,ǫ,δ 0 (52)
−ρr ρf cos ǫ
By differentiating, the relative linear velocity vector of each contact point can be
deduced as:
14
ṙ vcr = ṙ vRr + ω ϕ × r Rr cr
ṙvcf = ṙ vRf + ω vGf × rRf cf
By using equation (2), the contact point velocity vector is computed and the slip
variables are defined by the following equations:
j v · v ocr
αr = −atan (53)
iv · v ocr
j v · v ocf
αf = −atan + δ cos ǫ
iv · v ocf
Vc,i − rD,i ζ̇i
κi = −
max (Vc,i , rD,i ζ̇i )
where, Vc,i = kv oci k is the contact point speed, rD,i is the wheel’s dynamic radius
and i = r, f .
Finally, knowing that the contact point ci must always belong to the road surface
then, the vertical deformations δz,i of the wheels’ tire can be derived by:
These deformations help to determine the normal load necessary to maintain con-
tact with the road plane [29].
F T = Fx iT + Fy j T + Fz kT
M T = Mx iT + My j T + Mz kT + F T × r cR
( T T )
X ∂voRi ∂ωoRi
Qa,T = F T,i + M T,i (55)
∂ϑ ∂ϑ
i=f,r
15
5 Motorcycle model control
To simulate the motorcycle model, a convenient control is to be synthesized to track a
reference longitudinal speed profile and to stabilize the lateral dynamics around trim
trajectories. This is a hard task since the roll DoF is unstable and, for some aggressive
driving maneuvers conditions, system nonlinearities cannot be neglected. In addition,
it is impossible to exploit the whole nonlinear model to develop a suitable controller,
even by using a nonlinear control method. For these reasons, a linearized version of the
dynamics model representing small motions in the neighborhood of the straight motion
is considered.
In this paper, the speed and the roll controllers are based on Lyapunov method [30].
This technique avoids the gain-scheduling, necessary for the linear method control to
consider longitudinal speed variations. For each mode, longitudinal and lateral, the
controller acts on the speed error to produce a torque to be applied on the rear wheel
and on the roll angle error to generate a rider steering torque to be applied on the
motorcycle’s handlebar.
X
mi v̇x − vy ψ̇ = Fx,f + Fx,r − Fy,f sin δ (56)
iy,f ζ̈f = τb,f − ρf (Fx,f − Fy,f sin δ)
iy,r ζ̈r = τD + τb,r − ρr Fx,r
With the assumption of zero longitudinal slip, it goes that vx = ρr ζ̇r = ρf ζ̇f and
by replacing in equation (56) the simplified longitudinal mode becomes:
τb,f τD + τb,r X
meq v̇x = + + mi vy ψ̇ (57)
ρf ρr
P
where meq = mi + iy,f /ρ2f + iy,r /ρ2r is the equivalent mass. Let us introduce
the Lyapunov function V = meq e2 /2, where e = vx − vx,ref is the speed tracking
error, its derivative is given by:
τb,f τD + τb,r X
V̇ = meq eė = e + + mi vy ψ̇ − meq v̇x,ref (58)
ρf ρr
In order to the Lyapunov function derivative to be negative, the propulsion torque
τD must satisfy:
τb,f X
τD = ρr −kx ex − − mi vy ψ̇ + meq v̇x,ref − τb,r (59)
ρf
16
where kx is a positive constant. With this input control, the Lyapunov function
derivative is always negative (V̇ = −kx e2 ) and hence the system is also, always stable.
Remark With the control of equation (59), the error dynamics becomes ė = −kx e,
then, kx is just a proportional coefficient which is tuned until the desired tracking er-
ror level is achieved. However, a high value introduces an oscillatory behavior while
stability remains preserved.
17
(a)
24
22 reference
simulated
20
18
16
v (m/s)
14
x
12
10
4
0 5 10 15 20
time (s)
(b)
24
22
ρr ζ̇r
20
tire/road contact point speed (m/s)
ρf ζ̇f
18
16
14
12
10
4
0 5 10 15 20
time (s)
(c)
1
0.8
0.6
0.4
z, λ (cm) and θ, µ (°)
0.2
0
z
−0.2 λ
θ
−0.4 µ
−0.6
−0.8
−1
0 5 10 15 20
time (s)
(d)
1120
Fzf
F
zr
1100
1080
Fzf and Fzr (N)
1060
1040
1020
1000
0 5 10 15 20
time (s)
2.5
ax (m/s )
2
1.5
0.5
−0.5
0 5 10 15 20
time (s)
(b)
250
200
150
τD (N.m)
100
50
−50
0 5 10 15 20
time (s)
The following simulations are carried-out using the developed nonlinear model and
the controller of equation (59). Starting from 5ms−1 , the desired speed profile consists
of two main phases, an acceleration phase to reach 23m/s−1 followed by a deceleration
phase to keep a constant speed (see Figure 6-a). In Figure 6-b, the tire/road contact
point speed is shown to give an overview of the longitudinal slip where Figure 6-c
presents the simulation results of the in-plane mode variables (vertical displacement,
front suspension travel, pitch and swing angles). Figure 6-d, describes the evolution
of the nominal load applied on each tire w.r.t to speed variations (and so longitudinal
acceleration). Lastly, requested propulsion torque at the rear-wheel and the resulting
acceleration are sketched in Figure 7.
The second simulation aims to highlight the motorcycle behavior when subjected
to a braking torques applied separately on the front and the rear wheel. Here in, it is not
question to apply a particular braking model such as the ABS system. It will be shown
that when applying a braking torque, significant load transfers take place from the rear
wheel to the front one. If suspensions are assumed to be rigid, an approximation of the
vertical load on each wheel when braking [31] is given by:
19
P
mg (xGr − xRr ) + FB zGr
Fz,f = (60)
p
P
mg (p − xGr − xRr ) − FB zGr
Fz,r =
p
where, p is the motorcycle wheelbase and FB is the total braking force at the
tire/road contact. Based on these two equations, a limit braking force FB,lim which
causes the rear wheel lift can be computed.
Figure 8 represents the rear and the front brake profile applied separately on each
wheel. In the first scenario, a rear braking torque with a maximum value of -600 N.m
is applied between 4-7 seconds. The motorcycle dynamics behavior is demonstrated
in Figure 9. From an initial forward speed of 20m/s, the motorcycle brakes hardly up
to 2m/s with a deceleration of 0.4g. In that case, the vertical position z of point P is
lowered which introduces a clockwise and hence a negative pitch rotation θ. The same
vertical movement produces an anti-clockwise and hence a positive swing rotation µ.
It is clear that the rear wheel does not lift due to the applied vertical force (600N which
is equivalent to a payload of 60kg. See Figure 9-d).
τ
Br
−100 τ
Bf
−200
τB (N.m)
−300
−400
−500
−600
0 5 10 15 20
time (s)
In the second scenario, the same torque profile as before is applied but with minus
a half amplitude (-280Nm). The motorcycle dynamics behavior is shown in Figure
10. Initially the motorcycle was traveling at 20m/s and brakes up to 12 m/s with a
deceleration of 0.4g. A first conclusion concerns the amplitude of the applied torque
which is not sufficient to stop the motorcycle. In addition, such front brake can easily
create the limit braking force FB,lim and hence causes the loss of the rear wheel contact
with the ground. Indeed, the vertical force at the rear wheel is about 100N (a payload
of only 10kg. See Figure 10-d). With the same justification, the vertical position z of
point P goes up which introduces an anti-clockwise and hence a positive pitch rotation
θ and a clockwise and hence a negative swing rotation µ.
20
(a)
22
20
18
16
14
v (m/s)
12
x
10
2
0 5 10 15 20
time (s)
(b)
3
0
a (m/s2)
−1
x
−2
−3
−4
−5
0 5 10 15 20
time (s)
(c)
15
z
λ
10 θ
µ
z, λ (cm) and θ, µ (°)
−5
−10
−15
0 5 10 15 20
time (s)
(d)
1600
Fzf
1500
F
zr
1400
1300
1200
Fzf, Fzr (N)
1100
1000
900
800
700
600
0 5 10 15 20
time (s)
20
19
18
17
v (m/s)
16
x
15
14
13
12
11
0 5 10 15 20
time (s)
(b)
0.5
−0.5
−1
−1.5
ax (m/s )
2
−2
−2.5
−3
−3.5
−4
−4.5
0 5 10 15 20
time (s)
(c)
40
z
30 λ
θ
µ
20
z, λ (cm) and θ, µ (°)
10
−10
−20
−30
−40
0 5 10 15 20
time (s)
(d)
2000
Fzf
1800
F
zr
1600
1400
1200
Fzf, Fzr (N)
1000
800
600
400
200
0
0 5 10 15 20
time (s)
Figure 10: Simulation results according to only hard front brake torque
22
As is recognized in literature, this simulation presents the braking efficiency where
an optimal braking must be split between the two wheels with more brake ratio at the
front wheel (Figure 11).
(a)
22 1
20
0
18
16 −1
14
−2
ax (m/s2)
vx (m/s)
12
−3
10
8 −4
6
−5
4
2 −6
0 5 10 15 20 0 5 10 15 20
time (s) time (s)
1600
Fzf
Fzr
1400
1200
Fzf, Fzr (N)
1000
800
600
400
0 5 10 15 20
time (s)
Figure 11: Simulation results according to 70% front brake torque and 30% rear brake
torque of previous torques
ẋ = AL x + Bτr (61)
T
where x = [vy , ψ̇, ϕ̇, δ̇, ϕ, δ] is the state vector and τr is the rider torque input.
In Figure 12, the stability of the uncontrolled motorcycle model of equation (61) is
presented. This eigenvalues plot shows the existence of stable modes (in ◦ symbol)
and three instability modes, the first concerns the steering motion which appears from
8m/s (in ⋆ symbol), the second concerns the roll motion at very low speed until 7m/s
(in ▽ symbol) and finally, a coupled steering-roll instable motion generated at very low
speed (in + symbol). In [11], these instability modes are named capsize, weave and
wobble.
23
5
−10
−15
−20
−25
0 5 10 15 20 25 30 35 40
v (m.s−1)
x
Figure 12: Eigenvalues real part of the uncontrolled linearized motorcycle model w.r.t
the forward speed
ė = ϕ̇ − ϕ̇ref (62)
ϕ̈ = AL,(3,:) x + B 3 τr
where B 3 and AL,(3,:) are the third line of vector B and matrix AL . From equation
(62), it can be seen that ϕ̇ acts as a virtual input to the roll error dynamics. By using
backstepping technique [32], the required rider torque to track a reference roll angle is
given by:
where z = ϕ̇ − β is the error between the real roll rate and the virtual input β.
24
0.04
0.03
0.02
−0.01
−0.02
−0.03
0 10 20 30 40 50 60
time (s)
The following simulation are carried-out using the nonlinear model and the con-
trollers of equations (59,63) where Figure 13 shows the road curvature. From this
curvature, the reference roll profile is calculated. In Figure 14-a, the performance of
backstepping controller in reference tracking is described. Obviously, the linear model-
based control of equation (63) is able to stabilize the nonlinear dynamics. However, for
an aggressive cornering maneuver where the system nonlinearities are important, the
stability of the motorcycle is not guaranteed.
25
(a)
60
reference
simulated
40
20
φ (°)
0
−20
−40
−60
0 10 20 30 40 50 60
time (s)
(b)
350
reference
300 simulated
250
200
ψ (°)
150
100
50
−50
0 10 20 30 40 50 60
time (s)
(c)
5
2
δ (°)
−1
−2
−3
0 10 20 30 40 50 60
time (s)
(d)
15
10
0
τr (N.m)
−5
−10
−15
−20
−25
−30
0 10 20 30 40 50 60
time (s)
6 Conclusion
Based on the virtual power principle, called also Jourdain’s principle, a middle com-
plexity motorcycle dynamics model is proposed. The multi-body model has 11 DoF
and includes the main features of motorcycle subsystems. This approach is chosen for
its simplicity of synthesis and its fine analytical derivation with an acceptable calcu-
lation load, in particular, when the equations of motion are developed by using gener-
alized coordinates approach. This model exhibits many of interesting behaviors of a
real motorcycle such non-minimum phase steering response and load transfer resulting
from acceleration and braking.
Besides, a virtual rider is synthesized for longitudinal and cornering modes respec-
tively by using Lyapunov theory. This technique allows the calculation of speed and
roll controllers which are valid for a wide range of forward speeds and hence, avoids
controller scheduling.
From Figure 14.b, it is clear that using only roll stabilization cannot lead to a good
trajectory tracking. For this, additional cost functions must be included to minimize
the relative yaw angle and lateral deviation errors of the motorcycle from the road
center line. These issues will be the scope of future works. Moreover, as perspective,
experimental validation of the proposed motorcycle model and the simulation of more
limit riding maneuvers are expected.
References
[1] from www.preventionroutiere.asso.fr
[2] H. Arioui, L. Nehaoua, S. Hima, N. Sguy and S. Espi, Mechatronics, Design, and
Modeling of a Motorcycle Riding Simulator, IEEE/ASME Trans. on Mechatron-
ics, 15(5), (2010), pp. 805-818.
[3] ADvAnced Multi-body Simulation, http ://www.mscsoftware.com
[4] http :// www.mathworks.fr
[5] M. Tanelli, F. Schiavo, S.M. Savaresi and G. Ferretti, Object-Oriented Multibody
Motorcycle Modelling for Control Systems Prototyping, Proc. IEEE CACSD, Mu-
nich, Germany, 2006.
[6] R. Lot and M. Da Lio, A Symbolic Approach for Automatic Generation of the
Equations of Motion of Multibody Systems, j. of MSD, 12(2000), pp. 147-172.
27
[7] F.J.W. Whipple, Stability of the motion of a bicycle, Quarterly J. of Pure and App.
Math., 30 (1899).
[8] G.S. Bower, Steering and stability of single track vehicles, The Automob. Eng., 5
(1915).
[9] R.H. Pearsall, The stability of the bicycle, Proc. Instr. and Automotive Eng., 17
(1922), pp. 395-402.
[10] M. Kondo, A. Nagaok and F. Yoshimura, Theoretical study on the ruuning stabil-
ity of twowheelers, Trans. of Soc. Automotive Eng, 17:1 (1963).
[11] R.S. Sharp, The stability and control of motorcycles, J. Mech. Eng. Sc., 13 (1971),
pp. 316-329.
[12] R.S. Sharp, The influence of frame flexibility on the lateral stability of motorcy-
cles, J. Mech. Eng. Sc., 16:2 (1974).
[13] R.S. Sharp, Vibrational modes of motorcycles and their design parameter sensi-
tivities, Vehicle NVH and Refinement, Proc. of ICME (1994).
[14] R.S. Sharp and D.J.N. Limebeer, A Motorcycle Model for Stability and Control
Analysis, J. of MSD, 6 (2001), pp. 123-142.
[15] V. Cossalter and R. Lot, A motorcycle multibody model for real time simulation
based on the natural coordinates approach, J. VSD, 37:6 (2002), pp. 423-447.
[16] S. Hima, L. Nehaoua, N. Seguy and H. Arioui H, Motorcycle Dynamic Model
Synthesis for Two Wheeled Driving Simulator, IEEE ITSC, (2007), 812-817.
[17] S. Hima, L. Nehaoua, N. Seguy and H. Arioui, Suitable Two Wheeled Vehicle Dy-
namics Synthesis for Interactive Motorcycle Simulator, Proc. of IFAC Wor.Cong.,
Seoul, Korea, (2008).
[18] R. Lot, M. Massaro and V. Cossalter, Advanced motorcycle virtual rider, J. VSD,
46 (2008), pp. 215-224.
[19] R. Lot and V. Cossalter, A nonlinear rider model for motorcycles, Proc. of FISITA
Wor. Automotive Cong., Yokohama, Japan, (2006).
[20] R.S. Sharp, S. Evangelou and D.J.N. Limebeer, Advances in the Modelling of
Motorcycle Dynamics, J. of MSD, 12 (2004), pp. 251-283.
[21] F. Biral, R. Lot and M. Peretto, Optimization of the Layout of a Racing Motor-
cycle using the Optimal Maneuver Method, Symp. IAVSD, Berkeley, California,
(2007).
[22] A. Saccon, J. Hauser and A. Beghi, A virtual rider for motorcycles: An approach
based on optimal control and maneuver regulation, ISCCSP (2008).
28
[23] M. Corno, M. Massaro, R. Lot and S.M. Savaresi, On LPV Roll Angle Controller
Design for Two-Wheeled Vehicles, Proc. of Bicy. Moto. Dyn., Delft, The Nether-
lands, (2010).
[24] G. Rill, Simulation von Kraft-fahrzeugen, Vieweg, Braunschweig, Germany,
1994.
[25] S. Zhu, S. Murakami b and H. Nishimura, Motion analysis of a motorcycle taking
into account the rider’s effects, J. VSD, (2012).
[26] S. Zhu, H. Nishimura and S. Iwamatsu, Dynamical Analysis of Motorcycle by
Multibody Dynamics Approach, JSDD, 2 (2008), 703-714.
[27] H.B. Pacejka, Tire and Vehicle Dynamics, Butterworth Heineman Ed., 2005.
[28] S. Evangelou, D.J.N. Limebeer and M.T. Rodriguez, Influence of Road Camber
on Motorcycle Stability, ASME J. App. Mech., 75:6, (2008), pp. 231-236.
[29] J.G. Jalon and E. Bayo, Kinematic and Dynamic Simulation of Multibody Sys-
tems: The Real-Time challenge, Springer-Verlag, New-York, 1994.
[30] N. Dinga and S. Taherib, An adaptive integrated algorithm for active front steer-
ing and direct yaw moment control based on direct Lyapunov method, J. VSD, 48
(2010), pp. 1193-1213.
[31] V. Cossalter, A. Doria and R. Lot, Optimum Suspension Design for Motorcycle
Braking, J. VSD, 34 (2000), pp. 175-198.
[32] M. Krstic, I. Kanellakopoulos and P. Kokotovic, Nonlinear and Adaptive Control
Design, John Wiley and Sons, 1995.
Appendices
29
7 Notations
Table 1: Motorcycle geometric specification (lengths are in [m] and angles in [rad])
xGr zGr xGf zGf xGl zGl
0.1289 0.1116 0.0452 0.1237 0.0679 -0.263
xGs zGs xRr zRr xRf zRf
-0.353 -0.0495 -0.549 -0.0638 0.0474 -0.3655
xSl zSl xSu zSu hp lP B
-0.1047 -0.1826 -0.062 0.128 0.3608 0.6831
ρf ρr ǫ
0.324 0.297 -0.4189
31
9 Linear and angular velocity Jacobian matrices
Let be the set of coordinates ex = [1, 0, 0]T , ey = [0, 1, 0]T and ez = [0, 0, 1]T . The
expressions of the linear and angular Jacobian matrices for each body is given by:
∂v oGl
= ex ey ez × rvGl e′z ex × r vGl eθ × r P G l eδ × rBGl eδ 03,3
∂ϑ
∂ω oGl
= 03,2 ez 03 ex 03 eδ 03,4
∂ϑ
∂v oGs
= ex ey ez × rvGs e′z ex × rvGs 03,3 eµ × r P G s 03,2
∂ϑ
∂ω oGs
= 03,2 ez 03 ex 03,3 eµ 03,2
∂ϑ
eµ = eθ
∂v oRr
= ex ey ez × rvRr e′z ex × rvRr 03,3 e µ × r P Rr 03,2
∂ϑ
∂ω oRr
= 03,2 ez 03 ex 03,5 eζr
∂ϑ
eζr = eθ
32
• For the front wheel:
∂v oRf
= ex ey ez × rvRf e′z ex × rvRf eθ × r P Rf eδ × r BRf eδ 03,3
∂ϑ
∂ω oRf
= 03,2 ez 03,1 ex 03,1 eδ 03,2 eζf 03
∂ϑ
eζ,f = Rϕ,θ,ǫ,δ ey
33