Received June 4, 2020, accepted June 10, 2020, date of publication June 16, 2020, date of current version

June 29, 2020.

Digital Object Identifier 10.1109/ACCESS.2020.3002685

Towards an All-Wheel Drive Motorcycle:

Dynamic Modeling and Simulation
ANDREA BONCI , (Member, IEEE), SAURO LONGHI , (Senior Member, IEEE),
AND GIUSEPPE ANTONIO SCALA
Dipartimento di Ingegneria dell’Informazione (DII), Università Politecnica delle Marche, 60131 Ancona, Italy
Corresponding author: Andrea Bonci ([email protected])
This work was supported in part by the EU H2020 ENCORE Project ‘‘Energy aware BIM Cloud Platform in a COst-effective Building
Renovation Context’’, European Union’s Horizon 2020 research and innovation programme under Grant Agreement No 820434, in part by
the REACT Project ‘‘Methods and innovative instruments for REACTive Product Design and Manufacturing’’, Project from the Italian
Ministry of University and Research (MIUR)-in National Operative Plan (PON) for Research and Innovation 2014-2020, Project No
ARS01_01031, Grant Agreement D.D. July 13 2018, n. 1825-Code ARS01_01031, and in part by the HD3Flab Project ‘‘Human Digital
Flexible Factory of the Future Laboratory’’ EU ERDF (European Regional Development Fund), Regional Operative Plan (POR) MARCHE
Region FESR (Fondo Europeo di Sviluppo Regionale) 2014/2020, AXIS 1, Specific Objective 2, ACTION 2.1.

ABSTRACT Modern motorcycles are evolving more and more towards complex systems by the increasing
integration of mechanical, electrical and control disciplines. All-wheel drive (AWD) vehicles have proven
effective to improve vehicle’s performances and rider’s safety. Despite this, manufacturers have developed
few AWD motorcycles and little research has been devoted to them. Obvious difficulties concern torque
distribution to the front wheel because of steering system. Nowadays, the integration of technologies eases
the implementation of front wheel drive opening new research perspectives. In this work, the dynamic
model of an AWD motorcycle with an attached rider is proposed. It represents the first symbolic analysis
investigating the effects of front wheel traction on the dynamics of a motorcycle for supporting the design
of AWD motorcycles reducing trials and tests on prototypes. The proposed model is parametric with respect
to the motorcycle geometry, and it allows to simulate complex operating modes of the AWD, such as
cornering phenomena, taking into account coupling of lateral and longitudinal dynamics and tire-road
interactions. Unlike other works, here the authors include a full tire model by exploiting theoretical slips
of the brush model for tire’s aligning moment too, instead of applying a totally empirical representation less
suitable for a complete symbolic description. Besides, to simulate the equations of motion, the benefits and
disadvantages of using AWD with torque distribution have been pointed out introducing a new handling
ratio. Two verification procedures validate the model: one is performed theoretically, the other carries out
a comparison with a multibody software, whose model is more sophisticated, this latter embeds all main
motorcycle’s dynamics. Although radically different, being the first theoretical and the second numerical-
computational, both methods exhibit consistent behavior between them, and effectiveness of the former is
also consistent with the results of a multibody simulator under the assumptions made.

INDEX TERMS Vehicle dynamics, modeling, simulation, intelligent transportation systems, road vehicles.

I. INTRODUCTION driveline to transfer a fixed portion of the drive torque to the

During the last decades, despite a few motorcycle producers front wheel permanently or else when excessive slip at rear
have spent efforts into proposing solutions to front wheel wheel arises. These vehicles do not fully exploit the factual
drive of AWD motorcycles [1]–[3], researches on the topic potential of AWD motorcycle. Being not conceived for smart
are still lacking. Although different solutions have been management of the drive torque, they are unable to adapt
developed and even more can be foreseen by using electric the latter between both wheels in several complex dynamic
propulsion, these commercial AWD vehicles have all been conditions. Modulating torque distribution between wheels
conceived to overcome conditions of low traction due to will give the designer more control over vehicle’s perfor-
sandy or rough terrains. They use hydraulic or mechanical mances and handling characteristics; this will be proven in the
paper in a simulation scenario, in which analysis of a typical
The associate editor coordinating the review of this manuscript and cornering maneuver will show how the AWD advantages and
approving it for publication was Fabrizio Messina . drawbacks could be enhanced and mitigated, respectively,

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
VOLUME 8, 2020 112867
 A. Bonci et al.: Towards an AWD Motorcycle: Dynamic Modeling and Simulation

by torque modulation between wheels. Benefits of traction real prototype would be subject to during the development
available at the normally undriven wheel improve accelera- as a new product [4]; hence, the model behavior has been
tion and stability in cornering, reduces the tendency of the further compared with a most realistic multibody-simulator,
rear wheel to slip and applies useful drive in the direction the developed in MSC Adams, richer in terms of dynamics, rigid
rider wants to go. Evidences of experienced riders point out bodies and degrees of freedom. The symbolic model shows
these features. The availability of suitable mathematical tools consistent behavior with the multibody simulator under the
able to capture and describe these phenomena would facil- assumptions made. Finally, the benefits and disadvantages
itate engineering development of such vehicles and migra- of using AWD with torque distribution have been analyzed.
tion towards these new traction solutions. This work aims The paper is organized as follows: Section 2 introduces the
at being the first step of a broader research work, which related work; Section 3 and 4 are devoted to describing
develops a model that can lay the foundations for subsequent the AWD symbolic model. Section 5 describes the adopted
investigations of suitable control techniques from which the tire’s model. Section 6 introduces the multibody model,
AWD can benefit. Here, such a mathematical model has Section 7 is focused on the theoretical validation of the
been developed for supporting the design and analysis of model, Section 8 shows comparative simulation and results.
two-wheeled vehicles that intend to take advantage of AWD Section 9 concludes the paper.
features and to reduce trials and tests on prototypes. Regard-
ing technical solutions to deploy AWD potential, the electric II. RELATED WORK
wheel drive could help migration towards full AWD exploita- Over the years, many efforts have been devoted to developing
tion. It is becoming more appealing not only for cars but two-wheeled vehicles (2WVs) with rear wheel drive. The
also for two-wheeled vehicles. Nowadays, small and pow- joint progress between industrial development and mathe-
erful motors can be housed in the vehicle wheel assemblies matical models’ analysis had led to a meaningful techno-
bringing undoubted benefits such as precise and quick torque logical evolution of the 2WVs. The same did not happen
response, accurate drive torque control and the possibility for the AWD version of these vehicles, as instead occurred
of getting helpful information on the wheel angular velocity for cars. The reasons are twofold, technological and theo-
and torque by measuring the electric current absorbed by retical. To understand them better, in the following it will
the motor. Despite technological advances, at present AWD be introduced both the few industrial developments of AWD
motorcycles design still seems to rely on experience rather motorcycles and the main theoretical progress of the 2WVs,
than on an understanding of its phenomena, which instead up to the AWD counterpart. The idea of AWD is not new in
would allow a factual success. Before moving from an idea the field of motorcycles, indeed some early implementations
to a prototype and then to a product, development of newly of all-terrain vehicle date back to the 60s, see e.g. the Rokon
conceived vehicle requires an in-depth investigation of its motorcycles [1]. Then later over the years, some prototypes
distinguishing dynamics to cope in advance with limits and have been developed. However, only few commercial propos-
strengths. This requires high-fidelity dynamic model able to als have been made, mostly involving internal combustion
reproduce, although with approximations, the real vehicle engine (ICE) and with obvious mechanical difficulties in
over normal operating conditions. Once available the equa- distributing the torque through the front wheel. Without going
tions governing the system, by focusing on specific mod- into details of all the prototypes, we will briefly mention
eled dynamics, a suitably control-oriented model could be few notable products that have had commercial implications
arranged, facilitating model-based control with the purpose together with their main features. Rokon manufacturer is a
of directly influence vehicle dynamics not only by steer- long-time producer of AWD motorcycles. They are off-road
ing control but also through the torque distribution between and low speed motorcycles designed for use in the most
wheels. This offers to the designer more degrees of freedom rugged terrain. Rokon uses ICE of about 6 Hp and a combi-
to influence the handling, the driving comfort and safety in nation of belt, chain and shaft drives coupled to gear boxes
a variety of conditions and situations, thus relegating the to drive both front and rear wheel. Back in the early 90s,
necessity for expensive empirical testing only to the vali- Swedish suspension specialists Öhlins [2] started working
dation phase. This paper addresses the issue of defining a on a two-wheel drive system for the Japanese manufacturer
viable symbolic model for describing an AWD motorcycle’s Yamaha. The system is based on a hydraulic pump driven
behavior subject to torque distribution in different driving by the engine in the rear frame which forces oil in hoses to
conditions. The model is generally valid, namely it has been a small hydraulic motor in the front wheel. It activates in
developed regardless of the adopted driveline technology slip condition of the rear wheel; during this, the hydraulic
for the torque distribution to the front wheel. To strengthen pump transfers the exceeding energy of the spinning tire
model reliability, two verification procedures are adopted to to the hydraulic motor of the front wheel by generating a
validate it. The first one validates theoretically the symbolic smooth transmission torque. It involves an additional weight
model in steady-state cornering condition. The second one, of about 8 kg. In 2004 Yamaha produced in limited numbers
relies on multibody analysis, which as a matter of fact in the WR450 2-Trac off-road motorbike. In the same years the
the automotive field is reputed by now as a de-facto stan- two-wheel drive set-up from Öhlins has been deployed on the
dard, suitable to simulate the tests and the behaviors that a Yamaha R1 sportbike. Another noteworthy manufacturer is

112868 VOLUME 8, 2020

A. Bonci et al.: Towards an AWD Motorcycle: Dynamic Modeling and Simulation

the US firm Christini [3], it offers two-wheel drive conver- using a recursive Newton-Euler [21], whereas in [22] an
sions to experience the AWD for off-road bikes and military alternative modeling approach based on the Jourdain’s prin-
applications. The Christini’s AWD system delivers power ciple is presented to derive a mathematical description of
from the motorcycle transmission to the front wheel through a motorcycle vehicle; finally, more recently in [23] it has
a series of chains and shafts, it is complicated, but it is also been shown as the roll angle linearization assumption could
fairly light and can be applied to several bikes. The transferred be removed from rear wheel drive motorcycle models and
torque is predefined by design. The front wheel engages with it has been used for investigating critical motorcycle’s falls,
the drive system and starts to pull as soon as the rear loses such as the low side fall. As showed by the above overview,
traction. The above models have not been conceived to adapt the 2WVs has reached maturity in addressing many dynamic
the drive torque between both wheels in several dynamic aspects. By focusing instead on AWD motorcycle, at best
conditions without slip arising. Besides, a theoretical analysis of authors’ knowledge, a noteworthy multibody model is
did not follow jointly their development, probably because proposed in [24] where the authors analyze some handling
of the need to investigate behaviors mainly under slip condi- characteristics as well as the energy efficiency achievable by
tions. Conversely, over the years this has not happened for the using the torque distribution, this work shows as torque dis-
2WVs, for whom many efforts have been devoted to dynamic tribution does not provide significant improvement in terms
modeling and analysis. In this regard, the paper of Limebeer of energy saving, on the other hand it gives some clues on
and Sharp [5] represents a milestone about the development good handling characteristics of an AWD vehicle negotiating
of the subject concerning single track vehicles modeling and a cornering. While the dynamics of 2WVs has been investi-
control. They explored the subject since the mid-1850s when gated by means of the symbolic model formulation [25] or by
the landmark publications started to appear, and the refer- computer assisted multibody modeling (which overcomes the
ences therein also provide a comprehensive review concern- difficulties in handling complex symbolic equations), AWD
ing motorcycles, ranging from pioneering and simple models dynamics still appears to be in the early stage. For this topic
to more detailed ones. For the sake of brevity, only a few indeed few existing works exploit simple AWD models with
of many noteworthy works in [5] will be mentioned below, one rigid body, e.g. in [26] it has been proposed the AWD
the reader can refer to [5] for further details. Dates back to the slip control in presence of roll angle. However, a single
’70s the influential work of Sharp [6] to the theoretical anal- rigid body model is not suitable to investigate more complex
ysis of motorcycle in straight-running. A linearized model dynamics of AWD vehicle, and a more accurate model is
with minimum complexity that included for the first time sim- advisable. A work addressing practical low-cost solutions for
plified tire-road interaction without aerodynamic effects and the automatic engagement of the front wheel drive in an AWD
suspension motions had been investigated for lateral stability motorcycle can be found in [27]. In this general context, our
analysis together with influences of parameter uncertainties aim is to find a proper mathematical representation of the
on the stability. In the same years was shown the importance motorcycle which describes the essential dynamics issues of
of aerodynamic forces on the performance and stability of the AWD. The model is intended to facilitate the vehicle’s
motorcycles [7]. Some years later the same Sharp [8] tries dynamics control through the torque distribution between
to investigate the effects of acceleration and deceleration wheels.
on motorcycles’ stability with a rather simplified approach
introducing longitudinal equation of motion decoupled from III. AWD MOTORCYCLE MODELING
the lateral equations. Some decades later, in [9] those con- In this section, the model of AWD motorcycle is proposed
clusions were partially refuted by using a higher fidelity in terms of a set of symbolic equations of motion (EOMs),
multibody model but influence of acceleration or braking derived by using Lagrangian formulation. Generally, this
on a cornering machine is still an open issue. During ’90s provides fully nonlinear EOMs having higher complexity of
extensive use of multibody simulators to evaluate the per- expressions and wide operating range. Approaching the anal-
formance of motorcycles has started [10], [11], besides was ysis of newly conceived vehicle, rather than a full complexity
laid the foundations of motorcycle’s tire modeling [12], [13]. model, firstly requires investigation of its distinguishing
In the following decade the contemporary empirical tire mod- dynamics, albeit complex, over limited operating conditions
els for motorcycles was developed [14]. A comprehensive even with approximated models, as shown by vast litera-
overview of many of these models is given in [15]. During the ture [6], [28], [29]. Accordingly, by focusing here on certain
’2000s, Cossalter and Lot [16] present a motorcycle model basic operating conditions, is advisable to avoid the wide
described as a set of 6 bodies for FastBike simulation pro- range of vehicle’s trim conditions linked to a complex model.
gram. The literature concerning the stabilization of 2WV is Subsequently, once verified the model’s effectiveness in con-
quite recent [17], [18], a double loop controller is generally ditions deemed basic, the roll angle linearization assumption
proposed to follow a road path pre-computed by strategies could be removed to widen the operating range focusing
such as the optimal maneuver method [19]. Later, Kooijman on higher performance of the vehicle, as already showed
and Schwab [20] have presented a wide review on bicycle in [23], [30] for the investigation of rear wheel drive motorcy-
and motorcycle rider control. More recently for simulator cle’s falls. This would open the way to a further novel result
application a motorcycle dynamic model was derived by consisting of a symbolic model linearized around non-zero

VOLUME 8, 2020 112869

 A. Bonci et al.: Towards an AWD Motorcycle: Dynamic Modeling and Simulation

trim condition and parameterized with respect to the roll in the paper the subscript i ∈ {r, f } will refer to the rear and
angle, thus would allow it to perform a steady turning analysis the front body respectively. The rear frame, represented by its
at higher performance and investigate on suitable controllers. center of mass Gr includes the rear engine, the propellant stor-
Currently this was attempted with black-box approaches on age, either the petrol tank or the battery storage, the saddle,
rear wheel drive vehicles by acquiring the vehicle’s frequency the rear wheel with radius Rr and the rider. The front frame is
response in steady turning condition from multibody simu- represented by its center of mass Gf , it is constrained to the
lator, by assuming that it can reproduce realistic dynamics rear frame by means of a revolute joint, it includes the front
of the vehicle [31], [32]. The present paper focuses on the engine, the handlebars, the fork, the steering mechanism and
dynamic model of AWD motorcycle able to describe its basic the front wheel with radius Rf . Furthermore, Fig. 1 shows
and distinguishing dynamics; it represents the first stage of the vertical distances h and j of the two centers of mass,
the path above mentioned for developing newly conceived their longitudinal distance k and the wheel base l + b, i.e.
AWD vehicles. As first attempt of analysis of AWD vehi- the distance between the tire-road contact points P and S.
cles, the vertical trim condition has been chosen, afterwards, The steering mechanism is characterized by the steering head
excessive lean angles in cornering have been avoided by angle ǫ and the normal trail t. Numerical values of parameters
means of a virtual rider controller. Thus, nonlinear EOMs have been taken from a well established model [6] and are
have been linearized about the vertical position. The con- reported in Table 4 of Appendix B.
tributions of this work are different, firstly it proposes a
parametric model having the lowest allowed complexity, able B. MOTORCYCLE’S RELATIVE COORDINATES
to describe with enough accuracy the dynamics of AWD vehi- Relative coordinates define the position and orientation of
cle in complex operating modes such as cornering phenom- motorcycle’s bodies with respect to each other. The present
ena, taking into account coupling of lateral and longitudinal paper considers the right-handed coordinate systems (CSs)
dynamics and the tire-road interactions. Besides, it shows shown in Fig. 1: 60 is the inertial CS whose Xo Yo -plane
how to integrate into AWD symbolic model a full tire model represents the road surface with Zo -axis pointing downwards;
by exploiting theoretical slips of the brush model for tire’s 61 originates in A and rotates by the yaw angle ψ about
aligning moment too, instead of applying totally empirical Zo -axis; 62 originates in A and rotates by the roll angle φ
model less suitable for the symbolic modeling. In addition about X1 -axis; a rotation of 62 by the steering head angle
to simulate equations of motion, the model has proven to be ǫ about Y2 -axis and a translation to point B generates the
effective in highlighting effects of the front wheel drive on CS 63 ; a rotation of 63 by an angle δ and a translation to
motorcycle’s dynamics allowing to describe the benefits and Gf generates the CS 64 . The resulting kinematics chain has
disadvantages of using AWD with torque distribution. the following DoFs: the longitudinal and the lateral displace-
The tires model plays a major role in motorcycle modeling, ments of the moving point A in 6o , respectively xo and yo ,
in this regard a specific section will be devoted to describe a vehicle’s angular displacements ψ, φ, δ, respectively the yaw,
proper model and its integration with the symbolic environ- the roll and the steering angle, the front and the rear wheels
ment to provide the expressions of the contact forces and the spinning θr , θf .
contact moments acting on the tires.

A. ASSUMPTIONS AND NOTATIONS

The model derivation makes the following assumptions:
- the motorcycle moves on a flat road surface and the
vertical dynamics and the effects of the suspensions sys-
tem have been neglected. Accordingly, it will be deemed
acceptable for our purposes those traction torques avoid-
ing abrupt vehicle’s accelerations, so that no excessive
fore-and-aft load transfer is triggered. Also, the motor-
cycle tricky phenomena such as the wheelie and the
stoppie are out of the scope of the paper; FIGURE 1. Motorcycle geometric parameters and reference frames.
- the rider has been modeled as a body rigidly attached to
the rear frame;
- the direction and the velocity of the motorcycle is con- C. MOTORCYCLE’S DYNAMICS
trolled by the rider acting on the steering mechanism and The dynamics of the AWD motorcycle is described by a set
on the front and rear engine torques; of 2nd order nonlinear ordinary differential equations (ODEs)
- the contact forces and moments are accounted as exter- obtained by Euler-Lagrange’s formalism (1), where the vector
nal forces generated by a specific tire-road interaction of the generalized coordinates is q = [x1 , y1 , ψ, φ, δ, θr , θf ]T ,
model. and x1 , y1 represent the displacements of the reference point
The motorcycle’s geometry is shown in Fig. 1. It consists A in the CS 61 . The term T = T (q, q̇) is system’s kinetic
of two rigid bodies, the front and the rear one. From now on energy, V = V (q) is the potential energy. Appendix A-A

112870 VOLUME 8, 2020

A. Bonci et al.: Towards an AWD Motorcycle: Dynamic Modeling and Simulation

will provide the derivation of T and V needed to solve (1), tire contact moments. Therefore, the term Qψ is:
the results are summarised in (32), (37), (41), (42) and (50). X2
The term Qq is the vector of the generalized forces acting on Qψ = Qψj
j=1
the generalized coordinates. = (Yf cos γ + Xf sin γ )l − Yr b + (Mzf + Mzr )
d ∂T ≈ Yf l − Yr b + (Mzf + Mzr ) (3)
 
∂T ∂V
− + = Qq (1)
dt ∂ q̇ ∂q ∂q
where the same approximation used in (2) holds. The fourth
The full actuation of AWD motorcycle is obtained by generalized force Qφ is the sum of the moments about the
introducing in Qq three system’s inputs, i.e. the rear and the X1 -axis, exerted by the tires external forces Xi , Yi , Zi and by
front engine torques τwr , τwf , respectively applied on the the overturning moment Mxi internally generated by the tire
transverse axes of the rear and front wheel, and the torque itself, being always i ∈ {r, f }. The external forces acting on
τ applied on the handlebar by the rider. The vector Qq , can the rear wheel do not contribute to this moment because of
be defined as Qq = [Qx1 , Qy1 , Qψ , Qφ , Qδ , Qθr , Qθf ]T . their null arm, therefore Qφ is the sum of three contributes:
It is composed of the forces and moments acting on the X3
motorcycle’s generalized coordinates q. They mainly arise Qφ = Qφj = −tδ(Zf cos φ −Yf sin φ)+Mxr +Mxf cos γ
j=1
from the tire-road interaction and from the steering torque (4)
effects.
The first term Qφ1 is the X1 -component, linearized with
sin γ = γ and cos γ = 1, of the front external moment
Mef = re ∧ Fef expressed in 61 . Fef is the front force with
arm re (distance from the steering axis), both computed in 61 .
Notice that in the front tire CS it is Fef 6 = [Xf , Yf , Zf ]T
Tf
while in an auxiliary CS 65 centered in S and oriented as
64 is re65 = [−t, 0, 0]T hence these vectors converted to
61 allow to compute Mef . The terms Qφ2 and Qφ3 in (4) are
respectively the components about X1 -axis of the overturning
moments Mxr and Mxf , which will be derived in section V-B.
The fifth generalized force Qδ is the sum of the moments
about the steering axis Z3 of Fig. 1, which can be split into
FIGURE 2. External tire forces, moments and their CSs.
three contributes, Qδ1 , Qδ2 and Qδ3 :
X3
In Fig. 2 are shown the external forces Xi , Yi , Zi , i ∈ {r, f }, Qδ = Qδj
j=1
acting on the tire-road contact points P(i = r) and S(i = f ).
= (τ −K δ̇)+t{[(Yf sin φ −Zf cos φ) sin ǫ
Furthermore, P and S are the origins of the tires CSs 6Ti (XTi ,
YTi , ZTi ), i ∈ {r, f } with respect to which the longitudinal + (Xf −Yf γ ) cos ǫ]δ−(Yf cos φ +Zf sin φ)}
and the lateral tire forces Xi , Yi are defined. In the same CSs + Mzf cos ǫ cos φ (5)
we assume Zi as constant vertical loads. Referring to Fig. 2,
the sum of the external forces along X1 and Y1 axes of the The first contribute includes the torque τ which is applied
moving CS 61 , yields respectively the first two generalized by the rider on the steering axis and the torque −K δ̇ due to
forces Qx1 , Qy1 : the steering damper, where K is the damping constant. The
contribute Qδ2 includes the total moment acting along Z3 -
Qx1 = Xf cos γ − Yf sin γ + Xr ≈ Xf + Xr axis, generated by the external tire forces Xf , Yf , Zf applied
on the tire-road contact point S. It can be obtained by merely
Qy1 = Yf cos γ + Xf sin γ + Yr ≈ Yf + Yr (2)
converting the moment Mef previously considered to the
The approximations made in (2) hold under the assump- system 64 , and taking the third component, linearized with
tion of small δ, which also implies small effective steering sin γ = γ and cos γ = 1. The third component Qδ3 is
angle γ = δ cos ǫ. Further detail about γ can be found the aligning moment Mzf along Z3 -axis. Finally, the last two
in [33]. generalized forces are related to the rear and front wheels and
The third generalized force Qψ is the sum of the moments can be expressed as:
acting along Z1 -axis. It can be split into two contributes, Qψ1 Qθi = −τwi + Ri Xi i ∈ {r, f } (6)
and Qψ2 , as shown in (3). The first one includes the moments
arising about Z1 -axis because of the external tire forces Xi , Yi , where τw,i are the engine torques applied on the rear and front
i ∈ {r, f } acting at distances b and l from point A. The second wheels. The effects of the roll resisting moment are neglected
one includes the aligning moments Mz,i , i ∈ {r, f }, arising on since they are countered by wheel’s traction. As can be seen,
the rear and the front tire contact patch [15]. The aligning all the generalized forces depend on the longitudinal and the
moments will be derived later in section V-B devoted to the lateral tire forces Xi , Yi , on the vertical load Zi and on the

VOLUME 8, 2020 112871

 A. Bonci et al.: Towards an AWD Motorcycle: Dynamic Modeling and Simulation

moments acting on the tires. The section V-B, devoted to the [θ̈r ] (iry + iβr2 )θ̈r + (iry + iβr )ψ̇ φ̇ = −τwr + Rr Xr (15)
tire model, will explicit these forces and moments in terms of [θ̈f ] (ify + iβf2 )θ̈f + (ify + iβf )ψ̇ φ̇ − (ify + iβf ) cos ǫ φ̇ δ̇
the kinematic variables of the vehicle.
+ (ify + iβf ) sin ǫ ψ̇ δ̇ = −τwf + Rf Xf (16)
IV. THE AWD SYMBOLIC MODEL: where k = a cos ǫ +e cos ǫ −f sin ǫ and j = a sin ǫ +e sin ǫ +
EQUATIONS OF MOTION f cos ǫ. Despite the tricky symbolism, due mainly to the full
The seven EOMs representing the AWD motorcycle’s parametric representation of EOMs, the whole system is still
dynamics are reported in this section. They have been described by the seven variables in q. The only quantities still
obtained by applying to the nonlinear equations resulting missing to be able to integrate the seven EOMs are the tire
from (1) a first order Taylor expansion around vehicle’s forces Xi , Yi and the moments Mzi , Mxi , i ∈ {r, f } which will
vertical equilibrium point {φ, δ} = {0, 0} and taking into be provided in the following section.
account that γ = δ cos ǫ. The first three nonlinear equa-
tions are relative to longitudinal, lateral and yaw motions V. THE TIRE MODEL
and have been obtained by solving Lagrange’s equations as An effective simulation of motorcycle dynamics requires a
follows [30], [34]: proper tire model able to describe most of phenomena aris-
d ∂T ing from the tire-road interaction. This section is devoted
 
∂T
− ψ̇ = Qx1 (7) to compute the longitudinal and lateral tires forces as well
dt ∂ ẋ1 ∂ ẏ1
d ∂T
 
∂T as the aligning and the overturning moments acting on the
+ ψ̇ = Qy1 (8) tires contact points. For this purpose, the widely accepted
dt ∂ ẏ1 ∂ ẋ1
model conceived by Pacejka [15] has been adapted to the
d ∂T
 
∂T ∂T AWD symbolic model. Pacejka describes the tire forces and
− ẏ1 + x˙1 = Qψ (9)
dt ∂ ψ̇ ∂ ẋ1 ∂ ẏ1 moments in terms of the so-called Magic Formula [15],
The remaining nonlinear equations relative to the motions of which are functions of the tire slip. In Pacejka’s model
the roll, the steer and of the two wheels have been obtained the axes systems are in accordance with the standard SAE
by directly solving (1) with respect to the generalized coordi- J670 and 4976 (shown in Fig. 3a). In automotive, however,
nates φ, δ, θr , θf , whose generalized forces are, respectively, the ISO 8855 1991 (shown in Fig. 3b) is currently adopted
Qφ , Qδ , Qθr , Qθf . The following linearized EOMs have been as simulation-oriented standard for complex models. For this
found: reason, the latter has been here adopted in order to easily
compare the AWD symbolic model of the motorcycle with
[ẍ1 ] (Mf + Mr )(ẍ1 − ẏ1 ψ̇) − Mf k ψ̇ 2 − Mf e cos ǫ δ̇ 2
the model developed by means of a multibody software.
− 2(Mr h + Mf j)ψ̇ φ̇ − 2Mf eδ̇ ψ̇ − Xr − Xf = 0
(10)
[ÿ1 ] (Mf + Mr )(ÿ1 + ẋ1 ψ̇) + (Mr h + Mf j)φ̈
+ Mf eδ̈ + Mf k ψ̈ − Yr − Yf = 0 (11)
[ψ̈] Mf k ÿ1 + (Mf k 2 + Ifz cos2 ǫ + Ifx sin2 ǫ + Irz )ψ̈
 
+ (Ifz − Ifx ) cos ǫ sin ǫ − Crxz + Mf jk φ̈
+ (Ifz cos ǫ + Mf ek)δ̈ + (ify + iβf )θ̇f φ̇
+ (iry + iβr )θ̇r φ̇ + (ify + iβf ) sin ǫ θ̇f δ̇ + Mf k ẋ1 ψ̇
+ Yr b − Yf l − Mzr − Mzf = 0 (12)
[φ̈] (Mr h + Mf j)ÿ1 + [Mf jk + (Ifz − Ifx ) cos ǫ sin ǫ
FIGURE 3. Tire external forces and internal moments: a) std SAE, b) std
− Crxz )]ψ̈ + (Mr h2 + Mf j2 + Irx + Ifx cos2 ǫ ISO.
+ Ifz sin2 ǫ)φ̈ +(Mf ej + Ifz sin ǫ)δ̈ − (iry + βr i)θ̇r ψ̇
The modeling of the tire forces and moments requires two
− (ify + βf i)θ̇f ψ̇ + (Mr h + Mf j)ẋ1 ψ̇
essential wheel’s kinematic quantities, the longitudinal slip λ
− (ify + βf i) cos ǫ θ̇f δ̇ − (Mr h + Mf j)gφ and the side slip α [15]. Such quantities are defined as follow:
+ (tZf − Mf eg)δ − (Ify + iβf ) cos ǫδ θ¨f vx − ωRe vy
λ=− α = arctan (17)
− Mxr − Mxf = 0 (13) vx |vx |
[δ̈] Mf eÿ1 + (Mf ek + Ifz cos ǫ)ψ̈ + (Mf ej + Ifz sin ǫ)φ̈ where ω is the wheel’s angular velocity, Re is the effective
+ (Mf e2 + Ifz )δ̈ − (ify + iβf ) sin ǫ θ̇f ψ̇ + Mf eẋ1 ψ̇ rolling radius, vx and vy are the longitudinal and the lat-
eral velocities of the tire-road contact point. Equation (17)
+ (ify + iβf ) cos ǫ θ̇f φ̇ + (tZf − Mf eg)φ
is generally used by multibody software. As first analysis,
+ (tZf − Mf eg) sin ǫδ − tXf cos ǫδ + K δ̇ − t Yf the effective radius was simulated in both symbolic and multi-
− Mf e cos ǫδ x¨1 − Mzf cos ǫ = τ (14) body model by using the formulation defined by Pacejka.

112872 VOLUME 8, 2020

A. Bonci et al.: Towards an AWD Motorcycle: Dynamic Modeling and Simulation

Under the assumptions imposed on the motorcycle model, These Magic Formula are defined as:
it can be shown that the deformation of the tire is negligible
compared to the unloaded tire, as well as its effects. To further Xio (σi∗ ) = Dxi sin{Cx,i arctan[σi∗ Bxi
simplify the symbolic implementation, the effective radius − Exi (σi∗ Bxi − arctan(σi∗ Bxi ))]} (20)
Re is replaced with the unloaded radius Ri , i ∈ {r, f } while o ∗ ∗ ∗ ∗
Yi (σi , φi ) = Dyi sin{Cyi arctan[σi Byi − Eyi (σi Byi
the multibody model keeps the standard definition (17). This − arctan(σi∗ Byi ))] + Cφi arctan[φi∗ Bφi
choice simplifies the implementation without affecting the
model accuracy as proven later by comparing the two models − Eφi (φi∗ Bφi − arctan(φi∗ Bφi ))]} (21)
in simulation. In literature, the tire-road interaction has been The values of the parameters Bxi , Cxi , Dxi , Exi , Byi , Cyi , Dyi ,
described by two major models which differ in the way the Eyi , Bφi , Cφi , Eφi and Zo used in this paper are reported
longitudinal and lateral dynamics are coupled. The first one is in Table 5 of Appendix B. These parameters are described
the completely empirical model [15] which describes forces by H. B. Pacejka in his book, and are the same used by the
and moments by means of a set of Magic Formula, where the Adams Software. In particular, it has been adopted Pacejka’s
coupling effect is taken into account by multiplying the pure PAC-Motorcycle library, named PAC-MC (120/70R17-Dry).
slip formulation of each dynamics with a proper weighting In literature, such parameters are identified by means of
function. This implies the coupling is hidden in the formula. specific test bench and software, based on empirical approach
The second one is the semi-empirical model, it uses a subset and data acquisition. A detailed description of the acquisition
of Magic Formula in pure slip condition which are function procedures is described in Pacejka’s book.
of the theoretical slip computed by the Brush model [15]. The
majority of the multibody software use the former approach, B. TIRE CONTACT MOMENTS
however, the latter has been chosen in this paper coher- This subsection describes the overturning and the aligning
ently with its purpose of preserving the representation of the contact moments taken into account in the generalized forces
physical phenomena through a theoretical formulation, and (3), (4), (5) of the proposed symbolic model. The formulation
have proved more suitable for the symbolic formulation. The of the overturning moment is based on [15] and [35], the main
Brush model considers the coupling effects by introducing contributions taken into account are due to the lateral force
three theoretical slip quantities, the theoretical longitudinal and rolling effects, as shown below:
slip σxi , the theoretical lateral slip σyi and the total theoretical
Yi
slip σi , defined as follows: Mx,i = Ri Zi (qsx3 − qsx2 φ) i ∈ {r, f } (22)
Zo

1 λi
σxi =
 where Zi , Yi and Zo have been defined above. The parameters
1 + λi
q
σi = σxi2 + σyi2 i ∈ {r, f } (18) qsx2 and qsx3 are taken from the multibody tire model, their
1 tan αi
σyi = meanings are explained in [35] and their values are reported

1 + λi
in Table 5 of Appendix B; this formulation allows the pro-
where λi and αi are the slip quantities given in (17) specialized posed tire model to well approximate the multibody software
for the rear and the front wheel. These quantities are reported model, however for a more in-deep treatment, the reader can
in Appendix A-B. The total theoretical slip σi is precisely refer to [15]. The aligning moment is composed of three
the argument of the Magic Formula describing the contact terms, the first one depends on the lateral force effect, the sec-
forces and moments in the semi-empirical model, these are ond one is related to the longitudinal force effect, also known
described in the following subsections. as the coupling term, and the last one is due to the residual
aligning moment. The aligning moments have been expressed
A. TIRE CONTACT FORCES in compact form as follows [15]:
The longitudinal and the lateral contact forces acting on the Yi
tires, respectively Xi and Yi , are modeled as follows: Mz,i = −tri Yi + Xi (Ri (ssz2 ( ) + ssz3 φ)) + Mres,i (23)
Zo
σxi Zi o ∗
Xi = X (σ ) Zo As shown, the lateral effect is modulated by the terms tri , also
σi Zo i i σi∗ = σi
with Zi (19) known in the literature as the pneumatic trail [15]. This trail
σyi Zi o ∗ ∗ is described by a specific Magic Formula which is a function
Yi = Y (σ , φ ) i ∈ {r, f }
σi Zo i i i of the theoretical side slip σyi and the total slip quantities σi
where Zi represents the constant vertical load acting on the as follows:
front and rear tires while Zo is the nominal wheel load [15]. σyi Zi
tri = tr (24)
The lateral contact force also includes the effect of the incli- σi Zo Ro,i
nation of the tire by considering the tire camber angle φi∗ , trRo,i = Dti cos{Cti arctan[Bti σi∗ − Eti (Bti σi∗
where φr∗ = φ and φf∗ = φ + δ sin ǫ. Each force in (19) − arctan(Bti σi∗ ))]} cos(σyi ) (25)
is composed of the product of three terms: a coupling fac-
tor, a load factor and an empirical Magic Formula (Xio (σi∗ ) the values of the parameters Dti , Bti , Cti are reported in Table 5
and Yio (σi∗ , φi∗ )) describing the force in pure slip condition. of Appendix B. The second term of (23) is the coupling

VOLUME 8, 2020 112873

 A. Bonci et al.: Towards an AWD Motorcycle: Dynamic Modeling and Simulation

effect due to the longitudinal force Xi . As done for the the rider, and a second one representing the rear wheel. This
overturning moment (22), the formulation of the aligning has been done to allow the insertion of tire-road constraint
moment has been slightly simplified compared to the Pacejka in multibody environment. The same procedure has been
model [15], [35], i.e. only the lateral and the roll effects have performed for the front frame, by splitting the front wheel
been kept as well as the residual term Mres,i . The parameters from the front frame. The position of the centers of mass and
ssz2 and ssz3 are equivalent to those used in the multibody the inertia matrices of these new bodies have been suitably
tire model described in [35] and their values are reported rearranged (see Table 6 of Appendix B for the multibody
in Table 5 of Appendix B. Referring to the residual aligning model parameters). As regards to the body joints, the front
moment Mres,i , in order to compare its effect with the com- wheel is fixed to the suspension by a revolute joint which is
pletely empirical model used by multibody tools, the formula- fixed to the fork by a prismatic joint, the rear wheel is similar,
tion here proposed differs from the semi-empirical model that but a second revolute joint at distance b from the wheel fixes
is known in the literature [15]. Such difference consists in the the suspension to the rear frame; the front and rear frames are
use of a Magic Formula depending on the theoretical slip σi∗ , connected by a revolute joint about the steering axis. A steer-
σyi of the Brush model. The Magic Formula is derived from ing damper is also modeled by a rotational spring damper
the completely empirical model (see parameters in Table 5, acting between the two frames. The motorcycle is actuated by
Appendix B) and depends on the roll angle φ. A correction the torques applied to the revolute joints of both the wheels
factor (σyi /σi )(Zi /Zo ) is introduced to take into account the and to the steering mechanism. These torques will have the
side slip effect and vertical load normalization, therefore the same profile of the control inputs of the AWD symbolic
formula reads as: model. The MSC Adams tire model has been implemented
to model the tire-road interaction. It uses the fully empirical
σyi Zi
Mres,i = MRo,i (26) tire model, based on [15] and it is illustrated in [35]. Note that
σi Zo in Adams the CS, defining tire forces and moments, is defined
MRo,i = Dresφ,i cos(arctan(Bresi σi∗ )) cos σyi (27) according to the standard ISO 8855, nevertheless the Adams
Dresφ,i = ai φ(bi |φ| + ci ) (28) tire model is equivalent to the one from Pacejka [15], which
uses SAE standard, as previously shown in the tire section.
VI. THE AWD MULTIBODY MODEL In the multibody environment, additional CSs equivalent to
Taking into account that in many high-tech fields such auto- those from the symbolic model (60 , 61 , 62 , 63 , 64 ) were
motive, aerospace etc., the multibody analysis is a test-bench defined, so that the generalized coordinates could be mea-
to simulate tests and behaviors of real prototypes during sured along the same axes in both models, this allowing to
their development as a new product [4], the proposed model coherently compare such variables.
of this new kind of vehicle will be later compared with a
higher-realistic multibody model which is here introduced.
The MSC Adams software has been used to develop the AWD
motorcycle in a multibody environment (Fig. 4). In order
to have a test-bench as close as possible to the reality,
the vertical dynamics has been included in the Adams model,
thus, it includes both the front and rear wheel suspension
in order to pursue the best fidelity. The Adams model uses
the same linear dimensions, masses, inertia and tires model
used in the symbolic model. Therefore, under the same input
FIGURE 4. Multibody model of the AWD motorcycle.
torques, suitably chosen in order to not trigger the vertical
dynamics, it is expected a similar behavior of both symbolic
and Adams models. In order to simulate the full dynam-
ics of the AWD motorcycle in the multibody environment, VII. THEORETICAL AWD MODEL VERIFICATION
at least a model consisting of four rigid bodies is required. In order to validate the proposed AWD motorcycle model,
This model has 11 DOFs for describing the bodies relative a two steps procedure has been followed. This section con-
motion, 3 translational about the axes of the motorcycle CS cerns the first step and it provides the theoretical evaluation
(previously marked as 61 ), 3 rotational (Roll, Pich, Yaw), of the balance of forces and moments acting on the vehicle in
2 rotations about wheels revolution axes, 1 rotation about steady turning, moreover the power balance is also verified.
steering axis, 1 translational along the steering axis due to The next section will describe the second step of validation,
the front suspension system and 1 rotational about an axis where a comparison of the behavior of the AWD motorcycle
parallel to Y2 -axis due to the rear suspension system. The road model with the multibody software counterpart will be made.
is locked by a fixed joint to the ground CS (previously marked The first validation has followed the procedure described
as 60 ). The rear frame of the motorcycle has been split into in [25]. The vehicle in steady turning has been simulated and
two rigid bodies: a rear frame without the wheel, including the following three conditions have been verified in order to
the rear engine, the propellant storage system, the saddle and hold the vehicle in equilibrium:

112874 VOLUME 8, 2020

A. Bonci et al.: Towards an AWD Motorcycle: Dynamic Modeling and Simulation

- The sum of the external forces acting on the vehicle must Since we deal with sideslip value of 0.02 rad and side force
equal the sum of the inertial and gravitational forces; aerodynamic coefficient that approximates zero, the contri-
- The sums of the moments acting on the vehicle must bution of the aerodynamic side force is about 3% of the drag
equal zero; force. Besides, the computed side force is about 2% of the
- The power provided to the system by the motors must lateral force acting on the front tire-road contact point and 1%
equal the power losses. of the lateral force acting on the rear tire-road contact point.
The steady state condition for the vehicle has been achieved These results come from the evaluation of experimental data
by means of a simple rider model controlling the driving relating to a motorcycle running a curve at a constant speed
torque and the steering torque in order to have the motorcycle of 15 m/s and different yaw angles. They seem to fully agree
running along a circular trajectory with a radius of curvature with the assumptions made.
of 200 m at a constant velocity of 15 m/s. In this state other In Fig. 5.a is depicted the behavior of the norm of the
kinematic parameters are: roll angle φ = 7.4 deg, steering vector |1F|, which reaches a value of 0.4 N at steady state.
angle δ = 0.53 deg; yaw velocity ψ̇ = 0.075 rad/s. Such quantity is 0,04% of the overall force acting on the
The rider model follows a well known schema, e.g. the one vehicle in the steady state.
reported in [24] and it consists of two PI (Proportional Inte-
grative) controllers with values P1 = 2, I1 = 210, P2 =
50, I2 = 150. In order to verify the first condition the
following balance equation was considered that expresses the
force error vector 1F as:
X X
1F = Fi,tire + Mj (ωj ∧ vGj + g) + Faero (29)
i j

TABLE 1. Force balance - components. FIGURE 5. Fig (a): Time evolution of force and moment vectors’ norm;
zoomed graph: evolution in time frame 15 - 30 sec. Fig (b): Time evolution
of resulting power; zoomed graph: evolution in time frame 15 - 30 sec.

In order to verify the second condition related to the

moments balance, the following balance equation was con-
sidered that expresses the moment error 1M as:
X X
1M = M i,tire + M i,tireint
i i
X X
+ (M j,inertial + M j,grav ) + M i,gyro + M aero
j i
where the subscript i ∈ {r, f } stands for the rear and the
X X
= (Ci − A)61 ∧ Fi,tire + M i,tireint
front tire/wheel, while the subscript j ∈ {r, f } stands for the i i
rear and the front center of mass (COM) having mass Mj . +
X X
(Gj −A)∧Mj (ωj ∧vGj +g)+ M i,gyro +M aero
All the vectors of (29) are referred to the moving CS 61 j i
and their meaning are detailed in Table 1. In the term Faero
(30)
only the drag component is considered hence the lift force,
the side force and their relative moments have been neglected. again, the subscripts i, j maintain the meaning defined above.
The importance of aerodynamic forces on the performance The components of (30) are detailed in Table 2. In Fig. 5.a is
and stability of motorcycles at high speeds was demonstrated depicted the trends of the norm vector |1M|, at steady state
in [7] and pointed out in [5]. Generally, aerodynamic forces
are not negligible, in particular they are relevant at high
TABLE 2. Moments balance - components.
speeds. In our research, we refer to low accelerations and
not high speeds, resulting in low wind yaw angles and minor
crosswind effects, in that case much of aerodynamic effects
are negligible excepting for drag force. This is supported by
experimental data and theoretical assessment in [7]. It was
also found that in these conditions the lift force coefficient
was close to zero, therefore validating the assumptions made
in this work. Above evaluations are also confirmed by numer-
ical findings: aerodynamic side force is proportional to the
sideslip angle and the side force aerodynamic coefficient.

VOLUME 8, 2020 112875

 A. Bonci et al.: Towards an AWD Motorcycle: Dynamic Modeling and Simulation

is stable at 0.3 Nm, which is 0.06% of the resulting moments

acting on the motorcycle. Finally, the third condition regard-
ing the power balance has been verified by considering the
power error 1P, expressed as:
X X X
1P = Pi,drive + Pi,slip + Pi,tire + Paero (31)
i i i

where i ∈ {r, f }. Each component of (31) is detailed

in Table 3. In Fig. 5.b is shown how the power checksum
behave, and at steady state it reaches a mean value of 0.32 W
between 15 and 25 seconds. Such value represents a 0.022%
FIGURE 6. Rear and front wheel’s torques (τwr , τwf ) and minor and
of the power provided by the traction system of 1446 W. major steering torques (τmS , τMS ).

TABLE 3. Power balance - components. software and exported to Matlab in order to have the same
input profiles for both models.
As the multibody model has both vertical and suspension
dynamics, it is possible to quantify the fore-and-aft load trans-
fer under the acceleration generated by the torque applied
in Fig. 6. As shown in Fig. 7, the rear and front vertical
forces Fzr , Fzf have a load transfer of about 70 N from the
front to the rear frames. This clearly means a negligible force
compared with the total load of the vehicle of 2433.8 N, about
3%. It was tested that is acceptable to neglect such effects
VIII. COMPARATIVE VALIDATION AND RESULTS even at 10% of load transfer, as the two models still show
In this section, the AWD motorcycle model will be fur- compatible simulation results.
ther validated by means of comparative simulations with
the multibody software counterpart. Furthermore, simulation
tests highlighting AWD features pointed out by experienced
riders will be provided. The proposed symbolic model has
been implemented in Matlab Simulink environment. Com-
parative simulations, described in the next subsection, have
investigated model reliability and effectiveness in two note-
worthy conditions: maneuvers with roll and steering angles
exceeding the linearization range; acceleration in cornering
of AWD motorcycle when coupling dynamics is triggered.
A further subsection will provide AWD vehicle behaviors FIGURE 7. Trend of rear and front vertical forces (Fzr , Fzf ) under
compared with the rear-wheel-drive (RWD) vehicle. moderate vehicle acceleration.

A. CORNERING COMPARATIVE VALIDATION

Two simulations will be proposed here, named minor steering
torque (mS) and major steering torque (MS). Both aims to
estimate errors between symbolic and multibody models by
comparing the roll angle, the steering angle and the trajecto-
ries. Before detailing simulation results, it is worth noting that
the comparison was made with a higher-realistic multibody
model than the symbolic one. Adams model includes vertical
dynamics, more degrees of freedom, is free of linear approx-
imations, and uses Adams tire model. For our purposes, it is
closest to a real AWD byke.
Fig. 6 shows the wheels input torques applied in both
simulations, that are: τwr , τwf the rear and front respectively;
the steering torques named τmS , τMS are respectively used
in mS and MS simulations. A ratio of τMS = 3τmS will be
considered. In order to avoid the overlapping of torques tran- FIGURE 8. AWD models: comparisons of roll and steering angles.

sients, wheels torques act after the steering torques transient, Fig. 8 shows the comparison between the responses of
at t = 10s. These inputs have been generated in Adams AWD symbolic model versus AWD multibody model for

112876 VOLUME 8, 2020

A. Bonci et al.: Towards an AWD Motorcycle: Dynamic Modeling and Simulation

the two simulations scenarios. In mS scenario the roll angle

reaches 5.5 deg while the steering angle remains limited to
1 degree. The roll error magnitude between the two models
remains below 2%, confirming expected results under lin-
earization assumptions. In the MS scenario larger steering
torque is applied, causing the vehicle to roll up to 14 deg
and to steer up to more than 2 deg, nevertheless the roll error
remains limited under 8%. On the other hand, the different
steering angle behaviors (Fig. 8) produce a trajectory error
between the two models as shown in Fig. 9. In the first 10 sec-
onds the trajectory error is minimal, afterwards the difference FIGURE 10. Steering torque τs applied in both scenarios d1 , d2 , while
is stressed during the activation of the wheels torque. As an driving torque τd applied in d1 and τd applied in d2 .
1 2
overall result, the effect of the steering error on the trajectory
appears negligible and the symbolic model performs the same
as the multibody model both under the assumptions made of
small angles and beyond them. Similar overperformance in
the roll angle was outlined in a previous work [36] related to
single track vehicle.

FIGURE 11. Scenario d1 (entering the curve): roll angle and yaw speed
comparison between RWD and AWD.

in vertical trim, countering the desired turning, such behavior

resembles the oversteer effect. On the other hand, the AWD
FIGURE 9. AWD models: trajectories comparisons. simulation shows that the use of the front torque in the same
circumstances is even desirable to obtain a sharper cornering
trajectory, in fact yaw speed reaches a higher peak value than
that delivered by RWD. The same behavior affects the roll
B. AWD TESTING RESULTS
angle. This phenomenon resembles an understeer effect.
Specific simulations have been carried out in order to high-
light the effects of the front wheel drive on motorcycle’s
dynamics. Two relevant driving scenarios named d1 (entering
the curve) and d2 (exiting the curve) are here considered. The
same steering torque τs , shown in Fig. 10, is used in both
scenarios. In d1 , the motorcycle is accelerated by τd1 when
approaching a curve, namely between 5 and 11 seconds when
the steering torque is increasing. In scenario d2 the vehicle
is accelerated by τd2 when coming out of the curve, namely
between 11 and 17 seconds. The simulations start with the
motorcycle in vertical trim and constant speed of 8 m/s. For
each scenario, two simulations named RWD (Rear Wheel
FIGURE 12. Scenario d2 (exiting the curve): roll angle and yaw speed
Drive) and AWD are carried out by considering two differ- comparison between RWD and AWD.
ent driving torque distribution: in RWD the whole driving
torque is applied to the rear wheel (as in a standard motor- In scenario d2 , when the front torque is acting (see Fig. 12
cycle); in simulation AWD the 20% of the driving torque is at t = 11 s), AWD yaw speed decreases slower with respect
applied to the rear wheel and 80% to the front wheel. Fig. 11 to the RWD case, as well as the roll angle is held around the
shows the comparisons between the roll angle and yaw speed maximum value. Here the front torque affects the behavior
obtained in both simulations RWD and AWD for scenario d1 . negatively while the rear torque, as expected, slightly affects
As experienced by most riders, the RWD simulation shows the yaw speed as it appears steeper than the rising part of
that the use of a rear torque while approaching the curve is the trajectory (t = [5; 11] s). This phenomenon resembles
not advisable, because it affects the roll angle and yaw speed the behavior experienced by the few AWD riders. As they
negatively, indeed the rear tire contact forces keep the vehicle described, it’s harder to regain the vertical trim while exiting

VOLUME 8, 2020 112877

 A. Bonci et al.: Towards an AWD Motorcycle: Dynamic Modeling and Simulation

a curve by operating the front torque. To conclude, the sim-

ulations have shown that the use of a front drive eases the
maneuver when approaching a curve, while it may interfere
the coming out of a curve.

C. AWD HANDLING RESULTS

The front traction effect on motorcycle’s handling can be
considered advantageous or disadvantageous depending on
the rider’s intention. Indeed, if the rider accelerates with
the front torque when approaching a cornering maneuver,
a positive effect can be experienced facilitating the rider in FIGURE 14. Exiting the curve: steer/roll ratio and yaw speed comparison.
the maneuver. On the contrary, the effect is negative if the
front torque acts at the end of the curve, because more effort
is required to the rider to restore the vehicle’s vertical trim. In
this subsection, the vehicle’s handling is analyzed in various wheel and half to the front one, 0.8, indicating 80% of the
simulation tests characterized by a different value of the ratio torque on the front wheel and 20% on the rear one. The
between the steer and the roll angle. A feedback control time period when the steer torque acts is indicated by the
action ensures the vehicle to have the same longitudinal speed blue area, while the red area shows the traction action period.
along the trajectory. This control action guarantees that the When approaching the cornering maneuver at 7 seconds the
motorcycle dynamics generated by different traction ratios ratio in Fig. 13 has a value of 0.2, while it is 0.17 when
is easily comparable, since it is clearly dependent by the only the rear torque is acting. In other words, the steering
longitudinal speed [6]. In the simulations, different traction angle can be higher with the same rolling angle and hence
distribution between rear and front wheels are tested in two a smaller curvature radius can be achieved, if required by
different moments of the same cornering maneuver. The steer the rider. Conversely, in the scenario of Fig. 14, the front
torque acts in all simulations from 5 to 17 seconds with the traction acts when the curve ends, and the effects lead to
same trajectory shown in Fig. 10. The traction torque is gen- similar consideration as the ratio is higher. The reason to
erated by the feedback controller, which imposes the desired consider this last behavior as negative is based on the rider
longitudinal speed from 8 m/s to 12 m/s. Such speed variation intention to end the curve by reaching the vehicle’s vertical
takes place in entering the curve from 5 to 11 seconds, while trim. If the rider aims at reaching such trim with a straight
at the end of the cornering it occurs from 11 to 16 seconds. steer, this requires more effort when the front torque acts.
In all situations, the yaw rate differences are small (see Fig. 13 Because of the higher rolling angle reached halfway in the
and 14), nonetheless, a higher value is reached when the front cornering maneuver and the front torque action increasing
torque acts at the beginning of the cornering maneuver. the ratio δ/φ, the steer angle is delayed to reach the zero roll
angle. If the rider’s goal is to force the steer angle to reach
faster the zero trim as intended, this will require more effort
to the rider. By concluding, a handling ratio is introduced to
better understand the effect of the front torque, and such effect
proves to be bivalent as it eases the rider in approaching the
cornering maneuver, whilst it hinders the rider when exiting
a curve.

IX. CONCLUSIONS AND FUTURE WORKS

In this work, an AWD motorcycle model has been outlined
and the relevant steps needed to derive the EOMs, linearized
FIGURE 13. Entering the curve: steer/roll ratio and yaw speed
with respect to the roll and the steering angle have been
comparison. shown. The model has been integrated with an adapted ver-
sion, but nonetheless effective, of the tire model developed by
The effect of the front traction can be explained analyzing H. B. Pacejka which describes, by means of a semi-empirical
the ratio between steer and roll angle in Fig. 13. When the approach, the coupling between forces and moments arising
front torque is acting, this ratio is higher, which means that on the tire-road contact points. The AWD motorcycle model
lower rolling angle can be achieved with the higher steering thus obtained has been validated in two steps; the first one
angle. This can be considered as a positive or negative effect follows a theoretical approach commonly used in literature
depending on the intention of the rider. The Fig. 13 and 14 and it proves the modeled vehicle can be driven to an equi-
show three simulations with the following values of the rear- librium condition. This is done by evaluating the balance
on-front torque ratio respectively: 0, indicating that only rear of forces and moments acting on the vehicle in a steady
torque applied, 0.5, i.e., half torque is applied to the rear state cornering maneuver; in the second step the AWD model

112878 VOLUME 8, 2020

A. Bonci et al.: Towards an AWD Motorcycle: Dynamic Modeling and Simulation

has been compared in simulations against a richer model of By substituting (34), (35) and (36) in (33), the rear kinetic
the same motorcycle, implemented in a multibody software. energy becomes:
The comparison in cornering condition of the two models 1
have shown small differences. Finally, further simulations Tr = Mr [(ẋ1 − h sin φ ψ̇)2 + (ẏ1 + h cos φ φ̇)2
2
have been carried out to evaluate the effects of a torque 1 1
distribution on the vehicle dynamics. For this purpose an + (h sin φ φ̇)2 ] + Irx φ̇ 2 + Iry (sin φ ψ̇)2
2 2
appropriate handling ratio has been proposed. The benefits
+ Irz (cos φ ψ̇)2 − Crxz cos φ φ̇ ψ̇ (37)
and disadvantages of using full rear drive or AWD with torque
distribution have been highlighted and the results agree with The kinetic energy of the front frame Tf depends on the
factual behaviors of an AWD motorcycle experienced by velocity vf of Gf and its angular velocity ωf . Therefore,
riders. Future works will aim to investigate more in-depth by considering the speed vB of the point B and its angular
the best way to use the front traction in different scenarios, velocity ωB , the velocity vf is given by:
even when slip conditions occur; this will be functional to
the developing of control systems that will properly act on vf = vB + ωB ∧ (Gf − B)
ẋ1 − ψ̇(sin ǫ sin φ(a + e cos δ)+
 
the torque distribution in order to obtain improved and safer
vehicle’s behaviors without obstructing the driver. + e cos φ sin δ + f cos ǫ sin φ) − δ̇e cos ǫ sin δ 
 

 ẏ1 + ψ̇(cos ǫ(a + e cos δ) − f sin ǫ)+ 

APPENDIX A  + φ̇ cos φ sin ǫ(a + e cos δ) + δ̇e cos δ cos φ+ 
 
KINEMATIC MODEL = + φ̇f cos ǫ cos φ − φ̇e sin δ sin φ+ 

A. KINETIC AND POTENTIAL ENERGIES 
 − δ̇e sin δ sin ǫ sin φ 

In this appendix the kinetic energy T and the potential energy 
 sin φ φ̇ sin ǫ(a + e cos δ)+ 

V of the system are derived.  + δ̇e cos δ sin φ + φ̇e cos φ sin δ+ 
+ φ̇f cos ǫ sin φ + δ̇e cos φ sin δ sin ǫ
T = Tr + Tf + Tω (32) (38)
Tr and Tf are the kinetic energies of the rear and the front The angular velocity ωf of Gf is the angular velocity of 64
frame, they are given by: and it is given by:
1 1 ψ̇(sin δ sin φ − cos δ cos φ sin ǫ) + φ̇ cos δ cos ǫ
 
Ti = Mi vi 2 + ωi T Ii ωi i ∈ {r, f } (33)
2 2 ωf =  ψ̇(cos δ sin φ + cos φ sin δ sin ǫ) − φ̇ cos ǫ sin δ 
where Mi , Ii , are the inertial properties of the two frames, δ̇ + φ̇ sin ǫ + ψ̇ cos ǫ cos φ
i.e. their masses and their inertia matrices respectively, vi and (39)
ωi relate to the kinematic properties of the centers of mass
The inertia matrix If of the front frame, related to the principal
Gi and represent the square of their speed vector and their
axes of inertia X4 Y4 Z4 is given by:
angular velocity vector, respectively. The term Tω takes into
account the rotational extra terms not included in Ti such as
 
If = diag Ifx , Ify , Ifz (40)
the rotations of the wheels and flywheels.
The kinetic energy of the rear frame Tr , given by (33), The elements of If are reported in Table 4 of Appendix B.
depends on the speed vr of the rear mass center Gr and on By using the quantities (38), (39) and (40) the front kinetic
its angular velocity ωr . The vector vr is given by: energy Tf becomes:
1
ẋ1 − h sin φ ψ̇ Tf = Mf [(ẋ1 −e cos ǫ sin δ δ̇−(a sin ǫ sin φ +e sin δ cos φ
 
2
vr = vA + ωA ∧ (Gr − A) = ẏ1 + h cos φ φ̇  (34)
+ e sin ǫ cos δ sin φ + f cos ǫ sin φ)ψ̇)2
h sin φ φ̇
+ (ẏ1 + a sin ǫ cos φ φ̇ −e sin δ sin φ φ̇ + e cos δ cos φ δ̇
where vA and ωA are the speed and the angular velocity of 61 . + e sin ǫ cos δ cos φ φ̇ − e sin ǫ sin δ sin φ δ̇
The vector ωr is the angular velocity of 62 and holds:
+ f cos ǫ cos φ φ̇ +(a cos ǫ +e cos δ cos ǫ − f sin ǫ)ψ̇)2
 T
ωr = φ̇, sin φ ψ̇, cos φ ψ̇ (35) + (a sin ǫ sin φ φ̇ + e sin δ cos φ φ̇ + e cos δ sin φ δ̇
We assume Ir as the matrix of inertia of the rear frame whose + e sin ǫ cos δ sin φ φ̇ + e sin ǫ sin δ cos φ δ̇
1
inertia moments and inertia products are about axis parallel + f cos ǫ sin φ φ̇)2 ] + Ifx [(cos ǫ cos δ φ̇ + sin δ sin φ
to X2 Y2 Z2 through rear mass center. Their values are reported 2
1
in Table 4 of Appendix B: − sin ǫ cos δ cos φ)ψ̇]2 + Ify [− cos ǫ sin δ φ̇
2
+ (cos δ sin φ + sin ǫ sin δ cos φ)ψ̇]2
 
Irx 0 −Crxz
Ir =  0 Iry 0  (36) 1
−Crxz 0 Irz + Ifz [δ̇ + sin ǫ φ̇ + cos ǫ cos φ ψ̇]2 (41)
2
VOLUME 8, 2020 112879
 A. Bonci et al.: Towards an AWD Motorcycle: Dynamic Modeling and Simulation

In order to compute the total kinetic energy of the system (32) TABLE 4. Symbolic model - geometric and inertial parameters.
the extra kinetic energy Tω must be derived. The contribute
Tω is related to the rotation of the wheels and the engine
flywheels, it can be expressed by:
X X
Tω = Tωwi + Tflyi i ∈ {r, f } (42)
i i
The terms Tωwi are the kinetic energies of the rear and front
wheel not taken into account by Tr and Tf , whereas Tflyi are
the rotational energies of the engines flywheels included in
the vehicle. The contribute Tωwi can be derived by defining
ωTwi as the total angular velocity of the wheel, expressed as:

ωTwi = ωi + [0, θ̇i , 0]T i ∈ {r, f } (43)

where ωi was defined in (35) for i = r and in (39) for i = f .
The vector [0, θ̇i , 0]T is the rotational angular velocity of the
wheel, it is defined in 62 for i = r and in 64 for i = f .
The contribute Tωwi can be derived as a difference of kinetic
energies generated by ωTwi and ωi which yields:
1 T 1
Tωwi = ω ii ωTwi − ωi T ii ωi i ∈ {r, f } (44)
2 Twi 2 TABLE 5. Tire parameters.

where ii is the diagonal matrix of inertia of the wheel com-

pletely defined by the polar moment of inertia iry and the
camber inertia irx = irz , whose values are reported in Table 4
of Appendix B. Finally, by solving (44) for the rear and front
wheels it yields:
1 2
Tωwr = iry (sin φ ψ̇ θ˙r + θ˙r ) (45)
2
Tωwf = ify [(− cos ǫ sin δ φ̇ + (cos δ sin φ
1
+ sin ǫ sin δ cos φ)ψ̇)θ̇f + θ̇f2 ] (46)
2
Similarly, the engine flywheels contribute Tflyi , i ∈ {r, f },
is derived by applying (44) and taking care to replace Tωwi
with Tflyi , ii with iflyi and ωTwi with ωTflyi , the latter defined
as: where Vi is the potential energy of the mass center Gi , Mi
ωTflyi = ωi + [0, βi θ̇i , 0] T
i ∈ {r, f } (47) is the mass of the frame i, g is the gravitational acceleration
and zi is the height of Gi with respect to the ground level.
where βi is the gear ratio between the wheel and the engine The terms zi are easily derived by considering the component
flywheel and iflyi is the polar moment of inertia of the fly- along Z1 axis of the vector (Gi − A) computed in 61 .
wheel. Therefore the kinetic energies of the flywheels hold:
1 2 B. REAR AND FRONT SLIP
Tflyr = iflyry (βr sin φ ψ̇ θ̇r + βr2 θ˙r ) (48) The rear and front lateral slip αr and αf are formulated as
2
Tflyf = iflyfy (− cos ǫ sin δβf θ̇f φ̇ + cos δ sin φβf θ̇f ψ̇ follows [15]:
1 ẏ1 − bψ̇ ẏ1 + l ψ̇ − t δ̇
+ cos φ sin δ sin ǫβf θ̇f ψ̇ + βf2 θ̇f2 ) (49) αr = − αf = δ cos ǫ − (51)
2 ẋ1 ẋ1
The values of the polar moments iflyry and iflyfy are reported
in Table 4 of Appendix B. The longitudinal slip λr and λf are given by [15], [30]:
The potential energy of the system is the sum of the poten-
ẋ1 + Ri θ̇i − Ri sin φ ψ̇
tial energy of two mass centers Gr and Gf and it is given by λi = − i ∈ {r, f } (52)
X X ẋ1
V = Vi = Mi g zi = Mr g zr + Mf g zf
i i Considering the front wheel contact point S, the front longi-
= Mr gh cos φ + Mf g(a sin ǫ cos φ − e sin φ sin δ tudinal slip λi=f is approximated by putting δ = 0.
+ e sin ǫ cos δ cos φ + f cos ǫ cos ǫ) i = {r, f } (50) Equation (52) with i = f fixes the error appearing in [30].

112880 VOLUME 8, 2020

A. Bonci et al.: Towards an AWD Motorcycle: Dynamic Modeling and Simulation

TABLE 6. Multibody model - geometric and inertial parameters. [18] R. Lot, M. Massaro, and R. Sartori, ‘‘Advanced motorcycle virtual rider,’’
Vehicle Syst. Dyn., vol. 46, no. sup1, pp. 215–224, Sep. 2008.
[19] A. Saccon, J. Hauser, and A. Beghi, ‘‘A virtual rider for motorcycles:
An approach based on optimal control and maneuver regulation,’’ in
Proc. 3rd Int. Symp. Commun., Control Signal Process., Mar. 2008,
pp. 243–248.
[20] J. D. G. Kooijman and A. L. Schwab, ‘‘A review on bicycle and motorcycle
rider control with a perspective on handling qualities,’’ Vehicle Syst. Dyn.,
vol. 51, no. 11, pp. 1722–1764, Nov. 2013.
[21] S. Hima, L. Nehaoua, N. Seguy, and H. Arioui, ‘‘Suitable two wheeled
vehicle dynamics synthesis for interactive motorcycle simulator,’’ in Proc
IFAC World Cong., Seoul, South Korea, 2008, vol. 41, no. 2, pp. 96–101.
[22] L. Nehaoua, H. Arioui, N. Seguy, and S. Mammar, ‘‘Dynamic modelling
of a two-wheeled vehicle: Jourdain formalism,’’ Vehicle Syst. Dyn., vol. 51,
no. 5, pp. 648–670, May 2013.
[23] A. Bonci, R. De Amicis, S. Longhi, E. Lorenzoni, and G. A. Scala,
‘‘Comparison of dynamic models for a motorcycle during lowside fall,’’
WSEAS Trans. Appl. Theor. Mech., vol. 12, pp. 78–85, Jan. 2017.
[24] J. W. Griffin and A. A. Popov, ‘‘Multibody dynamics simulation of an all-
wheel-drive motorcycle for handling and energy efficiency investigations,’’
Vehicle Syst. Dyn., vol. 56, no. 7, pp. 983–1001, Jul. 2018.
[25] S. Evangelou, Control and stability Analysis of Two-Wheeled Road Vehicle.
London, U.K.: Univ. London, 2004.
[26] T. Abumi and T. Murakami, ‘‘Posture stabilization of two-wheel
APPENDIX B drive electric motorcycle by slip ratio control considering camber
VEHICLE PARAMETERS angle,’’ in Proc. IEEE Int. Conf. Mechatronics (ICM), Mar. 2015,
See Tables 4–6. pp. 353–358.
[27] J. Ben, B. V. Aswin, R. Adityan, and G. T. Basil, ‘‘Two wheel drive
motorcycle,’’ IOSR J. Mech. Civil Eng., vol. 11, no. 1, pp. 25–32,
REFERENCES 2014.
[1] Rokon Website. Accessed: Jun. 5, 2020. [Online]. Available: https://www. [28] S. Mammar, S. Glaser, and M. Netto, ‘‘Vehicle lateral dynamics estimation
rokon.com/ using unknown input proportional-integral observers,’’ in Proc. Amer.
[2] Öhlins Website. Accessed: Jun. 5, 2020. [Online]. Available: https://www. Control Conf., Jun. 2006, p. 6.
ohlins.com/ [29] Y. Sebsadji, S. Glaser, S. Mammar, and M. Netto, ‘‘Vehicle roll and road
[3] Christini Technologies Website. Accessed: Jun. 5, 2020. [Online]. Avail- bank angles estimation,’’ Proc. 17th World Congr. Int. Fed. Autom. Control
able: https://www.christini.com/awd-technology/about-the-technology Seoul, South Korea, 2008, pp. 7091–7097.
[4] M. Blundell and D. Harty, The Multibody System Approach to Vehicle [30] A. Bonci, R. De Amicis, S. Longhi, G. A. Scala, and A. Andreucci,
Dynamics. Oxford, U.K.: Butterworth-Heinemann, 2004. ‘‘Motorcycle lateral and longitudinal dynamic modeling in presence of tyre
[5] D. J. N. Limebeer and R. S. Sharp, ‘‘Bicycles, motorcycles and models,’’ slip and rear traction,’’ in Proc. 21st Int. Conf. Methods Models Autom.
IEEE Control Syst. Mag., vol. 26, no. 5, pp. 34–61, Oct. 2006. Robot. (MMAR), Aug. 2016, pp. 391–396.
[6] R. S. Sharp, ‘‘The stability and control of motorcycles,’’ J. Mech. Eng. Sci., [31] S. M. Savaresi and M. Tanelli, Active Braking Control Systems Design for
vol. 13, no. 5, pp. 316–329, Oct. 1971. Vehicles. London, U.K.: Springer-Verlag, 2010.
[7] K. K. Cooper, ‘‘The effect of aerodynamics on the performance and [32] S. M. Savaresi, C. Poussot-Vassal, C. Spelta, O. Sename, and L. Dugard,
stability of high speed motorcycles,’’ Nat. Aereonautical Establish- Semi-active Suspension Control Design for Vehicles. Oxford. U.K.:
ment, National Research Council of Canada, Ottawa, ON, Canada, Elsevier, 2010.
1974. [33] V. Cossalter, Motorcycle Dynamics, Morrisville, NC, USA: Lulù.com,
[8] R. S. Sharp, ‘‘The stability of motorcycles in acceleration and decelera- 2006.
tion,’’ in Proc. Inst. Mech. Eng. Conf. Braking Road Vehicles, London, [34] A. Bonci, R. De Amicis, S. Longhi, E. Lorenzoni, and G. A. Scala,
U.K., 1976, pp. 45–50. ‘‘A motorcycle enhanced model for active safety devices in intelligent
[9] D. J. N. Limebeer, R. S. Sharp, and S. Evangelou, ‘‘The stability of transport systems,’’ in Proc. 12th IEEE/ASME Int. Conf. Mech. Embedded
motorcycles under acceleration and braking,’’ Proc. Inst. Mech. Eng., C, Syst. Appl. (MESA), Aug. 2016, pp. 1–6.
J. Mech. Eng. Sci., vol. 215, no. 9, pp. 1095–1109, Sep. 2001. [35] MSC Adams Tyre Software Manual. Accessed: Jun. 27, 2018.
[10] R. S. Sharp and D. J. N. Limebeer, ‘‘A motorcycle model for stability [Online]. Available: https://simcompanion.mscsoftware.com/infocenter/
and control analysis,’’ Multibody Syst. Dyn., vol. 6, no. 2, pp. 123–142, index?page=content&channel=DOCUMENTATION&cat=1VMO50
2001. [36] A. Bonci, R. De Amicis, S. Longhi, E. Lorenzoni, and G. A. Scala,
[11] R. S. Sharp, ‘‘The application of multi-body computer codes to road vehicle ‘‘Motorcycle’s lateral stability issues: Comparison of methods for dynamic
dynamics modelling problems,’’ Proc. Inst. Mech. Eng., D, J. Automobile modelling of roll angle,’’ in Proc. 20th Int. Conf. Syst. Theory, Control
Eng., vol. 208, no. 1, pp. 55–61, Jan. 1994. Comput. (ICSTCC), Oct. 2016, pp. 13–15.
[12] H. B. Pacejka and R. S. Sharp, ‘‘Shear force development by pneumatic
tyres in steady state conditions: A review of modelling aspects,’’ Vehicle
Syst. Dyn., vol. 20, nos. 3–4, pp. 121–176, 1991.
[13] E. J. H. de Vries and H. B. Pacejka, ‘‘Motorcycle tyre measurements and
models,’’ Vehicle Syst. Dyn., vol. 29, no. 1, pp. 280–298, Jan. 1998. ANDREA BONCI (Member, IEEE) holds the
[14] Y. Tezuka, H. Ishii, and S. Kiyota, ‘‘Application of the magic formula tire
position of an Assistant Professor in automatic
model to motorcycle maneuverability analysis,’’ JSAE Rev., vol. 22, no. 3,
control and the Head of the Automation Labo-
pp. 305–310, Jul. 2001.
ratory, Università Politecnica delle Marche. His
[15] H. B. Pacejka, Tire Vehicle Dynamics. Oxford, U.K.: Butterworth-
Heinemann, 2012. research interests include modeling and control
[16] V. Cossalter and R. Lot, ‘‘A motorcycle multi-body model for real time of dynamic systems, system and control theory,
simulations based on the natural coordinates approach,’’ Vehicle Syst. Dyn., vehicle dynamics, automotive systems, model-
vol. 37, no. 6, pp. 423–447, Jun. 2002. ing and control of autonomous systems, mecha-
[17] R. Lot and V. Cossalter, ‘‘A nonlinear rider model for motorcycles,’’ tronic, robotics, embedded systems applications,
in Proc. FISITA World Automot. Cong., Yokohama, Japan, Jan. 2006, industrial automation, cyber-physical systems, and
pp. 1–11. smart manufacturing.

VOLUME 8, 2020 112881

 A. Bonci et al.: Towards an AWD Motorcycle: Dynamic Modeling and Simulation

SAURO LONGHI (Senior Member, IEEE) holds GIUSEPPE ANTONIO SCALA received the mas-
the positions of a Full Professor in robot tech- ter’s degree in computer and automation engineer-
nologies and a Rector at the Università Politecnica ing from the Università Politecnica delle Marche,
delle Marche. His main research interests include in 2018, where he is currently pursuing the Ph.D.
modeling, identification and control of linear and degree in automation engineering with the Dipar-
nonlinear systems, control of mobile robots, ser- timento di Ingegneria dell’Informazione. His main
vice robots for assistive applications supporting research interests include robotics, nonlinear sys-
mobility and cognitive actions, home and build- tem analysis and control, virtual simulations, and
ing automation, and automatic fault detection and computer science.
isolation.

112882 VOLUME 8, 2020