DOI 10.1007/s11012-010-9304-1
Received: 29 May 2009 / Accepted: 7 April 2010 / Published online: 8 June 2010
© Springer Science+Business Media B.V. 2010
Abstract Many scientific papers deal with motorcy- presence of three typical out-of-plane modes of vibra-
cle stability (weave and wobble modes) but very sel- tion: weave, wobble, and capsize. In particular, weave
dom they take into account the passive response of mode basically consists of yaw, roll and lateral oscil-
rider’s body. This paper aims at studying the interac- lations of the rear assembly, wobble mode is domi-
tion of the rider’s arms and torso with the handlebar nated by the oscillation of the front assembly around
and the frame. First the rider’s steering impedance is the steering axis, and capsize involves a roll motion
identified from experimental tests on a motorcycle rid- combined with a lateral displacement. A detailed re-
ing simulator, then this information is used on a motor- view of the existing papers is available in references
cycle model and the effect on straight-motion stability [2] and [3], however many authors consider neither the
is investigated by eigenvalues calculation. frame compliance nor rider mobility. A detailed exper-
imental study on rider’s properties was made in [4]: the
Keywords Motorcycle · Dynamics · Stability · authors measured the natural frequencies and damping
Rider · Identification ratios of the body lateral motion and upper body lean-
ing for several riders. The effect of the rider’s inter-
action with the handlebar very rarely is dealt with in
literature, probably owing to the lack of experimental
1 Introduction
data. In 2004 the coupled handlebar and rider’s torso
dynamics were taken into account by means of two
The modal and stability analysis of motorcycles gives
spring-damper systems connecting the upper rider’s
important information about riding safety, handling
torso with the handlebar and the rear frame [5]. The
capabilities and riding comfort. This is the main rea-
parameters were derived from a recent study on neuro-
son of the large research effort made over the last
muscular dynamics carried out on a car simulator [6].
40 years to investigate the free-modes of a motorcy- In the current work the impedance the rider exerts on
cle. The first systematic work on two-wheeled vehicle the handlebar is identified from the rider’s modal prop-
stability dates back from 1971 [1]: it underlines the erties measured on a motorcycle simulator [7, 8]. The
results are then used on a simple motorcycle model
and the effect on stability is discussed.
V. Cossalter · A. Doria · R. Lot · M. Massaro ()
The paper is organized as follows. The mathemat-
DIMEG—Department of Innovation in Mechanics and
Management, University of Padova, Via Venezia 1, ical model of the passive rider is briefly described in
35131 Padova, Italy Sect. 1; the identification method used for computing
e-mail: matteo.massaro@unipd.it biomechanical characteristics and to derive mean rider
280 Meccanica (2011) 46: 279–292
2 Mathematical model
Table 1 Riders properties range, the dispersion of data is large at high frequency
Rider Height [cm] Mass [kg] Torso inclination [◦ ] and around the resonance peak. For the phase the dis-
persion of data is large only at low frequency, below
1 1.76 72 9 1.5 Hz.
2 1.85 74 7 Even if the riders who participated the research pro-
3 1.76 76 11 gramme have different characteristics (see Table 1),
4 1.86 85 11 the calculated FRFs show some common features.
5 1.69 56 13 The modulus of the FRF always shows a peak at
6 1.83 72 7 about 1.5÷2.5 Hz, a minimum at about 5 Hz and in-
7 1.70 61 15 creasing values at high frequency. See for example
8 1.78 66 6
Fig. 3, which refers to rider 5. The phase at low fre-
quency is about 130÷180◦ , then reaches a minimum
9 1.88 86 6
at about 3 Hz and, finally, increases at high frequency
10 1.72 73 8
range (see Fig. 3). It is worth highlighting that in the
high frequency range both the graph of modulus and
but it is an inverse function (ratio between an effort the graph of phase differ from the graphs of the iner-
variable and a flow variable) and inverse functions, tance of a single degree of freedom system.
when used in the analysis and identification of systems
3.2 Identification
with many DOF may lead to confusion and errors [12].
A receptance type FRF (ratio between the handle-
The identification of the biomechanical properties of
bar angle and the steering torque) shows large values
the rider’s body was carried out by means of a two
in the very low frequency band, before the first res-
degrees of freedom model [11], see Fig. 4. Two rigid
onance. Since in this band there are the largest mea-
bodies rotate about two axes (fixed to the motorcycle
surement errors, the use of a receptance type FRF for mock-up) that represent the steer axis (rotation θs ) and
identification may lead to errors. the axis of rotation allowed by the waist (rotation θt ).
On the contrary, an inertance type FRF (ratio be- Lumped springs and dampers connect the rigid bodies
tween the handlebar acceleration and the steering and the motorcycle mock-up.
torque), which is obtained multiplying the measured The rigid body that rotates about the steer axis (mo-
amplitude by angular frequency squared, gives more ment of inertia Is ) represents the handlebar with possi-
weight to the high frequencies amplitudes that are af- bly a contribution of the inertia of the hands and fore-
fected by small errors, this manipulation helps best arms. The rigid body that rotates about the waist axis
fitting of data. (moment of inertia It ) represents the torso of the rider
The spectrum of angular acceleration was calcu- with a share of the inertia of the arms.
lated multiplying the spectrum of handlebar rotation The springs and dampers between the two rigid
by angular frequency squared. Finally to minimize the bodies (ka and ca ) represent the equivalent stiffness
effect of noise, the inertance was calculated as the ra- and damping of the arms.
tio between the cross spectrum of angular acceleration The spring (kt ) and the damper (ct ) about the waist
of handlebar and steering torque and auto spectrum of axis represent the actions of the hips and of saddle
the steering torque [12]. on the rider’s torso. If during the tests the rider holds
In order to assess the quality of the experimental firmly the tank of the mock-up by means of the legs
results some repeatability tests were carried out. The these terms represent mainly the action of the rest of
riders performed many tests in the same conditions the body on the torso, otherwise these terms take into
and for each value of frequency the mean value μ and account the contact stiffness and damping between the
the standard deviation σ of the modulus and phase rider and the saddle.
of the FRF were calculated. Figure 3 shows the re- The external action of the servomotor on the system
sults obtained with rider 5, who carried out eight tests. is represented by rotation θs imposed to the handlebar
For the modulus the band between μ + σ and μ − σ and by torque Ms . Since the torsiometer is very stiff,
(which has the probability of containing 68.3% of re- there is a negligible difference between handlebar ro-
sults) is sufficiently narrow in the whole frequency tation and servomotor rotation.
Meccanica (2011) 46: 279–292 283
Fig. 3 Modulus and phase of the FRFs of rider 5: mean value and standard deviation
θ̈s (ω)
= − (−ω2 It + iω(ca + ct ) + ka + kt )ω2
Fig. 4 2 DOF identification model Ms (ω)
× −(iωca + ka )2 + (−ω2 Is + iωca + ka )
−1
The correspondences between the identification × (−ω2 It + iω(ca + ct ) + ka + kt ) (6)
model and the rider model integrated with the motor-
cycle model are the following: the fixed frame of the The identification consists in determining the values
identification model coincides with the rear frame of of the inertia of the two bodies (Is and It ), the stiff-
the motorcycle; the rigid body with moment of inertia ness of the arms (ka ) and the stiffness about the waist
Is is replaced by the actual front frame of the motor- axis (kt ), the damping coefficients of the arms (ca ) and
cycle, but the share of the inertia of the arms (included about the waist axis (ct ).
in Is ) is added to the front frame inertia about the steer The identification is carried out in the Nyquist
axis. plane by minimizing the distance between the corre-
284 Meccanica (2011) 46: 279–292
Fig. 5 Comparison between experimental FRF and best fitting FRF for rider 2
1 0.1277 0.7626 1050.8 85.2 16.02 5.94 6.14 37.3 0.0697 7.34
2 0.1257 0.7587 946.9 91.7 21.75 7.30 5.89 51.8 0.0627 3.63
3 0.1250 0.8236 1436.5 45.7 40.69 2.59 6.75 61.9 0.0956 26.66
4 0.1253 0.7059 1851.8 166.2 22.04 9.25 8.51 41.4 0.1003 9.50
5 0.1266 0.5085 1358.2 81.0 16.35 4.98 8.47 39.4 0.1629 3.20
6 0.1250 0.9681 1444.7 126.8 47.22 7.07 6.41 69.6 0.0789 3.90
7 0.1256 0.6719 604.9 95.1 7.49 2.15 5.14 22.2 0.4028 9.28
8 0.1250 0.5080 1293.3 64.6 14.16 5.12 8.23 36.7 0.2843 9.79
9 0.1250 1.0167 1907.9 117.6 48.59 6.61 7.10 60.8 0.0471 14.57
10 0.2681 0.5644 502.4 81.0 29.27 2.36 5.12 87.2 0.1878 17.63
σ/μ 0.3220 0.2417 0.3790 0.3568 0.5505 0.4457 – – – –
Mean-rider 0.1250 0.6441 1053.4 75.8 19.28 4.79 6.66 44.6 0.1092 6.90
sponding points of the experimental and the theoreti- in which Mt (ω), Pt (ω) are the modulus and phase of
cal curves as shown in Fig. 5. the theoretical FRF and Me (ω), Pe (ω) of the experi-
The distance between the two curves at frequency
mental FRF.
ω is:
The algorithm is based on the lsqcurvefit function
d(ω) = (Mt (ω) cos(Pt (ω)) − Me (ω) cos(Pe (ω)))2
of MATLAB and the total distance between the theo-
+ (Mt (ω) sin(Pt (ω)) retical and experimental points from 0.5 Hz to 12 Hz
− Me (ω) sin(Pe (ω)))2 (7) is minimized.
Meccanica (2011) 46: 279–292 285
Table 2 shows the identified bio-mechanical para- This result is not surprising because there are large dif-
meters of some riders. ferences in physique, riding experience and behavior
The values of Is (with the exception of rider 10) are among the riders.
very close to 0.125 kg m2 , which is the experimental Since the main purpose of this research is to carry
value of the moment of inertia of the handlebar stand- out a general analysis of the influence of the passive
ing alone. This result indicates that the presence of the response of the rider’s body on motorcycle stability,
rider does not significantly increase the inertia about the definition of the biomechanical properties of an
the steer axis, but mainly increases the stiffness and mean rider is useful. In fact the study of the influence
damping of the steer. of the biomechanical characteristics of the mean rider
For all riders the inertia of the torso (It ) is larger on motorcycle stability gives more general results than
than the inertia of the handlebar. the study of the effect of the identified biomechanical
Arm stiffness ka is about one order of magnitude properties of a single test rider.
larger than waist stiffness kt . Similarly, damping coef- The characteristics of the mean rider were identi-
ficients ca is from 3 to 8 times ct . fied by calculating the mean FRF (modulus and phase)
It is worth highlighting that, if resulting stiffness of of the measured FRFs and carrying out the identifi-
springs ka and kt in series and the resulting damping cation using the mean FRF. The last row of Table 2
of dampers ca and ct in series are calculated, values
shows the biomechanical characteristics of the mean
similar to the ones of the single stiffness and damping
rider and the residuals of the identification.
identified in [6] (tense rider) are found.
The natural frequency νt refers to the torso vibra-
tion when the steer is locked and is calculated by
means of equation: 4 Rider’s impedance vs. steering dampers
1 (ka + kt ) The rider’s body generates torques on the handlebar
νt = (8)
2π It and on the rear frame that are functions of the steer
motion, like a steering damper does. There is an im-
It ranges from 5 to 9 Hz. portant difference: the steer damper generates equal
The corresponding damping ratio is: torques on the steer and on the rear frame, whereas the
(ca + ct ) rider’s body generates different torques on the steer
ζt = 100 √ (9) and on the rear frame owing to the presence of iner-
2 (ka + kt )It
tia It .
and it assumes values from 22 to 90%. The two systems may be compared by analyzing
The last columns deal with the quality of the identi- their transfer functions. In the frequency domain the
fication. The residual radius is the mean difference be- torque generated on the steer by rider’s body is:
tween the radii of the corresponding points of the mea-
sured and the best fitting FRF curves in the Nyquist Ma (ω) = −(iωca + ka )(θs (ω) − θt (ω)) (10)
plot. The values of the radius range from 0 to 4
((rad/s2 )/Nm) and the residual radius is always lower The FRF between this torque and the steer rotation (in-
than 0.41 ((rad/s2 )/Nm). The phase residual is the put) is derived by introducing in (10) the transfer func-
mean difference between the phases of the correspond- tion between torso rotation (output) and steer rotation
ing points of the measured and the best fitting FRF (input), which is easily derived from (4):
curves in the Nyquist plot. The values of the phase
range from 180 to 40◦ and the phase residual is lower Ma (ω) (iωca + ka )(−ω2 It + iωct + kt )
than 27◦ . These values highlight that the proposed =− (11)
θs (ω) (−ω2 It + iω(ca + ct ) + ka + kt )
model is well suited to fit the experimental results.
The analysis of the identification results, which are The torque generated on the rear frame by rider’s body
shown in Table 2, highlights that coefficient of varia- is
tion σ/μ (ratio of the standard deviation to the mean)
of some biomechanical characteristics is rather large. Mt (ω) = (kt + iωct )θt (ω) (12)
286 Meccanica (2011) 46: 279–292
Hence, the transfer function between the torque on the frames of the vehicle. The common amplitude of the
rear frame and the steer rotation (input) is: two body’s FRFs is larger than the amplitude of the
steering damper FRF with the largest cd . Near and af-
Mt (ω) (iωca + ka )(iωct + kt )
=+ (13) ter the resonance the dynamics of torso becomes im-
θs (ω) (−ω2 It + iω(ca + ct ) + ka + kt ) portant and the FRFs of rider’s generated torques be-
From the mathematical point of view the poles of (11) come more and more different in amplitude and phase.
and (13) are identical and depend on torso dynam- The FRF of the torque generated on the steer by the
ics. In (11) there is a zero related to arm stiffness and rider remains larger than the steering damper FRF
damping and a couple of zeros related to torso dynam- (cd = 7.5) and assumes the same slope. The FRF of
ics. In (13) the couple of zeros related to torso dynam- the torque generated on the rear frame by the rider
ics is replaced by a single zero that depends only on does not increase with frequency after the resonance
torso stiffness and damping. and, therefore, at high frequency it is smaller than the
As the steering damper is concerned, assuming (as steering damper FRF with cd = 0.5.
usual) a linear relationship between the torque and the
steer rate, the transfer function is simply a derivator,
whose slope is related to the damping coefficient cd 5 Numerical analysis
Md (ω)
= iωcd (14) In order to assess the effect of the rider’s passive
θs (ω)
impedance on the vehicle stability, numerical modal
Figure 6 makes a comparison between the transfer analysis has been carried out in straight-running mo-
functions of rider’s generated torques and the steering tion for speeds from 1 to 60 m/s. Figure 7 shows the
damper torque (three values of cd are considered: 0.5, comparison between the eigenvalues of the reference
4.0 and 7.5 Nm s/rad). model, which does not consider the rider’s impedance,
Figure 6 clearly shows that before the resonance and the presented model, for the mean rider defined in
the amplitudes of the two FRFs of rider’s generated Table 2.
torques are almost the same and the phases are 180◦ It should be highlighted that the stability results
apart. Actually, in this range of frequencies the rider’s presented herein are valid under the hypothesis that the
body behaves like a spring-damper system that gener- rider is completely passive, which is a good assump-
ates equal and opposite torques on the front and rear tion for studying stable riding and incipient instability.
Meccanica (2011) 46: 279–292 287
Fig. 10 Wobble for Mean rider (v = 1 m/s, f = 9.06 Hz) Fig. 11 Wobble for Mean rider (v = 60 m/s, f = 10.75 Hz)
Fig. 14 Root locus of reference model (gray circles), mean rider model (black circles) and reference model with increased steering
damping coefficient (black squares)
Fig. 15 Root locus of reference model (gray circles) and mean rider model with decreased arm damping (black circles)
If these results are compared with those of [5], pose of highlighting the effect of steer impedance;
which is the work that pioneered the effect of rider’s further developments of the model with the inclu-
steering impedance on stability, some differences ap- sion of body passive response to roll and pitch are
pear. In particular in [5] the wobble stabilization is less planned.
significant than the one obtained in this work and the However the authors think that the differences on
effect on weave is almost negligible: on the contrary wobble and weave stability depend mainly on the pa-
in this work the weave destabilization due to rider’s rameters chosen for the tuning of the passive rider
impedance is clearly visible. model. In particular in [5] the parameters for the rider
In both papers the passive rider models for yaw dy- yaw dynamics come from car literature [6], on the
namics are similar (i.e. the rider torso is connected contrary in this work the data for the rider yaw dy-
both to the rear frame and to the handlebar by means namics have been identified on a motorcycle mock-up,
of spring-damper elements), but [5] includes also a thus involving different muscles and a different pos-
model for the rider passive roll dynamics and simu- ture.
lates a different vehicle. In this paper only yaw dy- Before moving to the sensitivity analysis on rider’s
namics has been considered with the specific pur- properties, it is interesting to compare the effect of
290 Meccanica (2011) 46: 279–292
Fig. 16 Root locus of reference model (gray circles) and mean rider model with increased arm damping (black circles)
Fig. 17 Root locus of reference model (gray circles) and mean rider model with decreased waist damping (black circles)
steering damper on stability as compared to the effect pared with that of rider’s passive model, as expected
of the rider’s passive model. Indeed one may won- according to the consideration of Sect. 3. In partic-
der if it is the same to model the rider’s passive im- ular, when using an increased steering damper wob-
pedance with an additional steering damper instead ble mode increases its stability similarly to what hap-
of using the passive rider model presented in this pa- pens with the rider passive model (see rider-wobble in
per. In Fig. 14 the root-locus of the mean rider passive Fig. 14) but it does not reduce its frequency as it hap-
model (black circles) is depicted against the reference pens with the rider passive model (see rider-wobble).
model (gray circles, no passive rider model) and the As the weave is concerned, this vibration mode is
reference model with the steering damper coefficient slightly more stable than the one obtained with the
increased by 7.5 Nm s/rad, i.e. equal to 14.2 Nm s/rad rider passive model.
(black squares). Please note that in the 2–7 Hz range A sensitivity analysis has been carried out to an-
the magnitude of the transfer function of the steering alyze the influence of the biomechanical parameters
damper is close to that of the passive rider, see Fig. 6. that show the largest coefficients of variation. The fol-
When inspecting Fig. 14 it is clear that the effect of lowing figures show the differences between the eigen-
steering damper on stability is different when com- values obtained with the biomechanical parameters of
Meccanica (2011) 46: 279–292 291
Fig. 18 Root locus of reference model (gray circles) and mean rider model with increased waist damping (black circles)
the mean rider and the eigenvalues obtained with mod- of rider’s body causes a significant stabilization of the
ified biomechanical parameters. wobble mode, but may destabilize the weave mode at
As far as arm damping ca is concerned Figs. 15 high speed. This effect is similar to the one caused by
and 16 show a large influence on the wobble mode, the steering damper.
a minor influence on the rider-wobble and a negligible Future work will deal with the influence of posture
influence on the weave. The increase in arm damping on the dynamic behavior of the motorcycle-rider sys-
increases wobble stability, as expected. tem.
Figures 17 and 18 show that waist damping ct has
a small effect on stability, actually its value is much
smaller than the one of ca , see Table 2. References
10. Lot R, Da Lio M (2004) A symbolic approach for auto- tional conference on noise and vibration engineerig, Lo-
matic generation of the equations of motion of multibody vanio, September 18–20
systems. Multibody Syst Dyn 12:147–172 12. Ewins DJ (2000) Modal testing theory, practice and appli-
11. Cossalter V, Doria A, Fabris D, Maso M (2006) Measure- cation, 2nd edn. Research Studies Press Ltd., Baldock
ment and identification of the vibration characteristics of
motorcycle riders. In: Proc. of the ISMA 2006 interna-