Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Control of Motorcycles by Variable Geometry Rear Suspension

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

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

net/publication/221621212

Control of Motorcycles by Variable Geometry Rear Suspension.

Conference Paper · September 2010


DOI: 10.1109/CCA.2010.5611082 · Source: DBLP

CITATIONS READS

10 491

1 author:

Simos Evangelou
Imperial College London
63 PUBLICATIONS   886 CITATIONS   

SEE PROFILE

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

Green Adaptive Control for Future Interconnected Vehicles View project

Mechatronic Suspensions View project

All content following this page was uploaded by Simos Evangelou on 06 February 2015.

The user has requested enhancement of the downloaded file.


Control of motorcycles by variable geometry rear suspension
Simos A. Evangelou

Abstract— Cornering weave oscillations associated with high as the speed increases. The natural frequency rises from
performance motorcycles operating at high speeds, can be zero at very low speed to somewhere in the range 2 − 4
reduced by controlling the geometry of the rear suspension. Hz, depending on the mass and size of the machine; the
A conventional monoshock rear suspension arrangement is
modified and extended in a way that variations of the leverage lower frequencies corresponding to heavier motorcycles. The
ratio between the spring damper unit and road wheel are wobble mode involves mainly a caster shimmy type steering
possible by an actuator. The actuator varies the geometry oscillation. The frequency of this mode is normally in the
by controlling the displacement between two moving parts. range 6 − 9 Hz, it is relatively independent of speed, and
The design analysis makes use of a high fidelity mathematical it is governed primarily by the mechanical trail, the front
motorcycle model whose parameter set is based on a Suzuki
GSX-R1000 sports machine. A displacement control strategy is tyre cornering stiffness and the front frame steer inertia
developed based on classical Bode-Nyquist frequency response [7, 8]. In cornering, the above lateral modes and the in-
ideas. The controller further utilises an integrator anti-windup plane modes associated with tyre deflections and suspension
scheme to conform with the limited displacement space, and motions become coupled, as was first shown by Koenen [9].
to limit the maximum actuator force and power requirements. The motorcycle becomes prone to resonant forcing when
Simulation results are included to demonstrate that significant
improvements can be obtained with an actuator of practical regular road undulations produce displacement forcing that
dimensions. is tuned to lightly damped modal frequencies of the machine.
Moderate roll angles appear to represent the worst case
I. I NTRODUCTION conditions [10]. The participation of the rear suspension in
Conventional motorcycle rear monoshock suspension sys- weave oscillations implies the potential for an active variable
tems are designed to provide varying but predetermined geometry suspension to add damping to the weave mode.
leverage ratios between spring damper unit and road wheel, The broad purpose of this research is to investigate the
as the suspension travels to its limits. These characteristics practicality and performance of a variable geometry suspen-
are motivated and designed by static equilibrium considera- sion scheme in improving the cornering weave behaviour
tions. Further variations of the leverage ratio can be super- of modern high performance motorcycles. The suspension
posed by utilising a variable geometry arrangement. In this under consideration aims to incorporate four major charac-
suspension class the mechanical path along the suspension teristics. Firstly, due to packaging constraints, the operating
kinematic loop is opened at one location, by design, and it space of the variable geometry system should not be more
is subsequently closed by installing an appropriate actuator. than a few centimetres larger than that of the standard passive
The actuator will essentially act as a displacement con- system. Secondly, in order to reduce costs it should be
troller and will interact with the conventional passive force- possible to use much of the existing passive technology with
producing elements, spring and damper, to provide active easy retrofitting of any new parts. Thirdly, variation of the
control. The degree of their interaction will primarily be geometry should be done with low actuation force and power
determined by the geometric design, and it will prescribe the requirements so that actuators of reasonable size can be used.
actuator force and power requirements. Variable geometry Fourthly, fail safe operation should be easily achievable.
suspension ideas have already been explored in [1–5] for The paper is organised as follows. In Section II-A the
cars and in [6] for motorcycles. base motorcycle model is described, with the modelling of
In the case of two-wheeled vehicles two objectives have the variable geometry suspension described in Section II-B.
already been investigated: levelling under load changes and The displacement control design for the actuator is given in
manoeuvring, and high speed “cornering weave” stabilisa- Section III. Sections IV-A and IV-B present stability and sim-
tion [6]. The latter relates to the occurrence of dangerous ulation results and analyse the performance of the variable-
oscillations when motorcycles travel at high speeds. The geometry-suspension-controlled motorcycle. Conclusions are
main lateral oscillations in two-wheeled vehicles are “weave” drawn in Section V.
and “wobble”. The weave mode is a fishtailing motion that
II. M OTORCYCLE MODEL
involves mainly yawing, rolling and and to a lesser extent
steering and pitching of the vehicle. In straight running or A. Prior model
under moderate lean angle conditions, it is well damped The basis motorcycle model used in this work represents
at low and intermediate speeds, but becomes less damped the class of road-going sports machines, and has a parameter
set that is based on the Suzuki GSX-R1000. It has been
S. A. Evangelou is with Departments of Mechanical, and Electrical and
Electronic Engineering, Imperial College London, London SW7 2AZ, UK described in detail in [11, 12]. It is a multi-rigid body model
s.evangelou@imperial.ac.uk in which the constituent masses are introduced in a tree
structure. It involves a main frame that is allowed unrestricted generate C++ code that numerically integrates the nonlinear
motion with six degrees of freedom. The swinging arm, equations of motion, or it can be used to generate Matlab
upper body of the rider and front frame are attached to it code with a symbolic state-space representation that de-
with pitching, leaning and steering freedoms respectively. scribes small perturbations around a prescribed trim con-
Twisting of the front frame is allowed and it involves small dition. The machine’s trim state is controlled at constant
angular displacements at the steering head about an axis that forward speed and roll angle by speed-error to rear-wheel-
is perpendicular to the (inclined) steering axis. The model drive-moment and roll-angle-error to steering-torque PI and
also has spinning road wheels, standard rear monoshock PID feedback loops respectively. In the present work, the trim
suspension and telescopic front forks that allow linear in- states are found by nonlinear simulation and are used to set
line displacements. The mathematical model employed here up in numerical state-space form the open-loop linearised
also accommodates aerodynamic forces and moments which equations of motion. For the purposes of this study the rear
are proportional to the square of the speed. The road tires are suspension modelling had to be extended. These extensions
treated as ‘wide’, flexible in compression, and care is taken are described next.
to track dynamically their migration as the machine rolls,
pitches and steers. The tire forces and moments are generated B. Variable geometry rear suspension model
from the normal load, the camber angle and the combined It appears that a suitable location to implement design
slip using ‘magic formulae’, while the lateral compliance of modifications and install an actuator would be at point p21 .
the tires’ carcass is modelled using standard linear first-order This joint is replaced by a mechanism that is imagined to
relaxation length type tire models [13]. provide the kinematic relationship between spring/damper
The rear monoshock suspension system requires particular unit and swinging link shown in Fig. 2.
attention and it is presented here in sufficient detail. The
monoshock linkage system is attached to the main frame and
the swinging arm, and it is restrained by a parallel spring and Fb
damper as shown in Fig. 1.
X
p13
X Z
p13
Z spring/damper
p22
p13 to p21 is l;
unit
p7 p11 p19 to p20 is l1;
θ p20 to p22 is l2;
p19 p11 to p22 is l3;
p21
ζ δ p20 to p21 is l4;
p20 ϕ0 p13 to p21 is l0 ;
Fa
Fa p13 to p′21 is l
Fig. 1. Geometry of monoshock suspension of the GSX-R1000. Fs
p21
A geometric analysis of this suspension has been presented p′21
in [11, 14] in which the spring/damper unit length was Fb Fs
derived as a function of the swing arm displacement, θ, from da
nominal according to
p actuation
l = (x13 − x21 )2 + (z13 − z21 )2 . (1)
Fig. 2. Rear monoshock suspension variable geometry mechanism.
(x13 , z13 ) and (x21 , z21 ) are the coordinates of the upper
and lower attachment points p13 and p21 of the spring unit
respectively. The first point is attached on the main frame The lower side of the spring/damper unit, centred at point
and has fixed coordinates. The second point is attached to p′21 , slides on a new part which is fitted via bearings inside
the swinging link and its coordinates can be expressed in the joint p21 and can rotate relative to the swinging link.
terms of the swing arm angle [11]. By virtual work this pre- The sliding surfaces are assumed to be flat and to have
analysis yields an analytic relationship between the swing low friction. An actuator is imagined to be attached on the
arm angle change and the moment of the spring force about new part and to provide actuation force Fa between the two
the swing arm pivot, which is used directly in the model. sliding bodies. Initially points p21 and p′21 coincide but with
The overall model is written in LISP and makes use any actuation force they will displace relative to each other
of the multibody modelling code VehicleSim , R formerly by distance da . The spring/damper unit length is in this case
called AutoSim [15], also see1 , and can be obtained from not only a function of θ but also a function of the variable
the website2 . The VehicleSim model can be configured to geometry displacement and hence the following relationship
1 http://www.carsim.com
is derived q
2 http://www.imperial.ac.uk/controlandpower/motorcycles/ l = l0 2 − da 2 , (2)
where l0 is the distance from p13 to p21 and it is a function equilibrium of forces. The analysis is done in two steps.
of θ only (obtained from (1)). The leverage ratio between Firstly, we consider the equilibrium of the system between
spring/damper unit length and swing arm angle is obtained points p13 to p21 in Fig. 2 as one. The only forces that are
by differentiating (2), which yields of relevance in this case are the reaction forces Fb at the
∂l l0 ∂l0 two joints. The bearings at the joints are assumed frictionless
= . (3) and therefore each force Fb acts along the line joining points
∂θ l ∂θ
p13 and p21 . Secondly, we isolate the spring/damper unit and
It can be seen that it depends on l and therefore it will also consider its equilibrium alone. Forces Fb , Fa and Fs act on
depend on da . The change of l with respect to geometric it, where Fs is the spring/damper force. By considering the
displacements is given by force balance perpendicular to Fb we can show that
∂l da
=− . (4) da
∂da l Fa = Fs . (5)
l
Figs. 3 and 4 further illustrate the dependence of the
spring/damper unit length and leverage ratio to swing arm This expression shows that the actuation force will increase
angle and actuator displacement. Due to symmetry negative with actuator displacement and hence the magnitude of the
cases of da in Fig. 3 are not shown. The symmetry of displacement will have to be restricted appropriately to limit
the maximum actuation force. Fig. 5 illustrates how the actu-
ation force varies with swing arm and actuator displacement.
Leverage ratio dl/dθ Spring unit length (m)

0.31
The exact vertical point of application of the actuation force
0.305
da = 0, 10, 20, 30 mm
0.3 J
0.295
J
^ 500

400
0.29

Actuation force Fa (N)


0.05 0.06 0.07 0.08 0.09 0.1 0.11


300


200


0.32 100


0.315

0

0.31 da = 0, 10, 20, 30 mm −100 
−200


0.305
0.05 0.06 0.07 0.08 0.09 0.1 0.11
Swinging arm angle θ (rad) 
−300

−400
da = −30, −20, −10, 0, 10, 20, 30 mm
Fig. 3. Spring/damper unit length and leverage ratio variation with swinging −500
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
arm angle for 0, 10, 20 and 30 mm of variable geometry displacement. Swinging arm angle θ (rad)
The case of 0.0794 rad swinging arm angle and 15 mm variable geometry
displacement is shown with a + in each plot.
Fig. 5. Variable geometry static actuation force with swinging arm angle
for 7 cases of variable geometry position (-30 mm to 30 mm). The range
of swing arm angles from bump to rebound stop is shown. The case of
0.0794 rad swinging arm angle and 15 mm variable geometry displacement
Leverage ratio dl/dθ Spring unit length (m)

0.301
is shown with a +.

0.3005

0.3 Fa is a design choice and it does not affect its value. The
0.2995 longitudinal point of application of the spring/damper force
0.299 Fs can be found by considering the moment equilibrium of
−0.03 −0.02 −0.01 0 0.01 0.02 0.03
the spring/damper unit about point p13 say.
0.316
If a small change δθ in θ occurs, for which the corre-
0.3155 sponding change in l0 is δl0 , the moment M corresponding
0.315 to a spring/damper force Fs = f (l, l̇) is
0.3145
∂l0   l ∂l
0 0
  ∂l
0.314 M = Fb = f l, l̇ = f l, l̇ (6)
−0.03 −0.02 −0.01 0 0.01
Actuator displacement da (m)
0.02 0.03
∂θ l ∂θ ∂θ
by virtual work, where (3) was used to derive the right-
Fig. 4. Spring/damper unit length and leverage ratio variation with variable most expression. The properties of the spring/damper unit
geometry displacement for 0.0794 rad swinging arm angle. can thus be expressed in terms of an equivalent moment
M (θ, da , θ̇, d˙a ) about the swing arm pivot, as
the variable geometry mechanism about the da = 0 case 
∂l

∂l ˙ ∂l
is apparent from Fig. 4. M = f l, θ̇ + da , (7)
∂θ ∂da ∂θ
In order to estimate the magnitude of the actuator force
required to maintain constant displacement we consider the which is fully automated in the model.
It can easily be shown that when the moment expression frequency gain of the controller to approach zero so that the
in (7) is linearised about da = 0, it is identical to the moment actuator restores its default (offset) position when constant
expression corresponding to the conventional suspension in yaw rate cornering equilibrium conditions are reached. A
which da is not involved–for example, observe in the first simple second-order compensator scheme which achieves
plot of Fig. 4 that ∂l/∂da = 0 at da = 0. This means these objectives is da = K(s) sψ, where s is the Laplace
that small variations of the suspension geometry about the variable, sψ is the yaw rate and
actuator default position will bring insignificant benefits. It is kc s
therefore beneficial to operate the actuator about a nonzero K(s) = 2 . (8)
s + 2ζωs + ω 2
offset displacement at which the spring/damper unit length
will be sensitive to actuator displacement even for small The following parameter values were found to be suitable:
perturbations (∂l/∂da 6= 0). By considering space constraints kc = −5, ζ = 0.5 and ω = 10. The compensator design
a reasonable offset is 15 mm. This value is used in all methodology has been motivated by the approach used in
subsequent results. the design of mechanical steering compensators in [16–18].
The frequency response of the compensator is illustrated in
III. D ISPLACEMENT CONTROL DESIGN Fig. 7.
The actuator is considered to be a single-input single-
output displacement controller that reacts to the yaw rate of

Magnitude (dB)
−5
the machine. Yaw rate is chosen as the feedback control sig-
−10
nal because it is measurable in practice and because the yaw
−15
freedom participates systematically in weave oscillations.
−20
In order to design the feedback controller we make ref-
−25
erence to the Nyquist diagram of the open-loop linearised
−30
motorcycle model providing the signal transmission path 270

Phase (deg)
from actuator displacement to yaw rate, shown in Fig. 6. 225

This diagram should be interpreted according to a positive 180

feedback convention in which the important real axis point 135

is +1 rather than −1. The frequency response for a mild- 90


0 1 2
10 10 10
Frequency (rad/sec)

Fig. 7. Frequency response of variable geometry displacement controller,


K(s), in which yaw rate is the input and actuator displacement is the output.
1 +
Imaginary axis

0.5 The resulting controller is linear and will operate on the


assumption that displacements of any size are achievable by
0
the actuator. In practice this is not true and only a very
−0.5
limited displacement space is available. Based on simple
examination of the standard suspension, it is reasonable to
−1 assume that displacements of up to ± 25 mm away from the
centre position (point p21 in Fig. 2) are possible. In order
−1 −0.5 0 0.5
Real axis
1 1.5 2
to limit the response of the linear compensator in (8) inside
these boundaries, the controller is augmented with a standard
Fig. 6. Nyquist diagram for the open-loop motorcycle transfer function integrator anti-windup scheme, as illustrated in Fig. 8. The
from variable geometry actuator displacement, da , to yaw rate, ψ̇. The scheme utilises two smooth saturation functions, shown in
forward speed is 75 m/s and the roll angle is 15 deg. The frequency of Fig. 8 with larger blocks, so that upper and lower bounds on
27.5 rad/s, corresponding to weave, is shown with a +.
the values of the actuator displacement and actuator velocity
(va ) are not exceeded. The displacement saturation function
cornering, high-forward-speed condition is considered; at this is of the form shown in (9), with the one corresponding to
operating condition weave is likely to respond vigorously the velocity having a similar form. When the unconstrained
to road forcing [10]. The weave mode frequency is 27.5 displacement output, d˜a , is between damin and damax the
rad/s and the wobble mode frequency is 47.9 rad/s. The displacement output of the scheme, da , has the same value as
Nyquist diagram in Fig. 6 shows that, in order to improve d˜a , otherwise the saturation function returns damin or damax
the stability of the weave mode, one needs to phase com- depending on which limit has been reached. sc is a shaping
pensate the weave mode frequency (shown with a +) by constant that determines the sharpness of the transition
approximately +130◦. It is also beneficial to reduce the gain between non-saturation and saturation. It was established
of the frequency response at the wobble mode frequency by simulation trials that sc = 5000, kd = 10000 and
to reduce the controller bandwidth and economise on the kv = 3000 are suitable. The actuator velocity is constrained
actuator power. A further requirement is for the very low in order to reduce the peak forces and power required by the
kd
+ −
kv
2ζω

ψ̇ − + − va + − d˜a da
kc d R R
dt + −
+
ω 2 da 0

Fig. 8. Displacement controller with second order dynamics and integrator anti-windup. The yaw rate, ψ̇, is the input and the saturated variable geometry
displacement, da , is the output. The parameter values are kc = −5, ζ = 0.5, ω = 10, kd = 10000, kv = 3000 and da 0 = 0.015 m.

   
1 1 1 1
da = − arctan(sc (d˜a − damax )) + arctan(sc (d˜a − damin )) d˜a + arctan(sc (d˜a − damax )) + damax
π π π 2
 
1 1
− arctan(sc (d˜a − damin )) − damin (9)
π 2

actuator, and again, it was found by trial and error that it is 60

beneficial to restrict its value within ± 1.15 m/s without any


performance loss. 50

Imaginary axis
IV. R ESULTS 40

The main results of this paper are now presented and 30


comprise a) root-locus diagrams, derived by linearising the
high fidelity model presented in Sections II-A and II-B 20

and calculating its eigenvalues, and b) simulations with the


10
nonlinear model.
A. Root-locus results 0
−16 −14 −12 −10 −8 −6 −4 −2 0 2
Real axis
The first result, shown in Fig. 9, is a root-locus diagram
for the machine operating at 15 deg roll angle, with the speed Fig. 9. Root loci for 15 deg lean angle with forward speed the varied
increased from 7 m/s to 75 m/s. The nominal and controlled parameter. The speed is increased from 7 m/s () to 75 m/s (⋆). The ×
machine cases are considered. The nominal machine is fitted locus represents the nominal (uncontrolled) machine and ◦ represents the
suspension-controlled machine.
with the conventional rear suspension, while the controlled
machine is fitted with the variable geometry rear suspension
and the control is provided by the linear compensation
scheme (without saturation) given in (8). In the case of the geometry suspension on the small-perturbation behaviour of
nominal arrangement, Fig. 9 shows that the wobble-mode the vehicle at higher roll angles is studied in the root-locus
frequency varies between approximately 48 rad/s and 56 diagrams in Figs. 10 and 11. It is clear from Fig. 10 that
rad/s, while the weave mode frequency varies between 9 rad/s when the machine is operating at 30 deg roll angle the
and 27.5 rad/s. At moderate and high speeds the damping of damping of the weave mode is improved for all speeds, while
both the wobble and weave modes decreases with increased the wobble mode retains its prior behaviour, similarly to the
speed. These characteristics are familiar and have been 15 deg roll angle case. When the machine is operating at 45
reported in previous work [11, 12, 16]. Similar trends exist in deg roll angle, it can be seen from Fig. 11 that the weave-
the case of the controlled machine with the key observation mode damping at high speeds is increased, while the wobble
in Fig. 9 the increase in weave-mode damping. Although the mode is slightly destabilised at low speeds and remains
control design has been based on a single high-speed mild- relatively unaffected at all other speeds.
cornering linearised model, it is evident that, in comparison
with the nominal machine, improvements in the damping of B. Simulation results
the weave mode, for all speeds, have been achieved. The The large perturbation performance of the variable-
wobble-mode damping remains essentially unaffected, as geometry-controlled motorcycle is studied by simulation. In
would be expected by the relatively low compensator gain at this case the machine is controlled by the displacement
the wobble-mode frequencies. The influence of the variable compensation scheme with integrator anti-windup shown
15.3 2 30
uncontrolled uncontrolled uncontrolled
controlled 1.5 controlled 25 controlled

15.2
1 20
Roll angle (◦ )

Steer angle (◦ )

Yaw rate (◦ /s)


0.5 15
15.1
0 10

15 −0.5 5

−1 0
14.9
−1.5 −5

−2 −10
14.8
−2.5 −15

14.7 −3 −20
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (s) Time (s) Time (s)

Fig. 12. Transient behaviour of the roll and steer angles and the yaw rate for the uncontrolled (dashed line) and controlled (solid line) machine, in
response to sinusoidal road forcing that begins at t = 0 and ends at t = 0.623 s, and has a peak amplitude of 5 mm. The forcing frequency is tuned to
the weave mode. The forward speed is 75 m/s and the lean angle 15 deg.

in Fig. 8. All other controllers are removed. It is already


known that the resonant modes of motorcycles can become
60
vulnerable to displacement forcing from the road that is
only a few millimetres in amplitude [10]. Fig. 12 shows
50
the build up of oscillations in the roll and steer angles and
Imaginary axis

40 yaw rate in response to sinusoidal road forcing that is tuned


into the weave mode (27.7 rad/s). The forward speed is 75
30 m/s and the forcing lasts for 0.623 s (or 2 43 cycles) and
its amplitude is 5 mm. It is clear that 27.7 rad/s oscillations
20
build up in about 0.7 sec and decay afterwards, once the road
10
disturbance is removed. Approximately for the first 2 s there
is evidence of waveform distortion, a product of nonlinear
0
−16 −14 −12 −10 −8 −6 −4 −2 0 2
effects. Nevertheless, once the responses begin to decay,
Real axis one can estimate for the controlled machine a reduction in
the weave-mode half-amplitude decay time by about 36%
Fig. 10. Root loci for 30 deg lean angle with forward speed the varied from 0.74 s to 0.47 s. Fig. 13 shows the control action
parameter. The speed is increased from 7 m/s () to 75 m/s (⋆). The ×
locus represents the nominal (uncontrolled) machine and ◦ represents the provided by the actuator to the onset of oscillations, which
suspension-controlled machine. lasts approximately for 4 s. It is evident that the actuator

0.03
da
va
0.02

60 0.01

50 0
Imaginary axis

40 −0.01

30 −0.02

20 −0.03
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (s)
10

Fig. 13. Transient behaviour of the actuator displacement (solid line, m)


0 and velocity (dashed line, ×40 m/s), in response to sinusoidal road forcing
−16 −14 −12 −10 −8 −6 −4 −2 0 2
Real axis that begins at t = 0 and ends at t = 0.623 s, and has a peak amplitude
of 5 mm. The forcing frequency is tuned to the weave mode. The forward
speed of the machine is 75 m/s, the lean angle 15 deg, and the actuator
Fig. 11. Root loci for 45 deg lean angle with forward speed the varied nominal operating point is shown with a + in Figs. 3 and 5.
parameter. The speed is increased from 7 m/s () to 75 m/s (⋆). The ×
locus represents the nominal (uncontrolled) machine and ◦ represents the
suspension-controlled machine. displacement builds up in 0.2 s and subsequently, until t = 2
s, it is operating at “full capacity”–its response is “clipped”
by the anti-windup action when it exceeds ±25 mm. It is
also evident that the actuator velocity is similarly confined R EFERENCES
between its chosen limits, ±1.15 m/s. The actuator force and [1] P. J. T. Venhovens, A. C. M. Knaap van der, and H. B. Pacejka,
power are monitored in the simulation and are found not to “Semi-active attitude and vibration control,” Vehicle System Dynamics,
exceed 280 N and 265 W respectively. vol. 22, no. 5-6, pp. 359–381, 1993.
[2] N. J. Leighton and J. Pullen, “A novel active suspension system for
automotive application,” Proceedings of the Institution of Mechanical
V. C ONCLUSIONS Engineers Part D-Journal of Automobile Engineering, vol. 208, no. 4,
pp. 243–250, 1994.
This paper has introduced a variable geometry rear [3] D. Karnopp and S. G. So, “Energy flow in active attitude control sus-
pensions: A bond graph analysis,” Vehicle System Dynamics, vol. 29,
monoshock suspension system that can be used to control no. 2, pp. 69–81, 1998.
high performance motorcycles. The study includes the use [4] Y. Watanabe and R. S. Sharp, “Mechanical and control design of a
and extension of a high fidelity motorcycle simulation model variable geometry active suspension system,” Vehicle System Dynam-
ics, vol. 32, no. 2-3, pp. 217–235, 1999.
[11, 12] to demonstrate that performance benefits in cor- [5] W. J. Evers, A. Knaap van der, I. Besselink, and H. Nijmeijer,
nering weave mode damping can be obtained. A particular “Modeling, analysis and control of a variable geometry actuator,”
geometry has been chosen in which an actuator is assumed in 2008 IEEE Intelligent Vehicles Symposium (IV). Eindhoven,
Netherlands: IEEE, 2008, pp. 251–256.
to interact actively with the existing setup by varying the [6] R. S. Sharp, “Variable geometry active rear suspension for motorcy-
displacement between two rubbing surfaces. A modest offset cles,” in Proceedings of AVEC 2000, Ann Arbor, MI, August 2000,
in the actuator displacement has been introduced to increase pp. 389–395.
[7] S. Evangelou and D. J. N. Limebeer, “Lisp program-
the overall performance of the scheme. The actuator dis- ming of the “sharp 1971” motorcycle model,” 2000,
placement demand derives from machine yaw rate and a http://www.imperial.ac.uk/controlandpower/motorcycles.
compensator that links the two has been designed based [8] S. Evangelou, “The control and stability analysis of two-wheeled road
vehicles,” Ph.D. dissertation, Imperial College London, January 2004.
on classical frequency response methodologies. The control [9] C. Koenen, “The dynamic behaviour of motorcycles when running
system further utilises standard integrator anti-windup action straight ahead and when cornering,” Ph.D. dissertation, Delft Univer-
to confine the actuator displacement inside tight space limits sity of Technology, 1983.
[10] D. J. N. Limebeer, R. S. Sharp, and S. Evangelou, “Motorcycle steering
and, by restricting the maximum actuator speed and thus oscillations due to road profiling,” Journal of Applied Mechanics-
reducing the influence of damping forces and inertial effects, Transactions of the ASME, vol. 69, no. 6, pp. 724–739, 2002.
to abate the peak actuator force and power. [11] R. S. Sharp, S. Evangelou, and D. J. N. Limebeer, “Advances in
the modelling of motorcycle dynamics,” Multibody System Dynamics,
The results in Section IV-A show that the active sus- vol. 12, no. 3, pp. 251–283, 2004.
pension control improves the dynamic properties of the [12] ——, “Improved modelling of motorcycle dynamics (article no.
weave mode for a range of working regimes, including MB2003-029),” in ECCOMAS Thematic Conference on Advances in
Computational Multibody Dynamics, J. A. C. Ambrósio, Ed. Lisbon,
high-speed, low-lean operation in which weave is likely to Portugal: IDMEC, 1–4 July 2003, pp. 1–26.
respond vigorously to road forcing. Benchmark road forcing [13] H. B. Pacejka, Tyre and Vehicle Dynamics. Oxford: Butterworth
nonlinear simulation studies in Section IV-B have revealed Heinemann, 2002, ISBN 0-7506-5141-5.
[14] R. S. Sharp, S. Evangelou, and D. J. N. Limebeer, Advances in
the performance benefits in half-amplitude decay time likely Computational Multibody Systems. Dordrecht, The Netherlands:
to be obtained as compared to the nominal machine, and Springer-Verlag, 2005, vol. 2, ch. Multibody aspects of motorcycle
have quantified the requirements in actuator force and power. modelling with special reference to AutoSim, pp. 45–68, ISBN: 1-
4020-3392-3.
These requirements appear to be attainable by a standard [15] Anon., Autosim 2.5+ Reference Manual, Mechanical Simula-
torque motor of reasonable dimensions and gearing. With tion Corporation, 709 West Huron, Ann Arbor MI, 1998,
the power required often becoming negative there is benefit http://www.carsim.com.
[16] S. Evangelou, D. J. N. Limebeer, R. S. Sharp, and M. C. Smith, “Me-
in operating the actuator reversibly to recapture much of the chanical steering compensators for high-performance motorcycles,”
exported energy. Transactions of ASME, Journal of Applied Mechanics, vol. 74, no. 2,
One example of rear monoshock variable geometry sus- pp. 332–346, 2007.
[17] ——, “Steering compensation for high-performance motorcycles,” in
pension has been presented. Further optimisation of the Proceedings of 43rd Conference on Decision and Control; CDC 43.
mechanical design is possible, while optimisation of the Atlantis, Paradise Island, Bahamas: IEEE, 14-17 December 2004, pp.
control design that takes into account the entire machine 749–754, ISBN: 0-7803-8682-5.
[18] ——, “Control of motorcycle steering instabilities - passive me-
operating envelope is likely to yield improvements. Adaptive chanical compensators incorporating inerters,” IEEE Control Systems
control systems are also a practical possibility. The issue Magazine, vol. 26, pp. 78–88, 2006, ISSN: 1066-033X.
of practical implementation and integrated fabrication of the
variable geometry suspension is challenging, especially with
respect to the contact mechanics of sliding surfaces, and it
is the subject of ongoing investigation.

View publication stats

You might also like