State of The Art Survey Active and Semi Active Suspension Control

Vehicle System Dynamics
International Journal of Vehicle Mechanics and Mobility
ISSN: 0042-3114 (Print) 1744-5159 (Online) Journal homepage: https://www.tandfonline.com/loi/nvsd20
State of the art survey: active and semi-active

suspension control
H. Eric Tseng & Davor Hrovat
To cite this article: H. Eric Tseng & Davor Hrovat (2015) State of the art survey: active
and semi-active suspension control, Vehicle System Dynamics, 53:7, 1034-1062, DOI:
10.1080/00423114.2015.1037313
To link to this article: https://doi.org/10.1080/00423114.2015.1037313
Published online: 14 May 2015.
Submit your article to this journal
Article views: 6012
View related articles
View Crossmark data
Citing articles: 80 View citing articles
Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=nvsd20
Vehicle System Dynamics, 2015
Vol. 53, No. 7, 1034–1062, http://dx.doi.org/10.1080/00423114.2015.1037313
State of the art survey: active and semi-active suspension

control
H. Eric Tseng ∗ and Davor Hrovat
Ford Motor Company, 2101 Village Rd, Dearborn, MI 48121, USA
(Received 18 February 2015; accepted 31 March 2015 )
This survey paper aims to provide some insight into the design of suspension control system within
the context of existing literature and share observations on current hardware implementation of active
and semi-active suspension systems. It reviews the performance envelop of active, semi-active, and
passive suspensions with a focus on linear quadratic-based optimisation including a specific example.
The paper further discusses various design aspects including other design techniques, the decoupling
of load and road disturbances, the decoupling of pitch and heave modes, the use of an inerter as
an additional design element, and the application of preview. Various production and near produc-
tion suspension systems were examined and described according to the features they offer, including
self-levelling, variable damping, variable geometry, and anti-roll damping and stiffness. The lessons
learned from these analytical insights and related hardware implementations are valuable and can be
applied towards future active or semi-active suspension design.
Keywords: active suspension; optimal control; semi-active system; suspension control
1. Introduction
There are numerous important requirements that a typical automotive suspension must satisfy
[1,2] including
• maintaining proper vehicle posture when subject to various inertial and external forces and
moments caused by braking, turning, wind gust, and other events;
• providing ride comfort in view of the road roughness inputs which act as a major
disturbance to the vehicle;
• securing good road handling and overall vehicle agility;
• avoiding excessive suspension stroke or related hard stop/impact.
Additional opportunities exist with the more advanced active and semi-active suspensions
regarding various safety functions [3] and other new exciting functionalities.[4] Although
most current suspensions are of the conventional passive type, this survey will focus on active
suspensions where additional power sources – such as pumps and compressors – are needed
to accomplish the desired functionality (for a more precise definition of passive and active
*Corresponding author. Email: [email protected]

c 2015 Taylor & Francis
Vehicle System Dynamics 1035
suspensions in terms of the well-established mathematical notion of passive operators, see

[5–7] and references therein).
While advanced active suspensions offer opportunities for substantial improvements in
ride and handling as well as overall vehicle posture, stability, and added functionality, they
also typically imply higher energy consumption, cost, complexity, packaging, and addi-
tional operational requirements. The latter include smooth start-up and shutdown, robust
operations and diagnostics, fault containment and management, as well as the ability to
effectively adapt to different operating conditions including reconfiguration in case of
sensor/actuator/mechatronic system (partial) failure.
As a compromise between passive and active suspension performance, complexity, and
installation and operational costs, one sees more and more semi-active suspensions, which
are marketed under names such as continuously controlled damping, MagneRide, and oth-
ers. Semi-active suspensions typically include an adjustable damper or shock absorber [8]
where the damping parameter is modulated through a relatively small computer-controlled
actuator. Since it does not require much energy, the semi-active damper can be considered an
essentially passive device.
Complete development of an advanced suspension must consider all of the above require-
ments. This requires consideration of practically all vehicle degrees of freedom (DOF) within
the bandwidth of main interest, that is, in the case of vehicle main body mass or sprung
mass this would include all 6 DOF (vertical, longitudinal, lateral, pitch, roll, and yaw). How-
ever, the main emphasis of the present survey is on ride quality, which is a traditional main
attribute of interest when addressing advanced passenger vehicle suspensions. This would
typically lead to designs using a reduced vehicle sprung mass model with motion limited to
only vertical (or heave), pitch, and roll components.
Moreover, within the above somewhat limited context, the main focus will be on secur-
ing superior ride comfort in view of the always present road-induced disturbances caused by
road roughness and irregularities such as potholes and bumps. Thus, we will for the most part
separate the load-induced disturbance discussion – such as pitch and roll changes due to (sud-
den) braking, acceleration, turning, and similar external actions – from the ride performance
optimisation analysis. This, of course, does not mean that we are ignoring their importance,
but rather that the engineering task at hand has been ‘modularised’ or divided into more man-
ageable and insightful subtasks. In the case of road-induced disturbances, lots of insight can
be gained based on simple models which can later be integrated with other subsystems and
subtasks including the aforementioned load-induced dynamic effects.
2. Optimal suspension control with H2/linear quadratic synthesis/design
As stated in the introduction and as established from many suspension studies,[7,9–11] the
most significant and insightful conclusions for vehicle suspensions can be observed from a
simple quarter car model with an undamped tyre and a road disturbance described by Gaus-
sian white noise ground velocities or, equivalently,[7] a step in road displacement (or an
impulse in road velocity). Therefore, in this paper, we will focus our discussion on control
analysis and comparison using the quarter-car model.
The performance indices relating ride comfort and handling are often measured by the root-
mean-square (RMS) values of sprung mass acceleration and tyre deflection, respectively.[7]
Additionally, the RMS value of suspension displacement can be utilised to enforce the rattle
space (or available suspension displacement) constraint. As a result, the optimisation of the
suspension controller using the quarter-car model often lends itself to a well-defined H2 or
linear quadratic (LQ) optimisation problem.
1036 H. Eric Tseng and D. Hrovat
2.1. Performance index
The ground vehicle suspension design is influenced by a number of often conflicting factors
and attributes. These factors will be discussed and then combined into an overall performance
index for the H2 optimisation problem.
2.1.1. Performance index of ride comfort
One of the most popular and fitting ride measures is based on the RMS value of vertical
acceleration of the sprung mass, typically measured or projected at the driver’s or passengers’
seat locations. In a field study in 1978,[12] the authors involved 78 passengers in two different
vehicles and 18 different road sections to conclude that ‘excellent correlation was found to
exist between the subjective ride ratings and simple RMS acceleration measurement at either
the vehicle floorboard or the passenger/seat interface’.
Further refinements of the RMS ride measure are possible through adding the RMS value
of the sprung mass jerk (the derivative of acceleration) to the RMS value of sprung mass
acceleration. Some literature [13–15] advocates the inclusion of jerk as an added measure
that amplifies the contribution from high-frequency disturbances, which are important for
NVH (noise, vibration, and harshness).
To account for the frequency dependency of human sensitivity to vibrations and the length
of time of human exposure, a standard has been developed by the International Organi-
zation for Standardization as ISO 2631.[16,17] It noted that the region of greatest human
sensitivity to vertical vibration lies between 4 and 8 Hz, which roughly includes the various
resonances of human internal organs. The ISO-based metric can be easily incorporated into
the H2 optimisation formulation.
National Aeronautics and Space Administration (NASA) developed a more compre-
hensive measure that takes into account interdependency of various modes/direction
of vibrations.[18] A similar measure has been adapted and extended to automotive
applications,[19] considering the interdependence between heave, pitch, roll, and other
factors.
The ride metric of choice will depend on the context of its usage. While the more com-
plex ones can reflect more details and nuances, the simple one can better focus on the
system level and major benefits. This paper will utilise the simple RMS of sprung mass
acceleration.
2.1.2. Performance index for suspension displacement constraint
In practice, the available suspension displacement or so-called rattle space is limited. One can
include this constraint by adding the suspension displacement in the performance index. With
the attempt to minimise the corresponding cost, the H2 optimisation problem effectively soft
constrains the suspension displacement. Similar to the ride comfort metric, the suspension
displacement-related cost can be reflected on the RMS value of suspension displacement,
especially for random/stochastic road input.
2.1.3. Performance index for handling/road holding
As there exists a convex relationship between tyre cornering forces and normal forces
(reflected by tyre deflection), the variation of tyre deflection leads to the variation of cor-
nering forces. It is conceivable that a very large variation of tyre deflection may even lead to
a loss of contact with the ground. Therefore, the RMS value of tyre deflection variation can
be an explicit metric for handling characteristics. Note the direct relationship between tyre
deflection and cornering capabilities have been experimentally studied.[7]
Based on the aforementioned discussion, it is no surprise that the most widely used perfor-
mance index for active suspension study is the combination of the three previously discussed
RMS values. That is, using a weighted combination of RMS acceleration of sprung mass,
RMS suspension displacement, and RMS tyre deflection variation to represent the perfor-
mance index for optimisation. This performance index lends itself to an H2 norm of the
weighted states and input of the quarter-car model. Interestingly, the deterministic and the
stochastic interpretations of this H2 norm lend themselves to the road disturbance of a step in
road displacement, and that of Gaussian white noise ground velocities, correspondingly. The
equivalence of these two interpretations was described in more detail in [20].
2.2. H2 optimal control for 2 DOF quarter-car model
The mathematical state space representation of a 2 DOF quarter-car model (Figure 1) can be
found in [9]. Here, we choose the states as tyre deflection, unsprung mass velocity, suspen-
sion deflection, and sprung mass velocity (Figure 1). With this representation, we have the
performance index as
PI = E(r1 x21 + r2 x23 + u2 ), (1)
where x1 and x3 represent the tyre deflection and suspension stroke as illustrated in
Figure 1(a), u represents the sprung mass acceleration (u = U/ms ), and the expectation E
represents steady-state mean-square values. Figure 1(b) shows the corresponding limiting
case when unsprung mass is negligible, and Figure 1(c) illustrates a passive quarter car.
Figure 2 shows a quarter car equipped with various suspension elements including
adjustable spring and damper and different configurations such as springs in parallel to the
actuator and in series to the actuator.
The earlier four-state LQ optimisation was solved numerically,[7,9] and the results of
the corresponding global study for various weighting in the combination of performance
index are shown in Figures 3 and 4. The figures illustrate how the weighting factors serve as
tuning knobs, controlling the trade-off between acceleration, suspension stroke (design con-
straint), and tyre deflection (handling performance constraint). These plots were generated
with the ratio of sprung-to-unsprung mass set to 10 and the unsprung natural frequency (or
wheel-hop) as 10 Hz. The choice in optimisation with respect to different combinations of
Figure 1. (a) Two DOF quarter-car model, its states x, road velocity w, and actuator input U; (b) one DOF
quarter-car model; and (c) two DOF quarter-car model with passive spring and damper.
Figure 2. (a) Two DOF quarter-car model equipped with actuator, serial spring, and parallel adjustable spring and
damper and (b) the corresponding bond graph model.
weighting in the performance index can be predetermined as the vehicle design suspension
characteristic or adapted dynamically during on-road operation to enable the best driving
experience.
We see from Figures 3 and 4 that reducing r1 , the weighting factor on tyre deflection,
while maintaining r2 , the weighting factor on suspension deflection, results in increased tyre
deflection and decreased sprung mass acceleration, that is, better ride at the expense of some-
what worse handling. On the other hand, keeping r1 constant while reducing r2 leads to
increased suspension deflection and decreased sprung mass acceleration. In the extreme case
of assigning r2 to zero, the problem becomes a degenerative one where extremely large sus-
pension displacement would result. The corresponding optimal force of this degenerated case
would not attempt to support the static weight of the vehicle since there is no cost associated
with suspension stroke. As a rule of thumb, choosing r1 /r2 to about 10 leads to reasonable
results.[7] The shaded area in Figures 3 and 4 illustrates where most optimal designs for
automotive applications will reside due to the above-mentioned practical considerations. An
important observation is that no suspension control design, linear or nonlinear, can exhibit
behaviour in the area falling below the r1 near zero line (represented by r1 = 10−2 ) or the r2
near zero line (represented by r2 = 10−3 ).
Similar solutions can be obtained using linear matrix inequality (LMI)-based
approaches.[21–23] A desirable feature of the LMI-type controllers is that they can incor-
porate the slowly time-varying parameters, such as changes in vehicle sprung mass. Related
methods here include the linear parameter-varying techniques [22,24] that explicitly include
possible slowly varying system parameters, such as road roughness and vehicle speed
variations with results similar to gain scheduling investigated in [25].
Various observations about the suspension performance limitation from different
approaches as well as limited experimental work [7] have affirmed the performance limits
observed in Figures 3 and 4. For example, any improvement in one area (e.g. ride) would be
reflected through corresponding loss of performance in other complementary areas (e.g. han-
dling in wheel-hop mode and/or rattle space requirement). Essentially this is a consequence
of the fact that a single actuator is required to fulfil two conflicting requirements. This is also
reflected through the invariant properties reported in [26].
Figure 3. Normalised ride comfort versus suspension stroke (A as road roughness coefficient and V as vehicle
speed).
2.3. Illustrative example
To illustrate the performance in frequency response as well as where it resides on the global
‘carpet’ plot of Figures 3 and 4, a practical example of H2 optimal design active suspension
is described in the following.
Assume that for a particular vehicle ρ = 10 and ωus = 2π 10 rad/s. The road is described
by a road roughness coefficient A = 1.6 × 10−5 ft (4.9 × 10−6 m) which corresponds to a
medium quality road. The vehicle is traversing this road at a speed of V = 80 ft/s (88.5
km/h). It is reasonable and desirable to design a suspension such that the tyre deflection from
equilibrium under this road excitation would be maintained within 1 inch (2.54 cm) 99.7%
of the time. This would translate to a requirement on the normalised tyre deflection, x̃1 , to
be maintained within 0.3 s1/2 . Note that we use the accent ∼ to denote the normalised
√ RMS
value with respect to the road velocity excitation, for example, x̃1 := x1,rms / 2π AV , where
A is the road roughness coefficient and V is the vehicle travelling speed.[27]
From Figure 4 and a normalised RMS tyre deflection value of 0.3 (s1/2 ) in this illustrative
example, we see the minimum possible normalised RMS acceleration ũ is 10 (s−3/2 ). If we
accept the normalised value of 10.9 (which reflects to a 0.03 g sprung mass acceleration RMS,
Figure 4. Normalised ride comfort versus handling.
or the sprung mass acceleration under 0.1 g 99.7% of the time), we can arrive at this optimal
control law by picking the tuning factors r1 = 1100 and r2 = 100. This is represented as
design point A in Figures 3 and 4. Using these values for r1 and r2 in Figure 3 reveals a
normalised RMS secondary suspension deflection x3,rms = 0.605 s1/2 which ensures that the
suspension deflection will remain within ± 2 inch (5.08 cm) of the static value 99.7% of
the time. With r1 = 1000 and r2 = 100, the optimal control gains for full state feedback are
k 1 = − 6.084, k 2 = 0.548, k 3 = − 10.0, and k 4 = − 4.438 for tyre deflection x1 , unsprung
mass velocity x2 , suspension deflection x3 , and sprung mass velocity x4 , respectively. The
closed-loop eigenvalues are e1,2 = − 2.20 ± j2.26 and e3,4 = − 2.75 ± j62.9. The first set
of oscillatory eigenvalues corresponds to the vehicle heave mode with a natural frequency
of 0.5 Hz and damping ratio of ≈ 0.7. The second set corresponds to the wheel-hop mode
with a resonant frequency of 10 Hz and relatively small damping ratio of 4.4%. Whether or
not this small amount of damping is of consequence will depend on the particular hardware
implementation, which may include a combination of active and passive means to improve
closed-loop robustness. Also in practice, an excessively soft suspension setting may require
load levelling to keep the suspension static or steady-state deflection in the neighbourhood of
zero or some nominal value.
Figure 5. Normalised ride comfort versus handling – passive and active suspensions.
At this stage, it is instructive to compare the above active suspension results with a cor-
responding passive case with the structure shown in Figure 1(c). For this purpose consider
Figure 5 which displays normalised RMS acceleration versus tyre deflection for active and
passive suspensions. For simplicity, the optimal active suspension performance is shown via
the limiting curves r1 ∼ 0 and r2 ∼ 0 only, which are representative of the main trends.
Here, the design points with lower accelerations and higher tyre deflections are characterised
by softer, lower frequency body modes and less damped wheel-hop modes.
The added curves in Figure 5 correspond to the passive performance trade-offs with sprung
mass mode resonant frequencies f2 at 1 or 1.5 Hz (which corresponds to different passive
spring stiffness in Figure 1(c)), and damping ratios ς varying between 0.02 and 1 (which
corresponds to varying passive damping in Figure 1(c)). A typical passive suspension setting
is shown as design point P1 , where f 2 = 1 Hz and ς = 0.3. As can be seen from Figure 5,
this amount of damping is near optimal in terms of the ride and handling compromise for the
passive case. These values are within the range typically found in most of today’s vehicles
which have evolved through many iterations primarily based on intuition and decades of
experience.
The performance of an optimal active suspension with the same amount of wheel-hop
as the optimal passive is shown in Figure 5 as design point A1 . Comparing design points
P1 and A1 , it can be seen that in this case the active suspension results in only 11% lower
RMS acceleration levels with respect to P1 . However, it can also be seen that for some other
road/speed operating conditions (to the right of A1 ), there is a potential for substantial ride
improvement with active suspensions compared with their passive counterparts. Indeed, as
shown in Figure 5, for passive suspensions almost any deviation in operating conditions from
P1 will result in degradation of performance. On the other hand, for active suspensions, either
handling or ride can be improved by choosing the tuning parameters r1 and r2 such that the
resulting operation settles either to the left or to the right of A1 , respectively. For example, if
we pick a different active suspension design point A (Figure 5) by relaxing the tyre deflection
constraints, we see that the RMS acceleration at point A is as much as 67% smaller than that
at Point P1 .
Based on the above results and observations, it can be concluded that the full advantage
of active suspensions stems from possible adaptive tuning (or gain scheduling) of controller
parameters, depending on the driving condition.[7,25] For example, if the steering wheel
position or lateral acceleration sensors (possibly augmented with other road information)
indicate operation on a straight section of a road where handling is less critical, then it may
be possible to relax the wheel-hop constraint. Similarly, by knowing which sections of the
road the vehicle is or will be travelling on and their corresponding road roughness, the tyre
deflection constraints can be enforced differently, allowing more substantial improvement in
RMS acceleration. Note that the exact amount of wheel-hop that could be tolerated under
different driving conditions can be predetermined through appropriate vehicle tests.
The frequency responses of the three metrics with respect to road disturbance velocity offer
insightful information between the trade-offs in suspension design. Different attributes of the
performance at three operating points of interest (A, A1 , and P1 ) are illustrated in Figure 6
in their corresponding Bode plots. From the sprung mass acceleration frequency plot, it can
be seen that the main difference between the active suspension A1 and corresponding passive
counterpart P1 occurs at the sprung mass mode around 1 Hz, where the active suspension
brings much more damping through the damping on sprung mass velocity, the so-called
‘skyhook’ damping. The active suspension A results in further significant reduction of the
acceleration levels (except at the ‘invariant point’ at the wheel-hop frequency [26]), but at
the same time it deteriorates handling and rattle space performance due to a large, poorly
damped resonant peak around the wheel-hop frequency. Note that the ‘invariant point’ arises
from the structure of the suspension and results in the performance limit of sprung mass
acceleration at the wheel-hop frequency regardless of control design, assuming negligible
tyre damping. The reduction of acceleration at other frequency regions causing the rise of
suspension and tyre deflection can also be interpreted as the ‘waterbed’ effect, a term coined
in H infinity control analysis, that is, the improvement in one area (e.g. ride with respect to
sprung mass frequency) of the frequency domain would be reflected through corresponding
loss of performance in other complementary areas (e.g. handling in tyre-hop mode and/or
rattle space requirement).
The suspension deflection Bode plot of Figure 6(b) indicates that, unlike their passive
counterparts, the LQ optimal suspensions can result in a non-zero gain in very low frequency
between the rattle space and road velocity input.[7] However, in practice this may not be
detrimental since it is usually a standard practice to use appropriate signal processes, for
example, high-pass filters to eliminate the very low-frequency components of the ground
input signals resulting from hills and extended road grade.
Figure 6. (a) Tyre deflection Bode plot of points on Figure 5 and point A with dynamic absorber, (b) suspension
Bode plot of points on Figure 5 and point A with dynamic absorber; and (c) acceleration Bode plot of points on
Figure 5 and point A with dynamic absorber.
Moreover, Figure 6 shows the Bode plots of an active suspension control (designed with
the same weighting used in operating point A) combined with a dynamic vibration absorber.
We see that the sprung mass acceleration can be further improved in the wheel-hop frequency
since the introduction of the dynamic absorber (which interacts with only the unsprung mass)
presents a new structure.
2.4. Extension to semi-active optimal case
In semi-active suspension systems, the suspension force can be modulated through a very
small amount of supplied energy. In this case the suspension force cannot track an arbi-
trary desired force from the unconstrained LQ optimisation-derived control law. Rather, the
problem becomes a constrained LQ optimisation. As a compromise, the semi-active control
design typically follows its unconstrained active counterpart when it can, and operates along
the passivity envelope when it cannot. For example, the damping force is adjusted to follow
the desired suspension force derived from the optimal control law, and set to zero when a neg-
ative damping force is required. This control is therefore commonly referred to as ‘clipped
optimal’. This may result in 20% lower performance in ride comfort when compared to its
active suspension counterpart.[11]
The optimal control law for the semi-active system has been posed as a constrained LQ
optimisation and solved numerically in [28,29], involving the iterative solution of a time-
varying force constraint. A specific example [29] shows that a 10% advantage with respect
to clipped optimal can be achieved. However, it also finds that the amount of improvement
depends on driven scenarios and is usually very limited. A later work [30] leveraged the
explicit hybrid MPC to confirm analytically the previously numerical finding that clipped
optimal is not the optimal control for semi-active suspensions in general.
2.5. Extension to preview optimal control
There are numerous investigations of the potential performance of the suspension if the road
disturbance profile can be known in advance.[31–35] The solution consists of state feedback
terms that have the same gains as the non-preview case, and feedforward terms that consider
the road disturbance of the upcoming known duration.
It is worth noting that most of the papers place the origin of time (i.e. start of preview
time for future disturbances) at the quarter-car location [31,35,36] or at the location of the
front wheels.[32,37,38] In [34], this origin is simply moved to the location of preview,
thus transforming the preview problem into an equivalent delay-time problem, where an
extensive literature base already exists.[39] This approach is also used in [40] where the
authors developed a method for direct and efficient calculation of the related performance
index.
Comprehensive studies of the benefit of preview compared to its non-preview counterpart
have been demonstrated for full-bandwidth active suspensions [34] and limited bandwidth
ones.[41,42] The benefit for full-bandwidth active suspensions was illustrated on a plot sim-
ilar to Figures 3 and 4 in [34], relating to its non-preview counterpart. And the benefit for
limited bandwidth active suspensions in ride comfort and suspension stroke improvement
was shown in [42].
3. Special consideration of handling, loading, and suspension travel limit
Although not always explicitly mentioned, most of the studies with traditional simple ride
models (e.g. quarter-car, half-car, and full-car) focus on road roughness disturbances only,
neglecting external (inertia) forces and loads due to, for example, braking and turning. The
control damping for inertia load-induced excitations can be included later in a separate and
independent design procedure. In practice, this is accomplished using longitudinal and lateral
acceleration measurements as feedforward signals [43] indicating braking or turning actions.
This typically leads to appropriate ‘stiffening’ of the active or semi-active suspension to pre-
vent or counteract excessive pitch or roll. This stiffening action may reduce the ride quality,
but in practice this may be of secondary importance, especially during hard braking and/or
handling manoeuvres.
For the case when complete decoupling of the above road and load disturbance effects is
desired, Smith,[44] Smith and Wang,[45] and Wang and Smith [46] present an interesting
approach based on Youla parameterisation and LFT (linear fractional transformation). The
resulting controller can be designed independently – first for the road-induced disturbance
and then for the load counterpart. For example, the suspension can be made very stiff for
load inputs in order to hold a desired attitude, while at the same time it could be made very
soft for the ground inputs (of course, the latter has to be within some practical limits since
any prolonged ramp-like ground input could lead to suspension bottoming and undesirable
jerks and possibly even structural damage). This is somewhat similar to the above-mentioned
sequential design approach used in practice with the exception of total decoupling. In this
context, the ‘practical’ design uses longitudinal and lateral acceleration signals (in addition
to other heave-related measurements) as a feedforward signal or proxy for gas/brake pedal
and/or steering wheel applications.
A comprehensive and direct approach to simultaneous inclusion of road and load dis-
turbances has been recently presented in [47] based on the above more conventional LQ
stochastic optimisation. In addition to the usual road disturbance, which is assumed unmea-
sured, Brezas and Smith [47] also introduce the load disturbance but as a measured, constant
disturbance. An advantage of this approach is that it directly addresses optimisation of the
main attribute of interest, that is, ride comfort, which is introduced through an appropriately
formulated LQ performance index.
In the case of longitudinal dynamics, further extensions of this approach are possible by
introducing the force disturbance through a delay, which effectively presents a load preview
from the standpoint of the suspension controller similar to the preview modelling used in
[34]. This could be further refined by using a (simple) model of vehicle power train (brake)
paths, from gas (brake) pedal to actual wheel torque production. Similar comments apply for
load prediction due to steering inputs where ultimately some appropriate models of vehicle
handling can be used too.
A simple approach to load disturbance containment is through load levelling, that is,
applying an integral control to regulate the rattle space between sprung and unsprung
masses.[48–51] In particular, Hrovat and Hubbard [48] use jerk-optimal LQ control to end
up with an effective and systematic way of obtaining relatively fast load levelling with
some – mostly minor – sacrifice in sprung mass acceleration performance. It should be
pointed out that in practice a slow-acting load levelling is often used to centre the suspen-
sion against slow-changing loads. Davis and Thompson [49] use both integral and derivative
controls in a ride quality- and handling capability-focused 7 DOF full-car model to regulate
as well as maximise usable suspension deflection as the derivative action helps to decrease
overshoot.
As the suspension stroke is limited, practical design of all suspensions may introduce
additional feedback elements to stiffen up the suspensions as the rattle space approaches
its travel limits. In addition, actuator dynamics may contain delays, nonlinearities (which is
inherent in the semi-active case), or lags. Various control designs have been developed to
address these issues including backstepping design,[52] adaptive robust control,[53] learn-
ing automata,[54,55] and Pareto optimisation.[56] In [52], a nonlinear filter whose effective
bandwidth depends on the magnitude of the suspension travel was introduced to facili-
tate smooth shifting of the control objective between ride comfort and suspension stroke
utilisation. In [53], a two-loop controller with adaptive robust control was introduced to
ensure that the nonlinear actuator delivers the force requested from the LQ top (outer) loop
design. Gordon et al. [54] and Howell et al. [55] used an approach combining concepts from
stochastic optimal control with those of learning automata. And Valasek and Kortum [56]
used the Pareto optimisation procedure to determine the state-dependent feedback gains in a
predefined control structure.
It is worth noting that Gordon and Best [57] included non-quadratic terms for the optimi-
sation of semi-active suspensions in their optimisation and observed the performance with
the conclusion that the standard ‘clipped optimal’ technique is close to the one resulted
from a complex optimisation. In [58], the optimisation of semi-active suspension with pre-
view is numerically optimised and it was observed that only small improvement can be
achieved compared to the simple ‘clipped optimal’ preview control. These conclusions are
corroborated in [30] where analytical explicit model predictive control was used.
4. Other approaches and considerations
In addition to the above LQ optimisation-based approaches that are the main focus of the
present survey, there are many other possible approaches to advanced active and semi-active
suspension design and analysis. These include Youla parameterisation and LFT,[44–46] pole
placement,[59] LMI,[60] H2/H-infinity, and H-infinity.[61–65]
Furthermore, multi-objective optimisation (symbolically or numerically) has been widely
used to systematically manage nonlinearities and constraints.[66–68] It is particularly
prevalent for semi-active suspension applications.[69,70] As the suspension performance
index often includes conflicting terms, the solution of a multi-objective optimisation finds
the best trade-off among the various predefined control terms. This is also known as Pareto
optimisation used in systematic design procedures (note that a similar result for the linear
system was obtained using the previously mentioned global optimisation in Section 2 with
varying weights).
While ride comfort is the main focus for passenger vehicle suspensions, minimising road
damage is important for truck and heavy vehicle suspensions.[71–74] An approach that pays
particular attention to tyre force variation is an interesting extension of the well-known
skyhook concept,[8] proposed in [75] as the so-called ground-hook control. Specifically, it
amounts to augmenting the (semi-)active actuator desired force calculation with an addi-
tional, virtual damping term proportional to the relative velocity between the unsprung mass
and ground (Figure 7), emulating a fictitious damper between the unsprung mass and road.
The typical objective of such a suspension optimisation was to minimise the integral square
of the tyre force [76,77] by using a simple quarter-car vehicle model for the control synthe-
sis, which is similar to previously mentioned LQ-based RMS tyre deflection minimisation
except that the latter would typically be part of a performance index that includes additional
terms due to ride comfort and rattle space (design) constraint. The integral square sprung
mass acceleration term as a measure of ride comfort was later included in [78,79] where
the ground hook and other coefficients were determined from numerical optimisation with
specified cost function. Note that the above optimisation for the case of semi-active actuator
amounted to essentially nonlinear system parameter optimisation since the extended ground-
hook optimisation structure was fixed a priori.[76,77] Since the main nonlinearity is in the
semi-active actuator, state dependence of the related parameters was also introduced in the
process of optimisation.[56]
The ground-hook concept was found useful when minimising road damage [56,77] espe-
cially due to heavy trucks where the damage to pavement can be substantial and up to 10,000
times greater than due to passenger cars.[78] This leads to so-called road-friendly suspension,
Figure 7. (a) Skyhook, (b) ground hook, and (c) combination of skyhook and ground hook.
which was further extended to bridge-friendly [79–81] and ESC,[82] that is, Electronic Sta-
bility Control-friendly, applications. The concept was further tested through experimental
vehicle implementation [56,77] using a SKODA-LIAZ prototype truck.
The main emphasis in Section 2 was on advanced active and semi-active suspensions
using simplified (minimal component count) system models. In practice, however, any imple-
mentation of these advanced concepts will also include additional conventional, passive
components. For example, an active actuator may be positioned in parallel with a spring
for static support and a damper to dissipate energy without an active element. Thus, one must
eventually consider designing the whole suspension system consisting of both active and
passive elements.
In addition, as is typically the practice when implementing the LQ or any other linear
algorithm, one should first attempt (static) linearisation of any possible nonlinearities in actu-
ators, plants, and/or sensors. The combined controller-actuator subsystem should then be
further verified and as necessary refined through additional simulations based on a more
detailed performance-oriented model. The latter should include a detailed model of actu-
ator dynamics, which is often the most challenging and most important component of an
advanced, active, or semi-active vehicle suspension system design.
A reasonable starting point in an active or advanced suspension system design would be
a consideration of the passive system alone in order to establish the best passive benchmark
against which any advanced mixed passive–active system could then be compared. At this
stage there is a need to more precisely define passivity so that the whole passive suspension
optimisation problem can in turn be properly defined. One of the first attempts at this was in
[28] where the notions of passive operators from mathematics and electrical circuits are used.
In addition, Hrovat [28] also posed the corresponding optimal passive suspension problem.
However, at the time of its writing, much of the necessary mathematical Computer Aided
Engineering (CAE) tools and computer optimisation capability were missing. This class of
problems were effectively treated more than a quarter-century later [83] using LMIs and
other modern CAE tools. In [83] problems with both H2 and Hinf optimisation cost are
considered for the case of a quarter-car model with both ground and load inputs included.
This approach could likely be further extended to hybrid configurations where both passive
and active suspensions are combined.
Once the optimal passive transfer function is determined, there remains the problem of syn-
thesising or actually implementing it via standard passive elements such as springs, dampers,
and inertias. In the process, there is a need for an inertia element – called ‘inerter’ in [84]
– that acts on a relative velocity rather than the customary absolute velocity of relevant
masses or inertias. Perhaps the easiest way to see this is via bond graphs where one could
imagine a hydraulic piston–cylinder combination between sprung and unsprung masses [85]
(inerter-like structure was described in its third edition with more examples exhibited in its
later editions). Another example can be found in modelling hydraulic mounts.[86,87] An
effective and practical mechanical implementation of an inerter was shown in [83,84] con-
sisting of multiple gear sets with large gear ratios resulting in a large reflected inertia. The
latter facilitates relatively small overall weight for the device, which is an important practical
consideration. In addition to possible suspension applications, the device was also used to
mitigate steering instabilities of motorcycles.[88]
Another more constrained approach to optimisation was considered in [89] where the
underlying passive structure was first fixed and then the optimisation was performed for each
free design parameter and compared with the best tuning of the conventional passive suspen-
sion structure consisting of a parallel spring and damper. The results indicate around 10%
and more improvements in the comfort RMS acceleration metric, although it is not clear if
this included the same level of other constraints (tyre and rattle space).
Using the above-mentioned definition of passivity it is now possible to ask the follow-
ing question. For a given optimal suspension control law can we determine if it could be
implemented through a passive device? If the answer is positive this would have significant
practical implications since such a device could then be realised using only passive compo-
nents that are typically less expensive, less complex, and require no external power sources
and consumption to operate. It was first shown in [6,28] that for the quarter-car LQ opti-
mal control the resulting control law is necessarily active. A more comprehensive study was
subsequently undertaken in [5] including additional control laws applied to half- and full-car
cases.
In addition, Smith [44] and Smith and Walker [5] consider some other interesting aspects of
active and passive suspensions. These include the derivation of a complete set of constraints
for the road disturbance transfer functions for typical choices of measured outputs as well
as performance limitations and constraints for active and passive suspensions for quarter-,
half-, and full-car models. This complements the prior well-known results on the invariance
of certain transfer functions at special frequency points.[26]
While the quarter-car model largely describes the vehicle response to the road disturbance,
the heave and pitch experienced at various observed points along the vehicle longitudinal axis
can vary due to the ‘multi-input’ disturbance (coming through both front and rear wheels). For
passive vehicles, due to the fact that the excitation at the rear wheel is a time-delayed version
of the front wheel, there is a ‘wheelbase filtering’ effect [90] that changes the transfer function
to heave and/or pitch at particular frequencies as a function of vehicle speed and wheel base.
This effect is more perceptible for trucks with a long wheel base. The quarter-car suspension
design can be expanded to full car. One attractive approach to address the full-car and half-car
suspension design and leverage the extensive knowledge from the quarter-car study is to first
apply control design decoupling modes. The importance of decoupling between sprung mass
pitch and heave mode as well as between the equivalent front and rear quarter-car models
is described in [7,91]. In [91], an active suspension was designed to decouple sprung mass
heave and pitch mode through zeroing the coupling terms between pitch and heave relative
(suspension) deflections, and similarly those between pitch and heave relative (suspension)
velocities. This structurally decoupled active suspension minimises body heave motion to
road pitch disturbance, and similarly body pitch to road heave, to a secondary level. It was
exhibited [91] that a superior performance can be achieved by simply adding the skyhook
(or ‘absolute’) damping to the ‘decoupled’ active suspension design. This approach is further
verified in [92] through mathematical manipulation and state space transformation where the
heave and pitch modes and corresponding disturbances are decoupled.
Important aspects of state estimation are addressed in [93,94]. Analytical techniques for
fault detection and facilitating safe start-up and shutdown of active system are described
in [95]. Some recent attempts to introduce risk sensitivity into the LQ optimal suspension
problem were addressed in [47]. The approach has some practical appeal since it facilitates
tuning the balance between performance and associated risk due to plant uncertainties, mod-
elling limitations, and possible unmodelled dynamics, thus providing system designers with
additional flexibility through an added degree of tuning freedom.
5. Production and near production features and systems
5.1. Self-levelling feature
Various self-levelling systems have been introduced to the market including the Electronic
Air Suspension systems developed by Citroen’s hydropneumatic systems [96] and Ford
Motor Company’s Electronically Controlled Air Suspension on Lincoln,[97] allowing these

vehicles to maintain proper ride height and suspension stiffness over a wider range of vehi-
cle loading. This feature adjusts the vehicle ride height, usually at low frequencies, in order
to balance among (1) soft and comfortable ride stiffness from the passive spring, (2) proper
vehicle attitude/stance, and (3) ample rattle space for anticipated or unknown road distur-
bance ahead. For example, the system in the 1984 Lincoln Continental allows its spring to be
32–38% softer than a corresponding steel spring suspension while maintaining vehicle stance
and suspension travel.[97] In addition, this active suspension feature can enable the lower-
ing of the vehicle at highway speeds to improve aerodynamics. This feature is implemented
in the Lincoln Mark VIII [97] and most recently the Tesla Motor Model S,[98,99] provid-
ing improved aerodynamics, more consistent handling, better fuel economy, and extended
range.
Load levelling suspensions are often constructed using pneumatic springs whose equi-
librium length is adjusted by changing the volume of air in the spring. In the Lincoln
Continental, height sensors are used to trigger the adjustment of the air volume, and sig-
nals from the ignition, doors, and brake systems are also used to ensure levelling only at
appropriate times.
5.2. Variable suspension damping (semi-active suspensions) feature
Production semi-active suspension systems are generally constructed using an adjustable

damper in parallel with the primary suspension spring. These are typically constructed from
pneumatic and/or hydraulic piston/cylinder combinations with electromechanical control of
an orifice. Actuator bandwidth is primarily determined by the reaction time of the con-
trolling valve and associated pressure/force production dynamics. Semi-active suspension
systems have been introduced to the market since the 1980s’ Toyota Soarer’s TEMS (Toyota
Electronic Modulated Suspension).[100] A more recent system can be found in Lincoln’s
Continuous Controlled Damping system introduced in 2006. The Lincoln system uses a
suite of sensors that constantly monitor suspension motion, body movement, steering, and
braking inputs and adjusts the suspension in milliseconds, helping keep the car smoothly on
track. Specifically, it monitors up to 46 inputs and reacts on average within 20 milliseconds
[101,102] to reduce roll, pitch, and heave motions, to enhance driving comfort and dynamics,
and to isolate road harshness.
Another implementation of adjustable damping is through magneto-rheological (MR) flu-
ids. MR fluid characteristics can be changed electronically, allowing the force across the
actuator to change quickly. This method benefits from faster response time, although limited
fluid life can contribute to service concerns. One MR damper application is found in the 2002
Cadillac Seville STS and the 2003 Chevrolet Corvette whose MR fluid is co-developed by
Delphi and Lord Corporation. It is composed of suspended iron particles in a base fluid of
synthetic hydrocarbon. Its continuously variable damping is adjusted in milliseconds. The
fast response time of the MR damper enables closed-loop vehicle stability control through
the control of lateral and longitudinal load transfer characteristics of the suspension during
transient movements.[103,104]
Another method to increase the damping rate adjustment frequency when employing
limited bandwidth orifice control valves is to employ the resistance control semi-active damp-
ing method [105] demonstrated experimentally in [106] where a dual-acting damper with
two control valves is used. Each valve sets the damping rate in one direction (jounce or
rebound).
5.3. Variable suspension geometry (low-power low-bandwidth active suspensions)

feature
A variable geometry active suspension adjusts the ratio of wheel movement to the deflection
of the suspension spring in real time. By changing the leverage of the passive suspension
spring with wheel motion, it essentially controls the wheel rate or effective spring stiffness.
Various systems and hardware configurations that provide variable suspension geome-
try have been proposed in the literature [107–110] including the ‘Delft Active Suspension’
concept that was implemented as a prototype vehicle and demonstrated experimentally.[109]
Ideally, one would like to adjust the variable suspension geometry without requiring much
power. One way to do this is to adjust the suspension leverage by changing the suspension
spring mounting point which requires only actuation perpendicular to the (vertical) suspen-
sion load or vehicle weight (Figure 8).[108] Ideally, with the perpendicular configuration, the
mechanism would require very low power and low energy. In practice, however, the precisely
perpendicular arrangement would be compromised by suspension motion and deflection.
There may be additional trade-offs in mechanical design forcing some deviation from the
ideal geometry. The trade-offs may include vehicle stance/self-levelling, range of mecha-
nism motion (and resulting range of leverage ratio), packaging, effect of jounce/rebound
motion, and effects on wheel turning. A system design study with the trade-offs in mind
is discussed in [107].
Another form of leverage adjustment concept (Figure 9), the previously mentioned ‘Delft
Active Suspension’, is described in [107,111], and its practical realisation with a cone mech-
anism is illustrated in [112,113]. This mechanism connects the spring to the car body on one
end and to a rotatable crank on the other end. The crank is joint-connected to the suspen-
sion/wheel control arm and can be rotated at the joint around the base of the imaginary cone.
The cone mechanism serves two purposes: (1) the length of the spring remains the same as
the crank rotates and (2) the ratio of movement between the wheel/tyre control arm and the
crank changes as the crank rotates (Figure 10).
Due to suspension deflections and motions when experiencing road disturbances, the align-
ment of the ideal position of the cone mechanism will be affected in practice. In particular,
the base plane of the cone mechanism, that is, the plane of the crank motion, will rotate.
Nevertheless, this is a promising technology since it could provide some active suspension
characteristics with only a ‘semi-active’ type of actuator.
5.4. Roll control (semi-active or active) feature
A number of automobile manufacturers have been introducing active and semi-active sus-
pensions that focus on roll control to improve ride and handling with moderate cost and
Figure 8. Illustration of a variable geometry suspension concept.

Figure 9. Illustration of delft active suspension concept.
Figure 10. Illustration of delft active suspension realisation Fv cos(γ ) ∗ b = FD ∗ a.
complexity. It is well known that good ride comfort can be achieved with softly tuned pas-
sive or semi-active suspensions combined with soft roll bar settings. However, this will result
in large pitch and roll motions during more aggressive braking and cornering. The active
(or semi-active) roll control is then used to satisfy these conflicting constraints. In practice,
the concept is typically realised by inserting a small hydraulic jack and a gas spring accu-
mulator with a control solenoid, all placed in series with the anti-roll bar. The anti-roll bar
can be as stiff as in street race cars, without negatively impacting straight-line ride since the
solenoid connecting the hydraulic jack and the accumulator is then kept open, thus lowering
the effective stiffness of the roll bar/accumulator combination.
The performance of anti-roll/roll control devices varies as it may be achieved by a mechan-
ical regulator that depends on suspension motion [114,115] or by an electronic controller
that monitors additional non-suspension-related vehicle signals. It can be actuated with a
semi-active device that manipulates hydraulic valves or the like, or with an additional active
device such as a high-capacity pump. An idea similar to active tilt control has also been
explored in [116,117].
Citroen introduced hydropneumatic suspensions in 1955 with its model DS. The hydro-
pneumatic spring-absorber system incorporates the benefits of both the compressibility of
gas (nitrogen) and responsiveness of (incompressible) hydraulic fluid. As the load increases,
it compresses the suspension pistons and causes the gas volume to decrease and pressure to
rise; the gas spring becomes stiffer but keeps the sprung mass frequency roughly the same.
As the fluid passes through a two-way valve, damping is also provided. In addition to its self-
levelling feature that regulates ride height and resonance frequency (through stiffer spring
with increased load), the Citroen’s hydropneumatic system can be applied towards anti-roll
or roll control by manipulating the flow of hydraulic fluid between left and right corners.
Various improvements have been made in Citroen’s hydropneumatic suspensions. Most
notable are the changes from mechanical regulation to electronic control, the improved capa-
bility of the pump, and the associated increase in actuation bandwidth. For the roll control
feature in particular, there have been several incarnations ranging from semi-active (Hydrac-
tive I and II) to active (Activa). Citroen’s semi-active roll control (Hydractive) limits the roll
angle to about 3° in soft mode or 2.5° in hard mode for a 0.6 g constant turn, while its active
roll control (Activa) can eliminate roll angle completely due to the active and electronic
manipulation of hydraulic fluid.[96]
Roll control can be achieved with the help of a hydraulic rotary actuator inserted around
the middle of the stabiliser bar, as stated above, or it can be implemented through a rotary
actuator fitted with the anti-roll bar as in the BMW ‘Dynamic Drive’ concept.[118] The BMW
system monitors primarily the steering angle and lateral acceleration, among other signals,
and distributes the front and rear axle anti-roll torques. It controls the front and rear actuators
by means of two electronically regulated pressure control valves to minimise the roll motion
in corners and road input roll disturbances, thus improving response to vehicle steering for
vehicle dynamics handling and agility. Note that the roll angle will gradually increase at high
lateral acceleration as the vehicle approaches its stability limit.
It is understood that the benefits of active roll control in overall vehicle stability and active
safety can be further explored through coordinated action with other systems such as active
brakes, active steering, and electronically controlled all-wheel drive. This can be particularly
useful in emergency manoeuvres such as a sudden need for vehicle stabilisation and control
during evasive actions on slippery road surfaces.
5.5. Active suspensions with full features
In general, active suspensions can be categorised into narrow bandwidth and broad band-
width, referring to the control bandwidth of the actuation systems. This typically refers to
actuator force generation bandwidth for all possible motions across the actuator mounting
points. The narrow bandwidth systems typically exhibit actuator control bandwidth around or
slightly above sprung mass frequency and would use the passive spring, positioned in series to
the actuator, to address high-frequency disturbances and vibrations. As such, their bandwidth
is typically between 3 and 7 Hz.[119] Broad bandwidth systems, on the other hand, attempt
to address oscillations beyond the sprung mass frequency using high-bandwidth actuators
typically above 10 Hz.[119]
Since the broad bandwidth active suspensions (BBASs) attempt to control both the slower
body modes and the fast wheel-hop modes, it is generally expected that the broad bandwidth
systems will be more challenging to implement. It is expected that such suspensions will
incur higher costs, complexities, and possible fuel consumption.[120] However, one should
also observe that narrow bandwidth systems may experience similar fuel consumption due
to either their higher compliance and the corresponding increased pump flow/displacement
requirement [119] or higher control effort associated with the mismatch between the desired
and actual control bandwidth.[43]
5.5.1. Narrow bandwidth active suspensions
Some narrow bandwidth systems have been brought to the market with limited success. An
example is the Nissan Infinity Q45a [121] whose hardware included: (1) an accumulator (for
bandwidth limitation and NVH containment); (2) a mechanical supporting spring; and (3)
a relatively inexpensive pressure control valve. It had a single-acting actuator configuration
which is simple and inexpensive. It does, however, limit the system controllability on the
rebound stroke, which is in good part determined by the spring/accumulator settings.
Production hydraulic active suspension systems typically use pressure control valves to
regulate the desired force. Considering that a force-generating actuator is accompanied by
corresponding displacements and velocities of the in-series compliance in the total system,
one may consider the flow control valves to regulate the desired displacements/velocities.
Benefits and limitations of both force and displacement control approaches have been
discussed in [119,122–124].
DaimlerChrysler introduced the ‘Active Body Control’ (ABC) system in the 2000 CL
sedan. It consists of four low-bandwidth (5 Hz) hydraulic actuators providing active load
levelling, heave, pitch, and roll control. The ABC suspension system consists of [125,126]
hydraulic cylinders in series with steel coil springs (Figure 11). This cylinder sets the spring
preload and the series arrangement leads to the low-bandwidth system behaviour. In paral-
lel with the spring/cylinder is a twin-pipe gas shock absorber (damper). Hydraulic fluid is
supplied to each corner via a pump regulated to 200 bars, and an accumulator at each axle
provides pressure levelling during peak loads. High-pressure fluid is carried by two valve
blocks – one at each axle. Each block contains two 3-way proportional valves – one for each
corner – which are used to maintain plunger pressure and a 2-way shut-off valve for each cor-
ner to lock the cylinder when the vehicle is stopped and to provide failsafe operation. Return
fluid is damped through a return accumulator.
Figure 11. Schematic diagram of the Mercedes ‘Active Body Control’ actuator.
The ABC system utilises multiple sensors for mechanical state measurements. Sensors
include longitudinal and lateral accelerometers, three vertical accelerometers (for bump/
pitch/roll detections), as well as a level sensor and a plunger position sensor at each wheel.
Additionally, several hydraulic pressure sensors are used for pressure control and monitoring.
The ABC system employs various algorithm modules for vehicle dynamics control
depending on the vehicle states. Functions provided include speed-based load levelling and
ride height adjustment, skyhook control of the vehicle body heave mode, roll control for roll
angle response to vehicle lateral acceleration, and pitch control for pitch angle response to
longitudinal acceleration. The algorithms are further controlled by driver input via instru-
ment panel height and performance mode switches. ABC has continued to evolve since its
first introduction in 1998. For example, the wheel damping is now continuously adjustable
and the efficiency of the pump has been increased. In addition, computational processing
power is more than double that of the previous system.[127]
5.5.2. BBAS concept implementation at Ford
A broad bandwidth system was developed at Ford Research Laboratory in the early
1990s and was demonstrated by fitting the suspension concept hardware onto a 1989 Ford
Thunderbird.[128] The concept hardware and software verified the previous theoretical study
in ride quality improvement for rough road and for going over a speed bump or crossing
a crowned road. The effort also identified the shortcomings of the implemented hardware
structure to facilitate next generation active suspension hardware design.
This concept car development provided first-hand experience and useful answers to many
practical concerns such as actual power consumption, secondary ride harshness in the range
above 5 Hz, and actuator noise. It also enabled the identification of specific next steps to
address these issues. For example, some of the key items that have been identified to be
addressed in the future include the parasitic losses with the parallel spring structure and the
need for further refinement in the secondary ride quality.
The Ford Thunderbird BBAS research vehicle (Figure 12) used four high-fidelity electro-
hydraulic servo actuators, one at each corner of the vehicle. While a typical narrow bandwidth
suspension is controlled through three-way proportional pressure control valves, the BBAS
Figure 12. Ford BBAS concept implementation.

was controlled through four-way servo valves, which have much higher precision and speed
of response. In addition, the BBAS actuators were based on double-acting cylinders capable
of equally fast rebound and jounce strokes. The vehicle also had one central and four corner-
unit microprocessors for fast signal/control processing; four actuator displacement sensors
and four load cells for internal (force) loop calculations; and four air springs – one at each
corner placed in parallel with the BBAS actuators. The air springs support and self-centre
the vehicle sprung mass as a typical load levelling set would do and at the same time provide
lower sprung mass natural frequency for more comfortable basic ride, which is then appropri-
ately dynamically modified through the BBAS actuators. The system incorporated 26 various
sensors, including accelerometers, pressure sensors, vehicle speed sensors, and others.
The BBAS control strategy was based on coordinated individual wheel control and con-
sisted of two hierarchical levels.[43,128] The outer loop level operated at a 20 ms rate. It
calculated the desired corner forces for the four BBAS actuators, desired operating modes
(handling or ride dominated), and checked the overall system integrity. The ride-related cal-
culations were based on quarter-car vehicle models aimed at emulating skyhook damping
at each corner, which is often very close to the optimal possible ride benefit.[20] Different
effective spring and damping rates were used depending on prevailing operating modes, that
is, ride or handling. Table 1 summarises typical parameters used.[128]
The calculations also included roll and pitch attitude control during turning, braking, accel-
erating, and the like; this was based on lateral and longitudinal acceleration signals and
steering sensors. The commanded or desired levels of corner forces were communicated
to corner computers, which operated at higher sampling rates of 1 ms to implement the
inner-loop force control based on feedback provided by load cell sensors. The corner units
distributed the force requests between different servo valves. For example, the conventional
(relative) suspension damping and the passive segments of skyhook force were provided by
bypass or cross-port two-way servo valves that provided damping only when called for in a
manner similar to semi-active damping control.
The BBAS R&D vehicle accomplished many of its key objectives, especially with respect
to primary ride (up to circa 5 Hz) and handling. The typical results in terms of NASA ride
metrics are summarised in [128] where a smaller number on the scale of 10 implies better
ride. From this reference it can be seen that, on average, the active Thunderbird provided
the best ride among a dozen different vehicles under consideration. The only areas where the
BBAS had somewhat worse performance were on smooth roads, where Jaguar XJ6 and Lexus
LS 400 (both well known for their smooth ride) had an advantage. This may be attributable
to still less refined NVH (higher frequency) levels and related secondary ride quality that are
some of the remaining challenges for BBAS concepts. Based on their summarised results,
it should also be noted that the BBAS Thunderbird significantly outperformed both the
baseline passive suspension Thunderbird and its narrow bandwidth (NBAS) counterparts as
exemplified by production Infiniti Q45a and Toyota Soarer.
Table 1. Typical parameters used in Ford BBAS.
Ride mode Manoeuvring mode
Parameter Units Front Rear Front Rear
Skyhook damping Lb-s/in 23.5 10.0 23.5 10.0

N-s/m 4115.3 1751.2 4115.3 1751.2
Conventional damping Lb-s/in 3.0 3.0 8.0 4.0
N-s/m 525.4 525.4 1400.9 700.5
Suspension spring Lb/in 15.0 15.0 120.0 120.0
N/m 2626.8 2626.8 21,014 21,014
In the same paper, the handling characteristics of the active Thunderbird are compared with
its passive (production) version. The results of back-to-back tests show that the active vehicle
results in more agile handling with faster yaw response, less roll and pitch, and significantly
larger maximum lateral acceleration (0.895 vs. 0.81 g). Thus, in a direct comparison between
the two vehicles, the active car displayed significant improvements in both vehicle ride and
handling.
While achieving the full potential of a BBAS still remains an elusive goal, significant
progress has been achieved with the above early efforts. This holds particularly true for the
case of primary ride and handling. The challenging areas in need of further development
include secondary ride, NVH, and system power consumption, complexity, and cost. Possible
approaches here have been outlined in many of the previously mentioned references (e.g.
[7,11,88,107,119]).
In closing, it should also be mentioned that in addition to the above electro-hydraulic
implementation of the BBAS there was an additional BBAS effort focused on Electrical
Active Suspension (EAS), which was capable of power/energy regeneration.[129] The EAS
concept was also successfully implemented in a prototype vehicle with similar results and
challenges as for the electro-hydraulic counterpart.
5.6. Road preview feature – using road information for semi-active and active control
The concept of utilising the road profile information prior to road disturbances hitting the
wheel, investigated in [31–35], has been partially applied to production vehicles. The early
attempt includes the Nissan vehicles in 1990 with the Super Sonic Suspension [130] feature.
More recently, the preview feature can be found in the 2013 Mercedes S-Class with the
Magic Body Control feature.[131,132] The early Super Sonic Suspension was applied to a
semi-active suspension system and the recent Magic Body Control is equipped with an active
suspension.
The early Nissan Super Sonic Suspension had only three levels of damping that are auto-
matically selected to provide comfort and handling based on the road surface profile scanned
with ultrasonic sensors, as well as steering angle, vehicle speed, brake on/off status, and
measured accelerations.
The Mercedes S-Class detects road surface undulations in advance using a stereo camera
and adjusts the suspension to deal with them accordingly. The suspension system adjusts its
shock absorber damping in advance, making it stiffer or softer for each individual wheel, and
uses active hydraulic pistons to control the desired wheel loads.
The preview information for a three-dimensional road profile is obtained through a stereo
camera that monitors an area up to 15 m ahead during good visibility – preferably in the day
light with suitable road surface structures and at speeds up to 130 kph. The stereo camera
has an update rate of about 60 ms. Each section of the road can be measured several times
at various angles during vehicle approach, and statistical algorithms are used to provide the
final estimate of road height profile. It has been demonstrated that, under good visibility, the
system performs well, providing superior ride comfort [133] especially when going through
speed bumps and excellent handling during dynamic manoeuvres.
6. Conclusions
A 2 DOF quarter-car model, equipped with an actuator that generates force between the
car body/sprung mass and wheel/unsprung mass, has been widely studied in the literature
and in the process facilitated many useful insights and new concepts (sky hook, ground
hook, invariant points to name a few). Thus, this simple model served well as a base for
the comparison study of various suspension concepts, controller designs, benefits, and limi-
tations. The present survey first discusses the various performance index/cost functions that
are important for the design of a passenger vehicle. It then reviews controller design lim-
itations through the application of optimal control theory to a quarter-car model with an
LQ-based cost function. Figures are generated to illustrate the design limitations and trade-
offs in RMS measures of various vehicle attributes and their associated frequency contents.
These plots should help suspension designers to visualise the ability of active suspensions to
behave differently depending on the current (and forthcoming) road excitations and vehicle
manoeuvre-induced loading (e.g. braking, accelerating, and turning).
Based on more than 40 years of theoretical and practical developments in the industry and
academia, it can be concluded that one of the main challenges for widespread usage of active
suspensions still lies in the area of actuator design and implementation. It is expected that
superior outcome will eventually result from a synergetic interplay between control software
intelligence and ingenious hardware design, all led by model-based system engineering and
further facilitated by relentless developments in electronics, microcontrollers and associated
sensors, and vehicle-to-vehicle and vehicle-to-infrastructure connectivity. Indeed, an impor-
tant part of such an actuator development should include realisation of the desired actuator
force or, alternatively, displacement. As mentioned before, producing and tracking the desired
force (or displacement) is by no means a trivial task, especially when combined with prac-
tical constraints on cost, packaging, weight, and energy consumption. This important aspect
of advanced suspension design should be carefully evaluated through computer simulations,
bench testing, and eventually full vehicle on-road verification.
The above review of current state of the art reveals that the next advances in active
and semi-active suspension design will mainly come from two thrust areas. The first is the
increased efficiency in actuator design and implementation such as the usage of systematic
control software algorithm design combined with the previously mentioned innovative hard-
ware design measures such as inclusion of possibly variable in-series/parallel compliances
and fast load levelling. The second is more comprehensive usage of preview information from
camera, Global Positioning System (GPS), and electronic horizon such as vehicle-to-vehicle
communication, vehicle-to-infrastructure communication, vehicle localisation and real-time
accessing of cloud information, and crowd sourcing leading to up-to-date road profile maps.
Acknowledgement
The authors would like to thank their colleagues Li Xu, Vladimir Ivanovic, and Michael Fodor for their valuable
input and feedback on this survey.
Disclosure statement
No potential conflict of interest was reported by the authors.
References
[1] Galip UA, Peng H, Melih C. Automotive control systems. New York (NY): Cambridge University Press;
2012.
[2] Cao D, Song X, Ahmadian M. Editors’ perspectives: road vehicle suspension design, dynamics, and control.
Veh Syst Dyn. 2011;49(1–2):3–28.
[3] Lu J, Hrovat D, Thomas P, Engleman J, Tseng HE, Filev D. Adaptive crash height adjustment using active
suspensions. US patent 13/736,397. 2013.
[4] Tseng EH, Hrovat D, Lu J, McConnell M, Dehmel M, Seemann M, Owens FV. Suspension control system to
facilitate wheel motions during parking. US Patent 8,825,292. 2014 Sep 2.
[5] Smith MC, Walker GW. Performance limitations and constraints for active and passive suspension: a
mechanical multi-port approach. Veh Syst Dyn. 2000;33(3):137–168.
[6] Hrovat D. A class of active LQG optimal actuators. Automatica. 1982;18(1):117–119.
[7] Hrovat D. Survey of advanced suspension developments and related optimal control applications. Automatica.
1997;33(10):1781–1817.
[8] Karnopp D, Crosby M, Harwood R. Vibration control using semi-active force generators. J Manuf Sci Eng.
1974;96(2):619–626.
[9] Wilson D, Sharp R, Hassan S. The application of linear optimal control theory to the design of active
automobile suspensions. Veh Syst Dyn. 1986;15(2):105–118.
[10] Yue C, Butsuen T, Hedrick JK. Alternative control laws for automotive active suspensions. In American
Control Conference; 1988 June 15–17; Atlanta, GA.
[11] Sharp RS, Peng H. Vehicle dynamics applications of optimal control theory. Veh Syst Dyn. 2011;49(7):1073–
1111.
[12] Smith CC, McGehee DY, Healey AJ. The prediction of passenger riding comfort from acceleration data.
ASME J Dyn Syst Meas Control. 1978;100(1):34–41.
[13] Hedrick JK, Ravera RK, Amderes JR. The effect of elevated guideway construction tolerances on vehicle ride
quality. ASME J Dyn Syst Meas Control. 1975;97(4):408–416.
[14] Fearnsides JJ, Hedrick JK, Firouztash H. Specification of ride quality criteria for transportation systems: the
state of the art and a new approach. High Speed Ground Transp J. 1974;8(2):125–132.
[15] Hoberock LL. A survey of longitudinal acceleration comfort studies in ground transportation vehicles. ASME
J Dyn Syst Meas Control. 1977;99(2):76–84.
[16] Anonymous. ISO: Mechanical vibration and shock: evaluation of human exposure to whole-body vibration.
ISO 2631, International Standard Organization; 1997.
[17] Smith CC. On using the ISO standard to evaluate the ride quality of broad-band vibration spectra in
transportation vehicles. ASME J Dyn Syst Meas Control. 1976;98(4):440–443.
[18] Stephens GD. Comparative vibration environments of transportation vehicles. In: Proceedings of the Design
Engineering Technical Conference. Passenger Vibration in Transportation Vehicles; 1977 Sept 26–28;
Chicago, IL. p. 59–72.
[19] Hrovat D. Applications of optimal control to advanced automotive suspension design. ASME J Dyn Syst
Meas Control. 1993;115(2B):328–342.
[20] Hrovat D. Chapter 31: Active and semi-active suspension control. In: Mastinu G, Plöchl M, editors. Road and
off-road vehicle system dynamics handbook. Boca Raton (FL): CRC Press, Taylor and Francis Group; 2014.
p. 1179–1226.
[21] Boyd S, El Ghaoui L, Feron E, Balakishnan V. Linear matrix inequalities in system and control theory.
Philadelphia (PA): SIAM; 1994.
[22] Fialho I, Balas GJ. Road adaptive active suspension design using linear parameter-varying gain-scheduling.
IEEE Trans Control Syst Technol. 2002;10(1):43–54.
[23] Becker G, Packard A. Robust performance of linear parametrically varying systems using parametrically-
dependent linear feedback. Syst Control Lett. 1994;23(3):205–215.
[24] Chen H, Guo K. An LMI approach to multiobjective RMS gain control for active suspensions. In: Proceedings
of the 2001 American Control Conference; 2001 June 25–27; Arlington, VA. p. 2646–2651.
[25] Tran M, Hrovat D. Application of gain-scheduling to design of active suspensions. In: Proceedings of the
1993 IEEE Conference on Decision and Control. Vol. 2; 1993 Dec 15–17; San Antonio, TX. p. 1030–1035.
[26] Hedrick JK, Butsuen T. Invariant properties of automotive suspensions. Proc Inst Mech Eng D: J Automob
Eng. 1990;204(1):21–27.
[27] Hrovat D, Margolis D. Realistic road-track systems simulation using digital computers. In: Proceedings of the
Winter Simulation Conference; 1975 Dec; Sacramento, CA. p. 481–489.
[28] Hrovat D. Optimal passive vehicle suspensions [Ph.D. thesis]. Davis (CA): University of California;
1979.
[29] Tseng HE, Hedrick JK. Semi-active control laws-optimal and sub-optimal. Veh Syst Dyn. 1994;23(1):545–
569.
[30] Giorgetti N, Bemporad A, Tseng HE, Hrovat D. Hybrid model predictive control application towards optimal
semi-active suspension. Int J Control. 2006;79(5):521–533.
[31] Tomizuka M. Optimum linear preview control with application to vehicle suspension – revisited. ASME J
Dyn Syst Meas Control. 1976;98:309–315.
[32] Louam N, Wilson DA, Sharp RS. Optimal control of a vehicle suspension incorporating the time delay
between front and rear wheel inputs. Veh Syst Dyn. 1988;17(6):317–336.
[33] Louam N, Wilson DA, Sharp RS. Optimization and performance enhancement of active suspensions for
automobiles under preview of the road. Veh Syst Dyn. 1992;21(1):39–63.
[34] Hrovat D. Optimal suspension performance for 2D vehicle models. J Sound Vib. 1991;146(1):
93–110.
[35] Hac A. Optimal linear preview control of active vehicle suspension. Veh Syst Dyn. 1992;21:167–195.
[36] Hac A, Youn I. Optimal semi-active suspension with preview based on a quarter car model. J Vib Acoust.
1992;114(1):84–92.
[37] Hac A, Youn I. Optimal design of active and semi-active suspensions including time delays and preview. J
Vib Acoust. 1993;115(4):498–508.
[38] Van Auken RM. Active suspension control synthesis using reduced order models that account for the time
delay between road inputs. In: ASME 8th Biennial Conference on Engineering Systems Design and Analysis;
2006 July 4–7; Torino, Italy. p. 451–462.
[39] Jankovic M, Kolmanovsky I. Developments in control of time-delay systems for automotive powertrain
applications. In: Balachandran B, Kalmár-Nagy T, Gilsinn DE, editors. Delay Differential Equations. Berlin:
Springer; 2009. p. 55–92.
[40] Thompson AG, Pearce C, Millers E. Direct computation of the performance index for an optimally controlled
active suspension with preview applied to a half-car model. Veh Syst Dyn. 2001;35(2):121–137.
[41] Pilbeam C, Sharp RS. On the preview control of limited bandwidth vehicle suspensions. Proc Inst Mech Eng
D: J Automob Eng. 1993;207(3):185–193.
[42] Prokop G, Sharp RS. Performance enhancement of limited-bandwidth active automotive suspensions by road
preview. IEE Proc Control Theory Appl. 1995;142(2):140–148.
[43] Goran MB, Bachrach BI, Smith RE. The design and development of a broad bandwidth active suspension
concept car. In: XXIV FISITA Congress. Total Vehicle Dynamics. Vol. 2, no. IMECHE C389/418; 1992.
p. 231–252.
[44] Smith MC. Achievable dynamic response for automotive active suspension. Veh Syst Dyn. 1995;24(1):
1–33.
[45] Smith MC, Wang FC. Controller parameterization for disturbance response decoupling: application to vehicle
active suspension control. IEEE Trans Control Syst Technol. 2002;10(3):393–407.
[46] Wang FC, Smith MC. Disturbance response decoupling and achievable performance with application to
vehicle active suspension. Int J Control. 2002;75(12):946–953.
[47] Brezas P, Smith MC. Linear quadratic optimal and risk-sensitive control for vehicle active suspensions. IEEE
Trans Control Syst Technol. 2014;22(2):543–556.
[48] Hrovat D, Hubbard M. Optimum vehicle suspensions minimizing RMS rattlespace, sprung-mass acceleration
and jerk. J Dyn Syst Meas Control. 1981;103(3):228–236.
[49] Davis BR, Thompson AG. Optimal linear active suspensions with integral constraint. Veh Syst Dyn.
1988;17(6):357–366.
[50] Youn I, Im J, Tomizuka M. Level and attitude control of the active suspension system with integral and
derivative action. Veh Syst Dyn. 2006;44(9):659–674.
[51] Karnopp D. Force generation in semi-active suspensions using modulated dissipative elements. Veh Syst Dyn.
1987;16(5–6):333–343.
[52] Lin J, Kanellakopoulos I. Nonlinear design of active suspensions. IEEE Control Syst. 1997;17(3):
45–59.
[53] Chantranuwathana S, Peng H. Adaptive robust force control for vehicle active suspensions. Int J Adapt
Control Signal Process. 2004;18(2):83–102.
[54] Gordon TJ, Marsh C, Wu QH. A learning automaton methodology for control system design in active
vehicle suspensions. In: International Conference on Control ’94. Vol. 1; 1994 Mar 21–24; Coventry, UK.
p. 326–331.
[55] Howell MN, Frost GP, Gordon TJ, Wu QH. Continuous action reinforcement learning applied to vehicle
suspension control. Mechatronics. 1997;7(3):263–276.
[56] Valasek M, Kortum W. Nonlinear control of semi-active road-friendly truck suspension. In: Proceedings of
AVEC. Vol. 98; 1998 Sep 14–18; Nagoya, Japan. p. 275–280.
[57] Gordon TJ, Best MC. Dynamic optimization of nonlinear semi-active suspension controllers. In: International
Conference on Control ’94. Vol. 1; 1994 March 21–24, Coventry, UK. p. 332–337.
[58] Gordon TJ, Sharp RS. On improving the performance of automotive semi-active suspension systems through
road preview. J Sound Vib. 1998;217(1):163–182.
[59] Wang J, Wilson DA. Mixed GL2/H2/GH2 control with pole placement and its application to vehicle
suspension system. Int J Control. 2001;74(13):1353–1369.
[60] Abdellahi E, Mehdi D, M’Saad M. On the design of active suspension system by Hinf and mixed H2/Hinf:
an LMI approach. In: Proceedings of the 2000 American Control Conference; 2000 June 28–30; Chicago, IL.
p. 4041–4045.
[61] Lu J, DePoyster J. Multi-objective optimal suspension control to achieve integrated ride and handling
performance. IEEE Trans Control Syst Technol. 2002;10(6):807–821.
[62] De Jager AG. Comparison of two methods for the design of active suspension systems. Optim Control Appl
Methods. 1991;12(3):173–188.
[63] Yamashita M, Fujimoti K, Uhlik C, Kawatani R, Kumura H. Hinf control of an automotive active suspension.
In: Proceedings of the 29th IEEE Conference on Decision and Control; 1990 Dec 5–7; Honolulu, HI. p. 2244–
2250.
[64] Balas GJ, Doyle JC, Glover K, Packard A, Smith R. Mu-analysis and synthesis toolbox, Matlab user’s guide.
Boston (MA): MathWorks Inc; 1993.
[65] Doyle J, Packard A, Zhou K. Review of LFTs, LMIs, and Mu. In: Proceedings of the 30th IEEE Conference
on Decision and Control. Vol. 2; 1991 Dec 11–13; Brighton, UK. p. 1227–1232.
[66] Gobbi M, Mastinu G. Analytical description and optimization of the dynamic behaviour of passively
suspended road vehicles. J Sound Vib. 2001;245(3):457–481.
[67] Chatillon M, Jezequel L, Coutant P, Baggio P. Hierarchical optimization of the design parameters of a vehicle
suspension system. Veh Syst Dyn. 2006;44(11):817–839.
[68] Chen H, Sun P, Guo K. A multi-objective control design for active suspensions with hard constraints. Proc
Am Control Conf. 2003;5:4371–4376.
[69] Georgiou G, Verros G, Natsiavas S. Multi-objective optimization of quarter-car models with a passive or
semi-active suspension system 45. Veh Syst Dyn. 2007;45(1):77–92.
[70] Crews J, Mattson M, Buckner G. Multi-objective control optimization for semi-active vehicle suspensions. J
Sound Vib. 2011;330(23):5502–5516.
[71] Yi K, Hedrick J. Active and semi-active heavy truck suspensions to reduce pavement damage. SAE Technical
Paper 892486; 1989.
[72] Cole DJ, Cebon D. Truck suspension design to minimize road damage. Proc Inst Mech Eng D: J Automob
Eng. 1996;210(2):95–107.
[73] Kitching KJ, Cole DJ, Cebon D. Theoretical investigation into the use of controllable suspensions to minimize
road damage. Proc Inst Mech Eng D: J Automob Eng. 2000;214(1):13–31.
[74] Cole DJ. Fundamental issues in suspension design for heavy road vehicles. Veh Syst Dyn. 2001;35(4–5):
319–360.
[75] Novak M, Valasek M. A new concept of semi-active control of truck’s suspension. In: Proceedings of
AVEC’96; 1996 June 24–28; Aachen, Germany. p. 141–152.
[76] Valášek M, Novak M, Šika Z, Vaculin O. Extended ground-hook-new concept of semi-active control of truck’s
suspension. Veh Syst Dyn. 1997;27(5–6):289–303.
[77] Valášek M, Kortüm W, Šika Z, Magdolen L, Vaculın O. Development of semi-active road-friendly truck
suspensions. Control Eng Pract. 1998;6(6):735–744.
[78] Vaculın O, Valášek M, Kortüm W. Multi-objective control of semi-active truck suspensions. In: Proceedings
of AVEC’2000; 2000 August 22–24; Ann Arbor, MI. p. 405–412.
[79] Valasek M, Sveda J, Sika Z. Soil-friendly off-road suspension. Veh Syst Dyn. 1998;44(Suppl 1):
479–488.
[80] Valasek M, Kejval J, Maca J, Smilauer V. Bridge-friendly truck suspension. In: Proceedings of the 18th
IAVSD Symposium. Vol. 41; 2003 Aug 25–29; Atsugi, Japan. p. 13–22.
[81] MVO, Valášek KJ. Global chassis control: integration synergy of brake and suspension control for active
safety. In: Proceedings of the International Symposium on Advanced Vehicle Control; 2004 Aug 23–27;
Arnhem, The Netherlands. p. 495–500.
[82] van Zanten A. Control of horizontal vehicle motion. In: Mastinu G, Plöchl M, editors. Road and off-
road vehicle system dynamics handbook. Boca Raton (FL): CRC Press, Taylor and Francis Group; 2014.
p. 1093–1178.
[83] Papageorgiou C, Smith MC. Positive real synthesis using matrix inequalities for mechanical net-
works:application to vehicle suspension. IEEE Trans Control Syst Technol. 2006;14(3):423–435.
[84] Smith MC. Synthesis of mechanical networks: the inerter. IEEE Trans Automat Control. 2002;47(10):
1648–1662.
[85] Karnopp DC, Margolis DL, Rosenberg RC. System dynamics: a unified approach. 3rd/8th ed. New York:
Wiley; 2000 (3rd)/ 2012 (8th).
[86] Margolis D, Wilson R. Bond graph modeling of fluid filled engine mounts. Simulation Series 31: Proceedings
of the 1999 International Conference on Bond Graph Modeling and Simulation (ICBGM ’99): [held as part
of] 1999 Western Multi Conference; 1999 January 17–20; San Francisco, CA. p. 211–216.
[87] Singh R, Adiguna H, Tiwari M, Tseng H, Hrovat D. Transient response evaluation of a hydraulic engine
mount. J Sound Vibration. 2002;36(7):30–35.
[88] Evangelou S, Limebeer DJN, Sharp RS, Smith MC. Control of motorcycle steering instabilities. IEEE Control
Syst Mag. 2006;26(5):78–88.
[89] Smith MC, Wang FC. Performance benefits in passive vehicle suspensions employing inerters. Veh Syst Dyn.
2004;42(4):235–257.
[90] Gillespie TD. Fundamentals of vehicle dynamics. Warrendale (PA): Society of Automotive Engineers;
1992.
[91] Malek KM, Hedrick JK. Decoupled active suspension design for improved automotive ride quality/handling
performance. Veh Syst Dyn. 1986;15(Suppl 1):383–398.
[92] Hayakawa K, Matsumoto K, Yamashita M, Suzuki Y, Fujimori K, Kimura H. Robust H∞-output feed-
back control of decoupled automobile active suspension systems. IEEE Trans Automat Control. 1999;44(2):
392–396.
[93] Venhovens PJT. Optimal control of vehicle suspensions [Ph.D. Thesis]. Delft: Delft University of Technology;
1993.
[94] Best M, Gordon T, Dixon P. An extended adaptive Kalman filter for real-time state estimation of vehicle
handling. Veh Syst Dyn. 2000;34:57–75.
[95] Jeppesen BP, Cebon D. Analytical redundancy techniques for fault detection in an active heavy vehicle
suspension. Veh Syst Dyn. 2004;42(1–2):75–88.
[96] Nastasić Ž, Deák Jahn G. Citroen technical guide [Internet]. Available from: http://www.citroenkerho.fi/xantia
/pdf/tekniikka/Tekniikkaopas.pdf
[97] Chance BK. 1984 Continental mark VII/Lincoln continental electronically-controlled air suspension (EAS)
System. SAE Technical Paper 840342; 1984.
[98] Edmunds D. 2012 Tesla Model S signature performance suspension walkaround [Internet]. 2012 Sep 26 [cited
2015 Mar]. Available from: http://www.edmunds.com/car-reviews/track-tests/2012-tesla-model-s-signature-
performance-suspension-walkaround.html
[99] Korosec K. Potholes and Tesla’s model S: never the twain shall meet [Internet]. 2014 Sep 19 [cited 2015
Mar]. Available from: http://fortune.com/2014/09/19/tesla-model-s-suspension-upgrade/
[100] Kojima H, Nakano J, Nakayama H, Kawashima N, Fujimoto H. Development of new Toyota electronic
modulated suspension-two concepts for semi-active suspension control. SAE Technical Paper 911900;
1991.
[101] Nicolas R. Intelligent suspension system of Lincoln MKZ help mitigating pothole damage [Internet].
2014 Mar 20. Available from: http://www.car-engineer.com/intelligent-suspension-system-lincoln-mkz-help-
mitigating-pothole-damage/
[102] Lincoln Drive Control: CCD. Lincoln Motor Company [Internet]. Available from: http://www.lincoln.com/
cars/mks/features/?featID = 1#page = Feature1
[103] Bodie MO, Hac A. Closed loop yaw control of vehicles using magneto-rheological dampers. SAE Technical
Paper 2000-01-0107; 2000.
[104] Hac A, Bodie MO. Improvements in vehicle handling through integrated control of chassis systems. Int J Veh
Autonomous Syst. 2002;1(1):83–110.
[105] Fodor M, Redfield RC. Resistance control, semi-active damping performance. In: ASME Dynamic Systems
and Control Division, International Mechanical Engineering Congress and Exposition; 1995 Nov. 12–17; San
Francisco, CA. p. 161–169.
[106] Fodor M, Redfield RC. Experimental verification of resistance control, semi-active damping. Veh Syst Dyn.
1996;26(2):143–159.
[107] Sharp RS. Variable geometry active suspension for cars. Comput Control Eng J. 1998;9(5):217–222.
[108] Watanabe Y, Sharp RS. Mechanical and control design of a variable geometry active suspension system. Veh
Syst Dyn. 1999;32(2–3):217–235.
[109] Tumova G. Delft active suspension (DAS-II) [Internet]. Research report OND1281145. Available from:
http://www.narcis.nl/research/RecordID/OND1281145
[110] Venhovens PT, van der Knaap AM. Delft active suspension (DAS) background theory and physical realiza-
tion. In: Pauwelussen JP, Pacejka HB, editors. Smart vehicles. Amsterdam: Swets & Zeitlinger Publishers;
1995. p. 139–165.
[111] van der Knapp A. Design of a low power anti-roll/pitch system for a passenger car. Delft University of
Technology, Vehicle Research Laboratory, Report 89.3VT.2628.
[112] Venhovens P, van der Knaap A, Pacejka H. Semiactive vibration and attitude control. In: Proceedings of the
International Symposium on Advanced Vehicle Control, AVEC ’92; 1992 Sept 14–17; Yokohama, Japan.
p. 170–175.
[113] Evers WJ, Besselink IJM, van der Knaap ACM, Nijmeijer H. Analysis of a variable geometry active sus-
pension. In: Proceeding of International Symposium on Advanced Vehicle Control; 2008 Oct 6–9; Kobe
International Conference Center, Kobe, Japan. p. 350–355.
[114] Packer MB. Active ride control – a logical step from static vehicle attitude control. SAE Technical Paper
780050; 1978.
[115] Crolla DA, Horton DNL, Pitcher RH, Lines JA. Active suspension control for an off-road vehicle. Proc Inst
Mech Eng D: J Automob Eng. 1987;201(1):1–10.
[116] Raad J, Oliver J, Norton R, Kokotovic V. Method of improving response time in a vehicle active tilt control
system. US Patent 5,948,028. 1999.
[117] Raad J, Oliver J, Rhode D, Norton R, Hermann S, Watson J. Method of controlling a vehicle active tilt control
system. US Patent 6,076,027. 2000.
[118] Konik D, Bartz R, Barnthol F, Bruns H, Wimmer M. Dynamic drive – the new active roll stabilization sys-
tem from BMW group – system description and functional improvements. In: Proceeding of International
Symposium on Advanced Vehicle Control; 2000 Aug 22–24; Ann Arbor, MI. p. 413–420.
[119] Hrovat D, Tseng E, Fodor M, Asgari J. Chapter 22: Active and semiactive suspension systems. In: Mastinu
G, Plöchl M, editors. Road and off-road vehicle system dynamics handbook. Boca Raton (FL): CRC Press,
Taylor and Francis Group; 2014. p. 769–796.
[120] Hillebrecht P, Konik D, Pfeil D, Wallentowitz H, Zieglmeier F. The active suspension between customer
benefit and technological competition. In: Proceedings of XXlV FISITA Congress. Total Vehicle Dynamics.
Vol. 2; 1992 June 7–11; London. p. 221–230.
[121] Akatsu Y, Fukushima N, Talcahashj K, Satch M, Kawaranki Y. An active suspension employing an
electrohydraulic pressure control system. SAE Technical Paper 905123; 1990.
[122] Alleyne A, Liu R. On the limitations of force tracking control of hydraulic servosystems. ASME J Dyn Syst
Meas Control. 1999;121(2):184–190.
[123] Zhang Y, Alleyne A. A practical and effective approach to active suspension control. In: Proceedings
of the 6th International Symposium on Advanced Vehicle Control; 2002 Sep 9–13; Hiroshima, Japan.
p. 153–158.
[124] Peng H, Strathearn R, Ulsoy A. A novel active suspension design technique – simulation and experimental
results. In: Proceedings of the 1997 American Control Conference. Vol. 1; 1997 June 4–6; Albuquerque, NM.
p. 709–713.
[125] Merker T, Gaston G, Olaf T. Active Body Control (ABC) the DaimlerChrysler active suspension and damping
system. SAE Technical Paper 2002-21-0054; 2002.
[126] Merker T, Wirtz J, Hill M, Jeglitzka M. Das SL Fahrwerk-Dynamik und Komfort vereint (The SL chassis-
dynamics and comfort combined). ATZ/MTZ Magazine Special Issue; 2001. p. 84–91.
[127] Kruse J. Mercedes-Benz NEXT online magazine [Internet]. 2014 June 4. Available from: http://next.mercedes
-benz.com/en/mbc-eng/
[128] Goran M, Smith R. Insights gained from active suspension development. In: Proceedings of the 26th FISITA
Congress; 1996 June 16–23; Prague, Czech Republic. p. 2486–2514.
[129] Davis RI, Patil PB. Electrically powered active suspension for a vehicle. US Patent 5,060,959; 1991 Oct 29.
[130] Sugasawa F, Kobayashi H, Kakimoto T, Shiraishi Y, Tateishi Y. Electronically controlled shock absorber
system used as a road sensor which utilizes supersonic waves. SAE Technical Paper 851652; 1985.
[131] Streiter IR. Active preview suspension system. ATZ Worldwide. 2008;110(5):4–11.
[132] Rauh J, Ammon D. System dynamics of electrified vehicles: some facts, thoughts, and challenges. Veh Syst
Dyn. 2011;49(7):1005–1020.
[133] Mercedes-Benz S550. Consumer reports; 2014 Sep. p. 53.

State of The Art Survey Active and Semi Active Suspension Control

Uploaded by

Copyright:

Available Formats

State of The Art Survey Active and Semi Active Suspension Control

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

State of The Art Survey Active and Semi Active Suspension Control

Uploaded by

Copyright:

Available Formats

Vehicle System Dynamics

International Journal of Vehicle Mechanics and Mobility

ISSN: 0042-3114 (Print) 1744-5159 (Online) Journal homepage: https://www.tandfonline.com/loi/nvsd20

State of the art survey: active and semi-active

H. Eric Tseng & Davor Hrovat

To link to this article: https://doi.org/10.1080/00423114.2015.1037313

Published online: 14 May 2015.

Submit your article to this journal

Article views: 6012

View related articles

View Crossmark data

Citing articles: 80 View citing articles

Full Terms & Conditions of access and use can be found at

State of the art survey: active and semi-active suspension

(Received 18 February 2015; accepted 31 March 2015 )

Keywords: active suspension; optimal control; semi-active system; suspension control

*Corresponding author. Email: [email protected]

suspensions in terms of the well-established mathematical notion of passive operators, see

2. Optimal suspension control with H2/linear quadratic synthesis/design

2.1. Performance index

2.1.1. Performance index of ride comfort

2.1.2. Performance index for suspension displacement constraint

2.1.3. Performance index for handling/road holding

2.2. H2 optimal control for 2 DOF quarter-car model

2.3. Illustrative example

Figure 4. Normalised ride comfort versus handling.

2.4. Extension to semi-active optimal case

2.5. Extension to preview optimal control

3. Special consideration of handling, loading, and suspension travel limit

4. Other approaches and considerations

5. Production and near production features and systems

5.1. Self-levelling feature

Motor Company’s Electronically Controlled Air Suspension on Lincoln,[97] allowing these

5.2. Variable suspension damping (semi-active suspensions) feature

Production semi-active suspension systems are generally constructed using an adjustable

5.3. Variable suspension geometry (low-power low-bandwidth active suspensions)

5.4. Roll control (semi-active or active) feature

Figure 8. Illustration of a variable geometry suspension concept.

Figure 9. Illustration of delft active suspension concept.

Figure 10. Illustration of delft active suspension realisation Fv cos(γ ) ∗ b = FD ∗ a.

5.5. Active suspensions with full features

5.5.1. Narrow bandwidth active suspensions

5.5.2. BBAS concept implementation at Ford

Figure 12. Ford BBAS concept implementation.

Table 1. Typical parameters used in Ford BBAS.

Ride mode Manoeuvring mode

Parameter Units Front Rear Front Rear

Skyhook damping Lb-s/in 23.5 10.0 23.5 10.0

You might also like