General Theory of Skyhook Control and Its Application To Semi-Active Suspension Control Strategy Design
General Theory of Skyhook Control and Its Application To Semi-Active Suspension Control Strategy Design
General Theory of Skyhook Control and Its Application To Semi-Active Suspension Control Strategy Design
ABSTRACT A novel general theory of skyhook control is proposed and applied to the semi-active suspension
control strategy design to improve the performance of the vehicle suspension system. Based on this theory,
the mechanical impedance model of the general theory of skyhook suspension is established. To design the
suspension structure, the effect of the skyhook element and its parameters on suspension is analyzed. Then,
adaptive fish swarm algorithm based on nonlinear dynamic visual field is used to optimize the parameters
of the general theory of skyhook control. To realize the general theory of skyhook control and verify it,
a novel controllable inerter is designed and utilized into the semi-active suspension system. The simulation
results demonstrate that the semi-active suspension with a general theory of skyhook control can enhance
the suspension performance. Finally, the robustness of the general theory of skyhook control under different
spring stiffness and sprung mass is researched. The results indicate that the suspension with the general
theory of skyhook control has superior performance and robustness compared with the traditional skyhook
damper controlled suspension and passive suspension.
INDEX TERMS Semi-active suspension, skyhook control, mechanical impedance, controllable inerter.
I. INTRODUCTION those two design approaches, are all obtained a fixed structure
The function of vehicle suspension is to improve the ride and parameters of the suspension. Normally it cannot gain the
comfort and driving safety. Today the most widely used sus- satisfying performance in all of those three indexes or must
pension is passive suspension. It consists of spring, damper, rely on a complex structure [7], [8]. Semi-active suspension
and inerter [1]–[3]. A common approach used to design can adjust suspension parameters according to the control
the passive suspension is structure approach [4], [5], which strategy to achieve satisfying performance in different con-
bases on the passive elements (spring, damper, and inerter) dition of the vehicle and the road [9]–[11]. Skyhook (SH)
connected in parallel or in series. This approach can limit the control suspension is a semi-active suspension which is easy
complexity of the suspension structure. Whereas, some excel- to implement with little information about the vehicle state.
lent structures maybe be ignored as a consequence. To solve It sets a virtually damper between the vehicle body and the
this problem, in [1], [6], mechanical impedance approach was imaginary sky [12]. It is effective to enhance the ride comfort
used into suspension design. In this approach, a mechanical of the vehicle, but the dynamic tire load deteriorates at the
impedance function is established first. Then, the parameters same time. In [13], a solution was proposed to improve
of mechanical impedance function are optimized based on the performance of the skyhook control strategy by adding
the design objective. After that, network synthesis is used to the sliding mode and internal model theory. In term of the
realize the suspension structure passively. It is a more general performance of suspension, this solution is more superior
approach compare with the structure approach. However, to the traditional skyhook controller. In [14], Hu and Chen
designed a comfort-oriented vehicle suspension with skyhook
The associate editor coordinating the review of this manuscript and inerter. Semi-active inerter was used to realize the semi-
approving it for publication was Luigi Biagiotti. active skyhook inerter control and three different control laws
101552 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/ VOLUME 7, 2019
C. Liu et al.: General Theory of Skyhook Control and Its Application to Semi-Active Suspension Control Strategy Design
X2 2 T (s)kt s3
Hẍ2 ∼xr (s) = s = (4)
Xr (m1 m2 s + m1 s (T (s) + Y (s)) + m2 s T (s) + T (s)Y (s)s2 + kt m2 s2 + kt s(T (s) + Y (s))
4 3 3
Select the grade-A road as input model, the pavement In which, k1 is the skyhook spring stiffness, c1 is the
spectrum expressed by time frequency is skyhook damping coefficient, b1 is the skyhook inerter coef-
ficient. In special, when the Y (s) = Y2 (s), GSH sus-
G0 up−1
S(f ) = (7) pension is the traditional skyhook damper suspension. The
fp GSH suspension contains the traditional skyhook damper
In which, the road roughness coefficient G0 = 5 × suspension.
10−6 m3 /cycle., the slope of the spectral density curve in Select the small passenger car as the reference model. The
double logarithmic coordinates p = 2.5, u is the speed, f parameters of the quarter-car kinetic model in Fig. 1 (b) are
is the time frequency. Setting s = j2π f , the power spectral manifested in Table 1. The vehicle speed was set as 20 m/s.
density of the body acceleration, suspension working space, The time frequency f = 0.01.
and dynamic tire load are as follows:
2 TABLE 1. Suspension parameters.
Sẍ2 (f ) = Hẍ2 ∼xr (j2π f ) S(f ) (8)
2
Sx2 −x1 (f ) = H(x2 −x1 )∼xr (j2π f ) S(f ) (9)
2
S(x1 −xr )kt (f ) = H(x1 −xr )kt ∼xr (j2π f ) S(f )
(10)
The root-mean-square values are
sZ
∞
BA = Sẍ2 (f )df (11)
0
sZ
∞
SWS = Sx2 −x1 (f )df (12)
0 For the purpose of analyzing the effect of the one parameter
sZ
∞ of the skyhook element, suspension performance was studied
DTL = S(x1 −xr )kt (f )df (13) with changes parameters in skyhook element. The lower and
0
upper bounds of the parameters are demonstrated in Table 2.
B. THE EFFECT OF STRUCTURE AND PARAMETERS ON In which, the range of the parameter is large enough to design
GSH SUSPENSION the vehicle suspension. And the trends of the suspension
To GSH suspension, the best performance of suspension can performance with the change of the parameters of skyhook
be obtained by the appropriate structure and parameters. The element are demonstrated in Fig. 2.
high-order skyhook control strategy is always accompanied
with active suspension which stands for the high cost and TABLE 2. Range of optimized parameters.
FIGURE 2. The effect of suspension performance with changes of the parameters in skyhook element.
space to some extent. When the damping coefficient c1 is C. SUSPENSION PARAMETERS OPTIMIZED BY ADAPTIVE
less than 2500 N · s/m, the RMS value of dynamic tire load FISH SWARM ALGORITHM BASED ON NONLINEAR
will be reduced slightly. But when the damping coefficient DYNAMIC VISUAL FIELD
is more than 2500 N · s/m, the RMS value of dynamic Selecting the vehicle body acceleration RMS value BA,
tire load increases a lot. And with the increase of the sky- the dynamic tire load RMS value DTL, and the suspension
hook inerter coefficient, the RMS value of body acceleration working space RMS value SWS as the objective of ride com-
and the suspension working space plunges. Meanwhile, the fort, tire grounding, and body attitude separately to design
dynamic tire load soars significantly. The RMS value of body the suspension. Arranging related objective functions to find
acceleration plunges when the skyhook inerter coefficient b1 suspension parameters that focus on improving ride comfort
increases but less than 1000 kg. However, the suspension when the RMS value of the suspension working space and
working space and dynamic tire load deteriorate along with dynamic tire load is not greater than the traditional suspension
the variation at the same time. So, the skyhook inerter obviously. On account of the changing of those three indexes
coefficient b1 can maintain only a small value. Because of are different, especially the improvement of one index will
unsatisfied performance of the skyhook spring and the key cause the deterioration of other two indexes, the comprehen-
role of body acceleration in the passenger-vehicle suspension, sive evaluation index of suspension performance needs to
Y (s) should contain the damper characteristic and inerter be established. Because of the different units and orders of
characteristic. The Y4 (s) is set as c2 + b2 s, in which, c2 is the magnitude for body acceleration, suspension working space,
skyhook damping coefficient and b2 is the skyhook inerter and dynamic tire load, performance indexes of suspension
coefficient. are divided by the corresponding performance indexes of
passive suspension. The indexes of passive suspension are TABLE 3. Range of simulation parameters.
constant values because of the fixed parameters. This method
can normalize the data of objective function. Then, they are
combined by linear combination method to obtain a unified
objective function:
BA SWS DTL
min F(x) = w1 + w2 + w3
BA0 SWS0 DTL0
⇔ max −F(x)
s.t LB ≤ xi ≤ UB (17)
must measure the distance to the current optimal artificial fish
where F(x) is the fitness function of the adaptive fish swarm and use it as its own visual field. The evolution of the objec-
algorithm, w1 , w2 , and w3 are the weighting coefficients and tive function as demonstrated in Fig. 3. The optimization
set as 0.45, 0.35, 0.2. BA0 , SWS0 , and DTL0 are the RMS results of parameters in general theory of skyhook control are
of vehicle body acceleration, suspension working space, as follows: skyhook damping coefficient c2 is 1897 N · s/m,
dynamic tire load of the traditional suspension. xi is decision skyhook inerter coefficient b2 is 206 kg.
variables. xi = [c2 , b2 ]. UB is the upper bound value and LB
is the lower bound value.
The skyhook damping coefficient c2 and the skyhook
inerter coefficient b2 are optimized by adaptive fish swarm
algorithm based on nonlinear dynamic visual field. And the
ideal general skyhook should be realized by semi-active
control strategy, which contains controllable damper and
controllable inerter. To avoid the suspension damper c having
an adverse effect on the optimization progress, select
T2 (s) = k/s.
The fish swarm algorithm has strong global search ability.
For the optimization of suspension parameters, the fish FIGURE 3. Evolution of the objective function.
swarm algorithm can avoid local extrema and is not sensitive
to the initial value. It is a parallel search method with high
efficiency [18], [19]. However, the fixed visual field will lead III. PERFORMANCE EVALUATION OF SEMI-ACTIVE
the poor global search capability or poor local search ability. SUSPENSION SYSTEM WITH GSH CONTROL STRATEGY
This paper designs an adaptive fish swarm algorithm based A. QUARTER-CAR KINETIC MODEL OF
on nonlinear dynamic visual field. It can maintain a large SEMI-ACTIVE SUSPENSION
visual field in the early stage to ensure the strong global The ideal GSH suspension cannot be applied directly to
search capability. As the search progresses, the visual field engineering because the vehicle cannot be fixed to the sky-
can adjust and the algorithm evolves into a local search, hook. So the semi-active control was utilized to achieve
which effectively guarantees the accuracy and convergence. the functions of the ideal GSH suspension. For testing the
The visual field and moving step can be adjusted according performance of the suspension, the quarter-car kinetic model
to (18). of semi-active suspension based on the theory of general
skyhook control was established.
Visual = Visual × a + Visualmin
As demonstrated in Fig. 4, the quarter-car kinetic model
Step = Step × a + Stepmin (18) of semi-active suspension consists of the sprung mass,
s
a = exp(−30 × (t/Tmax ) ) the unsprng mass, suspension spring, tire stiffness equiva-
lent spring, a controllable damper and a controllable inerter.
In which, s is an integer which is greater than 1, t is the
The suspension spring, controllable damper, and controllable
current number of iterations.
inerter are paralleling between the sprung mass and the
The parameters range of the skyhook damping coeffi-
unsprng mass.
cient c2 and the skyhook inerter coefficient b2 are demon-
The kinetic equitation is
strated in Table 3.
The vehicle speed was set as 20 m/s on a road with an m1 ẍ1 + kt (x1 − xr ) − k(x2 − x1 )
unevenness coefficient of G0 = 5 × 10−6 m3 /cycle. Speed
−bctrl (ẍ2 − ẍ1 ) − cctrl (ẋ2 − ẋ1 ) = 0
white noise was selected as the road surface input. The artifi- (19)
m 2 ẍ 2 + k(x 2 − x1 ) + b ctrl (ẍ 2 − ẍ 1 )
cial fish, with a population size of 100, the maximum number
of iterations was set to 50, the maximum number of trials +cctrl (ẋ2 − ẋ1 ) = 0
was 100, the initial visual field was 2.5, the congestion factor In which, m2 is the sprung mass, m1 is the unsprung mass,
was 0.618 and the moving step was 0.3. Each artificial fish kt is the stiffness of the tire, x2 is the displacement of the
In this test, the vehicle speed u was 20 m/s and the road working space within a reasonable range, the suspension can
roughness coefficient G0 was ×10−6 m3 /cycle. To ana- absorb road vibration and avoid vibration transmit to the
lyze the performance of those GSH controlled suspension, vehicle body. Therefore, the suspension with GSH control
the traditional spring-damper structure suspension and the has better performance of body acceleration and dynamic
traditional skyhook damper control suspension (SH control tire load compared with the passive suspension and suspen-
suspension) were used as the comparative target and the sion with SH control. To quantify the performance of the
random road information was taken as input of it. The param- suspension with GSH control, the RMS value of the three
eters of the traditional suspension are k0 = 22000 N/m, performance indexes and the percentage of the improve-
c0 = 1200 N·s/m. The skyhook damper coefficient of the SH ment comparing with the passive suspension are recorded
control suspension csky0 = 1897 N · s/m. Other parameters in Table 4. The positive variation number means the value
are same as the traditional suspension. It is a special situation is increased, and the performance is deteriorated. The neg-
of GSH control suspension when the bctrl is set to 0. The ative variation number means the value is decreased, and
system output power spectral density of random response the performance is enhanced. Clearly, the suspension with
manifested in Fig. 6. The RMS values of the three indexes SH control improves the performance of body acceleration
were demonstrated in Table 4. (−5.6%) at the cost of deteriorated performance in dynamic
tire load (+7.9%). Whereas, the GSH controlled suspension
avoids this disadvantage. The body acceleration decreases
approximately 45.4% and the dynamic tire load decreases
approximately 4.2%. The Fig. 6 and Table 4 indicate that the
semi-active suspension based on the GSH control improves
the suspension performance significantly.
ACKNOWLEDGMENT
In general, the GSH controlled suspension has superior The authors would like to thank the associate editor and the
performance and robustness in different working conditions. anonymous reviewers for their careful reading and helpful
With the changing of the spring stiffness and the sprung mass, comments.
the GSH controlled suspension has better robustness than the
passive suspension and the SH controlled suspension. REFERENCES
[1] M. C. Smith, ‘‘Synthesis of mechanical networks: The inerter,’’ IEEE
IV. CONCLUSION Trans. Autom. Control, vol. 47, no. 10, pp. 1648–1662, Oct. 2002.
This paper proposed a novel general theory of skyhook [2] J.-Y. Li and S. Zhu, ‘‘Versatile behaviors of electromagnetic shunt damper
control and designed a novel controllable inerter which with a negative impedance converter,’’ IEEE/ASME Trans. Mechatronics,
vol. 23, no. 3, pp. 1415–1424, Jun. 2018.
can realize the continuous skyhook inerter control strategy.
[3] X. Wang, X. Liu, Y. Shan, Y. Shen, and T. He, ‘‘Analysis and optimization
It promoted the special theory of skyhook control at present. of the novel inerter-based dynamic vibration absorbers,’’ IEEE Access,
The testing results demonstrated that the suspension applied vol. 6, pp. 33169–33182, 2018.
[4] F. Scheibe and M. C. Smith, ‘‘Analytical solutions for optimal ride comfort CHANGNING LIU is currently pursuing the Ph.D.
and tyre grip for passive vehicle suspensions,’’ Vehicle Syst. Dyn., vol. 47, degree with the School of Automotive and Traffic
no. 10, pp. 1229–1252, Oct. 2009. Engineering, Jiangsu University, China. His main
[5] Y. Hu, M. Z. Q. Chen, Z. Shu, and L. Huang, ‘‘Analysis and optimisation research interests include the dynamic model-
for inerter-based isolators via fixed-point theory and algebraic solution,’’ ing and control of vehicle suspension, especially
J. Sound Vib., vol. 346, pp. 17–36, Jun. 2015. the structure design, performance evaluation, and
[6] C. Papageorgiou and M. C. Smith, ‘‘Positive real synthesis using matrix active/semi-active control of the ISD (inerter-
inequalities for mechanical networks: Application to vehicle suspension,’’
spring-damper) suspension.
IEEE Trans. Control Syst. Technol., vol. 14, no. 3, pp. 423–435, May 2006.
[7] M. Z. Q. Chen, K. Wang, Y. Zou, and G. Chen, ‘‘Realization of three-port
spring networks with inerter for effective mechanical control,’’ IEEE Trans.
Autom. Control, vol. 60, no. 10, pp. 2722–2727, Oct. 2015.
[8] K. Wang, M. Z. Q. Chen, and Y. Hu, ‘‘Synthesis of biquadratic impedances
with at most four passive elements,’’ J. Franklin Inst., vol. 351, no. 3,
pp. 1251–1267, Mar. 2014. LONG CHEN is currently a Professor with the
[9] X. Ma, P. K. Wong, J. Zhao, J.-H. Zhong, H. Ying, and X. Xu, ‘‘Design and Research Institute of Automotive Engineering,
testing of a nonlinear model predictive controller for ride height control Jiangsu University, China. His main research inter-
of automotive semi-Active air suspension systems,’’ IEEE Access, vol. 6,
ests include the dynamic modeling and control of
pp. 63777–63793, 2018.
vehicle engineering.
[10] X. Tang, H. Du, S. Sun, D. Ning, Z. Xing, and W. Li, ‘‘Takagi–Sugeno
fuzzy control for semi-active vehicle suspension with a magnetorheologi-
cal damper and experimental validation,’’ IEEE/ASME Trans. Mechatron-
ics, vol. 22, no. 1, pp. 291–300, Feb. 2017.
[11] M. Z. Q. Chen, Y. Hu, C. Li, and G. Chen, ‘‘Application of semi-Active
inerter in semi-active suspensions via force tracking,’’ J. Vib. Acoust.,
vol. 138, no. 4, pp. 1–11, Aug. 2016.
[12] D. Karnopp, M. J. Crosby, and R. A. Harwood, ‘‘Vibration control using
semi-Active force generators,’’ J. Eng. Ind., vol. 96, no. 2, pp. 619–626,
May 1974.
XIAOFENG YANG is currently an Associate
[13] A. Jayabalan and N. K. S. Kumar, ‘‘Vibration suppression of quarter car
Professor with the School of Automotive and
using sliding-mode and internal model-based skyhook controller,’’ J. Vib.
Eng. Technol., vol. 6, no. 2, pp. 117–126, Jun. 2018. Traffic Engineering, Jiangsu University, China.
[14] Y. Hu, M. Z. Q. Chen, and Y. Sun, ‘‘Comfort-oriented vehicle suspen- His main research interests include the dynamic
sion design with skyhook inerter configuration,’’ J. Sound Vib., vol. 405, modeling and control of vehicle engineering.
pp. 34–47, Sep. 2017.
[15] M. C. Smith and F.-C. Wang, ‘‘Performance benefits in passive vehicle
suspensions employing inerters,’’ in Proc. 42nd IEEE Int. Conf. Decis.
Control, Dec. 2003, pp. 2258–2263.
[16] M. Z. Q. Chen and M. C. Smith, ‘‘Restricted complexity network realiza-
tions for passive mechanical control,’’ IEEE Trans. Autom. Control, vol. 54,
no. 10, pp. 2290–2301, Oct. 2009.
[17] L. Chen, C. Liu, W. Liu, J. Nie, Y. Shen, and G. Chen, ‘‘Network synthesis
and parameter optimization for vehicle suspension with inerter,’’ Adv.
XIAOLIANG ZHANG is currently a Professor
Mech. Eng., vol. 9, no. 1, pp. 1–7, Jan. 2016.
[18] W. Zhao, C. Du, and S. Jiang, ‘‘An adaptive multiscale approach for iden- with the Research Institute of Automotive Engi-
tifying multiple flaws based on XFEM and a discrete artificial fish swarm neering, Jiangsu University, China. His main
algorithm,’’ Comput. Methods Appl. Mech. Eng., vol. 339, pp. 341–357, research interests include the dynamic modeling
Sep. 2018. and control of vehicle engineering.
[19] A. M. A. C. Rocha, M. F. P. Costa, and E. M. G. P. Fernandes, ‘‘A shifted
hyperbolic augmented Lagrangian-based artificial fish two-swarm algo-
rithm with guaranteed convergence for constrained global optimization,’’
Eng. Optim., vol. 48, no. 12, pp. 2114–2140, Apr. 2016.
[20] A. C. Zolotas and R. M. Goodall, ‘‘New insights from fractional order
skyhook damping control for railway vehicles,’’ Vehicle Syst. Dyn., vol. 56,
no. 11, pp. 1–24, Feb. 2018.
[21] C.-Y. Hsieh, B. Huang, F. Golnaraghi, and M. Moallem, ‘‘Regenera-
tive skyhook control for an electromechanical suspension system using
a switch-mode rectifier,’’ IEEE Trans. Veh. Technol., vol. 65, no. 12,
YI YANG is currently pursuing the Ph.D. degree
pp. 9642–9650, Dec. 2016. with the School of Automotive and Traffic Engi-
[22] S. B. A. Kashem, M. Ektesabi, and R. Nagarajah, ‘‘Comparison between neering, Jiangsu University, China. Her main
different sets of suspension parameters and introduction of new modified research interests include the dynamic model-
skyhook control strategy incorporating varying road condition,’’ Vehicle ing and control of vehicle suspension, especially
Syst. Dyn., vol. 50, no. 7, pp. 1173–1190, Jul. 2012. the structure design, performance evaluation, and
[23] A. Shamsi and N. Choupani, ‘‘Continuous and discontinuous shock active/semi-active control of the ISD (inerter-
absorber control through skyhook strategy in semi-active suspension sys- spring-damper) suspension.
tem (4DOF model,’’ Int. J. Mech., Ind. Aerosp. Eng., vol. 2, no. 5,
pp. 697–701, 2008.