Article

Integrated Comfort-adaptive Cruise and Semi-active

Suspension Control for an Autonomous Vehicle: An
LPV Approach
Gia Quoc Bao Tran 1 , Thanh Phong Pham 2 * , Olivier Sename 1 , Eduarda Costa 1 , and Peter
Gaspar 3
1 Univ. Grenoble Alpes, CNRS, Grenoble INP⊥ , GIPSA-lab, F-3800 Grenoble, France
⊥ Institute
of Engineering Univ. Grenoble Alpes
2 Faculty of Electrical and Electronic Engineering, The University of Danang - University of Technology and
Education, 550000 Danang, Vietnam
3 Systems and Control Laboratory, Institute for Computer Science and Control, Hungarian Academy of
Sciences, Kende u. 13-17, H-1111 Budapest, Hungary
* Correspondence: ptphong@ute.udn.vn

Version February 20, 2021 submitted to Electronics

1 Abstract: This paper presents an integrated linear parameter-varying (LPV) control approach of
2 an autonomous vehicle with an objective to guarantee driving comfort, consisting of cruise and
3 semi-active suspension control. First, the vehicle longitudinal and vertical dynamics (equipped with
4 a semi-active suspension system) are presented and written into LPV state-space representations.
5 The reference speed is calculated online from the estimated road type and the desired comfort level
6 (characterized by the frequency weighted vertical acceleration defined in the ISO 2631 norm) using
7 precomputed polynomial functions. Then, concerning cruise control, an LPV H2 controller using
8 an LMI-based polytopic approach combined with the compensation of the estimated disturbance
9 forces is developed to track the comfort-oriented reference speed. To further enhance passengers’
10 comfort, a decentralized LPV H2 controller for the semi-active suspension system is proposed,
11 minimizing the effect of the road profile variations. The interaction with cruise control is achieved
12 by the vehicle’s actual speed being a scheduling parameter for suspension control. To assess the
13 strategy’s performance, simulations are conducted using a realistic nonlinear vehicle model validated
14 from experimental data. The simulation results demonstrate the proposed approach’s capability to
15 improve driving comfort.

16 Keywords: autonomous vehicle; advanced driver-assistance system; LPV approach; cruise control;
17 semi-active suspension control; passenger comfort

18 1. Introduction
19 Autonomous vehicles always remain an interesting research topic thanks to their numerous
20 advantages, including collision avoidance and fuel consumption reduction capabilities, satisfying
21 traffic safety and environmental objectives.
22 There has been a considerable amount of research work conducted on either cruise or suspension
23 control of autonomous vehicles. Cruise control refers to the control of the vehicle speed, which is
24 related to longitudinal dynamics, for multiple purposes such as collision avoidance [1,2]. For this,
25 different control strategies (optimal, robust, LPV, etc.) have been proposed [3–7]. Recently, cruise
26 control has been linked to a comfort objective [8–10], which opens the field to the coordination between
27 longitudinal and vertical controllers.

Submitted to Electronics, pages 1 – 19 www.mdpi.com/journal/electronics

 Version February 20, 2021 submitted to Electronics 2 of 19

28 Indeed, the suspension system is a key subsystem that allows us to improve driving comfort
29 and road holding performance of the vehicle [11,12]. It has a remarkable ability to limit the vertical
30 oscillations of the vehicle body caused by road displacements at the four wheels. From recent years, it
31 is known that the semi-active suspension system provides better performance than the passive one
32 while being less energy-consuming than the active one [12]. Existing work on semi-active suspension
33 control includes model predictive control or state-feedback with various observers, from robust, LPV
34 to unified ones [11–14].
35 However, there has not been much work combining cruise and suspension control into an
36 integrated problem, considering their interaction. Besides, very few studies do consider the driving
37 comfort level in a cruise control problem. To improve driving comfort, a potential strategy is to relate
38 the speeds at which the vehicle should travel to the desired comfort level and w.r.t specific road profiles.
39 Such speed values are determined using criteria formed by examining the human body, including
40 which range of frequency is most absorbed by humans. Our group has conducted a study [15] into
41 relating the vehicle speed with the comfort level measured using the ISO 2631 standard [16] and the
42 international roughness level (IRI) [17] for each road type from A to D (defined in [18]). Recent research
43 about road profile estimation using adaptive observers allows us to detect which road type the vehicle
44 is traveling on [19], thus enabling this strategy.
45 The purpose of this paper is to bring further results and to introduce a comfort-oriented strategy
46 of integrated cruise-suspension control of an autonomous vehicle. There has been some existing
47 work combining these problems [9,20], that either often requires too many assumptions and much
48 information from the environment (therefore being challenging to embed in reality) or does not
49 integrate a comfort-oriented relationship between the longitudinal and vertical motions. This work
50 proposes a more realistic approach, handling unknown inputs using a robust LPV control approach. We
51 analyze both longitudinal and vertical dynamics and their interaction through the road displacement
52 at each of the four wheels. The H2 condition is used as it is suited for the type of noise we are faced
53 with in this suspension control case where one of the sensors is an accelerometer. For cases where
54 the variation of a specific parameter(s) significantly affects the system, we model the parameter(s)
55 into an LPV problem, which is solved as a set of linear matrix inequalities (LMIs). We also show
56 how driving comfort is evaluated through the measurement of the vertical acceleration transmitted
57 to passengers, from which we propose a way to relate the current speed to comfort level using the
58 ISO 2631 standard. This allows us to determine which speed the vehicle should travel at in order to
59 guarantee that the acceleration felt by one passenger does not exceed a predefined value. Combining
60 the cruise and suspension controllers with a comfort-oriented reference speed generation leads to the
61 proposed integrated comfort-oriented vehicle control. The integrated control scheme is then tested
62 using simulations on a realistic nonlinear vehicle model validated from experimental data.
63 This paper is organized as follows. In Section 2, we present the vehicle longitudinal and vertical
64 dynamics (quarter-car model) then the integrated dynamics model. The general scheme of the strategy
65 is presented in Section 3, which consists of comfort-guaranteeing speed calculation (described in detail
66 in Section 4) and integrated cruise-suspension control (discussed in Section 5). Finally, simulation
67 results are presented in Section 6, which shows the effectiveness of our strategy.

68 2. Vehicle dynamics modeling in LPV form

69 This section introduces and discusses the vehicle longitudinal and vertical dynamics considered
70 in this paper. The integrated full vehicle simulation model is also briefly presented.

71 2.1. Longitudinal dynamics

72 In this part, we present the vehicle longitudinal dynamics and the corresponding system’s LPV
73 state-space representation. First, let us make some assumptions for the longitudinal dynamics system:
74 • The vehicle mass is considered to be time-varying. It is measured online thanks to multiple
75 built-in sensors that detect the additional load (the mass of the empty vehicle is the known
 Version February 20, 2021 submitted to Electronics 3 of 19

76 nominal mass). This is the most crucial assumption as it allows for gain-scheduling based on
77 mass. The vehicle speed is also directly measurable using a speedometer.
78 • The road slope is known/estimated in real-time thanks to algorithms such as in [21–23]. Such an
79 assumption allows us to implement road slope compensation using a feed-forward term in the
80 cruise control input.

81 Suppose we have a vehicle of mass m traveling at the speed of v, as shown in Figure 1. Let F be
82 the longitudinal control force on the vehicle, and Fd the total disturbance force.

Figure 1. Longitudinal forces on the vehicle.

We have the following equation of motion [3]

mv̇ = F − Fd . (1)

The disturbance force consists of three components: the rolling friction supposed to have a
constant value, the drag by gravity supposing the road’s slope θ to be sufficiently small (between ±10◦
which is a realistic assumption for real roads), and the aerodynamic drag that adds nonlinearity to the
system, respectively
Fr = mgCr cos(θ ) ≈ mgCr , (2)

Fg = mgsin(θ ) ≈ mgθ, (3)

1
Fa = Cv Da Sv2 , (4)
2
where Cr is the rolling friction coefficient, Cv is the aerodynamic drag coefficient, Da is the air’s density,
and S denotes the vehicle’s frontal area. The disturbance force thus has the following equation

1
Fd = mgCr + mgθ + Cv Da Sv2 . (5)
2
Finally, the vehicle’s motion equation is formulated as

1
mv̇ = F − mgCr − mgθ − Cv Da Sv2 . (6)
2
 Version February 20, 2021 submitted to Electronics 4 of 19

The input force F is composed of two parts

F = Ff f + Fl , (7)

83 where Ff f = mgĈr + mgθ̂ is the feed-forward term that compensates for the rolling friction and the
84 road slope and Fl is the longitudinal control force. Here Ĉr is an estimated nominal value for Cr
85 (constant) and θ̂ is the road slope estimated in real-time by the methods in [21–23]. Because this
86 compensation is not perfect, i.e., Ĉr 6= Cr and θ̂ 6= θ, we model all this imperfection by replacing θ with
87 ∆wl where ∆ is a constant bound and |wl | ≤ 1 is noise.
The system is then written in LPV form, with xl = v being the state variable, ul = Fl being the
h i> h i>
cruise control input, yl = v being the measured output, and ρl = ρl1 ρl2 = 1/m v being the
varying parameter of the longitudinal control case, as
(
ẋl = Al (ρl ) xl + Bl1 wl + Bl2 (ρl )ul
Σl (ρl ) : (8)
yl = Cl xl ,

where
Al (ρl ) = [− 12 Cv Da Sρl1 ρl2 ], Bl1 = [− g∆], Bl2 (ρl ) = [ρl1 ], Cl = [1].

88 2.2. Vertical dynamics

89 The suspension control design is carried out using the quarter-car suspension system [11]. Indeed,
90 this model is simple enough to catch the comfort objective w.r.t the bounce motion and to cope with
91 the requirements about reducing the complexity of an embedded controller. For pitch/roll control, a
92 full vehicle model would be needed, which is not the case here.
93 We use the quarter-car model with a semi-active Magneto-Rheological (MR) suspension system
94 to model the vehicle vertical dynamics, as shown in Figure 2. This consists of the sprung mass ms , the
95 unsprung mass mus , and the suspension components positioned between them including a spring
96 with stiffness k s and the damper part. Let us denote zs and zus as the sprung and unsprung masses’
97 displacements, respectively.

Figure 2. The quarter-car model for illustration of vehicle vertical dynamics.

 Version February 20, 2021 submitted to Electronics 5 of 19

98 In Figure 3, the considered MR damper force - deflection velocity (żde f = żs − żus ) characteristic
99 is shown, from the MR damper available at ITESM, Mexico (refer to [24]).

Figure 3. MR damper force - deflection velocity characteristic.

From Newton’s second law of motion, we obtain

(
ms z̈s = − Fspring − Fdamper
(9)
mus z̈us = Fspring + Fdamper − Ftire ,

where Fspring = k s (zs − zus ) is the spring force, Ftire = k t (zus − zr ) is the tire force; the damper force
Fdamper is defined as
Fdamper = k0 (zs − zus ) + c0 (żs − żus ) + Fv , (10)

100 where c0 and k0 are constant parameters. We take Fv as the input for suspension control.
To link with longitudinal dynamics, we here consider benefiting from some knowledge of the
road displacement input model zr , which is related to the current vehicle speed according to [25] as

żr + a · v · zr = b · v · wv , (11)

101 where wv is white noise, a and b are coefficients that depend on the road type according to International
102 Organization for Standardization (ISO) classification [18].
103 Remark: Using a road profile model is indeed possible since the information on the type of
104 road profile may be obtained using some adaptive road profile estimator, as proposed in [19], or a
105 frequency-wise approach [25].
h i>
From (9) and (11), by selecting the system states as xl = zs żs zus żus zr ∈ R5 , the
h i>
measured variables yl = z̈s zs − zus ∈ R2 , the control input ul = Fl , and by choosing the
scheduling variable ρv = v to link with longitudinal dynamics, the extended quarter-car system can be
written in LPV form as
(
ẋv = Av (ρv ) xv + Bv1 (ρv )wv + Bv2 uv
Σv (ρv ) : (12)
yv = Cv2 xv + Dv21 wv + Dv22 uv ,
 Version February 20, 2021 submitted to Electronics 6 of 19

where
     
0 1 0 0 0 0 0
− k
 ms − mc0s k
ms
c0
ms 0 

 0 
 
− 1 
 ms 
Av (ρv ) =  0 0 0 1 0 , Bv1 (ρv ) =  0  , Bv2 =  0 ,
     
c0
− km+uskt − mcus kt
 k 0
    1 
m mus mus
  0  m 
us us
0 0 0 − a · ρ"v # b · ρv 0
" # 0 " #
− mks − mc0s k
ms
c0
ms 0 0 − m1s
Cv2 = , Dv21 = , Dv22 = ,
1 0 −1 0 0 0 0

106 with k = k s + k0 . In this work, we take the coefficients a and b as coherent with those of a road of type
107 B in [18].

108 2.3. Full-car dynamics

109 In this paper, the full vehicle model presented in [26,27] is used for simulation and validation
110 purposes. This model and its parameters have been validated on a real Renault Megane vehicle1 . For
111 illustration, the model is presented in Figure 4, but the interested readers should refer to [26] for more
112 details.

Figure 4. Full vehicle model.

1 Thanks to M. Basset, from the MIAM research team.

 Version February 20, 2021 submitted to Electronics 7 of 19

113 Note that the main interest in using the full nonlinear vehicle model is that it allows us to consider
114 nonlinear load transfer, fast nonlinear dynamics entering the tire force description, and consequently,
115 in the global chassis dynamic. It reproduces the longitudinal (xs ), lateral (ys ), vertical (zs ), roll (θ), pitch
116 (φ), and yaw (ψ) dynamics of the chassis. It also models the vertical and rotational motions of the
117 wheels (zusij and ωij respectively), the slip ratios (λij ), and the center of gravity side slip angle (β cog )
118 dynamics, as a function of the tires and suspensions forces.

119 3. Integrated cruise - suspension LPV control of an autonomous vehicle for comfort: Structure
120 and objectives
121 The proposed strategy is illustrated in the figure below:

Figure 5. Integrated cruise - suspension LPV control for comfort.

122 Our strategy consists of three main parts closely connected to each other and the full vehicle
123 dynamics. Note that the vehicle speed connects the longitudinal and vertical dynamics due to the
124 relationship (11).
125 The road type is assumed to be known/estimated in real-time thanks to algorithms such as in
126 [19]. This is the condition that enables the making of the proposed reference speed generation strategy,
127 which gives suitable speed values based on the road profile and comfort objective. In the reference
128 speed calculation part, given a road type detected thanks to an estimation algorithm and the desired
129 comfort level specified by the driver, a suitable reference speed value is determined so as to guarantee
130 this level. How we quantify driving comfort and calculate the reference speed is presented in Section
131 4.
132 In the cruise control part, given the calculated reference speed value, the cruise control part drives
133 the vehicle speed to track this value. This uses not only the feedback measured by the speedometer
134 but also road information such as road slope in order to compensate for this, providing a smoother
135 response. How we design this part is discussed in Part 5.2.
136 In the semi-active suspension control design method, a semi-active suspension control strategy is
137 used to further improve driving comfort. How we design this part is discussed in Part 5.3.
138 Combining the three mentioned parts constitutes what we propose in this paper as the integrated
139 cruise - suspension control of an autonomous vehicle with a comfort objective.
 Version February 20, 2021 submitted to Electronics 8 of 19

140 4. Comfort-oriented reference speed calculation

141 4.1. Comfort evaluation using the ISO 2631 standard

142 First, the road types are characterized by the ISO norm [18]. Below, we examine the road
143 displacement profiles of types from A to D described in the ISO standard, with the vehicle’s speed
144 being 5 m/s. As show in [28] such profiles do change w.r.t the speed as we considered in the modeling
145 step (see (11))

0.1

0.05

0 5 10 15 20 25

Figure 6. Road displacement profiles of road types A - D for a given speed.

This displacement is then transmitted to the passengers through the vehicle vertical dynamics.
What affects driving comfort is the acceleration felt by passengers, as analyzed in the ISO 2631 norm
[16]. In order to characterize human comfort, i.e., the effect of exposure to vibration, a filter is applied
on the sprung mass acceleration [29]. The filter’s transfer function is

81.89s3 + 796.6s2 + 1937s + 0.1446

WISO (s) = . (13)
s4 + 80s3 + 2264s2 + 7172s + 21196

146 Driving comfort is then assessed according to the following scale of the ISO 2631 standard using
147 the root mean square (RMS) value of the vertical acceleration as [16]

Table 1. Vertical acceleration RMS value and comfort level.

RMS value of acceleration Comfort level

Less than 0.315 m/s2 Not uncomfortable
0.315 - 0.63 m/s2 A little uncomfortable
0.5 - 1 m/s2 Fairly uncomfortable
0.8 - 1.6 m/s2 Uncomfortable
1.25 - 2.5 m/s2 Very uncomfortable
Greater than 2 m/s2 Extremely uncomfortable

148 4.2. Modeling of the vehicle speed/ comfort interaction

149 It is known that, as the vehicle travels on certain different road types with the same speed, it
150 experiences different road displacement profiles. Thus, the felt human comfort varies according to the
151 vehicle speed [28,30].
 Version February 20, 2021 submitted to Electronics 9 of 19

152 Our objective is to propose a comfort-oriented reference speed profile to link the comfort level
153 to the vehicle’s speed. This is carried out using a vertical vehicle model performing simulation with
154 different speed values and computing the comfort criterion. This allows us to first evaluate the human
155 comfort (for the RMS of the vertical acceleration) as seen in Figure 7.

40
4

30
3

2 20

1
10
0
5 10 15 20 25 30 35 40 0 1 2 3 4 5

Figure 7. Comfort evaluation for different Figure 8. Polynomial functions: Speed vs.
road types. comfort level for different road types.

156 Then, following our previous study in [15], we define the comfort-oriented reference speed values
157 from a polynomial fitting method, which are illustrated in Figure 8 and Table 2.

Table 2. Calculation of comfort-oriented reference speed values.

Road type Comfort-oriented reference speed values vre f ( x ) where x = RMS

−281058.82x7 + 616932.65x6 − 553287.74x5 + 259269.77x4 − 67006.34x3 +
A
+9126.1x2 − 490.82x + 17.03
−2918.46x7 + 12566.60x6 − 22133.93x5 + 20420.08x4 − 10438.88x3 +
B
+2839.03x2 − 319.40x + 20.36
−20.58x + 176.82x6 − 620.95x5 + 1140.85x4 − 1158.9x3 +
7
C
+623.06x2 − 134.38x + 17.84
−0.19x + 3.22x6 − 22.33x5 + 81.06x4 − 162.92x3 +
7
D
+174.09x2 − 76.47x + 19.57

158 These polynomials are precomputed and programmed into the autonomous vehicle’s computer.
159 In practice, the vehicle detects the current road type by performing the road estimation algorithm, then
160 from a given desired comfort level, the corresponding reference speed is calculated thanks to these
161 polynomials.

162 5. LPV H2 cruise and semi-active suspension control

163 In this section, both longitudinal and vertical controllers are synthesized using the H2 control
164 approach for LPV systems. A short background is presented herein, followed by the two design
165 applications.

166 5.1. Preliminaries on LPV H2 control design [26]

Throughout this paper, we design two LPV controllers for cruise and semi-active suspension
control. This part presents briefly the LPV approach including the optimization problem to be solved.
An LPV system is represented as
    
ẋ A(ρ) B1 (ρ) B2 (ρ) x
Σ(ρ) :  z  = C1 (ρ) D11 (ρ) D12 (ρ) w , (14)
    
y C2 (ρ) D21 (ρ) D22 (ρ) u
 Version February 20, 2021 submitted to Electronics 10 of 19

167 where x is the state, z is the controlled output, y is the measured output, w is the disturbance, u is the
h i>
168 control input, and ρ = ρ1 ρ2 ... ρ N ∈ Ω is the vector of varying parameters (Ω is a convex set).
169 The assumptions on ρ are:

170 • ρ varies in the set of continously differentiable parameter curves and is known or measurable.
171 • ρ is bounded, i.e., ρ j ∈ [ρ j , ρ j ], ∀ j.
172 • The system matrices A(·), etc. are continuous on Ω.

The vector of parameters evolves inside a polytope represented by 2 N vertices ωi , as

ρ ∈ Co{ω1 , . . . , ωZ }, (15)

It is then written as the convex combination

2N 2N
ρ= ∑ α i ωi , αi ≥ 0, ∑ αi = 1, (16)
i =1 i =1

173 where the vertices are defined by a vector ωi = [νi1 , . . . , νiN ] where νij equals ρ j or ρ j .
Therefore, we consider a polytopic model of the LPV system above, represented as

2N 2N
" #
A ( ωi ) B ( ωi )
Σ ( ρ ) = ∑ αi ( ρ ) , αi (ρ) ≥ 0, ∑ αi (ρ) = 1, (17)
i =1
C ( ωi ) D ( ωi ) i =1
" #
A ( ωi ) B ( ωi )
174 where is the linear time-invariant (LTI) system corresponding to one of the system’s
C ( ωi ) D ( ωi )
175 2 N vertices.
An LPV controller has the following structure
" # " #" #
ẋc Ac (ρ) Bc (ρ) xc
K: = . (18)
u Cc (ρ) Dc ( ρ ) y

Solving for an LPV controller using the H2 condition is here carried out using the polytopic
approach so computing the controllers Ki ∀i, at each vertex of the parameter polytope, such that a
single, global performance γ2 is minimized. For a given parameter value ρ, the controller is then
determined as
2N 2N
K (ρ) = ∑ α i ( ρ ) Ki , αi (ρ) ≥ 0, ∑ αi (ρ) = 1. (19)
i =1 i =1

Proposition 1. A dynamical output feedback controller K (ρ) (18) that solves the control problem is obtained by
solving the following LMIs in (X(ρ), Y(ρ), A
e ( ρ ), B
e ( ρ ), C e (ρ)) at the 2 N vertices ωi of the polytope ,
e (ρ) and D
while minimizing γ2 ,  
M11 (∗)> (∗)>
 M21 M22 (∗)>  ≺ 0, ∀i,
 
M31 M32 M33
 
N11 (∗)> (∗)> (20)
 N21 N22 (∗)>   0, ∀i,
 
N31 N32 N33
Trace(Z) < γ2 ,
 Version February 20, 2021 submitted to Electronics 11 of 19

where,
M11 = A(ωi )X(ωi ) + X(ωi ) A(ωi )> + B2 C e ( ωi ) > B > ,
e ( ωi ) + C
2
= A >
e ( ωi ) + A ( ωi ) + C D > > >
M21 2 ( ωi ) B2 ,
e
M22 = Y( ωi ) A ( ωi ) + A ( ωi ) > Y( ωi ) + B e (ωi )C2 + C > B e > ( ωi ),
2
M31 = B1 (ωi )> + D21 (ωi )> D e ( ωi ) > B > ,
2
M32 = B1 (ωi ) Y(ωi ) + D21 (ωi )> B
> e ( ωi ) > ,
M33 = − Inu ,
N11 = X ( ωi ),
N21 = In ,
N22 = Y ( ωi ),
N31 = C1 (ωi )X(ωi ) + D12 (ωi )C e ( ωi ),
N32 = C1 (ωi ) + D12 (ωi )D e (ωi )C2 ,
N33 = Z.
Then, the reconstruction of the controller K is obtained by the following equivalent transformation,


 Dc ( ω i ) = D
e ( ωi )

 Cc (ωi )

 e (ωi ) − Dc (ωi )C2 (ωi )X(ωi )) M (ωi )−>
= (C
Bc (ωi ) = N ( ωi ) −1 (B
e (ωi ) − Y(ωi ) B2 (ωi ) Dc (ωi )) (21)


 A c ( ωi ) −
= N ( ωi ) (A1 e (ωi ) − Y(ωi ) A(ωi )X(ωi ) − Y(ωi ) B2 (ωi ) Dc (ωi )C2 (ωi )X(ωi )

− N (ωi ) Bc (ωi )C2 (ωi )X(ωi ) − Y(ωi ) B2 (ωi )Cc (ωi ) M(ωi )> ) M(ωi )−> ,



176 where M (ωi ) and N (ωi ) are defined such that M(ωi ) N (ωi )> = In − X (ωi )Y (ωi ) (that can be solved through
177 a singular value decomposition plus a Cholesky factorization).

178 5.2. Application of the LPV H2 approach to cruise control

179 5.2.1. Cruise control design

The approach above is here applied to the LPV longitudinal model presented before as
(
ẋl = Al (ρl ) xl + Bl1 wl + Bl2 (ρc )ul
Σl (ρl ) : (22)
yl = Cl xl ,

180 where ρl includes the vehicle mass and speed.

To use the polytopic approach, the control input matrix has to be independent of the scheduling
parameter. Therefore, following [31] the system is extended with the following filter at the input
variable " # " #" #
ẋ f Af Bf xf
Wf : = , (23)
ul Cf 0 uf

where A f , B f , and C f are constant matrices and kW f k∞ ≤ 1 so as to ensure that the saturation
constraint on u is kept for the new control input u f . Here, we choose A f = −1/τ f , B f = 1/τ f , and
C f = 1 where τ f is a small constant. The extended state-space representation is therefore
 " # " #" # " # " #
 ẋl Al (ρl ) Bl2 (ρl )C f xl Bl1 0
 ẋ = + wl + uf


0 Af xf 0 Bf

f
Σe (ρl ) : h
"
i x
# (24)
l

 yl = Cl 0 x .



f
 Version February 20, 2021 submitted to Electronics 12 of 19

To synthesize the controller, we have chosen the following weighting functions (see in Figure 9)
in order to ensure tracking performances and to cope with the actuator limitations

0.5s + 2 1
We (s) = , Wu (s) = , Wd (s) = 0.01. (25)
s + 0.0002 100

Figure 9. LPV cruise control scheme.

According to [31], since the system matrices Al (ρl ), Bl1 (ρl ) are affine in ρl and since the scheduling
parameter ρl varies in a polytope of four vertices ρl1 ∈ [ρl1 , ρl1 ] and ρl2 ∈ [ρl2 , ρl2 ], the generalized
plant in Figure 9 can be expressed as a polytopic system composed of four vertices

4 4
Σl (ρl ) = ∑ α li ( ρ l ) Σ li , αli (ρl ) ≥ 0, ∑ αli (ρl ) = 1, (26)
i =1 i =1

where Σl1 = Σl (ρl1 , ρl2 ), Σl2 = Σl (ρl1 , ρl2 ), Σl3 = Σl (ρl1 , ρl2 ), and Σl4 = Σl (ρl1 , ρl2 ). Solving the LMIs
in Proposition 1, the LPV controller Kl (ρl ) with the scheme as shown in Figure 9 is defined as
" # " #" #
ẋcl Acl (ρl ) Bcl (ρl ) xcl
Kl ( ρl ) : = . (27)
ul Ccl (ρl ) Dcl (ρl ) yl

The controller Kl (ρl ) can be transformed into a convex interpolation as follows

" #
4
Acl i Bcl i
Kl ( ρl ) = ∑ α li ( ρ l ) Ccl i Dcl i
. (28)
i =1

181 5.2.2. Cruise control simulation

182 We verify by simulation that the vehicle speed can track a given reference value in the presence of
183 disturbance and noise in the form of inexact road slope compensation.
 Version February 20, 2021 submitted to Electronics 13 of 19

4000
35

2000
30

0
25

-2000
20

0 500 1000 1500 2000 2500 3000 3500 0 500 1000 1500 2000 2500 3000 3500

Figure 10. Reference and real speed. Figure 11. Longitudinal control force.

184 We see that the tracking performance is guaranteed, and tracking is achieved after a few hundreds
185 of meters, with a control input force smaller than 4000 N.

186 5.3. Application of the LPV H2 approach to semi-active suspension control

187 5.3.1. Semi-active suspension control design

First, let us recall the LPV system where the extended quarter-car system is described by
(
ẋv = Av (ρv ) xv + Bv1 (ρv )wv + Bv2 uv
Σv (ρv ) : (29)
yv = Cv2 xv + Dv21 wv + Dv22 uv ,

188 where ρv is the vehicle speed.

189

In order to guarantee the comfort and road holding objectives, we define the performance output
h i>
vector as zv = zus z̈s ∈ R2 and obtain the generalized system

 ẋv = Av (ρv ) xv + Bv1 (ρv )wv + Bv2 uv

Σ g (ρv ) : zv = Cv1 xv + Dv11 wv + Dv12 uv (30)

 y =C x +D w +D u ,
v v2 v v21 v v22 v

where " # " # " #

0 0 1 0 0 0 0
Cv1 = , Dv11 = , Dv12 = .
− mks − mc0s k
ms
c0
ms 0 0 − m1s
In order to obtain the performance objectives, the parameter-dependent weighting function
Wz̈s (ρv ) and the weighting function Wzus are shaped in order to reduce the amplification of the sprung
mass acceleration z̈s and zus depending on the vehicle speed. Ww and Wn model white noise (ws ) and
measurement noise, respectively (see in Figure 12). These functions can be chosen as

s2 +2ζ 11 Ω11 s+Ω211 s2 +2ζ 21 Ω21 s+Ω221

Wz̈s (ρv ) = ρv · k z̈s · s2 +2ζ 12 Ω12 s+Ω212
, Wzus = k zus · s2 +2ζ 22 Ω22 s+Ω222
, Ww = 10−4 , Wn = 10−3 .
(31)
190 The parameters of the weighting functions are coherent with those in [32].
 Version February 20, 2021 submitted to Electronics 14 of 19

Figure 12. LPV semi-active suspension control scheme.

According to [31], since the system matrices Av (ρv ), Bv1 (ρv ) are affine in ρv and since the
scheduling parameter ρv varies in a polytope of two vertices ρv ∈ [ρv , ρv ], the generalized plant
in Figure 12 can be expressed as a polytopic system composed of two vertices

2 2
Σv (ρv ) = ∑ α vi ( ρ v ) Σ vi , αvi (ρv ) ≥ 0, ∑ αvi (ρv ) = 1, (32)
i =1 i =1

where Σv1 = Σv (ρv ) and Σv2 = Σv (ρv ). Solving the LMIs in Proposition 1, the LPV controller Kv (ρv )
with the scheme as shown in Figure 12 is defined as
" # " #" #
ẋcv Acv (ρv ) Bcv (ρv ) xcv
Kv ( ρ v ) : = . (33)
uv Ccv (ρv ) Dcv (ρv ) yv

The controller Kv (ρv ) can be transformed into a convex interpolation as follows

" #
2
Acvi Bcvi
Kv ( ρv ) = ∑ α vi ( ρ v ) Ccvi Dcvi
. (34)
i =1

191 5.3.2. Frequency-domain analysis

192 The Bode diagrams given in Figure 13 show the system frequency performance according to the
193 varying ρv . Compared with the passive suspension, the closed-loop system provides efficient vibration
194 mitigation (attenuation) in the whole frequency range of 10−3 –104 Hz. The effects of road vibrations
195 on the performance output (sprung mass acceleration which is directly linked to driving comfort) are
196 shown, which are effectively attenuated.
 Version February 20, 2021 submitted to Electronics 15 of 19

Bode Diagram

-50

-100
Magnitude (dB)

-150

-200

-250

-300
10 -3 10 -2 10 -1 10 0 10 1 10 2 10 3 10 4
Frequency (Hz)

Figure 13. Bode diagram (from wv to z̈s ) of the closed-loop systems corresponding to the values of
ρvmin and ρvmax , and of the passive system.

197 5.3.3. Semi-active suspension control simulation

198 In order to highlight the performance of the LPV H2 approach, simulation in the time domain is
199 performed in this part. In this simulation, the sprung mass acceleration at one corner of the vehicle is
200 considered. We let the vehicle speed rise from its minimum to maximum value to test the proposed
201 suspension control strategy. From Figure 14 and Figure 15, the LPV H2 control for the semi-active
202 suspension provides better comfort than the passive one.

600
2
400
1 200
0
0
-200
-1 -400
-600
-2
0 100 200 300 400 500 0 100 200 300 400 500

Figure 14. Acceleration felt by passengers Figure 15. Damper force at the front right
(filtered by (13)). corner.

203 6. Simulation of the integrated control strategy

204 In this part, we perform simulations using the full-car model presented in [27], following the
205 scheme presented in Figure 16. It should be noted that the parameters used for simulations are chosen
206 according to a real Megane vehicle. Because of this, the speed is limited to 10 - 35 m/s. This vehicle is
207 equipped with four independent semi-active suspension systems controlled with a sampling period of
208 0.005 s. Since we perform simulations with a full-car model, there is a varying delay of L/v (where
209 L is the distance between the front and rear wheels, i.e., L = L f + Lr in Figure 4) in the road profile
210 zr at the rear wheels compared to the front wheels. Driving comfort is evaluated using the RMS
211 value of the acceleration at the vehicle’s center filtered by (13). We combine the two control strategies
212 into an integrative case where the reference speed varies according to the road type and the desired
213 comfort level to guarantee driving comfort. In this part, we perform simulations with various road
214 types and desired RMS acceleration values to test the reference speed generation and integrated
215 cruise-suspension vehicle control strategies.
 Version February 20, 2021 submitted to Electronics 16 of 19

Figure 16. Simulation scheme.

216 6.1. Simulation scenario

217 The road profile inputs are chosen according to the road model (11) for the given road types,
218 where the input is white noise. The total simulation time is 54 s. The scenario is the following (see
219 Figure 17 and Figure 18):

220 • At 18 s, the desired comfort level (characterized by the given RMS acceleration) changes from 0.4
221 to 0.3 m/s2 .
222 • At 36 s, the road type (characterized by the estimated road roughness) changes from type A to
223 type B.

0.4 10-5
2

0.38
1.5
0.36

0.34
1
0.32

0.3 0.5
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200

Figure 17. Integrated cruise-suspension Figure 18. Integrated cruise-suspension

control: Desired comfort level. control: Road roughness.

224 The resulting road displacement (at the front right corner of the vehicle) is as follows:
 Version February 20, 2021 submitted to Electronics 17 of 19

0.015

0.01

0.005

-0.005

-0.01
0 200 400 600 800 1000 1200

Figure 19. Integrated cruise-suspension control: Road input at the front right corner.

225 We note a significant increase in magnitude in zr after 36 s, due to the change in road type.

226 6.2. Simulation results

4
30
2

25 0

-2
20

-4
15
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200

Figure 20. Resulting reference and vehicle Figure 21. Acceleration felt by passengers
speed. (filtered by (13)).

227 We see that each time the road type or the desired RMS value changes, a new reference speed is
228 calculated, and the cruise control effectively tracks this value. The resulting RMS acceleration values are
229 1.2360 m/s2 for the passive suspension case and 0.4301 m/s2 for the LPV H2 semi-active suspension
230 case, which shows that the latter further improves driving comfort by limiting the acceleration
231 transmitted to passengers.

232 7. Conclusion
233 This work presents an integrated strategy for comfort-oriented vehicle cruise and suspension
234 control with a robust/LPV approach in the H2 framework. Indeed, the integrated control approach is
235 adapted to the comfort requirement and the vehicle speed. It is worth mentioning that we rely on basic
236 assumptions and a reasonable amount of knowledge of the environment, which makes this strategy
237 realistic.
238 Funding: This research received no external funding.
239 Conflicts of Interest: The authors declare no conflict of interest.

240 Abbreviations
241 The following abbreviations are used in this manuscript:
242
 Version February 20, 2021 submitted to Electronics 18 of 19

IRI international roughness index

ISO International Organization for Standardization
LMI linear matrix inequality
243 MR Magneto-Rheological
LPV linear parameter-varying
LTI linear time-invariant
RMS root mean square

244 References
245 1. Hedrick, J.K.; Tomizuka, M.; Varaiya, P. Control Issues in Automated Highway Systems. IEEE Control
246 Systems Magazine 1994, 14, 21–32.
247 2. Ioannou, P.A.; Chien, C.C. Autonomous Intelligent Cruise Control. IEEE Transactions on Vehicular Technology
248 1993, 42, 657–672.
249 3. Gáspár, P.; Szabó, Z.; Bokor, J.; Nemeth, B. Robust Control Design for Active Driver Assistance Systems.
250 Springer, DOI 2016, 10, 978–3.
251 4. Rajamani, R.; Zhu, C. Semi-autonomous Adaptive Cruise Control Systems. IEEE Transactions on Vehicular
252 Technology 2002, 51, 1186–1192.
253 5. Kayacan, E. Multiobjective H∞ Control for String Stability of Cooperative Adaptive Cruise Control Systems.
254 IEEE Transactions on Intelligent Vehicles 2017, 2, 52–61.
255 6. Németh, B.; Gáspár, P. LPV-based Control Design of Vehicle Platoon Considering Road Inclinations. IFAC
256 Proceedings Volumes 2011, 44, 3837–3842.
257 7. Öncü, S.; Ploeg, J.; Van de Wouw, N.; Nijmeijer, H. Cooperative Adaptive Cruise Control: Network-aware
258 Analysis of String Stability. IEEE Transactions on Intelligent Transportation Systems 2014, 15, 1527–1537.
259 8. Du, Y.; Liu, C.; Li, Y. Velocity Control Strategies to Improve Automated Vehicle Driving Comfort. IEEE
260 Intelligent Transportation Systems Magazine 2018, 10, 8–18.
261 9. Wu, J.; Zhou, H.; Liu, Z.; Gu, M. Ride Comfort Optimization via Speed Planning and Preview Semi-active
262 Suspension Control for Autonomous Vehicles on Uneven Roads. IEEE Transactions on Vehicular Technology
263 2020, 69, 8343–8355.
264 10. Tran, G.Q.B.; Sename, O.; Gaspar, P.; Nemeth, B.; Costa, E. Adaptive Speed Control of an Autonomous
265 Vehicle with a Comfort Objective. VSDIA 2020 - 20th International Conference on Vehicle System Dynamics,
266 Identification and Anomalies 2020.
267 11. Savaresi, S.M.; Poussot-Vassal, C.; Spelta, C.; Sename, O.; Dugard, L. Semi-active Suspension Control Design
268 for Vehicles; Elsevier, 2010.
269 12. Poussot-Vassal, C.; Spelta, C.; Sename, O.; Savaresi, S.M.; Dugard, L. Survey and Performance Evaluation
270 on Some Automotive Semi-active Suspension Control Methods: A Comparative Study on a Single-corner
271 Model. Annual Reviews in Control 2012, 36, 148–160.
272 13. Murali Madhavan Rathai, K. Synthesis and Real-time Implementation of Parameterized NMPC Schemes
273 for Automotive Semi-active Suspension Systems. PhD thesis, Université Grenoble Alpes, 2020.
274 14. Pham, T.P.; Sename, O.; Dugard, L. Unified H∞ Observer for a Class of Nonlinear Lipschitz Systems:
275 Application to a Real ER Automotive Suspension. IEEE Control Systems Letters 2019, 3, 817–822.
276 15. Costa, E.; Pham, T.P.; Sename, O.; Tran, G.Q.B.; Do, T.T.; Gaspar, P. Definition of a Reference Speed of an
277 Autonomous Vehicle with a Comfort Objective. VSDIA 2020 - 20th International Conference on Vehicle System
278 Dynamics, Identification and Anomalies 2020.
279 16. International Organization for Standardization. ISO 2631-1:1997, Mechanical Vibration and Shock —
280 Evaluation of Human Exposure to Whole Body Vibration — Part 1: General Requirement 1997.
281 17. Pawar, P.R.; Mathew, A.T.; Saraf, M. IRI (International Roughness Index): An Indicator of Vehicle Response.
282 Materials Today: Proceedings 2018, 5, 11738–11750.
283 18. International Organization for Standardization. ISO 8608:2016, Measurement and Evaluation of Mechanical
284 Vibration and Shock as Applied to Machines, Vehicles and Structures 2016.
285 19. Tudón-Martínez, J.C.; Fergani, S.; Sename, O.; Martinez, J.J.; Morales-Menendez, R.; Dugard, L. Adaptive
286 Road Profile Estimation in Semiactive Car Suspensions. IEEE Transactions on Control Systems Technology
287 2015, 23, 2293–2305.
 Version February 20, 2021 submitted to Electronics 19 of 19

288 20. Basargan, H.; Mihály, A.; Gáspár, P.; Sename, O. Adaptive Semi-Active Suspension and Cruise Control
289 through LPV Technique. Applied Sciences 2021, 11, 290.
290 21. Vahidi, A.; Stefanopoulou, A.; Peng, H. Recursive Least Squares with Forgetting for Online Estimation of
291 Vehicle Mass and Road Grade: Theory and Experiments. Vehicle System Dynamics 2005, 43, 31–55.
292 22. Yin, Z.; Dai, Q.; Guo, H.; Chen, H.; Chao, L. Estimation Road Slope and Longitudinal Velocity for
293 Four-wheel Drive Vehicle. IFAC-PapersOnLine 2018, 51, 572–577.
294 23. Li, B.; Zhang, J.; Du, H.; Li, W. Two-layer Structure Based Adaptive Estimation for Vehicle Mass and Road
295 Slope under Longitudinal Motion. Measurement 2017, 95, 439–455.
296 24. Tudon-Martinez, J.C.; Hernandez-Alcantara, D.; Amezquita-Brooks, L.; Morales-Menendez, R.;
297 de J Lozoya-Santos, J.; Aquines, O. Magneto-rheological Dampers—model Influence on the Semi-active
298 Suspension Performance. Smart Materials and Structures 2019, 28, 105030.
299 25. Unger, A.; Schimmack, F.; Lohmann, B.; Schwarz, R. Application of LQ-based Semi-active Suspension
300 Control in a Vehicle. Control Engineering Practice 2013, 21, 1841–1850.
301 26. Poussot-Vassal, C. Robust LPV Multivariable Automotive Global Chassis Control. PhD thesis, Institut
302 National Polytechnique de Grenoble-INPG, 2008.
303 27. Poussot-Vassal, C.; Sename, O.; Dugard, L.; Savaresi, S.M. Vehicle Dynamic Stability Improvements
304 through Gain-scheduled Steering and Braking Control. Vehicle System Dynamics 2011, 49, 1597–1621.
305 28. Loprencipe, G.; Zoccali, P.; Cantisani, G. Effects of Vehicular Speed on the Assessment of Pavement Road
306 Roughness. Applied Sciences 2019, 9, 1783.
307 29. Zuo, L.; Nayfeh, S. Low Order Continuous-time Filters for Approximation of the ISO 2631-1 Human
308 Vibration Sensitivity Weightings. Journal of Sound and Vibration 2003, 265, 459–465.
309 30. Ahlin, K.; Granlund, N.J. Relating Road Roughness and Vehicle Speeds to Human Whole Body Vibration
310 and Exposure Limits. International Journal of Pavement Engineering 2002, 3, 207–216.
311 31. Apkarian, P.; Gahinet, P.; Becker, G. Self-scheduled H∞ Control of Linear Parameter-varying Systems: A
312 Design Example. Automatica 1995, 31, 1251–1261.
313 32. Do, A.L.; Sename, O.; Dugard, L.; Soualmi, B. Multi-objective Optimization by Genetic Algorithms in
314 H∞/LPV Control of Semi-active Suspension. IFAC Proceedings Volumes 2011, 44, 7162–7167.

315 c 2021 by the authors. Submitted to Electronics for possible open access publication
316 under the terms and conditions of the Creative Commons Attribution (CC BY) license
317 (http://creativecommons.org/licenses/by/4.0/).