sensors

Article
Multi-Objective Evolutionary Design of an Electric
Vehicle Chassis
Pablo Luque 1 , Daniel A. Mántaras 1 , Álvaro Maradona 1 , Jorge Roces 1 ,
Luciano Sánchez 2, * , Luis Castejón 3 and Hugo Malón 3
1 Department of Transportation Engineering, University of Oviedo, 33203 Gijón, Spain; [email protected] (P.L.);
[email protected] (D.A.M.); [email protected] (Á.M.); [email protected] (J.R.)
2 Department of Computer Science, University of Oviedo, 33203 Gijón, Spain
3 Department of Mechanical Engineering, University of Zaragoza, 50018 Zaragoza, Spain;
[email protected] (L.C.); [email protected] (H.M.)
* Correspondence: [email protected]; Tel.: +34-985-182-130




Received: 26 May 2020; Accepted: 22 June 2020; Published: 28 June 2020


Abstract: An iterative algorithm is proposed for determining the optimal chassis design of an electric
vehicle, given a path and a reference time. The proposed algorithm balances the capacity of the battery
pack and the dynamic properties of the chassis, seeking to optimize the tradeoff between the mass of
the vehicle, its energy consumption, and the travel time. The design variables of the chassis include
geometrical and inertial values, as well as the characteristics of the powertrain. The optimization
is constrained by the slopes, curves, grip, and posted speeds of the different sections of the track.
Particular service constraints are also considered, such as limiting accelerations due to passenger
comfort or cargo safety. This methodology is applicable to any vehicle whose route and travel
time are known in advance, such as delivery vehicles, buses, and race cars, and has been validated
using telemetry data from an internal combustion rear-wheel drive race car designed for hill climb
competitions. The implementation of the proposed methodology allows to reduce the weight of the
battery pack by up to 20%, compared to traditional design methods.

Keywords: chassis optimization; electric vehicle (EV); energy consumption; genetic algorithm;
electric powertrain

1. Introduction
Every year, the European Union, USA, Japan, among others, set stricter targets for CO2 emissions
for passenger cars and commercial vehicles. This fact, combined with the rising price of petrol per
barrel, social awareness regarding global warming, and many other factors, has led to a growing
demand for alternatives to traditional internal combustion engine vehicles [1], such as electric, hybrid,
and fuel cell vehicles.
There are numerous examples in which an electric vehicle can be used in a very efficient way;
for example, in delivery trucks, whose acceleration and deceleration patterns make the use of internal
combustion engines inefficient. Therefore, this kind of transport can be optimized using electric
trucks [2]. Electric urban buses produce zero tailpipe emissions and reduce pollution and noise [3].
However, range is probably the major limiting factor for the mass use of electric vehicles in these
cases. In order to improve the range of electric vehicles, the key points to be considered are energy
consumption and vehicle performance.
According to the literature, numerous attempts have been made to develop different methodologies
for optimizing the powertrain in electric vehicles [4], taking into account one or more of their components
(battery pack, electric machine, transmission, electric devices for vehicles, and control system).

Sensors 2020, 20, 3633; doi:10.3390/s20133633 www.mdpi.com/journal/sensors

Sensors 2020, 20, 3633 2 of 22

Different design variables have been analyzed in relation to the different purposes of the
optimization of the powertrain. Lei et al. [5] considered power performance, energy consumption,
and ride comfort using an in-wheel motor vehicle model as the basis for their multi-objective problem.
They used two approaches to carry out the optimization, first a weighted objective method to transform
the multi-objective optimization into a single-objective optimization, and a second approach to obtain
the optimal solution from the Pareto front. In this paper, energy consumption was evaluated by the
urban driving cycle (ECE-15).
Domingues-Olavarría et al. [6] also took into account the cost of the powertrain as an additional
constraint. To do it they used a model, simulated over a given drive cycle, that takes into account
the required size, performance, efficiency, and thermal characteristics of the main components in an
electric powertrain.
Kulik et al. [7] estimated the requirements for a hybrid electric powertrain based on the analysis
of a city vehicle GPS track together with accelerometer data. From the acceleration and velocity data of
the track, the instantaneous power developed by the car is estimated and areas where regeneration is
possible during braking.
Gearboxes are used for reasons of efficiency [8], in addition to improving the vehicle’s range and
performance [9]. Different geared systems over the NEDC (New European Driving Cycle) were tested
and the energy consumption was calculated.
Transmissions with a gear ratio (i.e., a gearbox) have been optimized by Dagci et al. [10], among
others. Standard drive cycles are used.
Guo et al. [11] proposed a method for controlling the gearshift command in multispeed EV, reaching
a reduction of 3–5% in energy consumption for city driving cycles. They used an algorithm, combining
Pontryagin’s minimum principle and a numerical method, bisection method, for determining the gear
positions and gearshift points. The algorithm was tested using the New European drive cycle (NEDC)
and Urban Dynamometer Driving Schedule (UDDS).
Yu et al. [12,13] minimized the time consumed by a vehicle to travel along a given route or track,
both for two independent wheel driving and four independent wheel driving. Describing a very
detailed vehicle model, they solve a large-scale nonlinear optimization problem constrained to a certain
track. The base speed, the constant power speed ratio, the static braking force distribution and the gear
ratio were selected as the design parameters of the vehicle. The methodology does not optimize the
vehicle mass and do not consider the energy consumption.
Xin [14] developed an optimal design of the powertrain system in order to minimize the curb
mass of the vehicle, due to the sensitivity of vehicle mass for energy consumption. To carry out the
simulation the Urban Driving Cycle (UDC), the NEDC cycle, and a constant speed cycle were chosen.
However, powertrain optimization is not the only way to improve the range of a vehicle. In order
to improve the driving range of electric vehicles, especially for fleets of electric commercial vehicles,
many studies [15] focus their efforts on planning the most efficient route. These studies use an average
driver and are programmed to follow a driving cycle that include speed variation, topography, time
stopped, and changes in payload. They are focused on the track, not on the vehicle or the travel time.
Other authors [16] focused their research on defining the best acceleration profile so that the
energy consumed is minimized. Both a constant value of acceleration and a multi-step acceleration
approach are considered in this last reference.
Energy consumption is a recurrent issue in this field, although battery life is not always regarded
as an important variable. Liu et al. [17] optimized an acceleration profile for an electric vehicle,
taking into account the energy consumption and percentage of battery capacity loss per kilometer.
Yu et al. [18] also considered the battery life in the optimization process of an electric vehicle. In their
paper, the sizing and energy management of the battery pack arranged in a hybrid racing vehicle is
addressed as a multi-objective optimization problem.
Multi-objective optimization algorithms have also been applied to the design of complex
mechanical systems [19] and, in particular, to the design of the powertrain of a hybrid vehicle [20].
Sensors 2020, 20, 3633 3 of 22

Among multi-objective algorithms, evolutionary algorithms are widely used methods because of
their flexibility and ease of use. Depending on the number of objectives, many different types of
multi-objective evolutionary algorithms (MOEAs) exist. For problems with a moderate number of
objectives, Pareto optimization algorithms such as the well-known NSGA2 (Non-dominated Sorting
Genetic Algorithm–2) [21] and SPEA2 (Strength Pareto Evolutionary Algorithm–2) [22], among others,
are effective and allow a set of solutions to be obtained in a single run (see [23] for a recent review of
many-objective optimization algorithms). There are also scalarization strategies that transform the
multi-objective problem into a scalar-valued optimization problem [24]. These last techniques are
either used to rank incomparable solutions in the optimization algorithm or to select the most suitable
candidate from the set of solutions produced by the MOEA.
Evolutionary multi-objective algorithms have also been used in problems related to the case at
hand, such as the definition of operational strategies. For instance, MOEAs were used in [25] to reduce
the operational costs of fleets of hybrid electric trucks.
As can be seen, many research methods have been applied to optimize the powertrain of a specific
vehicle and were validated according to a standard driving test cycle [26]. Therefore, the need to
develop an optimization method that defines the optimal chassis design for an electric vehicle, not
only for the powertrain system, but also for a specific route is evident.
The methodology proposed in this paper is intended not only to optimize the powertrain, but also
to minimize energy consumption by determining the optimal chassis design for a given track, setting
the travel time as an objective. This enables adapting the chassis design to the path selected by the
driver, being applicable to transport vehicles that cover one or more fixed routes. On the other hand,
a standard driving cycle is not used. A driving mode is defined that fully approves the performance of
the vehicle based on the set path. This allows for the optimization of acceleration and braking cycles
and speeds to minimize energy consumption.
In addition to other resistance forces normally considered, the vehicle model takes into account
the effects of curves in order to improve the accuracy of the evaluation of energy consumption [27].
A previously set path and time allows the designer to find a solution adapted to the specific needs
of the vehicle. The use of multi-objective genetic algorithms to solve this issue will serve as a reliable
generic method for finding the optimal configuration of the chassis on any given route.
As a particular case, it will allow the design of gearboxes with an optimal relation for each route
or the definition of an automatic change strategy depending on the chosen route.
The case study in this paper is based on an electric race car, in which the values of the track and
time were obtained from previous tests.

2. Materials and Methods

2.1. Optimization Methodology

An optimization methodology has been developed to achieve an optimal solution to the problem.
The objective is to minimize the consumed energy with a maximum (reference) time on a given track.
The starting information is as follows:

• With regard to the vehicle chassis. Each of these parameters can be assumed to be given data
(fixed parameter) or may be optimized (design variables):

# Geometrical and inertial values, such as vehicle mass, mv , radius of the wheels, rw ,
wheelbase, L, and position of the center of gravity, l1 , as shown in Figure 2.
# Powertrain characteristics: related to the electrical motor (defined by its maximum power,
pmax , maximum torque, Tmax , and maximum rotation speed, nmax . It is assumed that
the engine always runs at maximum performance), mass of the battery pack, mb , and
transmission ratio, it .
Sensors 2020, 20, 3633 4 of 22

• With regard to the track/road. The track is divided into several sections with constant values of
different parameters. The information related to the track and related to each section consists
of the:

# Number
Sensors 2020, 20, xof sections.
FOR PEER REVIEW 4 of 20
# Initial and final point measured from the start point.
o Radius of the curves.
# Radius ofo the curves.
Slope.
# Slope. o Grip or friction coefficient (longitudinal, , and lateral, ).
# o Posted
Grip or friction speed.
coefficient (longitudinal, µx , and lateral, µ y ).
• Related to the optimization case:
# Posted speed.
o Reference time, .
• • the
Related to Particular service constraints
optimization case: (limit accelerations due to passenger comfort or cargo safety,
for instance).
The data related to the track act as constraints of the problem and need to be obtained prior to
# Reference time, t f .
the optimization of the re
chassis. Additional information, such as initial or final speeds, stops or delays
• can be included
Particular service inconstraints
the numerical(limit
definition of the problem.due to passenger comfort or cargo safety,
accelerations
As to the subject of this paper, a multi-objective genetic algorithm is put forward to optimize the
for instance).
chassis of any kind of electric vehicle that performs a pre-established, known route, such as that
shown in Figure 1.
The data This
related to the track act as constraints of the problem and need to be obtained prior to the
algorithm will generate the initial values of the design variables, which will be introduced
optimization of the chassis. Additional
in the calculation sequence information,
(Simulator), obtaining a such asthe
value of initial
erroror final
with speeds,
respect to thestops or delays can
references
be included in the numerical definition of the problem.
(time, minimum of energy, etc.). Due to its genetic character, it will take the best values of each
As togeneration
the subject as aof
reference to create
this paper, the initial values ofgenetic
a multi-objective the next algorithm
generation. is put forward to optimize the
The process starts with the definition of the boundaries of the problem. This will give the
chassis of any kind of electric vehicle that performs a pre-established, known route, such as that shown
algorithm a set of points within which to find the solution. In this part of the process, some settings
in Figure 1.
of the algorithm are established such as the number of iterations, the tolerance for the error, and
so on.

FigureFigure 1. General
1. General flowchart of
flowchart ofthe
theproposed
proposedmethodology.
methodology.

Each specimen in the population represents a set of designs variables of the chassis. The
This algorithm will generate the initial values of the design variables, which will be introduced in
simulator evaluates this configuration of chassis and calculates the final time ( ) and the
the calculation sequence
consumed energy(Simulator), obtaining
( ). These interim resultsa will
value passof to
theanerror
error with respect
function, wheretothey
the will
references
be (time,
minimum compared
of energy, etc.).
with Due totime
a reference its (genetic
) andcharacter, it will( take
the stored energy ) in the best values
the battery pack. of each generation as
a reference to The
create
finalthe
timeinitial values
obtained of the next generation.
in the simulation, , is related to the reference time set as an objective
for the vehicle. This reference time is subtracted
The process starts with the definition of the boundaries from the result time.
of the problem. This will give the algorithm
a set of points within which to find the solution. = −In this part of the process, some settings (1) of the
algorithm are established such as the number of iterations, the tolerance for the error, and so on.
Each specimen in the population represents a set of designs variables of the chassis. The simulator
evaluates this configuration of chassis and calculates the final time (t f inal ) and the consumed energy
Sensors 2020, 20, 3633 5 of 22

(ec ). These interim results will pass to an error function, where they will be compared with a reference
time (tre f ) and the stored energy (eb ) in the battery pack.
The final time obtained in the simulation, t f inal , is related to the reference time set as an objective
for the vehicle. This reference time is subtracted from the result time.

εt = tre f − t f inal (1)

where εt is the error of the final time.

The consumed energy, ec , is compared with the available stored energy in the battery pack, eb , to
check whether there is enough energy for the track. The stored energy in the battery pack for a certain
mass is calculated as follows:
eb = mb ·β (2)

where eb is the total stored energy in the battery pack and β is the energy density of the battery pack.
The energy error, εe , is calculated as:
εe = |ec − eb | (3)

The objective of the methodology is to minimize the overall error, E. This overall total error is
calculated as a weighted sum of the two aforementioned errors and the consumed energy with its own
weighting (kc ):
E = kt ·εt + ke ·εe + kc ·ec (4)

where kt is the value of the weighting associated with the time error and ke is the error associated with
the energy. Once the total error has been calculated for every specimen in the population, the best
values are selected and crossed over to define the next generation.
If the stopping criteria are reached, the best values of the last generation will be the optimization
result. The objective of the methodology is to obtain a set of non-dominated designs with respect
to the three aforementioned criteria: εt , εe and ec . The NSGA2 algorithm [21] is used to solve this
vector-valued optimization problem. A scalarization strategy [24] is used to select a suitable candidate
from the final set of Pareto-optimal solutions produced by the MOEA.

2.2. Mathematical Model of Longitudinal Dynamics

The definition of a mathematical model of the behavior of the vehicle is subject to the following
first-order dynamic constraints:
.
x(t) = f[x(t), u(t), t, p] (5)
.
where x is the first-order derivative of the state variables, f is the dynamic model, and x, u, p are,
respectively, the state, input, and design vectors.
The state vector, x, includes the variables used to describe the speed, v(t), and position, x(t), from
the start point of the track:
x(t) = [v(t), x(t)] (6)

The input vector includes the longitudinal forces acting on the vehicle:
h i
u = Rr , R g , Ra , Rc , F (7)
 
where the rolling resistance, (Rr ), grade resistance, R g , aerodynamic drag, (Ra ), curve resistance,
(Rc ) [27], and traction/braking force, (F) are considered. The values and equations are discussed later.
The considered design variables are related to the design of the chassis, including geometrical,
inertial and powertrain characteristics. These variables are included in the design vector, (p), as follows:

p = [mv , rw , L, l1 , pmax , Tmax , nmax , mb , it ] (8)

 = · sin (12)
The value of the ramp will be determined at each time instant depending on the position of the
vehicle on the road under study: = ( ) .
, : Tractive or braking effort of the front and rear tires, respectively:
Sensors 2020, 20, 3633 6 of 22
( )= + (13)
The value of ( ), will be discussed later. The first-order dynamic constraints (5),
The derivation ofthe
theforce,
first-order dynamic constraints is proposed according to Newton’s Second
including all the forces applied to the vehicle, can be expressed as follows:
Law [28], in which: “The sum of the external forces acting on a body in a given direction is equal to the
product of its mass( ) and( the
) ( ) ·
acceleration ( )
in ·that direction ( )
(assuming the mass is fixed)”:
=
∀ /0 ≤ ( ) ≤ (14)
( )
= (X) dv(t)
Fext = mt · (9)
dt
where is the gravitational acceleration, is the density of the air, is the aerodynamic
wherecoefficient of the vehicle
Fext represents and forces
the external is the and
frontal
mt ,area
the of the mass
total vehicle.
of A mass
the factor, , is introduced to
vehicle.
take external
The into account the acting
forces effect ofon
thea inertia
two-axleof the rotating
vehicle parts.
[29] are shown in Figure 2. These include:

Figure2.2.Forces
Figure Forces acting
acting on
onaavehicle.
vehicle.

To predict howresistance:
Ra : Aerodynamic the vehicle will behave on the route, the first-order dynamic constraints will be
integrated in a given path for every set of values
1 of the design
2 variables. The above differential
equations system is integrated dependingRa = onρCthe (v(t))and
x A froad the vehicle characteristics so as to (10)
2
determine its position, ( ), and speed, ( ).
Rr f , Rrr : Rolling resistance of the front and rear tires, respectively, where fr represents the rolling
The force, ( ), should be calculated according to numerous variables as discussed later.
resistance coefficient and mt = mv + mb :
( )= ( ), ( ), , , , , , , , , , ℎ, , , (15)
On each section of the track, the R r = Rr(f ),
force, +may
Rrr =
be m t g· fr or braking, as will be seen in Section (11)
tractive
2.3 Track Characterization. This is linked to the possibility of the vehicle accelerating, maintaining a
R g : Grade
constant resistance:
speed or braking to a certain final speed.
R g = mt g· sin θ (12)

The value of the ramp will be determined at each time instant depending on the position of the
vehicle on the road under study: θ = θ(x(t)).
F f , Fr : Tractive or braking effort of the front and rear tires, respectively:

F(t) = F f + Fr (13)

The value of the force, F(t), will be discussed later. The first-order dynamic constraints (5),
including all the forces applied to the vehicle, can be expressed as follows:

F(t)−(Rc (v(t))+mt g· fr +mt g· sin θ(x(t))+ 12 ρCx A f (v(t))2 )


dv(t)
dt = mt γ

∀t/0 ≤ x(t) ≤ x f inish (14)

dx(t)
= v (t)

dt

where g is the gravitational acceleration, ρ is the density of the air, Cx is the aerodynamic coefficient of
the vehicle and A f is the frontal area of the vehicle. A mass factor, γ, is introduced to take into account
the effect of the inertia of the rotating parts.
To predict how the vehicle will behave on the route, the first-order dynamic constraints will
be integrated in a given path for every set of values of the design variables. The above differential
Sensors 2020, 20, 3633 7 of 22

equations system is integrated depending on the road and the vehicle characteristics so as to determine
its position, x(t), and speed, v(t).
The force, F(t), should be calculated according to numerous variables as discussed later.
 
F(t) = F x(t), v(t), µx , µ y , rw , it , Tmax , pmax , nmax, L, l1 , h, mt , ηb , ax (15)

On2020,
Sensors each20,
section track, the force, F(t), may be tractive or braking, as will be seen in Section72.3
of theREVIEW
x FOR PEER of 20
Track Characterization. This is linked to the possibility of the vehicle accelerating, maintaining a
constantThe two factors
speed thatto
or braking limit the maximum
a certain final speed.tractive effort are _ and _ . _ is related to the
friction coefficient,
The two factors that, and
limitthethenormal load on
maximum the driven
tractive effort axle.
are FT_µ_ and is F
related to the
T_m . FT_µ characteristics
is related to the
friction coefficient, µx , and the normal load on the driven axle. FT_m is related to the characteristics of
of the vehicle’s powertrain:
the vehicle’s powertrain: = min n _ , _o (16)
FT = min FT_m , FT_µ (16)
In some cases, however, the tractive effort could result in unsuitable acceleration for passenger
In
comfort someor cases,
cargo however,
safety. Hence,the tractive effort could
depending on theresult
caseinunder
unsuitable
study,acceleration
a maximum forlongitudinal
passenger
comfort or cargo, safety.
acceleration, should be Hence, depending
established that isoncompared
the case with
underthe study,
valueaobtained
maximum longitudinal
in Equation (14).
acceleration,
The minimum a , should be established that is compared with the value obtained
x value is introduced in the first-order dynamic constraint (5) and the demanded tractive in Equation (14).
The minimum
effort value is introduced in the first-order dynamic constraint (5) and the demanded tractive
is calculated.
effort is calculated.
On the other hand, it is necessary to check whether the vehicle passes the speed limit in each
On the other hand, it is necessary to check whether the vehicle passes the speed limit in each section.
section.
InInorder
ordertotoperform
performthis thiscalculation,
calculation,the
thefirst-order
first-orderdynamic
dynamicconstraints
constraintsare areintegrated
integratedfrom
fromthe the
beginning
beginning to the end of the track. The position of the vehicle with respect to the road is obtainedvia
to the end of the track. The position of the vehicle with respect to the road is obtained via
numerical
numericalintegration
integrationtotoidentify
identifywhich
whichsection
section the
thevehicle
vehicle is is
inin
atat
any
any instant
instantinintime.
time.
The
Thewhole
wholetracktrackisissimulated,
simulated,obtaining
obtainingthe theoverall
overalltime
timeandandoverall
overallenergy
energyconsumed
consumedbybythe the
vehicle. The evaluation of the tractive or braking force is discussed in the following
vehicle. The evaluation of the tractive or braking force is discussed in the following subsection. subsection.

2.2.1. Tractive Effort due to the Powertrain

2.2.1. Tractive Effort due to the Powertrain
The electric motor is responsible for converting electrical energy into mechanical energy, consuming
The electric motor is responsible for converting electrical energy into mechanical energy, consuming
the stored power in the battery in order to move the transmission set. The behavior (Figure 3) of the
the stored power in the battery in order to move the transmission set. The behavior (Figure 3) of the motor
motor defines the speed, torque and other important variables for the performance of the vehicle.
defines the speed, torque and other important variables for the performance of the vehicle.

Figure 3. Torque and power curves of an electric motor.

Figure 3. Torque and power curves of an electric motor.

To mathematically define the powertrain of an electric vehicle, the starting point is the maximum
torque and power values of the motor. The motor power, , is defined as:
= · (17)
where is the motor torque and , the rotational speed. The changeover point between maximum
torque and constant power (cut-off speed, _ ) is defined as:

= · _ (18)
Sensors 2020, 20, 3633 8 of 22

To mathematically define the powertrain of an electric vehicle, the starting point is the maximum
torque and power values of the motor. The motor power, p, is defined as:

p = T·nm (17)

where T is the motor torque and nm , the rotational speed. The changeover point between maximum
torque and constant power (cut-off speed, nm_c ) is defined as:

pmax = Tmax ·nm_c (18)

and can be calculated via the following expression:

pmax
nm_c = (19)
Tmax

Thus, the characteristic torque versus speed curve for the motor can be defined as:

Tmax , nm < nm_c

(
T (nm ) = pmax (20)
nm , nm ≥ nm_c

Knowing the rotational speed of the motor, it is possible to calculate the vehicle’s longitudinal
speed (v):
nm
v= ·rw (21)
it
where rw is the effective radius of the wheels on the drive axles. The slip phenomena are neglected in
this model. The top speed (vtop ) due to the powertrain can be calculated via the following expression:

nmax
vtop = ·rw (22)
it

The torque applied at the wheels, Tw , by the powertrain can be calculated as follows:

Tw = T·it (23)

The tractive force of the motor due to wheel torque, FT_m , is defined as:

Tw
FT_m = (24)
rw

This force will be constant, until the vehicle reaches the speed at which the torque stops being
constant, which is called the cut-off speed or vc . This speed is related to the variables of the motor via
the expression:
nm_c rw pmax
vc = ·rw = · (25)
it it Tmax
Bearing this last equation in mind, a new traction force equation can be described in terms of the
cut-off speed. This can be expressed as:


 Tmax · riwt , v < vc
FT_m (v) =  (26)

pmax it pmax
nm · rw = v , v ≥ vc



These are the maximum tractive forces related to the powertrain, which will later be compared
with the maximum efforts due to the grip of the track.
Another element to consider in order to obtain a reliable model of the vehicle is the battery.
The existence of efficiencies in the transfer of power and its subsequent use by the motor should
Sensors 2020, 20, 3633 9 of 22

be taken into account. These efficiencies have to be considered so that the necessary power is not
underestimated when modeling the vehicle.
Assuming pb to be the total demanded power needed to drive the vehicle and pm to be the output
power of the battery, then its efficiency, ηbat , is defined as:
pm
ηbat = (27)
pb

Furthermore, pm is usually modeled as a function of the vehicle’s mechanical power and the
machine or motor efficiency, ηm :
T· nm
pm = (28)
ηm (T, nm )
As can be seen, the motor efficiency is a function of the speed and torque output multiplied by
itself. This efficiency normally reaches its maximum in lower speed regions compared to an internal
combustion engine [11]. A polynomic expression can be used to obtain a description of the behavior of
the power demanded by the vehicle. A two-dimensional polynomial expression can be used to obtain
this value, as follows:
2 X2
T·nm X
pb = ≈ pij Ti nm j (29)
ηm ·ηbat
i=0 j=0

where pij are the fitting coefficients or tuning values. These values are set to effectively represent the
motor power [30]. The value of these fitting coefficients can be obtained by running a set of tests.
It is easier to work with this approximate closed-form (29) expression, instead of using the analytic
version (28).

2.2.2. Tractive Effort due to the Grip of the Track

In order to define the maximum tractive effort due to the grip in a rear/forward wheel-drive
vehicle, the first step is to determine the normal loads on the vehicle’s axles. The normal load on the
rear axle is related to the sum of the moments acting on that axle:

mt gl1 cos θ + Ra ha + h amt + mt gh sin θ

Wr = (30)
L
where l1 is the distance between the front axle and the vehicle’s center of gravity and, in this equation,
the sign of θ is positive when the vehicle is climbing a hill.
The normal load on the front axle is related to the sum of the moments acting on that axle:

mt gl2 cos θ − Ra ha − h amt + mt gh sin θ

Wf = (31)
L
where l2 is distance from the rear axle to the vehicle’s center of gravity and, in this equation, the sign
of θ is negative when the vehicle is climbing a hill. Assuming that the point of application of the
aerodynamic resistance (ha ) is near the height of the center of gravity (h), the last two equations may be
rewritten as:
l2 h
W f = mt g − (Ra + amt + Rd + sin θ(x(t))) (32)
L L
and:
l h
Wr = 1 mt g − (Ra + amt + Rd + sin θ(x(t))) (33)
L L
Substituting Equation (11) in the last equation, we have that:

l1 h
Wr = mt g + (F − Rr ) (34)
L L
Sensors 2020, 20, 3633 10 of 22

and:
l2 h
Wf =mt g − (F − Rr ) (35)
L L
Using this last equation, the maximum tractive force of the vehicle can be determined for a rear
wheel-drive vehicle as: " #
l1 h
FT_µ_r = µx Wr = µx mt g + (Fmax − Rr ) (36)
L L
and:
µx W (l1 − fr h)/L
FT_µ_r = (37)
1 − µh/L
For a front wheel-drive vehicle, the maximum tractive force is:
" #
l2 h
FT_µ_ f = µx W f = µx mt g + (Fmax − Rr ) (38)
L L

and:
µx W (l2 + fr h)/L
FT_µ_ f = (39)
1 − µh/L

2.2.3. Braking Dynamics

While studying the behavior of the vehicle on each section of the route, when the speed is higher
than the final speed of the section, it will be necessary to brake in order to get to the end of the section
at the specified speed. No regenerative braking is considered. To do so, it is necessary to determine the
braking point. This point is related to the braking performances of the vehicle. These performances are
controlled by the braking force. To determine this force, the values of the maximum grip force in the
longitudinal direction, µx , and the braking efficiency, ηb , need to be obtained. The braking force of a
vehicle can be determined as:
Fb = mt gµx ·ηb (40)

The braking acceleration can be calculated as:

−Fb − mt g· f + mt g· sin θ + 12 ρCx A f (v(t))2

 
ab_µx (t) = (41)
mt γ

In some cases, however, the braking acceleration (41) might not be suitable for passenger comfort
or cargo safety. Hence, depending on the case, a maximum longitudinal acceleration (ax ) should
be established that is compared with the value obtained in Equation (41). The minimum value is
consequently chosen: n o
ab (t) = min ab_µx (t) , |ax | (42)

Considering the braking acceleration as constant in every section, the braking distance, db , to
decrease the actual speed, v(t), to a final speed, v f inal , can be calculated as follows:

v f inal 2 − v(t)2
db (v(t)) = (43)
ab

2.3. Track Characterization

A road route can have a complex three-dimensional geometry, with straight and curved alignments
as well as elevation. It is therefore necessary to apply the dynamics equations to the entire route.
In the proposed methodology, the Geographic Information System (GIS) data is pre-processed
before modeling. The road centerline (Figure 4) is determined by analyzing the GIS data and
Sensors 2020, 20, 3633 11 of 22

orthophotos. Subsequently, the precision of the route is improved and adapted to the terrain by
combining
Sensors thex FOR
2020, 20, horizontal alignment and elevation profile.
PEER REVIEW 11 of 20

Figure 4. 3D
Figure 4. 3D characterization
characterization of
of the
the track.
track. The
The road
road centerline
centerline (red)
(red) is
is shown
shown in
in aa Geographic
Geographic
Information System using a digital model (yellow) of the study area
area and
and road.
road.

The geometrical
Following data onofthe
this analysis theroute
track,isthe
exported and divided
result thus obtainedinto set of ( ) sections.
is a parametrized sections with
Each their
section
(corresponding
) is characterizedlength, by curvature slope, and grip values.
several parameters:
•Following this analysis
Longitudinal of theoftrack,
coordinate the result(thus
the initial, ), and
obtained
final, is a set ( ),of points.
(n) sections. Each section
(i) is• characterized
Elevation above by several parameters:
sea level of the initial, ( ), and final, ( ), points.
• Radius of curvature, ( ).
• Longitudinal coordinate of the initial, xini (i), and final, x f inal (i), points.
• Section grip ( ( ) , ( )), which is characterized by the vehicle dynamics limit. This
• Elevation above
parameter mayseabelevel of the initial,
calculated zini (i),from
or obtained and final, z f inal (i), points.
an instrumented vehicle.
• Radius
The finalofpoint
curvature, rc (i). (i) matches the initial point of the next one (i+1). This means that each
of a section
•
section Section
startsgrip x (i), µ ythe
right(µwhere (i))previous
, which isone characterized
ends; that is: by the vehicle dynamics limit. This parameter
may be calculated or obtained from an instrumented vehicle.
( + 1) = () (44)
The final point of a section (i) matches the initial point of the next one (i + 1). This means that each
It will be assumed that the start of the track has a value of (1) = 0 and the finish, a value of
section starts right where the previous one ends; that is:
( + 1) = = length of the track.
This allows the length of each section, (( ), to be characterized:
xini i + 1) = x f inal (i) (44)
()= ( )− () (45)
It will be assumed that the start of the track has a value of xini (1) = 0 and the finish, a value of
xini (nHence,
+ 1) =when
x f inishthe vehicle,ofatthe
= length time step , is located on section , the following holds:
track.
This allows the length of each section,( l)(≤ i), to( be
) ≤characterized:
() (46)
The slope of each section, ( ), which
l(i) is=assumed
x (i) −tox be(iconstant,
) can be calculated as: (45)
f inal ini
( )− ()
()= · 100 (47)
( )− ()

Each section is characterized by a maximum speed ( ( )) that is related to the vehicle and
the track. The vehicle presents a top speed, , related to the maximal rotational speed of the motor.
For each section , the track itself imposes another speed limit, ( ). This value is related to the
maximum lateral acceleration due to particular limitations (passenger comfort, cargo safety, etc.),
Sensors 2020, 20, 3633 12 of 22

Hence, when the vehicle, at time step t, is located on section i, the following holds:

xini (i) ≤ x(t) ≤ x f inal (i) (46)

The slope of each section, θ(i), which is assumed to be constant, can be calculated as:

z f inal (i) − zini (i)

θ(i) = ·100 (47)
x f inal (i) − xini (i)

Each section is characterized by a maximum speed (vmax (i)) that is related to the vehicle and
the track. The vehicle presents a top speed, vtop , related to the maximal rotational speed of the motor.
For each section i, the track itself imposes another speed limit, va y (i). This value is related to the
maximum lateral acceleration due to particular limitations (passenger comfort, cargo safety, etc.),
a y_pc (i), and the lateral friction limit, µ y (i). The maximum speed, va y (i), of section i is calculated
as follows: q q 
vay (i) = min a y_pc rc (i), µ y (i) grc (i) (48)

An additional limitation may appear, depending on traffic rule constraints on each section of
the track, denoted as vposted (i). The maximum speed, vmax (i), of section i is the minimum of these
three values: n o
vmax (i) = min vtop , va y (i), vposted (i) (49)

Once the sections of the track have been defined, the behavior of the vehicle as it passes through
each one needs to be analyzed.

2.4. Driving Dynamics and Energy

The proposed algorithm takes the initial speed of a section (i) as the final speed of the immediately
previous one (i − 1).
As a rule of thumb, when the maximum value among the tractive forces is the one related to
the powertrain, the powertrain will be used to the maximum, but the maximum speed, vmax (i), can
never be passed. Moreover, if the vehicle is passing through a section (i) at a higher speed than the
predefined final speed, v f inal (i), it needs to brake.
To understand the vehicle response in each section, a series of rules must be considered:

• The initial speed of section i is known, vini (i), determined by the algorithm as the final speed of
the previous section (i − 1).
• If the speed is below the maximum for this section and there is no need to brake, the case will
be ‘Traction’.

if [v(t) < vmax (i) ∧ not[case : BRAKING]] then 0 case : TRACTION0 (50)

• At the end of a section (i), the speed cannot be greater than the maximum speed, for the next
section, vmax (i + 1).
if [x(t) = xini (i + 1)] then v(t) ≤ vmax (i + 1) (51)

The finish speed of a section, v f inal (i), cannot be higher than the speed limit for the next section,
vlim (i + 1).

• If the speed v(t) > vmax (i + 1), the vehicle must brake. The braking process must ensure that
the final speed at the end of the section (i), is also the maximum speed for the next section
(i + 1), namely:
v f inal (i) = vmax (i + 1) (52)
Sensors 2020, 20, 3633 13 of 22

Considering that the braking acceleration, ab (i), is constant in the section, the braking distance, db ,
can be calculated as follows:
Sensors 2020, 20, x FOR PEER REVIEW
v f inal (i)2 − v(t)2 13 of 20
db (i, v(t)) = (53)
ab ( i )
1. calculation
The The vehiclealgorithm
accelerates
mustfrom ( ) toeach time
check, during at constant
step, the torque.
distanceThe totalthe
between force
end in this
of the
section andsubsection is: position of the vehicle in order to activate the braking case, or not:
the current
h  = TRACTION ∧ ( ( i) < )
if x f inal (i) − x(t) ≤ db (i, v(t)) ∧ v(t) > v f inal (i) then 0 case := BRAKING0 (54)
(55)
=
In the case of maximal longitudinal acceleration, as shown in Figure 5, the drive through a section
2.defined
can be The vehicle
via fouraccelerates from scenarios:
different dynamic to ( ) at constant power. The total force in this
subsection is:
1. The vehicle accelerates from vini (i) to vc at constant torque. The total force in this subsection is:
= TRACTION ∧ < ( )< ()
T i (56)
if [case = TRACTION ∧ = (v(t) < vc )] then F = max t (55)
R
3. The vehicle circulates at constant speed, ( ). In this case, the force will be that needed to
2. The vehicle accelerates from vc to vmax (i) at constant power. The total force in this subsection is:
overcome the forces of resistance.
1 pmax
if [case · +
= = TRACTION c < (v()t)
∧ (· vsin +< vmax (i))] then( )F = (56)
(57)
2 v

3. 4.
The The vehicle
vehicle brakes at
circulates constant(speed,
from ) to v. The total force of this subsection is:
max (i). In this case, the force will be that needed to
overcome the forces of resistance. = BRAKING
(58)
F = mt g· fr + mt g· sin θ=
1
 
(i) + ρCx A f (vmax (i))2 (57)
2
A way to calculate the energy consumed by the vehicle on its route is needed in order to achieve
optimization.
4. This
The vehicle energy
brakes fromcan be (calculated
vmax as:total force of this subsection is:
i) to v f . The

= ( )] then F = Fb
if [case = BRAKING (59)
(58)

where is the overall consumed energy.

Figure 5. Vehicle dynamic scenarios on a section of the track.

Figure 5. Vehicle dynamic scenarios on a section of the track.

3. Results
A study case was devised to implement and validate the proposed methodology. The case is
defined by the track and the vehicle. The vehicle used as a reference for the implementation of the
algorithm is an ICE (internal combustion engine) rear-wheel drive race car (as shown in Figure 6),
designed for hill climb competitions (Table 1). The hill climb discipline is perfect for the use of electric
vehicles due to the track length [30].
Sensors 2020, 20, 3633 14 of 22

A way to calculate the energy consumed by the vehicle on its route is needed in order to achieve
optimization. This energy can be calculated as:
Z tf
Sensors 2020, 20, x FOR PEER REVIEW ec = pb (t)dt 14 of 20
(59)
0
Table 1. Vehicle parameters.
where ec is the overall consumed energy.
( ) Vehicle Mass without the Battery 475
3. Results ( ) Wheelbase 2.5
( )
A study case was devised to implement Wheel
andradius
validate the proposed 0.264
methodology. The case is
defined by the track and ( the
) vehicle. The
Length of the
vehicle usedvehicle 3.750
as a reference for the implementation of the
( ) combustion
algorithm is an ICE (internal Height of the
engine) vehicle drive race1.030
rear-wheel car (as shown in Figure 6),
( )
designed for hill climb competitions Width
(Table of the
1). The hillvehicle
climb discipline is1.750
perfect for the use of electric
vehicles due to the track( length
) [30]. Frontal area of the vehicle 1.210
Aerodynamic coefficient 0.3

Figure 6. ICE vehicle.

Table 1. Vehicle parameters.

The optimization target is to run the chosen race track in the same time (reference time) as the
ICE vehicle, consuming
mv (kg) the minimum amount
Vehicle of without
Mass energy.the
To Battery
do so, an electric motor
475 was previously
selected (Table 2). L (m) Wheelbase 2.5
r (m) Wheel radius 0.264
Lv (m) TableLength
2. Motor parameters.
of the vehicle 3.750
Hv (m) Height of the vehicle 1.030
Wv((m)) Motor maximum
Width power
of the vehicle 120
1.750
A f( m2 ) Motor maximum torque
Frontal area of the vehicle 350
1.210
( Cx ) Motor maximum
Aerodynamicrotational speed
coefficient 12,000
0.3
The energy density of the chosen batteries cells is = 128 Wh/kg . According to these
conditions, the design variables
The optimization can
target is to bethe
run expressed
chosen in this
race casein
track as:the same time (reference time) as the
ICE vehicle, consuming the = minimum
475, 0.264,amount
2.5, 1.431,
of 120,000,
energy. To350,
do12,000, , , motor was previously
so, an electric (60)
selected (Table 2).
where several variables are fixed and assumed as parameters. In this particular case, the mass of
battery packs, , and the transmission ratio,
Table , are
2. Motor the design variables to be optimized.
parameters.
The chosen track is a scoring event for the European Hill Climb Championship. The track
characterizationpismax (kW)on the three-dimensional
based Motor MaximumanalysisPower 120
of a LIDAR dataset (2008-2015), reference
 
ETRS89, from theTmax N
Spanish National Geographic
Motor Institute
maximum (26,355,144
torque points have
350 been processed).
m
nm (rpm
The density of points is )0.5 points/m2Motor maximum
and the rotational
altimetry speed
precision is better than12,000
20 cm RMSE(z) (root
mean square error).
Istram®BIM software (Buhodra Ingeniería S.A., Llanera, Asturias, Spain) was used to import and
process the LIDAR (light detection and ranging) data and for 3D characterization of the route
geometry (Figure 7). The main horizontal route and its elevation were defined, based on terrain data,
an orthophoto and Spanish road regulations IC-3.1.
Sensors 2020, 20, 3633 15 of 22

The energy density of the chosen batteries cells is β = 128 Wh/kg. According to these conditions,
the design variables can be expressed in this case as:

p = [475, 0.264, 2.5, 1.431, 120, 000, 350, 12, 000, mb , it ], (60)

where several variables are fixed and assumed as parameters. In this particular case, the mass of
battery packs, mb , and the transmission ratio, it , are the design variables to be optimized.
The chosen track is a scoring event for the European Hill Climb Championship. The track
characterization is based on the three-dimensional analysis of a LIDAR dataset (2008–2015), reference
ETRS89, from the Spanish National Geographic Institute (26,355,144 points have been processed).
The density of points is 0.5 points/m2 and the altimetry precision is better than 20 cm RMSE(z) (root
mean square error).
Istram® BIM software (Buhodra Ingeniería S.A., Llanera, Asturias, Spain) was used to import
and process the LIDAR (light detection and ranging) data and for 3D characterization of the route
geometry (Figure 7). The main horizontal route and its elevation were defined, based on terrain data,
Sensors
an 2020, 20, x FOR
orthophoto and PEER REVIEW
Spanish road regulations IC-3.1. 15 of 20

Figure 7. Map of the track.

Figure 7. Map of the track.

The track, as shown in Figure 7, has a length of 5.260 km with a maximal gradient of 6.65% and an
The track, as shown in Figure 7, has a length of 5.260 km with a maximal gradient of 6.65% and
average gradient of 5.96%. The altitude above sea level is 46 m at the beginning and 359 m at the end,
an average gradient of 5.96%. The altitude above sea level is 46 m at the beginning and 359 m at the
as shown in Table A1—Appendix A. During the optimization process, the algorithm decomposed this
end, as shown in Table A1 – Annex A. During the optimization process, the algorithm decomposed
track into 107 sections.
this track into 107 sections.
Real data recorded in this event (telemetry of the ICE vehicle) was used in the optimization
Real data recorded in this event (telemetry of the ICE vehicle) was used in the optimization
procedure. According to these data, the braking acceleration presents a minimum constant value
procedure.2According to these data, the braking acceleration presents a minimum constant value of
of −13 m/s , the reference time is 168.5 s and the initial speed is 0 km/h. There is no restriction for
–13 m/s2, the reference time is 168.5 s and the initial speed is 0 km/h. There is no restriction for
maximum longitudinal or lateral acceleration due to passenger comfort or cargo safety, so high values
maximum longitudinal or lateral acceleration due to passenger comfort or cargo safety, so high
are defined (above 6 g).
values are defined (above 6 g).
The parameters of the genetic algorithm are as follows:
The parameters of the genetic algorithm are as follows:
• •Population:
Population:
o Population Size: 30
# Population Size: 30[5–100, 1–10]
o Initial Range:
• Selection:
o Selection Function: Tournament.
o Tournament Size: 2.
• Crossover:
o Crossover Fraction: 0.8
Sensors 2020, 20, 3633 16 of 22

# Initial Range: [5–100, 1–10]

• Selection:

# Selection Function: Tournament.

# Tournament Size: 2.
• Crossover:

# Crossover Fraction: 0.8

# Crossover Function: Intermediate.
• Mutation:

# Mutation Function: Constraint dependent.

• Migration:

# Direction: Forward.
# Fraction: 0.2
# Interval: 20
• Multi-objective Options:

# Pareto Fraction: 0.35

• Stopping Criteria:

# Maximal Generation: 400

# Stall Generation: 100
# Function Tolerance: 1 × 10-4

The final values of the optimization after 102 generations are shown in Table 3.

Table 3. Optimization solution.

tfinal (s) Total Time 168.49999999999471

mb (kg) Battery pack mass 35.12
it Transmission ratio 4.011
ec (kWh) Consumed energy 2.73

A design methodology based on standard driving cycles, without taking full advantage of
the powertrain’s performance, obtains mass values of the battery pack greater than 50 kg. If the
methodology does not take into account the precise values of the route, but instead consider average
values, the mass values of the battery pack would be greater than 42 kg. The implementation of
methods of optimization of the consumed energy, without taking into account the multivariable
analysis, does not allow obtaining values lower than 40 kg of the mass of the battery pack.
Figure 8 shows the speed profile of the model with the optimized powertrain generated by the
simulator. The new profile meets the requirements imposed on the model in terms of both energy
consumption and elapsed time.
values, the mass values of the battery pack would be greater than 42 kg. The implementation of
methods of optimization of the consumed energy, without taking into account the multivariable
analysis, does not allow obtaining values lower than 40 kg of the mass of the battery pack.
Figure 8 shows the speed profile of the model with the optimized powertrain generated by the
Sensors 2020, 20,
simulator. The3633 17 of 22
new profile meets the requirements imposed on the model in terms of both energy
consumption and elapsed time.

Figure 8. Speed vs. distance using the optimized chassis.

Figure 8. Speed vs. distance using the optimized chassis.
Figure 9 shows the plot of acceleration and speed versus distance, on the first kilometer of the
track.Figure 9 showsvehicle
The different the plot of acceleration
dynamic scenariosand
are speed versus distance,
also identified on the
in the figure (1: first kilometer ofmax;
traction–torque the
track.
Sensors The
2020, different
20, x FOR vehicle
PEER dynamic
REVIEW scenarios are
2: traction–power max; 3: constant speed; and 4: braking).also identified in the figure (1: traction – torque
17 of 20
max; 2: traction–power max; 3: constant speed; and 4: braking).

Figure 9. Acceleration
Figure 9. Acceleration and
and speed
speed vs. distance on
vs. distance on the
the first
first kilometer
kilometer of
of track.
track.
4. Discussion and Conclusion
4. Discussion and Conclusion
Most published methods optimize the powertrain of electric vehicles in standard driving tests.
Most
In this published amethods
contribution, general optimize the powertrain
methodology of electricthat
has been proposed vehicles in fine-tuning
allows standard driving tests.
the chassis
In this contribution, a general methodology has been proposed that allows fine-tuning the chassis of
electrical vehicles in an arbitrary route or set of routes. An approximate mathematical model has been
developed that takes into account many different parameters of both the vehicle and track, such as
geometries, inertial values, forces, resistances, grip, and powertrain, among others. This set of
information enables accurate evaluation of the consumed energy and at the same time is simple
Sensors 2020, 20, 3633 18 of 22

of electrical vehicles in an arbitrary route or set of routes. An approximate mathematical model has
been developed that takes into account many different parameters of both the vehicle and track, such
as geometries, inertial values, forces, resistances, grip, and powertrain, among others. This set of
information enables accurate evaluation of the consumed energy and at the same time is simple enough
for being computationally efficient, thus, it can be embedded in the fitness function of a multi-objective
genetic algorithm.
This methodology has been used and validated in the optimization of a race car. The chosen track
was a scoring event for the European Hill Climb Championship. In this case, the design variables were
the battery pack mass and the transmission ratio, both optimized to ensure optimal performance on
the track. The multi-objective model was validated with the use of a real-life case that confirms its
functionality and accuracy. The implementation of the proposed methodology allows to reduce the
weight of the battery pack by up to 20%, compared to traditional design methods.
For a more robust design, in future work the mathematical model should be able to cope with
uncertainties in the acquired data and the state of the vehicle. We intend to deploy algorithms for
multi-criteria optimization under uncertainty and obtain solutions are that are insensitive to small
variations in weight of the loaded vehicle or the capacity of the batteries, to name some examples.

Author Contributions: Conceptualization: P.L. and D.A.M.; methodology: P.L. and L.S.; software: P.L., J.R. and
L.S.; validation: Á.M., D.A.M., and L.S.; formal analysis: L.C.; resources: J.R. and H.M.; writing—original draft
preparation: P.L., D.A.M., and Á.M.; writing—review and editing: J.R., L.S., L.C., and H.M.; project administration:
P.L. All authors have read and agreed to the published version of the manuscript.
Funding: This research was funded by the University Institute of Industrial Technology of Asturias, grant number
SV-19-GIJÓN-1-08 and by the Ministry of Economic Affairs and Digital Transformation from Spain/FEDER under
grant TIN2017-84804-R.
Conflicts of Interest: The authors declare no conflict of interest.

Appendix A

Table A1. Track Parameters.

ID(i) xini (i) Rcurve (i) z(i)

1 0 ∞ 46,125
2 127.467 150 48.6441558
3 180.666 −100 52.2207885
4 210.155 −200 54.203416
5 277.484 70 58.729799
6 335.355 −50 62.6206238
7 385.251 ∞ 65.9751298
8 418.889 40 68.2363408
9 451.944 ∞ 70.3709868
10 546.883 50 74.596315
11 565.302 ∞ 75.0700769
12 608.156 −25 75.736674
13 650.325 ∞ 76.0347263
14 671.923 30 76.1826726
15 700.137 −800 76.3759385
16 911.484 200 77.8236654
17 969.517 −60 78.2211915
18 1023.979 ∞ 78.5942562
19 1195.154 150 79.9948911
20 1258.969 ∞ 81.8667499
Sensors 2020, 20, 3633 19 of 22

Table A1. Cont.

ID(i) xini (i) Rcurve (i) z(i)

21 1493.249 200 96.4353708
22 1545.19 ∞ 99.9014325
23 1639.779 40 106.214248
24 1660.423 −25 107.592235
25 1692.078 60 109.704603
26 1733.973 ∞ 112.500698
27 1777.712 −80 115.41939
28 1807.215 80 117.388601
29 1859.316 300 120.866343
30 1932.884 −70 125.776007
31 1976.292 ∞ 128.672491
32 2045.448 −25 133.287654
33 2075.643 50 135.30317
34 2126.936 −35 138.725978
35 2170.978 ∞ 141.665782
36 2235.555 50 145.975296
37 2256.985 ∞ 147.405749
38 2301.196 20 150.355833
39 2317.233 ∞ 151.426303
40 2370.216 −50 154.962407
41 2424.61 −150 158.592718
42 2486.11 150 162.696843
43 2571.955 −40 168.425996
44 2609.579 20 170.936919
45 2637.789 −70 172.819416
46 2682.104 80 175.777337
47 2702.238 20 177.120387
48 2747.368 ∞ 180.132814
49 2785.704 −30 182.690742
50 2801.271 ∞ 183.729824
51 2829.93 25 185.663205
52 2866.397 ∞ 188.227783
53 2882.906 −20 189.430497
54 2912.449 150 191.656971
55 2969.646 −30 196.206993
56 3000.088 150 198.715247
57 3048.231 −25 202.681234
58 3073.721 ∞ 204.78161
59 3112.699 40 207.993398
60 3143.962 ∞ 210.569469
61 3166.457 −10 212.422734
62 3173.294 ∞ 212.985761
63 3193.212 60 214.626669
64 3227.409 −40 217.444502
65 3252.349 ∞ 219.499558
66 3297.097 −70 223.186793
67 3349.735 350 227.523677
68 3487.385 120 238.865524
69 3537.84 ∞ 243.022124
70 3570.729 60 245.73207
71 3621.916 ∞ 249.949878
72 3635.727 −35 251.087905
Sensors 2020, 20, 3633 20 of 22

Table A1. Cont.

ID(i) xini (i) Rcurve (i) z(i)

73 3682.136 ∞ 254.912006
74 3751.879 40 260.65783
75 3780.272 −20 262.997413
76 3784.487 ∞ 263.344729
77 3804.28 −40 264.975672
78 3851.058 70 268.830179
79 3930.629 −80 275.38583
80 3959.784 ∞ 277.788202
81 3974.473 100 278.998575
82 4006.034 ∞ 281.599202
83 4024.854 −20 283.14997
84 4047.718 ∞ 285.033565
85 4085.93 20 287.68453
86 4119.758 60 288.976
87 4146.509 ∞ 289.74911
88 4167.222 −20 290.348077
89 4191.544 ∞ 291.051622
90 4349.483 50 295.619059
91 4432.747 ∞ 298.027026
92 4449.532 35 298.512475
93 4498.898 −120 299.952233
94 4629.658 50 309.736896
95 4656.788 −200 312.044819
96 4740.748 −50 319.187617
97 4784.98 ∞ 322.950549
98 4841.944 −60 327.796337
99 4880.016 60 331.035361
100 4918.913 −100 334.343551
101 4951.792 150 337.14091
102 4995.102 −100 340.82518
103 5023.51 ∞ 343.241701
104 5095.255 12 348.287763
105 5122.771 −400 349.169281
106 5208.882 600 353.500725
107 5271.667 −400 358.800695

References
1. Chan, B.C.C. The State of the Art of Electric, Hybrid, and Fuel Cell Vehicles. Proc. IEEE 2007, 95, 704–718.
[CrossRef]
2. Zhao, Y.; Onat, N.C.; Kucukvar, M.; Tatari, O. Carbon and energy footprints of electric delivery trucks: A
hybrid multi-regional input-output life cycle assessment. Transp. Res. Part D Trans. Environ. 2016, 47,
195–207. [CrossRef]
3. Qin, N.; Gusrialdi, A.; Paul Brooker, R.; T-Raissi, A. Numerical analysis of electric bus fast charging strategies
for demand charge reduction. Transp. Res. Part A Policy Pract. 2016, 94, 386–396. [CrossRef]
4. Alagarsamy, T.; Moulik, B. A review on optimal design of hybrid electric vehicles and electric vehicles.
In Proceedings of the 3rd International Conference for Convergence in Technolgy (I2CT), Pune, India,
6–8 April 2018; pp. 1–5. [CrossRef]
5. Lei, F.; Bai, Y.; Zhu, W.; Liu, J. A novel approach for electric powertrain optimization considering vehicle
power performance, energy consumption and ride comfort. Energy 2019, 167, 1040–1050. [CrossRef]
6. Domingues-Olavarria, G.F.J. Marquez-Fernandez, G.; Fyhr, P.; Reinap, A.; Andersson, M.; Alakula, M.
Optimization of electric powertrains based on scalable cost and performance models. IEEE Trans. Ind. Appl.
2019, 55, 751–764. [CrossRef]
Sensors 2020, 20, 3633 21 of 22

7. Kulik, E.; Tran, X.T.; Anuchin, A. Estimation of the requirements for hybrid electric powertrain based on
analysis of vehicle trajectory using GPS and accelerometer data. In Proceedings of the 25th International
Workshop on Electric Drives: Optimization in Control of Electric Drives (IWED), Moscow, Russia,
31 January–2 February 2018; pp. 1–5. [CrossRef]
8. Ren, Q.; Crolla, D.A.; Morris, A. Effect of transmission design on Electric Vehicle (EV) performance.
In Proceedings of the 5th IEEE Vehicle Power and Propulsion Conference, Dearborn, MI, USA,
7–10 September 2009; pp. 1260–1265. [CrossRef]
9. Abdelrahman, A.S.; Algarny, K.S.; Youssef, M.Z. Optimal gear ratios selection for a Nissan Leaf: A case study
of ingear transmission system. In Proceedings of the IEEE Energy Conversion Congress and Exposition,
Cincinnati, OH, USA, 1–5 October 2017; pp. 2079–2085. [CrossRef]
10. Dagci, O.H.; Peng, H.; Grizzle, J.W. Hybrid electric powertrain design methodology with planetary gear sets
for performance and fuel economy. IEEE Access 2018, 6, 9585–9602. [CrossRef]
11. Guo, L.; Gao, B.; Liu, Q.; Tang, J.; Chen, H. On-line optimal control of the gearshift command for multispeed
electric vehicles. IEEE/ASME Trans. Mechatron. 2017, 22, 1519–1530. [CrossRef]
12. Yu, H.; Castelli-Dezza, F.; Cheli, F. Optimal powertrain design and control of a 2-IWD electric race car.
In Proceedings of the International Conference of Electrical and Electronics Technologies for Automotive,
Torino, Italy, 15–16 June 2017; Volume 1, pp. 1–7. [CrossRef]
13. Yu, H.; Cheli, F.; Castelli-Dezza, F. Optimal design and control of 4-IWD electric vehicles based on a 14-DOF
vehicle model. IEEE Trans. Veh. Technol. 2018, 67, 10457–10469. [CrossRef]
14. Xin, X.; Chengning, Z. Optimal design of electric vehicle power system with the principle of minimum curb
356 mass. Energy Procedia 2017, 105, 2629–2634. [CrossRef]
15. Basso, R.; Kulcsár, B.; Egardt, B.; Lindroth, P.; Sanchez-Diaz, I. Energy consumption estimation integrated
into the Electric Vehicle Routing Problem. Transp. Res. Part D: Trans. Environ. 2019, 69, 141–167. [CrossRef]
16. Vaz, W.S.; Nandi, A.K.; Koylu, U.O. A multiobjective approach to find optimal electric-vehicle acceleration:
361 simultaneous minimization of acceleration duration and energy consumption. IEEE Trans. Veh. Technol.
2016, 65, 4633–4644. [CrossRef]
17. Li, L.; Liu, Q. Acceleration curve optimization for electric vehicle based on energy consumption and battery
life. Energy 2019, 169, 1039–1053. [CrossRef]
18. Yu, H.; Cheli, F.; Castelli-Dezza, F.; Cao, D.; Wang, F.-Y. Multi-objective optimal sizing and energy management
of hybrid energy storage system for electric vehicles. Syst. Control 2019, arXiv:abs/1801.07183.
19. Mastinu, G.; Gobbi, M.; Miano, C. Optimal Design of Complex Mechanical Systems, 1st ed.; Springer:
Berlin/Heidelberg, Germany, 2006; pp. 47–98. [CrossRef]
20. Callejo, A.; García de Jalón, J.; Luque, P.; Mántaras, D.A. Sensitivity-based, multi-objective design of vehicle
suspension systems. J. Comput. Nonlinear Dyn. 2015, 10, 31008. [CrossRef]
21. Deb, K.; Pratap, A.; Agarwal, S.; Meyarivan, T. A fast and elitist multiobjective genetic algorithm: NSGA-II.
IEEE Trans. Evol. Comput. 2002, 6, 182–197. [CrossRef]
22. Zitzler, E.; Thiele, L. Multiobjective evolutionary algorithms: A comparative case study and the strength
Pareto approach. IEEE Trans. Evol. Comput. 1999, 3, 257–271. [CrossRef]
23. Li, B.; Li, J.; Tang, K.; Yao, X. Many-objective evolutionary algorithms: A survey. ACM Comput. Surv. (CSUR)
2015, 48, 1–35. [CrossRef]
24. Eichfelder, G. Adaptive Scalarization Methods in Multiobjective Optimization; Springer: Berlin, Germany, 2008.
25. Huang, B.; Wang, Z.; Xu, Y. Multi-objective genetic algorithm for hybrid electric vehicle parameter
optimization. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems,
Beijing, China, 9–15 October 2006; pp. 5177–5182. [CrossRef]
26. Yin, Q.; Wu, Z.; Rui, X. Parameter design and optimization of electric vehicle. In Proceedings of the IEEE
Conference and Expo Transportation Electrification, Asia-Pacific, Beijing, China, 31 August–3 September
2014; pp. 1–7.
27. Allonca, D.; Mantaras, D.A.; Luque, P.; Alonso, M. A new methodology to optimize a race car for inertial
sports. Proc. Inst. Mech. Eng. Part P J. Sports Eng. Technol. 2019, 233, 312–323. [CrossRef]
28. Gillespie, T.D. Fundamentals of Vehicle Dynamics; Society of Automotive Engineers: Warrendale, PA, USA,
2002; ISBN 978-1-56091-199-9. [CrossRef]
Sensors 2020, 20, 3633 22 of 22

29. Genta, G. Motor Vehicle Dynamics: Modeling and Simulating; World Scientific: Singapore, 1997; pp. 123–125.
ISBN 978-981-02-2911-5.
30. Guo, L.; Gao, B.; Chen, H. Online shift schedule optimization of 2-speed electric vehicle using moving
horizon strategy. IEEE/ASME Trans. Mechatron. 2016, 21, 2858–2869. [CrossRef]

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).