Article

Model Predictive Control with Powertrain Delay Consideration for

Longitudinal Speed Tracking of Autonomous Electric Vehicles
Junhee Lee 1 and Kichun Jo 2, *

1 Department of Smart Vehicle Engineering, Konkuk University, Seoul 05029, Republic of Korea;
[email protected]
2 Department of Automotive Engineering, Hanyang University, Seoul 04763, Republic of Korea
* Correspondence: [email protected]

Abstract: Accurate longitudinal control is crucial in autonomous driving, but inherent delays and
lag in electric vehicle powertrains hinder precise control. This paper presents a two-stage design
for a longitudinal speed controller to enhance speed tracking performance in autonomous electric
vehicles. The first stage involves designing a Model Predictive Control (MPC) system that accounts
for powertrain signal delay and response lag using a First Order Plus Dead Time (FOPDT) model
integrated with the vehicle’s longitudinal dynamics. The second stage employs lookup tables for the
drive motor and brake system to convert control signals into actual vehicle inputs, ensuring precise
throttle/brake pedal values for the desired driving torque. The proposed controller was validated
using the CarMaker simulator and real vehicle tests with a Hyundai IONIQ5. In real vehicle tests, the
proposed controller achieved a mean speed error of 0.54 km/h, outperforming conventional PID and
standard MPC methods that do not account for powertrain delays. It also eliminated acceleration
and deceleration overshoots and demonstrated real-time performance with an average computation
time of 1.32 ms.

Keywords: autonomous driving; longitudinal control; model predictive control; delay compensation

Citation: Lee, J.; Jo, K. Model

Predictive Control with Powertrain
Delay Consideration for Longitudinal 1. Introduction
Speed Tracking of Autonomous
As autonomous driving technology advances, the increasing importance of accurate
Electric Vehicles. World Electr. Veh. J.
speed tracking in real driving scenarios such as adaptive cruise control and lane merging
2024, 15, 433. https://doi.org/
has led to a growing demand for improvements in longitudinal control systems. Further-
10.3390/wevj15100433
more, with the advent of Electric Vehicles (EVs) and the shift from internal combustion
Academic Editor: Yujie Wang engines to motor-driven powertrains in recently produced vehicles, significant research
and development has focused on longitudinal control methods suited to these new char-
Received: 9 August 2024
Revised: 19 September 2024
acteristics [1–3]. As shown in Figure 1, the longitudinal control systems covered in this
Accepted: 24 September 2024
paper output throttle or brake pedal commands to track a given speed profile created by
Published: 25 September 2024 the high-level planning system. A key requirement for these control systems is to reduce
speed tracking errors without resorting to aggressive brake or throttle operations.

Reference
speed

speed

Measured speed
Copyright: © 2024 by the authors.
Throttle
Published by MDPI on behalf of the High-level time Speed Tracking
World Electric Vehicle Association. speed planning Reference speed Controller
Brake
time
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and Figure 1. Longitudinal speed tracking control flow. Speed tracking controller determines appropriate
conditions of the Creative Commons throttle and brake values to follow the reference speed from high-level speed planning.
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/ Traditionally, Proportional-Integral-Derivative (PID) control has been widely utilized
4.0/). for longitudinal speed tracking due to its simplicity and ease of implementation without

World Electr. Veh. J. 2024, 15, 433. https://doi.org/10.3390/wevj15100433 https://www.mdpi.com/journal/wevj

World Electr. Veh. J. 2024, 15, 433 2 of 14

relying on a system model [4–6]. However, PID control does not account for system
dynamics and external disturbances, which can lead to significant overshoot or persistent
oscillations when subjected to noise or delays [7,8]. To address these limitations, advanced
control techniques such as Sliding Mode Control (SMC) [9–11], Linear Quadratic Regulator
(LQR) [12,13], and Model Predictive Control (MPC) [14–16] that directly consider vehicle
dynamics have been developed. SMC, for instance, drives the system states toward a
predefined sliding surface, ensuring rapid convergence to the desired state based on the
designed dynamics. However, these methods are not inherently designed to minimize a
given cost function or performance metric and thereby lack the ability to consider multiple
objectives such as tracking performance and steering stability in the determination of
control actions. In contrast, techniques like LQR and MPC solve optimization problems
based on a multivariable objective function, enabling the derivation of optimal control
actions that balance tracking accuracy and stability. Notably, MPC has emerged as one
of the most widely studied model-based approaches in the field of autonomous driving
control [17]. MPC allows for the establishment of goals considering a variety of variables,
the definition of constraints on system inputs and states during the initial design phase, and
the facilitation of real-time control through a receding horizon technique that continuously
updates the prediction horizon [18].
Given the nature of MPC, which predicts and controls the vehicle’s state based on a
model, using a model that accurately represents the longitudinal characteristics of actual
vehicles can significantly improve tracking performance. Existing MPC studies [14,19,20]
have focused primarily on improving vehicle dynamics models. However, to fully replicate
the longitudinal behavior of real vehicles, it is crucial to consider factors such as power-
train delays in addition to dynamics. These additional factors are vital because control
signals take time to be executed in the actual vehicle and actuator lag further delays the
response, regardless of the vehicle model. This delay means that the calculated longitudinal
commands will not be executed instantly, leading to a model mismatch and deteriorated
performance. As a result, control oscillation and instability can occur.
In longitudinal driving, any vehicle will experience delays in response to control inputs.
As shown in Figure 2, the overall signal delay observed in our experimental electric vehicle
was approximately 0.2 s. Additionally, the powertrain gradually responds after a certain
period when a step wheel torque input is applied. Therefore, powertrain delays can be
decomposed into pure delay and actuation lag. Pure delay arises from CAN communication
and computation processes between the control module and the drive/brake motor system,
and driveline freeplay, where gaps in mechanical components cause delays in torque
transmission, can also contribute to pure delay. Furthermore, actuation lag occurs because
the drive motor or brake actuator requires time to respond to the vehicle, and this lag
is influenced by the hardware capability of the actuator and the design of the low-level
control system. Moreover, mechanical filtering in the driveline, caused by the elasticity of
components such as drive shafts and tires, can exacerbate this lag.
To address these challenges, this paper presents a novel MPC approach for longitudinal
control in autonomous electric vehicles that explicitly incorporates powertrain delays
into the control model using a First Order Plus Dead Time (FOPDT) model [21]. Unlike
existing MPC methods that primarily focus on improving vehicle dynamics models without
accounting for actuator delays, our approach models both the pure time delay and actuation
lag inherent in electric vehicle powertrains. This integration allows the controller to
anticipate and compensate for delays, significantly enhancing speed tracking accuracy and
control stability.
Additionally, we develop detailed torque maps for both the motor and brake sys-
tems, capturing the non-linear torque characteristics of the electric vehicle’s powertrain.
By integrating these torque maps into the control framework, the controller can accu-
rately translate the desired drive force into appropriate throttle and brake inputs, further
improving control precision.
World Electr. Veh. J. 2024, 15, 433 3 of 14

Figure 2. Response of drive motor in the experimental EV. When comparing the measured driv-
ing wheel torque with the commanded driving wheel torque, it can be observed that there is an
approximate delay of 0.2 s, which consists of a pure time delay and an actuation lag.

The key contributions of this work are as follows:

1. By incorporating a FOPDT model into the MPC, the controller effectively handles the
inherent delays in electric vehicle powertrains, which has been largely overlooked in
previous studies.
2. The creation of detailed torque maps for the motor and brake systems, enabling
precise control over throttle and brake inputs in response to the target wheel torque.
3. A comprehensive evaluation of the proposed controller through both simulation and
real-world experiments, demonstrating significant improvements in speed tracking
accuracy and control smoothness compared to conventional MPC and PID controllers.
The remainder of this paper is organized as follows. Section 2 introduces the proposed
system architecture for longitudinal speed tracking control. Section 3 delves into the MPC
considering powertrain delays, while Section 4 discusses the torque map for the throttle
and brake pedals. Section 5 presents the experimental results and analysis from both
simulations and real vehicles. Finally, Section 6 concludes the paper.

2. Overall System Architecture

As illustrated in Figure 3, the proposed longitudinal control system for Autonomous
Electric Vehicles (AEVs) tracks the reference speed profile planned by the higher-level
module and outputs the throttle and brake pedal values for the vehicle. This control system
is composed of two key modules: (1) a longitudinal MPC module considering powertrain
delays and (2) a wheel torque table module for drive/brake torque and throttle/brake
pedal values.
The MPC module first receives a reference speed profile, vre f , composed of target
speeds over a fixed horizon as input, and outputs the target drive force, F. The currently
measured longitudinal speed of the vehicle is used as the initial state of the MPC. The MPC
solves an optimization problem using a model that considers powertrain delays, aiming
to minimize the difference between the given reference speed profile and the predicted
vehicle longitudinal speed.
The wheel torque table module then receives the F output from the MPC module
and converts it to the wheel torque Twheel . Based on the Twheel input, the current vehicle
speed, and the throttle torque table during acceleration, or the brake torque table during
deceleration, it determines the throttle and brake values. These throttle and brake values
are then communicated to the AEV.
This wheel torque table module is designed by acquiring driving data from the vehicle
CAN, which consists of pedal positions, motor torque, acceleration, and speed, and then
interpolating and smoothing the data to accurately reflect the torque characteristics of the
drive motor and brakes.
World Electr. Veh. J. 2024, 15, 433 4 of 14

1) MPC Module 2) Wheel Torque Table Module

Reference
Speed profile Powertrain delayed Throttle torque table

Drive force to wheel torque

𝑣𝑟𝑒𝑓,0 vehicle model
𝑣𝑟𝑒𝑓,1
⋮
𝑣𝑟𝑒𝑓,𝑇
Solve optimization 𝐹𝑑𝑟𝑖𝑣𝑒 𝑇𝑤ℎ𝑒𝑒𝑙
problem
min 𝒥
Δ𝐹(𝑘)
subject to:
Brake torque table
x(𝑘 + 1) = 𝑓 x(𝑘), Δ𝐹(𝑘)
x(0) = xො (0)
⋮

𝑢𝑡ℎ𝑟𝑜𝑡𝑡𝑙𝑒
𝑣𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑
𝑢𝑏𝑟𝑎𝑘𝑒

Experimental AEV

Figure 3. Overall longitudinal control system architecture. The proposed system consists of two main
modules: (1) a longitudinal MPC module that accounts for powertrain delays and (2) a wheel torque
table module for converting drive force to throttle and brake pedal values.

3. MPC Module
3.1. Baseline Vehicle Model
The longitudinal model used in the proposed MPC system is primarily based on
the vehicle’s dynamic model, incorporating factors such as aerodynamic drag, rolling
resistance, and gravitational forces [22] as the following equation:

mv̇ = F − mg sin θ − f mg cos θ − 0.5ρAc Cd v2 , (1)

where v̇ is the vehicle’s longitudinal acceleration; m is the vehicle’s mass; F is the drive
force; θ represents the vehicle’s pitch (substituting for the ground slope); f is the rolling
resistance coefficient; and ρ, Ac , Cd , v represent air density, frontal area of the vehicle, drag
coefficient, and longitudinal velocity, respectively. The primary objective of this model
is to accurately capture the vehicle’s longitudinal acceleration behavior under varying
conditions of force and resistance. To achieve this, the model focuses exclusively on the
longitudinal dynamics of the vehicle, which involves a single degree of freedom related
to the longitudinal velocity. This approach ensures that the model effectively represents
the vehicle’s acceleration response under the influence of longitudinal forces, without
incorporating additional complexities such as lateral dynamics, suspension components,
driveline, or tire models.
Based on Equation (1), a discrete-time model can be established as follows:
" #
Fk −mg sin θ − f mg cos θ −0.5ρACd v2k
x k +1 = kv + m dt , (2)
Fk + ∆Fk dt

where the change in drive force, ∆F, is the input, u, and the speed, v, and drive force, F, are
the states, x. The model update interval is represented by dt, and xk+1 denotes the model
state at time step k + 1.

3.2. Powertrain Delayed Vehicle Model

To account for the delay in the vehicle’s powertrain within the MPC model itself, a
FOPDT model was integrated with the existing baseline model. This integration approxi-
mates the signal delay as dead time and the actuator lag as a first-order system. Although
the time delay and lag for the drive motor and brake system can vary, and different time
World Electr. Veh. J. 2024, 15, 433 5 of 14

delays and lag parameters may be applied depending on acceleration or deceleration

situations, we assume that the delay and lag elements applied to each system are identical.
The FOPDT model can be expressed in the differential equation form of a continuous-time
model as follows:
dy(t)
τ + y ( t ) = u ( t − D ). (3)
dt
In the FOPDT model, u(t) represents the system’s input and y(t) represents the
system’s output at time t. The dead time, D, represents pure signal delay, and the first-
order system with time constant τ corresponds to the actuator’s lag response.
To integrate the FOPDT model into the discrete-time baseline vehicle model, a state
augmentation technique was utilized. To implement the dead-time and first-order system,
the state was augmented from x = [v, F ] to x = [v, Flag , FNd −1 , FNd −2 , . . . , F0 , F ]. The state
Flag was added to represent the first-order system, while F0 to FNd −1 were added to account
for the dead time. After state augmentation, the longitudinal vehicle model considering
powertrain delays is as follows:
F −mg sin θ − f mg cos θ −0.5ρACd v2k
 
v k +1 vk + lag,k dt
 
  −F m +F  
 Flag,k+1   Flag,k +
lag,k Nd ,k
dt 
   τ 
 FN −1,k+1   
 d
x k +1 =  =
  FNd −2,k 
(4)
.
.. ..

   

 F
 
  . 

0,k +1  Fk 
Fk+1
Fk + ∆Fk dt
In this augmented model, the states added for dead time simply update with the
drive force values from the previous steps, ensuring that the control values affect the actual
baseline model after a delay of Nd · dt time. Furthermore, the state Flag , added for the
first-order system, ensures that the control values are gradually applied to the baseline
model according to the time constant τ.
Finally, to transform this model into a Quadratic Programming (QP) form and solve it
as a linear MPC problem, the model was linearized using a Taylor first-order expansion
around the operating point x̂ based on the reference speed profile and the driving force
from previous solved states. This can be expressed mathematically as follows:

∂ f ( x, u) ∂ f ( x, u)
xk+1 ≈ x̂k+1 + ( xk − x̂k ) + (uk − ûk )
∂x x = x̂,u=û ∂u x = x̂,u=û
= Ax (k) + Bu(k) + d
where
!
∂f
A = I + dt (5)
∂x x̂,û
∂f
B = dt
∂u x̂,û
!
∂f ∂f
d= f ( x̂, û) − x̂ − û dt
∂x x̂,û ∂u x̂,û

Applying the powertrain delayed vehicle model to the above equation can be ex-
pressed as follows:
World Electr. Veh. J. 2024, 15, 433 6 of 14

xk+1 = Axk + Buk + d,

2
 
ρA C v̄ d v̄
  
dt −dt − ρAC
1 − dt m d 0 + g sin + f g cos
 
m 0 0 2m θ θ
0 1 − dt dt
0 0
   
 τ τ    0 
0 0 1 0 0 (6)
   
0
    
A= , B =  , C =

..  ..  
.


 . 0
 .
 

 .. 


0 0 0 0 1
 0  
  0 
0 0 0 0 1 dt 0

3.3. MPC Problem Formulation

The longitudinal MPC controller aims to minimize the error between the reference
speed and the amount of change in drive force, solving a multivariable optimization
problem as follows:
" #
T −1 T

u=∆F0 :∆FT
min Q ∑ (xk,ref − xt )2 + R ∑ (uk )2
t =0 k =0
s.t. (7)
xk+1 = Axk + Buk + d,
Fmin < F < Fmax
where Q is a diagonal weight matrix for state error and R is a weight for the change in
driving force, used to find the optimal driving force that minimizes the objective func-
tion. The MPC model utilizes the linearized delayed powertrain vehicle model, with
constraints on the driving force to prevent excessively large acceleration or deceleration.
The resulting quadratic problem was addressed using the gradient-based, interior-point
solver HPIPM [23], which uses the BLASFEO [24] linear algebra software framework. The
HPIPM solver was chosen for its efficiency in handling structured QP problems typical
in MPC applications, particularly due to its ability to exploit problem sparsity for faster
computation, and as an open-source tool, it offers flexibility and accessibility.

4. Wheel Torque Table Module

The output of the MPC module is the driving force, F, while the actual control input
of the AEV is the values of the throttle and brake pedal values. Therefore, the wheel torque
table module is designed to effectively convert F into throttle and brake pedal values. As
illustrated in Figure 4, the F is multiplied by the wheel radius, rwheel , to convert it into the
driving wheel torque, Twheel , in the initial stage of the module.

Twheel = F · rwheel (8)

This torque is then used as the input for the wheel torque tables for throttle and brake.

Wheel Torque Table Module

Drive force to

Throttle
wheel torque

MPC torque table 𝑢𝑡ℎ𝑟𝑜𝑡𝑡𝑙𝑒 CAN

𝐹𝑑𝑟𝑖𝑣𝑒 𝑇𝑤ℎ𝑒𝑒𝑙
𝑇𝑤ℎ𝑒𝑒𝑙 Gateway
Module
Brake 𝑢𝑏𝑟𝑎𝑘𝑒
torque table Experimental AEV

Figure 4. Process of control commands moving from the MPC module to the AEV. The driving force,
Fdrive , output by the MPC module, passes through the wheel torque table module to be converted
into the vehicle’s inputs uthrottle and ubrake .

In the wheel torque table module, the amount of brake or throttle pedal input is
determined based on the target wheel torque input, Twheel . Therefore, modeling is required
World Electr. Veh. J. 2024, 15, 433 7 of 14

for the drive motor used during acceleration and the brake system used during deceleration.
To prevent simultaneous throttle and brake inputs, we use the regenerative braking torque
value when both the throttle and brake pedals are at zero as a neutral value. If Twheel
exceeds this neutral value, the throttle torque table is used; if it is below, the brake torque
table is employed to determine the control inputs.

4.1. For the Throttle Pedal

In conventional internal combustion engine vehicles, converting torque values to
throttle pedal position required torque tables for each gear, taking into account engine RPM
and the current gear ratio [4,25]. However, electric vehicle powertrains mostly operate with
a single gear without a transmission, eliminating the need for gear shifting and thereby
requiring only one torque table for the drive motor. Given that motor torque decreases
as motor RPM increases, achieving a specific torque requires different throttle positions
depending on the speed of the vehicle. Therefore, in this study, motor torque data relative
to vehicle speed and throttle position were collected from both on-road natural driving and
experimental driving on the proving ground.
As shown in the data acquired section of Figure 5, the orange data obtained from the
proving ground tests with constant throttle inputs and the blue data distribution obtained
from normal on-road driving are presented. The necessary vehicle speed, throttle position,
and motor torque data for the table generation were collected through the vehicle CAN.
Then, as illustrated in the wheel torque table of Figure 5, these data were interpolated into
continuous values using the cubic spline approximation method [26] to create a lookup
table. Through this process, when the current speed and target wheel torque values are
input, the appropriate amount of throttle pedal is determined.

Figure 5. Acquired drive wheel torque data from on-road driving and proving ground tests, and
generated drive motor torque table for the throttle. The drive wheel torque data were obtained by
converting the drive motor torque from the vehicle CAN, and these data were interpolated and
smoothed to create a lookup table.

4.2. For the Brake Pedal

The vehicle’s brake system is designed based on hydraulic brakes, which typically
provide braking force proportional to the brake pedal input. Therefore, when determining
the required brake pedal input for a target drive wheel torque, speed information is not
necessary, unlike in the case of the throttle pedal, and the modeling can be conducted based
on the relationship between the brake pedal input and the drive wheel torque. However,
unlike the motor torque used for the throttle pedal, the torque values during braking are
not provided by the vehicle’s CAN bus. Thus, to convert the longitudinal acceleration
data obtained from the CAN into wheel torque, we inversely applied Equations (1) and
World Electr. Veh. J. 2024, 15, 433 8 of 14

(8) inversely to convert the longitudinal acceleration value obtained from the CAN into
wheel torque. Subsequently, as illustrated in Figure 6, this relationship was modeled using
a cubic polynomial, such that when a target wheel torque below the neutral value is input,
the appropriate brake pedal input is determined.

Figure 6. Generated brake torque table. Data were collected by increasing the brake input by 10%
increments from an initial speed of 100 km/h. From 70% brake input onwards, similar brake torque
was observed due to the intervention of the Anti-lock Braking System (ABS).

5. Experiment
The speed tracking performance of the proposed MPC, which considers powertrain
delay, is analyzed against a conventional MPC that does not consider delay and a classical
PID control approach. The main parameters of the proposed MPC are shown in Table 1. A
wheel torque table module is used uniformly across all controllers to ensure consistency
in the evaluation. All these control algorithms were implemented in C++ on the Ubuntu
ROS1 Noetic middleware and were run on an Intel i9-13900TE CPU (Intel, Santa Clara, CA,
USA) with 64 GB of RAM.

Table 1. Main parameters of proposed MPC.

Parameter Symbol Value Units

Model update interval time dt 0.02 s
Number of prediction steps T 100 -
Delayed steps Nd 5 -
Time constant τ 0.15 s
Weight matrix for states Q diag(300, 0, . . . , 0) -
Weight matrix for input R 0.0001 -
Max driving force Fmax 10,819.0 N
Min driving force Fmin −14,485.0 N

5.1. Comparative Controllers

(1) Conventional MPC: This MPC without a delay consideration approach utilizes the
proposed MPC framework but omits the FOPDT model, simplifying the system model
to include only vehicle speed and drive force as a state X without considering dead time
and lag, i.e., X = [v; F ]. The inputs remain the same, u = [∆F ], as in the powertrain delay
considered MPC. The discrete linearized model for this controller is defined as follows:

xk+1 = Axk + Buk + d,

2
" # "  #
ρACd v̄ d v̄ (9)
dt −dt − ρAC
 
1 − dt m m
0 2m + g sin θ + f g cos θ
A= , B= , d=
0 1 dt 0

and all parameters are the same as in the delay considered MPC.
(2) PID: The PID controller is a widely used control technique across various industries
due to its simplicity and effectiveness in a range of applications. The method operates
on the principle of feedback control, where the control action is based on the difference
World Electr. Veh. J. 2024, 15, 433 9 of 14

between the current speed and the reference speed from the speed profile. The speed
profile, part of the vehicle’s reference trajectory, provides target speeds assigned to the
closest trajectory points relative to the vehicle. Through careful tuning of the P, I, and D
gains, the PID controller calculates the appropriate target wheel torque, Twheel , to accurately
follow the target speed.

5.2. Simulation Validation

To evaluate the performance of the proposed MPC considering powertrain delay, we
first compared the proposed MPC with powertrain delay, a conventional MPC, and a PID
controller in the IPG Automotive CarMaker vehicle simulator. The simulation used a
Hyundai IONIQ5 electric vehicle provided by CarMaker, and a wheel torque map module
was specifically designed for the simulated vehicle using data acquired from the simulator.
The vehicle parameters used in this paper were based on the IONIQ5 data from CarMaker,
as shown in Table 2. The powertrain delay in the simulator was modeled as 0.1 s using a
queue buffer, while the lag was modeled as 0.15 s using a first-order system, both applied
just before the throttle and brake inputs. The performance of the proposed MPC with
delay consideration was compared against an MPC without delay consideration and a PID
controller, and validation was performed under two scenarios: (1) a step speed profile from
30 km/h to 50 km/h and (2) a trapezoidal speed profile with acceleration and deceleration
ranging from −4 m/s2 to 4 m/s2 .

Table 2. Vehicle parameters for validation and experimental AEV.

Parameter Symbol Value Units

Vehicle mass m 2300.0 kg
Rolling resistance coefficient f 0.015 -
Mass density of air ρ 1.21 kg/m3
Frontal area of the vehicle A 2.88 m2
Aerodynamic drag coefficient Cd 0.35 -
Wheel radius rwheel 0.32 m

First, the driving results for the step speed profile are shown in Figure 7. In the
subfigure for target wheel torque and commanded throttle, the MPC-based methods began
accelerating earlier than the PID controller, owing to their ability to predict future states
through the prediction horizon. As a result, the PID controller showed the slowest response
to the step speed profile and, therefore, exhibited the lowest speed tracking accuracy. The
conventional MPC, which did not account for powertrain delay, demonstrated slower
convergence compared to the proposed MPC and responded later to the step speed profile.
As shown in Table 3, the proposed MPC exhibited the highest speed tracking accuracy
among the three controllers, with a maximum speed error of 11.48 km/h and an average
speed error of 0.68 km/h.
Second, when examining the tracking performance for the trapezoidal speed profile
shown in Figure 8, similar to the step speed profile scenario, it can be observed that the
proposed MPC, conventional MPC, and PID controller responded quickly to the accel-
eration and deceleration speed profiles in that order. Also, the proposed MPC had the
smallest control overshoot and thus required the smallest amount of control input. Through
this, the proposed MPC’s longitudinal acceleration converged to a stable value without
oscillation during both acceleration and deceleration phases. As shown in Table 4, the
maximum absolute speed error was largest for the PID controller (8.93 km/h), followed
by the conventional MPC (2.19 km/h), and smallest for the proposed MPC (0.77 km/h).
This result indicates that the proposed longitudinal controller quickly responds to the
given speed profile in the presence of powertrain delays and exhibits excellent speed
tracking performance.
World Electr. Veh. J. 2024, 15, 433 10 of 14

Figure 7. Simulation result of step speed profile in CarMaker simulator. This scenario involves
tracking a reference speed that increases from 30 km/h to 50 km/h.

Figure 8. Simulation result of the trapezoidal speed profile in the CarMaker simulator. This scenario
involves tracking a reference speed with acceleration and deceleration ranging from −4 m/s2 to
4 m/s2 .
World Electr. Veh. J. 2024, 15, 433 11 of 14

Table 3. Speed tracking performance of step speed profile in the simulation.

Max Speed Error (km/h) Mean Speed Error (km/h)

PID 20.69 2.13
Conventional MPC 14.68 1.09
Proposed MPC 11.48 0.68

Table 4. Speed tracking performance of trapezoidal speed profile in the simulation.

Max Speed Error Mean Speed Error Mean Acceleration

(km/h) (km/h) Error (m/s2 )
PID 8.93 1.43 0.59
Conventional MPC 2.19 0.47 0.45
Proposed MPC 0.77 0.29 0.18

5.3. Real-World Vehicle Experiment

For the real vehicle experiments, the same speed profile used in the simulator veri-
fication was used, with additional acceleration and deceleration phases added towards
the end. As shown in Figure 9, these tests were conducted at a highway section of the
K-City, the autonomous driving testbed of the Korea Automotive Technology Institute.
The experimental vehicle was a Hyundai IONIQ5 AWD model, and the wheel torque
map module was also developed based on real vehicle experimental data. Vehicle pose
estimation utilized results from a Novatel CPT7 dual antenna RTK-GNSS/INS solution,
and vehicle control inputs, including throttle and brake pedal positions, were applied
through a CAN/CAN-FD gateway.

Figure 9. Experimental AEV Platform, Hyundai IONIQ5, and the test site, K-City highway section.

In the real-vehicle experiments, the same three controllers used in the simulation
validation were tested for speed tracking performance, and the experimental results are
presented in Figure 10. In the speed error subplot, the PID controller shows high errors
during rapid acceleration and deceleration due to its delayed response in following the
reference speed. Additionally, while the conventional MPC reduced the speed error with a
faster response than the PID controller, thanks to its predictive characteristics, the target
wheel torque subplot shows that it converged more slowly than the proposed MPC. As
shown in Table 5, the maximum absolute speed error was highest for the PID controller
(8.53 km/h), followed by the proposed MPC (1.98 km/h), and the conventional MPC
(1.71 km/h). The ranking for mean absolute speed error differed slightly, with the PID
controller showing the highest error (2.12 km/h), followed by the conventional MPC (0.88
km/h), and the proposed MPC exhibiting the lowest error (0.54 km/h).
Compared to the simulation, tracking errors increased due to external disturbances
and approximations in the actual model. However, the proposed MPC still showed the best
performance in terms of mean error. The PID controller exhibited significant overshoots in
acceleration and deceleration due to its delayed response. While the conventional MPC
demonstrated speed tracking performance similar to the proposed MPC, it showed larger
overshoots in the target wheel torque during the acceleration and deceleration phases
World Electr. Veh. J. 2024, 15, 433 12 of 14

compared to the proposed algorithm. As confirmed in Figure 11 and Table 6, although

the computation time increased due to the additional number of states, the proposed
MPC sufficiently maintained real-time performance, with an average computation time of
1.32 ms, satisfying the CAN data cycle requirement of 10 ms.

Table 5. Speed tracking performance in the real-vehicle experiment.

Max Speed Error Mean Speed Error Mean Acceleration

(km/h) (km/h) Error (m/s2 )
PID 8.53 2.12 0.48
Conventional MPC 1.71 0.88 0.32
Proposed MPC 1.98 0.54 0.22

Figure 10. Real-vehicle experiment result of the trapezoidal speed profile in K−City. This scenario
involves tracking a reference speed with acceleration and deceleration ranging from −4 m/s2 to
4 m/s2 .

Figure 11. Computational time of each comparative controller.

World Electr. Veh. J. 2024, 15, 433 13 of 14

Table 6. Mean computational time of each comparative controller.

Mean Computation Time (millisec)

PID 0.02
Conventional MPC 0.74
Proposed MPC 1.32

6. Conclusions
This paper proposes a longitudinal control system that considers powertrain delay
to enhance speed tracking performance in autonomous electric vehicles. The longitudinal
controller models the actuator lag and signal delays as a FOPDT model, incorporating them
into the MPC module. The resulting wheel torque output is converted into the throttle and
brake pedal positions using a torque table. The designed controllers were implemented and
tested on a Hyundai IONIQ5, demonstrating reduced speed tracking errors and smoother
control behaviors. Despite an increase in the number of states, real-time performance is
still maintained with a response time of less than 10 ms. The proposed controller has been
validated through both simulator and real-vehicle experiments, offering superior speed
tracking performance and eliminating acceleration and deceleration overshoots compared
to traditional PID and MPC methods, which do not consider powertrain delays. Future
work will focus on online estimation of powertrain delays, which is expected to reduce the
time and effort required for system calibration and further improve the control system’s
adaptability and performance.

Author Contributions: Conceptualization, J.L.; methodology, J.L.; software, J.L.; validation, J.L.;
formal analysis, J.L.; investigation, J.L.; resources, J.L.; data curation, J.L.; writing—original draft
preparation, J.L.; writing—review and editing, J.L.; visualization, J.L.; supervision, K.J.; project
administration, K.J.; funding acquisition, K.J. All authors have read and agreed to the published
version of the manuscript.
Funding: This research was supported by a Korea Institute for Advancement of Technology (KIAT)
grant funded by the Korean Government (MOTIE) (P0020536, HRD Program for Industrial Innova-
tion) and a National Research Foundation of Korea (NRF) grant funded by the Korean government
(MSIT) (No.RS-2023-00209252).
Data Availability Statement: The original contributions presented in the study are included in the
article; further inquiries can be directed to the corresponding author.
Acknowledgments: This research was originally presented as an oral poster at the EVS37 conference
held in Seoul, Republic of Korea, on 25 April 2024. We extend our heartfelt gratitude to the EVS37
program committee for the esteemed opportunity to contribute this paper to the Special Issue of the
World Electric Vehicle Journal.
Conflicts of Interest: The authors declare no conflicts of interest.

References
1. Geamanu, M.S.; Mounier, H.; Niculescu, S.I.; Cela, A.; LeSolliec, G. Longitudinal control for an all-electric vehicle. In Proceedings
of the 2012 IEEE International Electric Vehicle Conference, Greenville, SC, USA, 4–8 March 2012; pp. 1–6.
2. He, H.; Han, M.; Liu, W.; Cao, J.; Shi, M.; Zhou, N. MPC-based longitudinal control strategy considering energy consumption for
a dual-motor electric vehicle. Energy 2022, 253, 124004. [CrossRef]
3. El Majdoub, K.; Giri, F.; Ouadi, H.; Chaoui, F. Nonlinear cascade strategy for longitudinal control of electric vehicle. J. Dyn. Syst.
Meas. Control. 2014, 136, 011005. [CrossRef] [PubMed]
4. Hamersma, H.A.; Els, P.S. Longitudinal vehicle dynamics control for improved vehicle safety. J. Terramechanics 2014, 54, 19–36.
[CrossRef]
5. Nie, L.; Guan, J.; Lu, C.; Zheng, H.; Yin, Z. Longitudinal speed control of autonomous vehicle based on a self-adaptive PID of
radial basis function neural network. IET Intell. Transp. Syst. 2018, 12, 485–494. [CrossRef]
6. Lu, C.; Gong, J.; Lv, C.; Chen, X.; Cao, D.; Chen, Y. A personalized behavior learning system for human-like longitudinal speed
control of autonomous vehicles. Sensors 2019, 19, 3672. [CrossRef] [PubMed]
7. Xu, S.; Peng, H.; Song, Z.; Chen, K.; Tang, Y. Accurate and smooth speed control for an autonomous vehicle. In Proceedings of
the 2018 IEEE Intelligent Vehicles Symposium (IV), Changshu, China, 26–30 June 2018; pp. 1976–1982.
World Electr. Veh. J. 2024, 15, 433 14 of 14

8. MacAdam, C.C. Application of an optimal preview control for simulation of closed-loop automobile driving. IEEE Trans. Syst.
Man Cybern. 1981, 11, 393–399. [CrossRef]
9. Nouveliere, L. Experimental vehicle longitudinal control using a second order sliding mode technique. Control. Eng. Pract. 2007,
15, 943–954. [CrossRef]
10. Ferrara, A.; Incremona, G.P. Sliding modes control in vehicle longitudinal dynamics control. In Advances in Variable Structure
Systems and Sliding Mode Control—Theory and Applications; Springer: Berlin/Heidelberg, 2018; pp. 357–383.
11. Jo, A.; Lee, H.; Seo, D.; Yi, K. Model-reference adaptive sliding mode control of longitudinal speed tracking for autonomous
vehicles. Proc. Inst. Mech. Eng. Part D J. Automob. Eng. 2023, 237, 493–515. [CrossRef]
12. Perng, J.W.; Lai, Y.H. Robust longitudinal speed control of hybrid electric vehicles with a two-degree-of-freedom fuzzy logic
controller. Energies 2016, 9, 290. [CrossRef]
13. Jung, H.; Jung, D.; Choi, S.B. LQR control of an all-wheel drive vehicle considering variable input constraint. IEEE Trans. Control.
Syst. Technol. 2021, 30, 85–96. [CrossRef]
14. Matute, J.A.; Marcano, M.; Zubizarreta, A.; Perez, J. Longitudinal model predictive control with comfortable speed planner.
In Proceedings of the 2018 IEEE International Conference on Autonomous Robot Systems and Competitions (ICARSC), Torres
Vedras, Portugal, 25–27 April 2018; pp. 60–64.
15. Sun, X.; Cai, Y.; Wang, S.; Xu, X.; Chen, L. Optimal control of intelligent vehicle longitudinal dynamics via hybrid model
predictive control. Robot. Auton. Syst. 2019, 112, 190–200. [CrossRef]
16. Qiu, W.; Ting, Q.; Shuyou, Y.; Hongyan, G.; Hong, C. Autonomous vehicle longitudinal following control based on model
predictive control. In Proceedings of the 2015 34th Chinese Control Conference (CCC), Hangzhou, China, 28–30 July 2015;
pp. 8126–8131.
17. Yu-Geng, X.; De-Wei, L.; Shu, L. Model predictive control—Status and challenges. Acta Autom. Sin. 2013, 39, 222–236.
18. Huang, Y.; Wang, H.; Khajepour, A.; He, H.; Ji, J. Model predictive control power management strategies for HEVs: A review. J.
Power Sources 2017, 341, 91–106. [CrossRef]
19. He, Z.; Gong, L.; Zhou, E.; Wei, B.; Li, E.; Huang, J. Lateral and Longitudinal Coordinated Control of Intelligent Vehicle Based on
High-Precision Dynamics Model under High-Speed Limit Condition. IEEE Trans. Intell. Veh. 2024.. [CrossRef]
20. Chen, S.p.; Xiong, G.m.; Chen, H.y.; Negrut, D. MPC-based path tracking with PID speed control for high-speed autonomous
vehicles considering time-optimal travel. J. Cent. South Univ. 2020, 27, 3702–3720. [CrossRef]
21. Bagheri, P.; Sedigh, A.K. Analytical approach to tuning of model predictive control for first-order plus dead time models. IET
Control. Theory Appl. 2013, 7, 1806–1817. [CrossRef]
22. Rajamani, R. Vehicle Dynamics and Control; Springer Science & Business Media: Berlin/Heidelberg, Germany, 2011.
23. Frison, G.; Diehl, M. HPIPM: A high-performance quadratic programming framework for model predictive control. IFAC-
PapersOnLine 2020, 53, 6563–6569. [CrossRef]
24. Frison, G.; Kouzoupis, D.; Sartor, T.; Zanelli, A.; Diehl, M. BLASFEO: Basic linear algebra subroutines for embedded optimization.
ACM Trans. Math. Softw. TOMS 2018, 44, 1–30. [CrossRef]
25. Dekraker, P.; Barba, D.; Moskalik, A.; Butters, K. Constructing Engine Maps for Full Vehicle Simulation Modeling; Technical Report,
SAE Technical Paper; SAE International: Warrendale, PA, USA, 2018.
26. De Boor, C.; De Boor, C. A Practical Guide to Splines; Springer: New York, NY, USA, 1978; Volume 27.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.