Article

Research on Active Obstacle Avoidance of Intelligent Vehicles

Based on Improved Artificial Potential Field Method
Jing Tian 1 , Shaoyi Bei 2, *, Bo Li 2, * , Hongzhen Hu 2 , Zhenqiang Quan 2 , Dan Zhou 2 , Xinye Zhou 2
and Haoran Tang 2

1 College of Mechanical Engineering, Jiangsu University of Technology, Changzhou 213001, China;

tj_0506@163.com
2 College of Automobile and Traffic Engineering, Jiangsu University of Technology, Changzhou 213001, China;
huhongzhen7171@163.com (H.H.); quanzhenqiang@163.com (Z.Q.); zd13952018530@163.com (D.Z.);
e2090391825@163.com (X.Z.); gelare594054103@163.com (H.T.)
* Correspondence: bsy1968@126.com (S.B.); bolifly311@gmail.com (B.L.)

Abstract: In the study of autonomous obstacle avoidance of intelligent vehicles, the traditional
artificial potential field method has the problem that the vehicle may fall into the local minima and
lead to obstacle avoidance failure. Therefore, this paper improves the traditional potential field
function. Based on the vehicle dynamics model, a strategy of jumping out of local minima based
on smaller steering angles is proposed. By finding a smaller steering angle and setting a suitable
jump out step length, the intelligent vehicle is enabled to jump out of the local minima. Simulation
experiments by MATLAB show that the improved method can jump out of the local minima. By
comparing the planned trajectories of the traditional method and the improved method in static
and dynamic obstacles situations, the trajectory planned by the improved method is smooth and the
curvature is smaller. The planned trajectory is tracked by the Carsim platform, and the test results
show that the improved method reduces the front steering wheel angle while the intelligent vehicle
Citation: Tian, J.; Bei, S.; Li, B.; Hu, satisfies the vehicle dynamics constraints during active obstacle avoidance, which verifies the stability
H.; Quan, Z.; Zhou, D.; Zhou, X.; and rationality of the improved method.
Tang, H. Research on Active Obstacle
Avoidance of Intelligent Vehicles Keywords: artificial potential field; local minima; path planning; intelligent vehicle
Based on Improved Artificial
Potential Field Method. World Electr.
Veh. J. 2022, 13, 97. https://doi.org/
10.3390/wevj13060097
1. Introduction
Academic Editor: Joeri Van Mierlo Intelligent vehicles are advanced cars that integrate technologies such as computer
Received: 22 April 2022
science, sensors, and data processing. It has received extensive attention and research in
Accepted: 23 May 2022
academia and has been applied to some extent. Currently, most of the intelligent vehicles
Published: 28 May 2022
use advanced sensing technology to obtain vehicle location, speed, and other data through
data fusion technology to achieve the extraction and analysis of characteristic data infor-
Publisher’s Note: MDPI stays neutral
mation between “human-vehicle-road”, and finally making the vehicle environmentally
with regard to jurisdictional claims in
aware, enabling autonomous decision-making and motion planning capabilities [1].
published maps and institutional affil-
Path planning, as an important technology for intelligent vehicles to achieve au-
iations.
tonomous driving, frequently refers to motions of a vehicle in a 2D or 3D world that
contains obstacles [2]. The path planning technology of intelligent vehicles is to analyze
and process the data collected by vehicle sensors and plan the vehicle trajectory inde-
Copyright: © 2022 by the authors.
pendently without a human driver through a certain algorithm. This technology can not
Licensee MDPI, Basel, Switzerland. only plan the appropriate trajectory, but also reduce the traffic accidents caused by road
This article is an open access article environment changes and driver’s operation errors, which is important to build a safe,
distributed under the terms and efficient, and convenient driving environment.
conditions of the Creative Commons According to the degree of grasp of environmental information, path planning can
Attribution (CC BY) license (https:// be divided into global path planning based on a priori complete information and local
creativecommons.org/licenses/by/ path planning based on sensor information [3,4]. The path planning techniques applied
4.0/). in automated driving scenarios can be roughly divided into four categories: graph search

World Electr. Veh. J. 2022, 13, 97. https://doi.org/10.3390/wevj13060097 https://www.mdpi.com/journal/wevj

World Electr. Veh. J. 2022, 13, 97 2 of 21

based planners, sampling based planners, interpolating curve planners, and numerical
optimization approaches [5]. Here are common path planning algorithms used in early
applications for intelligent vehicles: (1) Grid method [6,7]: The grid method was earlier
applied to the path planning of robots. It divides the robot’s workspace into a grid, and the
size of the grid cells is determined by the smallest rectangular space in which the robot can
move freely, and the path is planned by calculating the shortest distance between the grids.
(2) Dijkstra algorithm: Dijkstra’s algorithm was proposed by the Dutch computer scientist
Dijkstra in 1959 [8]. It is a classic graph searching algorithm; however, it has disadvantages
such as high data computation and low efficiency. (3) A* algorithm [9,10]: It is an extension
of Dijkstra’s graph search algorithm. Its most important design aspect is the determination
of the cost function, which defines the weights of the nodes [5]. (4) RRT (rapidly exploring
random trees) algorithm [11,12]: RRT algorithm is a random sampling planning algorithm.
It permits fast planning in semi-structured spaces by performing random searches in the
navigation area [13].
The classical graph searching algorithms and sampling-based algorithm mentioned
above have shortcomings and drawbacks. Vega-Brown et al. [14] proposed some asymp-
totically optimal algorithms for motion planning problems, the ideas in these algorithms
can likely be combined with the ideas in many other sampling-based motion planning and
graph search algorithms. Other scholars have also proposed some novel path planning
algorithms based on advanced sensors and natural sciences: Nakrani et al. [15] designed a
fuzzy-based obstacle avoidance navigation controller, which obtains information from an
ultrasonic sensor array. Chen Y. et al. [16] proposed padding mean neural dynamic model
(PMNDM), planning paths by transmitting nerve impulses in a topologically organized
network. Jafari M. et al. [17] proposed a novel biologically inspired approach based on a
computational model of emotional learning in mammalian limbic system, and it is applied
for the first time in a synthetic UAV (Unmanned Aerial Vehicles) path planning scenario.
In real driving scenarios, when the drivers or sensors detect obstacles, they will slow
down or steer to avoid the obstacles. This shows that the obstacle avoidance behavior of the
vehicle depends on whether the location of the obstacles will influence the driving safety of
the vehicle. Although the vehicle does not contact the obstacles during obstacle avoidance,
the vehicle’s motion is altered. This indicates that there is at least one virtual force acting
on the vehicle, and this virtual force belongs to a field force [18]. This means that there is a
potential field that changes the motion of the vehicle during the obstacle avoidance process.
Therefore, we can explain the phenomenon of vehicle obstacle avoidance process in terms
of fields and field forces.
The artificial potential field (APF) method has received much attention and research
for its simplicity of calculation and high real-time performance. The APF method was first
proposed by Khatib [19] in 1986 and it is commonly used for local path planning. The
principle of APF method is to transform the real environment into a virtual potential field,
which is the attractive field generated by the target and the repulsive field generated by the
obstacle, and the vector superposition of the attractive field and repulsive field forms the
combined potential field. Ultimately, the intelligent vehicle plans its path under the action
of the combined potential field.
However, the traditional APF method has many drawbacks, the most typical of which
is the problem of local minima and unable to reach target. To address the problem of
local minima, scholars at home and abroad have proposed several improved methods.
Choe, T.S. et al. [20] proposed the concept of steering potential fields, using overlapping
integrated force fields to avoid obstacles and follow the planned path. Huang, Y. et al. [21]
generate a diamond-shaped grid on a local map, add a voltage source between the starting
point and the target, assign the resistance value of each edge on the grid by the APF method,
and plan the local path by calculating the current maximum. Fan, S. [22] optimized the
gravity model by setting tracking target points and optimized the potential field function
by adjusting factors to eliminate local minima. Li, E. [23] proposed the SOPC-APF method
World Electr. Veh. J. 2022, 13, 97 3 of 21

that introduces the idea of collision prediction and makes decisions before the robot enters
the trap area or the minimum point problem.
APF method can also be combined with other intelligent algorithms. Bounini, F. et al. [24]
find the
World Electr. Veh. J. 2022, 13, x FOR PEER REVIEW
actual path in the potential field according to a potential gradient3 descent of 21
algorithm
and add a repulsive potential field to the current state with a local minimum. Cummings,
M.L. [25] studied the path planning problem by combining the ant colony algorithm with
the APF algorithm.
APF method can alsoLuo, D. et al. [26].
be combined proposed
with other to use
intelligent the gridBounini,
algorithms. method F. to
et establish the
planning
al. [24] findspace andpath
the actual use in
thetheant colonyfield
potential algorithm
according astothe global path
a potential search
gradient strategy to search
descent
algorithm and addpath.
for the optimal a repulsive potentialdue
In addition, fieldtotothe
the early
current state with a of
application local
theminimum.
APF method to robot
Cummings,
kinematics,M.L. the [25] studied constraints
restrictive the path planningof theproblem
intelligentby combining
vehicle mustthe antbe colony
considered when it
algorithm with the APF algorithm. Luo, D. et al. [26]. proposed to use the grid method to
is referenced as a local path planning algorithm for intelligent vehicles.
establish the planning space and use the ant colony algorithm as the global path search
To address the above problems, this paper first improves the potential field functions.
strategy to search for the optimal path. In addition, due to the early application of the APF
method tothe
Second, robot road boundary
kinematics, potentialconstraints
the restrictive field is added to constrain
of the intelligent themust
vehicle motion
be region and
limit the when
considered lateral it ismotion of as
referenced the intelligent
a local vehicle.
path planning Finally,
algorithm combining
for intelligent with the vehicle
vehicles.
dynamics
To addressmodel, a strategy
the above problems, for this
jumping
paper outfirst of local minima
improves based
the potential onfunc-
field smaller steering
tions.
angles Second, the road when
is proposed boundary thepotential field vehicle
intelligent is added falls
to constrain the motion
into local minima. regionThe intelligent
and limitwill
vehicle the lateral
searchmotion of the intelligent
for potential jump-out vehicle.
pointsFinally,
in thecombining
direction with
of the vehicle
smaller steering angles
dynamics model, a strategy for jumping out of local minima based on smaller steering
within the maximum steering angle range of the intelligent vehicle. The improved APF
angles is proposed when the intelligent vehicle falls into local minima. The intelligent ve-
method
hicle will is verified
search and analyzed
for potential jump-out bypoints
MATLAB in the and Carsim
direction platform,
of smaller andangles
steering the improved APF
method can successfully jump out of the local minima,
within the maximum steering angle range of the intelligent vehicle. The improved APF at the same time, comparing the
method is verified and analyzed by MATLAB and Carsim platform, and the improved environment.
trajectory curvature before and after the improvement in the same obstacle
The method
APF trajectories planned byjump
can successfully the out
improved APF
of the local method
minima, hassame
at the smaller
time,curvature
comparingand are easier
the trajectory curvature before and after the improvement
to be tracked under the same Carsim tracking model. This paper provides in the same obstacle environ-a new idea for
ment. The trajectories planned by the improved APF
intelligent vehicle path planning based on potential field theory. method has smaller curvature and
are easier to be tracked under the same Carsim tracking model. This paper provides a new
idea for intelligent vehicle path planning based on potential field theory.
2. APF Method
2.1.
2. Traditional
APF Method APF Method
The principle
2.1. Traditional of APF is to assume that the vehicle is in a virtual space; the attractive
APF Method
potential field generated
The principle of APF is toby the target
assume that theand the isrepulsive
vehicle in a virtualpotential
space; thefield generated by the
attractive
obstacles.
potential fieldAttractive
generated byfield
the has
targetlow
andpotential energy
the repulsive andfield
potential both start point
generated by theand repulsive
obstacles.
field have Attractive field has low
high potential potential
energy. Theenergy and both start
superposition point and
of these tworepulsive
potentialfield
fields forms an
have high potential
undulating spaceenergy.
as shownThe in
superposition of these two potential fields forms an un-
Figure 1 [27].
dulating space as shown in Figure 1 [27].

Figure
Figure1.1.Potential fields.
Potential fields.

InInattractive
attractivefield, the the
field, magnitude
magnitudevariesvaries
with the distance
with between between
the distance the vehiclethe
posi-
vehicle position
tion and the target. The potential field function in this paper is referenced to Zhu, W.
and the target. The potential field function in this paper is referenced to Zhu, W. [28] and[28]
and the attractive potential field is defined as:
the attractive potential field is defined1as:
𝑈 (𝑋) = 𝐾 𝜌 (𝑃, 𝑃 ) (1)
2
1 2


where 𝑈 (𝑋) is the attractive potential Ufield, ) =is the
att ( X𝐾 Ka ρ P, Pg (1)
2 attractive field coefficient, 𝜌(𝑃, 𝑃 )
is a vector whose magnitude is the Euclidean distance between the vehicle position 𝑃 and
the target
Uposition 𝑃 , the direction is the vehicle position toward the target position.


where att ( X ) is the attractive potential field, K a is the attractive field coefficient, ρ P, Pg
is a vector whose magnitude is the Euclidean distance between the vehicle position P and
the target position Pg , the direction is the vehicle position toward the target position.
World Electr. Veh. J. 2022, 13, 97 4 of 21

The attractive force is the negative gradient of the attractive potential field which can
World Electr. Veh. J. 2022, 13, x FOR PEER REVIEW 4 of 21
be calculated from the following equation:


Fatt ( X ) = −∇Uatt ( X ) = Ka ρ P, Pg (2)
The attractive force is the negative gradient of the attractive potential field which can
be calculated
Repulsive fields arefrom the following
virtual potentialequation:
fields generated by obstacles. Each obstacle has
its own range of influence. When the vehicle 𝐹 (𝑋)is=not −∇𝑈 within
(𝑋) the
= 𝐾influence
𝜌(𝑃, 𝑃 ) range of an obstacle,(2)
the potential energy magnitude of the vehicle is zero; when the vehicle enters the influence
Repulsive fields are virtual potential fields generated by obstacles. Each obstacle has
range of the obstacle, the potential energy of the vehicle varies with the distance between
its own range of influence. When the vehicle is not within the influence range of an obsta-
the vehicle position and the obstacles. The repulsive potential field is defined as:
cle, the potential energy magnitude of the vehicle is zero; when the vehicle enters the in-
fluence range of the(obstacle, the potential 2energy of the vehicle varies with the distance
1 1
between 2 Kr ( ρand − ρ10 ) 0 < ρ( P, Pobs ) ≤potential
ρ0
Uthe vehicle position
rep ( X ) =
the
( P,Pobs ) obstacles. The repulsive field is defined
(3) as:
0 ρ( P, Pobs1) > ρ0 1 1
𝐾( − ) 0<𝜌(𝑃, 𝑃 )≤𝜌
𝑈 (𝑋) = 2 𝜌(𝑃, 𝑃 ) 𝜌 (3)
where Urep ( X ) is the repulsive potential field,0 𝜌(𝑃,K𝑃r is )the> 𝜌repulsive field coefficient, ρ( P, Pobs )
is a vector whose magnitude is the Euclidean distance between the vehicle position P and
the obstaclewhere 𝑈 P (𝑋), the
position is the repulsive
direction potential
is from field, 𝐾to the
the obstacle is the repulsive
vehicle, ρ0 is field coefficient,
the radius of
obs
𝜌(𝑃, 𝑃 ) is a vector whose magnitude is the Euclidean distance between the vehicle po-
the obstacle’s influence range.
sition 𝑃 and the obstacle position 𝑃 , the direction is from the obstacle to the vehicle,
The repulsive force is the negative gradient of the repulsive potential field which can
𝜌 is the radius of the obstacle’s influence range.
be calculated from the following equation:
The repulsive force is the negative gradient of the repulsive potential field which can
be calculated from the following
(  equation: 
1
Kr ρ( P,P − ρ10 ρ2 ( P,P
1
0 <ρ( P, Pobs ) ≤ ρ0
Frep ( X ) = −∇Urep ( X ) = obs ) 1 1 obs ) 1 (4)
0 ρ ( P, P𝐾 () > ρ − ) 0<𝜌(𝑃, 𝑃 ) ≤ 𝜌
𝐹 (𝑋) = −∇𝑈 (𝑋) = obs 𝜌(𝑃, 𝑃 ) 𝜌 𝜌 (𝑃, 𝑃 )
0 (4)
0 𝜌(𝑃, 𝑃 )>𝜌
Vehicle is often affected by the repulsive field of multiple obstacles while moving
Vehicle
toward the target. is oftenthe
Therefore, affected by theforce
combined repulsive
fieldfield
is a of multiple obstacles
superposition of anwhile moving
attractive
toward the target. Therefore,
field and multiple repulsive fields. the combined force field is a superposition of an attractive
field and multiple repulsive fields.
The combined force potential field function can be expressed as:
The combined force potential field function can be expressed as:
n
( X )=+𝑈 ∑(𝑋)
Utotal ( X ) =𝑈 Uatt(𝑋) Urep
+ ( X )𝑈 (𝑋) (5) (5)
i =1

In the aboveIn equation,

the above equation, (𝑋)
Utotal ( X )𝑈is the is the combined
combined potential
potential field field n is 𝑛the
and and is number
the number
of
of obstacles. obstacles.
The combinedThe combined
force is the force is the negative
negative gradient gradient
of the of the potential
force force potential
fieldfield
andandcancan
be be
expressed as:
expressed as:
Ftotal ( X ) = 𝐹−∇U (𝑋) =(−∇𝑈
total X ) = Fatt(𝑋)
( X=) 𝐹
+ F(𝑋)
rep (+X )𝐹 (𝑋) (6) (6)
Ultimately, Ultimately,
as shown in as shown
Figurein
2,Figure 2, the combined
the combined forces forces control
control the motion
the motion of the
of the vehicle.
vehicle.

Frep
Ftotal
Ftotal
Frep
Fatt
Fatt
Fatt Fatt

Target
Obstacle

The range of
obstacle influence

Figure
Figure 2. Forces on 2.the
Forces on the vehicle.
vehicle.
World Electr. Veh. J. 2022, 13, 97 5 of 21
World Electr. Veh. J. 2022, 13, x FOR PEER REVIEW 5 of

2.2. Shortcomings of Traditional APF Method

2.2. Shortcomings of Traditional APF Method
1. Because the attractive force is proportional to the distance between vehicle and target,
1. attractive
an excessive Because the attractive
force forcethe
will cause is vehicle
proportional
to hittoa the distance
nearby between
obstacle whenvehicle
the and ta
get, an excessive
vehicle is far from the target. attractive force will cause the vehicle to hit a nearby obstacle whe
2. the vehicle is far from the target.
The attractive force generated by the target is zero when the vehicle reaches the target.
2. The attractive force generated by the target is zero when the vehicle reaches the ta
Assuming there is an obstacle near the target, the vehicle will still be repelled away
get. Assuming there is an obstacle near the target, the vehicle will still be repelle
from the target by the repulsive force generated by the obstacle at this time, which
away from the target by the repulsive force generated by the obstacle at this tim
causes the vehicle to oscillate near the target and cannot reach it;
which causes the vehicle to oscillate near the target and cannot reach it;
3. In practical urban driving scenarios, vehicles must move on a defined road and cannot
3. In practical urban driving scenarios, vehicles must move on a defined road and ca
move outside the road boundaries. Therefore, adding a road boundary potential field
not move outside the road boundaries. Therefore, adding a road boundary potenti
to limit the lateral motion of the vehicle is necessary;
field to limit the lateral motion of the vehicle is necessary;
4. Assume that the vehicle is at a point where the combined force on the vehicle is zero.
4. Assume that the vehicle is at a point where the combined force on the vehicle is zer
The vehicle will Thefall into awill
vehicle local
fallminima if theminima
into a local vehicleifdoes not reach
the vehicle doesthe
nottarget.
reach It
theistarget. It
the most common fault of the traditional APF method.
the most common fault of the traditional APF method.
3. Vehicle Dynamics Model
3. Vehicle Dynamics Model
The vehicle dynamics
The vehiclemodel is verymodel
dynamics essential
is veryfor essential
designing forfeasible
designingtrajectory.
feasible The
trajectory. Th
planned trajectory should be limited within the vehicle dynamics constraints
planned trajectory should be limited within the vehicle dynamics constraints [29], and [29], and th
the two-degree-of-freedom vehicle dynamics
two-degree-of-freedom model established
vehicle dynamics in this paper
model established in thisispaper
shown in
is shown in Fi
Figure 3. In this
urefigure: Fy f figure:
3. In this and Fyr𝐹represent
and 𝐹 the lateral force
represent of theforce
the lateral frontofwheel and wheel
the front the and th
rear wheel; Vx rear Vy are the
and wheel; 𝑉 andlongitudinal
𝑉 are thespeed and thespeed
longitudinal lateralandspeed, respectively;
the lateral ϕ is
speed, respectively;
the yaw angle;isβ the
is the lateral deflection angle of the center of mass;
yaw angle; 𝛽 is the lateral deflection angle of the center α f and α r ,
of mass; denote
𝛼 and the 𝛼 , deno
slip angle of the
thefront
slip and
anglerear axle,
of the respectively;
front l f isrespectively;
and rear axle, the distance 𝑙from thedistance
is the center of mass
from the center
to the front axle; l is the distance from the center of mass to the rear axle; represents
mass to the front axle; 𝑙 is the distance from the center of mass to the rear axle; 𝛿 rep
r δ the
front tire angle.resents the front tire angle.

Y Vf
y lf αf
lr
V δ
Vy x
Vr β
φ Vx
αr Fyf

Fyr X

Figure 3. 2-DOFFigure 3. 2-DOF(degree-of-freedom)

(degree-of-freedom) vehicle
vehicle dynamics dynamics model.
model.

Based on
Based on the balance the balance
of forces in the of forces
y-axis in the y-axis
direction, direction,
the vehicle the vehicle
dynamics dynamics equili
equilibrium
rium equation
equation can be expressed as:can be expressed as:
Fy f cosδ + Fyr =𝐹 ma
𝑐𝑜𝑠𝛿
y + 𝐹 = 𝑚𝑎 (7) (
The relationship
Thebetween lateral
relationship force of
between the front
lateral force wheel and the
of the front rearand
wheel wheel, lateral
the rear wheel, later
deflection angle and lateral deflection stiffness can be expressed as:
deflection angle and lateral deflection stiffness can be expressed as:

Fy f = C f α f 𝐹 = 𝐶 𝛼

(8) (
Fyr = Cr αr 𝐹 =𝐶 𝛼
where 𝐶 and 𝐶 are the front wheel lateral deflection stiffness and the rear wheel later
where C f and Cr are the front wheel lateral deflection stiffness and the rear wheel lateral
deflection stiffness, respectively. According to the coordinate system, the lateral deflectio
deflection stiffness,
anglerespectively. According
of the front and to the
rear wheels of coordinate system,
the vehicle can the lateral
be expressed as:deflection
angle of the front and rear wheels of the vehicle can be expressed as:
World Electr. Veh. J. 2022, 13, 97 6 of 21

 .
 α = ϕl f +Vy − δ
f Vx. (9)
 α = Vy − ϕlr
r Vx

Assuming the front tire angle δ at a small angle, we can regard cosδ ≈ 1. According to
the Equations (8) and (9), Equation (7) can be transformed as:
. . ! . . 
.. .
ϕl f + y y − ϕlr
m y + Vx ϕ = C f − δ + Cr (10)
Vx Vx

Note that the velocity and acceleration in the y-axis direction can be expressed as:
 .
Vy = y
.. . (11)
ay = y + Vx ϕ

Finally, the vehicle dynamics model based on the front tire angle δ can be expressed as:

C f + Cr . l f C f − lr Cr Cf
 
.. .
y= y+ − Vx ϕ − δ (12)
mVx mVx m

4. Improved APF Method

Based on the previous analysis of the traditional APF method, the details of the
improved APF method are as follows:

4.1. Potential Field Functions

To avoid the problem of excessive initial attractive force of the vehicle at the starting
point, the attractive field function could be modified. The improved attractive potential
field function can be expressed as:



εKa ρ P, Pg
ρ P, Pg
≥ d0
Uatt ( X ) = 1 2 (13)
2 K a ρ P, Pg ρ P, Pg < d0

The corresponding attractive force can be expressed as:


ρ( P,Pg )

−εKa P− P ρ P, Pg ≥ d0

Fatt ( X ) = | g| (14)
 −K ρ P, P
ρ P, P
< d
a g g 0

where d0 is the threshold value for the distance between the vehicle and the target; ε is the
attractive field modulation factor. The most significant difference after the improvement is
that when the distance between the vehicle and the target is greater than d0 , the gravitational
force is considered constant.
To address the problem of vehicle oscillation near the target, an adjustment factor
ρn P, Pg is added to the repulsive field function. The improved repulsive field function

can be expressed as:

1 2 n
(
1 1


Urep ( X ) = 2 Kr ( ρ( P,Pobs ) − ρ0 ) ρ P, Pg 0 < ρ ( P, Pobs ) ≤ ρ0 (15)
0 ρ( P, Pobs ) > ρ0

The corresponding repulsive force can be expressed as:

( * *
no,v Frep1 + nv,g Frep2 ρ( P, Pobs ) ≤ ρ0
Frep ( X ) = (16)
0 ρ( P, Pobs ) > ρ0
 should be in motion as shown in the Figure 4. The road boundary potential field function is
expressed by segmentation function, which can make vehicles move steadily within the
road boundary. The road boundary potential field function can be expressed as:
1 𝐷 𝑊 𝐷
⎧ 𝐾 (𝑃 + ) − 𝐷 + ≤ 𝑃 < −
⎪3 2 2 2
World Electr. Veh. J. 2022, 13, 97 7 of 21
⎪ 𝜆 𝐷 𝐷
⎪− 𝐾 (𝑃 + ) − ≤ 𝑃 < 0
3 2 2
𝑈 (𝑃) = 𝜆 𝐷 𝐷 (18)
where the direction of repulsive force⎨ 𝐾 Frep1 (𝑃 is−from ) 0the <𝑃 ≤
obstacle toward the vehicle, and the
direction of repulsive force Frep2 ⎪3is from the vehicle 2 toward2the target. The magnitude of
⎪ 1 𝐷 𝐷 𝑊
Frep1 and Frep2 is determined by the ⎪ following equation:
⎩− 3 𝐾 𝑃 −
2 2
<𝑃 ≤𝐷−
2
  ρn P,P
( g)

The corresponding road boundary 1 1
 Frep1 = Kr ρrepulsive
( P,Pobs )
− ρforce can be expressed as:
ρ2 ( P,Pobs )
0 (17)
1 𝐷2 n−1
n 1
P, Pg𝑊 𝐷


rep2 = 2 Kr ( ρ( P,P ) − ρ0 ) ρ
 F
⎧𝐾 obs (𝑃 + ) − 𝐷 + ≤ 𝑃 < −
where n is an arbitrary non-zero constant, ⎪ taken as n2 =𝐷2 in this 2
𝐷 paper.
2
⎪
The road boundary potential field is−𝜆𝐾 mainly (𝑃 used + to)restrict
2
− ≤ 𝑃 <0
2 the driving area of intelli-
𝐹 (𝑃) = −∇𝑈 (𝑃) = (19)
gent vehicles, which is essentially a type⎨of repulsive force. 𝐷 𝐷
In this paper, taking a two-lane
road as an example, the two green zones 𝜆𝐾 (𝑃 − ) 0 < 𝑃 ≤
⎪ represent the 2 area where2the intelligent vehicle
should be in motion as shown in the Figure ⎪ 4. The road 𝐷 boundary
𝐷 𝑊 field function
potential
is expressed by segmentation function, ⎩ −𝐾 can (𝑃 make− ) vehicles < 𝑃 move
≤ 𝐷 −steadily within the
which 2 2 2
road boundary. The road boundary potential field function can be expressed as:
where 𝑊 is the vehicle width, 𝐷 is the lane width, 𝐾 is the road boundary potential
field coefficient, 𝜆 is the modulation factor, and 3𝑃 is the vertical coordinate of the vehi-
  
cle center of mass position in the



1
3 Kroad
road Py + D2 system
coordinate − D + 𝑋𝑂𝑌. W
2 ≤ P The − D2
y <direction of repulsion

   3
of the road boundary potential  field is
 − Kroad Py +
 λ perpendicularD Dto the
− 2 ≤ Py < 0 road boundary. When the
3 2
intelligent vehicle Ucan avoid
road ( P ) =the obstacle  without  changing
3 lanes, the intelligent vehicle(18)
is
D D
restricted to remain in the current
λ
K
 3 lane,


 P
road unless
y − 0 < P
2the longitudinal y ≤ 2 repulsive force on the ve-
 3
hicle is greater than the maximum  −repulsive forceD of the D road boundary W without changing
 
1
− 2 2 < Py ≤ D − 2

lanes to reduce the driving risk of changing3 Kroad Pylanes.

Y
D
D−W/2

W/2
X
0
−W/2

−D+W/2
−D

Figure 4. Intelligent vehicle driving areas.

Figure 4. Intelligent vehicle driving areas.
The corresponding road boundary repulsive force can be expressed as:
By superposition of the above three potential fields, the equilibrium equation for the
vehicle subjected to the potential fieldcan be obtained:
  
D 2

 K road Py + 2 −D+ W D
2 ≤ Py < − 2
  
 −λKroad Py + D 2 − D ≤ Py < 0


Froad ( P) = −∇Uroad ( P) =  2 2
(19)
D 2 D


 λK road Py − 2 0 < P y ≤ 2
  
 −Kroad Py − D 2 D < Py ≤ D − W


2 2 2

where W is the vehicle width, D is the lane width, Kroad is the road boundary potential field
coefficient, λ is the modulation factor, and Py is the vertical coordinate of the vehicle center
of mass position in the road coordinate system XOY. The direction of repulsion of the
road boundary potential field is perpendicular to the road boundary. When the intelligent
vehicle can avoid the obstacle without changing lanes, the intelligent vehicle is restricted to
remain in the current lane, unless the longitudinal repulsive force on the vehicle is greater
than the maximum repulsive force of the road boundary without changing lanes to reduce
the driving risk of changing lanes.
World Electr. Veh. J. 2022, 13, 97 8 of 21

World Electr. Veh. J. 2022, 13, x FOR PEER REVIEW 8 of 21

By superposition of the above three potential fields, the equilibrium equation for the
vehicle subjected to the potential field can be obtained:

U𝑈
total ==
U𝑈att ++
U𝑈rep ++
U𝑈road (20)

The equilibrium
The equilibrium equation
equation for
for the
the forces
forces on
on the
the vehicle
vehicle can
can be
be expressed
expressed as:
as:
𝐹 =𝐹 +𝐹 +𝐹 (21)
Ftotal = Fatt + Frep + Froad (21)

4.2. Strategies for Jump out of Local Minima Based on Smaller Steering Angles Angles
The intelligent
intelligentvehicle
vehiclewill
willstall
stallororoscillate
oscillate when
when it falls
it falls intointo local
local minima.
minima. To solve
To solve this
this problem, this paper proposed a strategy, that is searching for
problem, this paper proposed a strategy, that is searching for potential jump potential jump out points
within
within the steering angle
the steering angle range
range asasshown
shownin inFigure
Figure5.5.The
Thespecific
specificsteps
stepsare
areasasfollows:
follows:

40° n=5
Potential jump-out
point B

Forward step
length Sf n=1
Current location A
θ
θ
Virtual backward step Obstacle Target
length Sb Forward step
n=1
length Sf

Potential jump-out
point C
−40° n=5
Figure 5. Strategy for jumping out of local minima.

When the vehicle has not reached

reached the
the target,
target, the
the vehicle
vehicle satisfies
satisfies the
the equation:
equation:

P𝑃−−Pg𝑃 ≥≥step
𝑠𝑡𝑒𝑝length
𝑙𝑒𝑛𝑔𝑡ℎ (22)
(22)
The local minima are defined as:
The local minima are defined as:
|𝑃 − 𝑃 | ≤ 0.1 ∗ 𝑠𝑡𝑒𝑝 𝑙𝑒𝑛𝑔𝑡ℎ || 𝐹 =0 (23)
| Pi+2 − Pi | ≤ 0.1 ∗ step length || Fsum = 0 (23)
where 𝑖 is the serial number of step. When the intelligent vehicle falls into the local min-
ima, the
where i iscoordinates of the of
the serial number intelligent
step. Whenvehicle at that moment
the intelligent vehiclearefallsnoted 𝐴(𝑥 ,minima,
as local
into the 𝑦 ), the
direction
the of vehicle
coordinates of themotion at that
intelligent moment
vehicle at thatismoment
taken asare
thenoted
reference
as A(line.
x0 , y0Initializing the
), the direction
number
of vehicleofmotion
angle changes 𝑛 = 1, and
at that moment picking
is taken the reference
as the angles online.
bothInitializing
sides of thethe reference
numberline
of
direction,
angle changes n = 1, 𝜃and are
the angles determined
picking by Equations
the angles (24) of
on both sides andthe(25):
reference line direction,
the angles θth are determined by Equations (24) 1 and (25):
⎡ ⎤
1⎢1024⎥

1024 1
 1⎢ ⎥
 256
1⎢ 256⎥

1

⎢ ⎥ 1 · · · n

nth = ±
 64 (24)
1
= ±16
𝑛  ⎢ 64  1 ⋯ 𝑛
⎥ (24)
1
4⎢ 1 ⎥

 
1⎢ 16 ⎥
⎢ 1q ⎥
n
θth = th⎢ ∗4 |⎥320 ∗ nth | (25)
|nth⎣| 1 ⎦
𝑛
𝜃 = ∗ |320 ∗ 𝑛 | (25)
|𝑛 |
The magnitude of the front tire angle 𝛿 of a vehicle can be expressed as:
World Electr. Veh. J. 2022, 13, 97 9 of 21

The magnitude of the front tire angle δ of a vehicle can be expressed as:

θth Fsumy
δ = 2π ∗ + arctan + π ∗ ( Fsumx < 0) (26)
360 Fsumx

At the same time, according to Equation (27) to obtain the former step length S f .

 0.5 ∗ step length ∆U X Ai − X Ai−2
≤ 0.8



Sf = step length 0.8 <∆U X Ai − X Ai−2
< 1.2 (27)

1.5 ∗ step length ∆U X Ai − X Ai−2 ≥ 1.2


where ∆U X Ai − X Ai−2 is the potential field magnitude difference between the point A

and the position twice the length steps before the point A. Based on the front tire angle δ
and the former step length S f , the vehicle gets two potential jump-out points, sit marked
as B( x1 , y1 ) and C ( x2 , y2 ).
Calculate the value of the potential field when the vehicle is at positions A, B, and
C respectively and write U ( x0 ), U ( x1 ), and U ( x2 ). Comparing the magnitude of U ( x0 ),
U ( x1 ) and U ( x2 ):
If the potential field value satisfies both U ( x0 ) < U ( x1 ) and U ( x0 ) < U ( x2 ), it means
that the potential field value at point A is the lowest. Neither point B nor point C can be the
suitable jump-out point. Then increase the number of angle changes once and perform the
above step operation again. Meanwhile, the steering angle of the intelligent vehicle should
satisfy its dynamics constraints in the actual driving scenario. The maximum steering angle
of the vehicle does not exceed 40◦ and n should be no more than 5. If n is greater than
5, then the vehicle is set back two times the virtual step length to obtain a critical local
minima, and performed again from the beginning.
If it does not satisfy U ( x0 ) < U ( x1 ) or U ( x0 ) < U ( x2 ), it means that point B and
point C have at least one point with a lower potential field value than point A. Then
comparing the magnitude of U ( x1 ) and U ( x2 ), the vehicle moves to point C if potential
field value satisfied U ( x1 ) > U ( x2 ) and moves to point B if potential field value satisfied
U ( x2 ) > U ( x1 ).
When determining the potential jump-out points, attention should be paid to whether
the line between the current position of the intelligent vehicle and the potential jump-out
point will collide with the obstacle. If a collision occurs, the potential jump-out point should
be discarded and searched again. The planned trajectory and front tire angle should satisfy
Equation (12). The flowchart of the strategy is shown in Figure 6.
World
WorldElectr.
Electr.Veh.
Veh. J.J. 2022,
2022,13,
13,97
x FOR PEER REVIEW 10 of
10 of 21

Start

The position of the vehicle A(x0,y 0) is the

center of the circle, and the direction of
the current position of the vehicle is the
reference line

Initialize the number of angle changes

n=1

No
The vehicle backs up on the original path
n<5?
with 2 times the virtual step length

Yes

According to the formula to get the front

tire angle δ

The coordinates of potential jump-out

points are obtained as B(x1,y 1)and
C(x2,y 2) , respectively, by taking the value
of step Sf forward in the δ direction

Calculate U(x0)，U(x1)，U(x2)

Yes
U(x0) < U(x1) and
n=n+1
U(x0) < U(x2)?

No

No
U(x1) > U(x2)?

Yes

C(x 2,y 2) is the potential jump-out point B(x1,y 1)is the potential jump-out point

Does the smart car Yes Yes Does the smart car
collide with obstacles in collide with obstacles in
the planned path? the planned path?

No No

The vehicle moves to C(x 2,y 2) The vehicle moves to B(x1,y 1)

End

Figure 6.
Figure 6. Flowchart
Flowchart of
of the
the strategy
strategy for
for jump
jump out
out of
of the
the local
local minima.
minima.
World Electr. Veh. J. 2022,
World13, 97 Veh. J. 2022, 13, x FOR PEER REVIEW
Electr. 11 of 21 11 of

5. Simulation and Analysis and Analysis

5. Simulation
5.1. Simulation 5.1.
Environment
SimulationConstruction
Environment Construction
To verify the feasibility
To verify ofthethe improved
feasibility APF
of the method APF
improved in local path in
method planning of planning
local path intelli- of inte
gent vehicles, simulations are carried out on the MATLAB 2021a platform. The traditional
ligent vehicles, simulations are carried out on the MATLAB 2021a platform. The trad
and improved tional
APF methods
and improvedare compared to deal
APF methods arewith the problem
compared to dealof local
with theminima;
problemtheof local mi
planning trajectories of the traditional and improved APF methods are compared
ima; the planning trajectories of the traditional and improved APF methods in the
are compare
static obstaclesinenvironment and the
the static obstacles dynamic obstacles
environment environment.
and the dynamic Theenvironment.
obstacles effectivenessTheof effectiv
the improved method is evaluated
ness of the improvedby the curvature
method of theby
is evaluated trajectory. The main
the curvature of theparameters
trajectory. The ma
of the simulation experiments
parameters of theare shown inexperiments
simulation Table 1. are shown in Table 1.

Table
Table 1. Simulation 1. Simulation
parameters parameters
setting table. setting table.
Parameter
Parameter Name and Symbol Name and Symbol Representation
Representation Value/Unit Value/Unit
Attractive field action coefficient 𝐾 15
Attractive field action coefficient Ka 15
Repulsive
Repulsive field action coefficient Kr field action coefficient 𝐾 10 10
Road boundary potentialRoadfield boundary
coefficient potential
Kroad field coefficient 𝐾 20 20
Radius
Radius of the influence range of the
of the obstacle ρ0 range of the obstacle 5𝜌m
influence 5m
Vehicle length L Vehicle length 𝐿 4.7 m 4.7 m
Vehicle width W Vehicle width 𝑊 1.8 m 1.8 m
Lane width D 3.5 m
Lane width 𝐷 3.5 m
Step length 0.1 m
Step length 0.1 m

5.2. Analysis of 5.2.

Simulation
AnalysisResults
of Simulation Results
5.2.1. Simulation Analysis of Local Minima Problems
5.2.1. Simulation Analysis of Local Minima Problems
Set the startingSetposition of the
the starting main vehicle
position as (0vehicle
of the main m, 0 m)asand the0 m)
(0 m, target
andposition asposition
the target
(50 m, 0 m). The obstacle coordinates are (25 m, 0 m) in the single obstacle case and (25 m,
(50 m, 0 m). The obstacle coordinates are (25 m, 0 m) in the single obstacle case and (25 m
3.5 m) (25 m, −3.5
3.5 m)
m) in
(25the
m,multiple obstacles
−3.5 m) in case. The
the multiple simulation
obstacles results
case. The of the traditional
simulation results of the trad
APF method are shown in Figure 7.
tional APF method are shown in Figure 7.
4
Obstacle
Target
2
Starting Point
Path
0

-2

-4
0 10 20 30 40 50 60

(a)
4
Obstacle
Target
2
Starting Point
Path
0

-2

-4
0 10 20 30 40 50 60

(b)
Figure 7.
Figure 7. Local minima Local minima
formed formed APF
by traditional by traditional
method. APF method.
(a) Single (a) Single
obstacle; (b)obstacle;
multiple(b)obstacles.
multiple obstacles.

From Figure 7,From Figure

it can 7, it can
be seen thatbethe
seen that theall
vehicles vehicles all fall
fall into intominima,
local local minima,
and theand the coo
coordinates ofdinates of the
the local local minima
minima are (19 are m, (19
0 m)m,and
0 m)(19.8
and (19.8
m, 0m,m).0 m).
TheThe results
results ofofpath
path plannin
are failures
planning are failures becausebecause they
they fail to fail to achieve
achieve their targets.
their targets. The simulation
The simulation resultsresults
of the of the im
proved APF method are
improved APF method are shown in Figure 8. shown in Figure 8.
From Figure 8, it can be seen that the vehicle jump out of the local minima by using
the improved APF method, the planned trajectory is continuous and smooth. As can be
seen from Figure 9, the absolute values of the curvature of the trajectories planned by
the improved APF method are all less than 0.4 m−1 , which satisfies the requirements of
path planning. It also indicates that the strategy of jump out of the local minima based on
smaller steering angle is effective.
 4
Obstacle
Target
World Electr.
2
Veh. J. 2022, 13, x FOR PEER REVIEW Starting Point 12 of 21
Path

World Electr.0 Veh. J. 2022, 13, x FOR PEER REVIEW 12 of 21

-2

-4 4
0 10 20 30 40 50 60
Obstacle
Target
World Electr. Veh. J. 2022,
World13, 97 Veh. J. 2022, 13, x FOR PEER REVIEW
Electr. 42
(a) Starting Point 12 of 21 12 of
Path
Obstacle
Target
4 20
Starting Point
PathObstacle
-2
0 Target
2
Starting Point
Path
-4
-2
0 0 10 20 30 40 50 60
4

-2
2
-4
0 10 20 (a) 30 40 50
Obstacle
Target
60

4 Starting Point

-4
0
0 10 20
(a) 30 40 50
Path
Obstacle
Target 60
42 Starting Point
Path
Obstacle
-2
20 (b) Target
Starting Point
Path
-4
-2
0 0 10 20 30 40 50 60

-4
-2
improved APF method. (a) Single obstacle; (b) multiple obstacl
Figure 8. Trajectory planned by the(a)
0 10 20 30 40 50 60
4
-4
(b) 0 10 20 30 40 50 Obstacle 60
2
From Figure 8, it can be seen that the vehicle jump out of the local minima by us Target
Starting Point
(b) Path
0
Figure 8. Trajectory
the improved planned bythe
APF method, the planned
improved APF method.is
trajectory (a)continuous
Single obstacle;and(b) multiple
smooth.obstacles.
As can
-2
Figure 8. Trajectory planned by the improved APF method. (a) Single
seen from Figure 9, the absolute values of the curvature of the trajectories planned by obstacle; (b) multiple obstacles.
From Figure 8, it can be seen that the vehicle jump out of the local minima by using
improved APF method are all less than 0.4 m−1 , which satisfies the requirements of p
-4
0 10 20 30 40 50 60

the improved
From Figure APF8, method,
it can be(b) the
seen planned
that thetrajectory is continuous
vehicle jump out of theandlocalsmooth.
minimaAs bycan be
using
planning. It also indicates
seenimproved
from Figure
that the strategy of jump out of the local minima based
the APF9,method,
the absolute values oftrajectory
the planned the curvature of the trajectories
is continuous and smooth.planned by the
As can be
smaller
Figure
Figure 8. Trajectory
seen steering
planned
from
improved by angle
8. Trajectory
Figure
APF
is effective.
planned
themethod
improved
9, the by APF
the
all improved
absolute
are method.
lessvalues APF
0.4
(a)
than of −1 obstacle;
method.
m
Single
the (a) Single
curvature
, which theobstacle;
of (b) multiple
satisfies (b) multiple
obstacles.
trajectories planned
the requirements obstacle
bypath
of the
−1
planning. APF
improved It also indicates
method that
are all less thethan 0.4 mof ,jump
strategy whichout of thethe
satisfies local minima based
requirements on
of path
0.4 0.04
Fromsteering
smaller
planning. Figure 8,
It alsoangleit can bethat
seen
is effective.
indicates the that the vehicle
strategy jump
of jump outoutof of
thethe local
local minima
minima by usin
based on
0.3
the improved
smaller steeringAPF method,
angle the0.03
is effective. planned trajectory is continuous and smooth. As can b
0.4 seen from Figure 9, the absolute0.02 values
0.04 of the curvature of the trajectories planned by th
0.2 0.4
improved APF method are all less
0.010.04
than
0.03
0.4 m−1 , which satisfies the requirements of pa

Curvature/m -1
Curvature/m-1

0.3

0.1 0.3
planning. It also indicates that the 0 0.02
0.03
strategy of jump out of the local minima based o
0.2 0.01
smaller steering angle is effective. -0.010.02

Curvature/m -1
Curvature/m-1

0.2 0.010
0 -0.02

-1
0.1
-1

Curvature/m
Curvature/m

0.4 -0.01
0
0.04
0.1
-0.03
-0.1 0 -0.02
-0.01
0.03
-0.04
0.3
0 -0.03
-0.02
0.02
-0.2 -0.1 -0.05
0 5 10 15 20 25 30 35 40 45 50 0
-0.04 5 10 15 20 25 30 35 40 45 50
0.2 -0.03
0.01
-0.1 Longitudinal displacement / m Longitudinal displacement / m
Curvature/m -1
Curvature/m-1

-0.2 -0.05
-0.04
0 5 10 15 20 25 30 35 40 45 50 0
0.1
-0.2
(a)
Longitudinal displacement / m -0.05
0 5 10 15 20
(b) 25
Longitudinal displacement / m
30 35 40 45 50

0 5 10 15 20 25 30 35 40 45 50 -0.01 0 5 10 15 20 25 30 35 40 45 50

Figure 9.0 Curvature the(a)

FigureLongitudinal
9. displacement / m
ofCurvature trajectory of the trajectory
planned planned
by the improved
-0.02
Longitudinal(b)
by the
APF improved displacement / m
method; APF method; (a)
(a) trajectory trajectory curvat
curvature
for single
for single obstacleFigure
case; (b) (a)
obstacle case; (b)
trajectoryofcurvature
9. Curvature trajectory curvature
for multiple
the trajectory for multiple
obstacles
planned by
-0.03
(b)
the improved obstacles case.
case. APF method; (a) trajectory curvature
-0.1
for single
Figure obstacle case;
9. Curvature (b)trajectory
of the trajectory-0.04
curvature
planned by for
the multiple
improved obstacles case. (a) trajectory curvature
APF method;
5.2.2. Simulation ofsingle
Pathobstacle
5.2.2.forSimulation Planning of case;
Path
-0.2
in(b)
Static
Planning Obstacles
trajectory in StaticSituations
curvature
-0.05 Obstacles
for multipleSituationsobstacles case.
0 5 0 510 10 15
15 20 20
25 30 25
35 40 3045 50 35 40 45 50
5.2.2.
Taking two Taking Simulation
lanes intwo both of Path
directions Planning in Static Obstacles Situations
Longitudinal displacement / m
lanes in bothasdirections
an example, as an inexample,
the initialin/state,
Longitudinal displacement m the position
the initial of position
state, the
5.2.2.
the subject vehicle Simulation
(a)vehicle
isTaking
(0 m, − 1.75 ofm)Pathand Planning in Static
thedirections
target asObstacles is (b) Situations
the subject two islanes
(0 m, in−1.75
both m) andposition
the an
targetexample,(100
positionm, 1.75
in the m). m,
is initial
(100 Assuming
state,
1.75the
m).position
Assum o
that the subject the
vehicle
thatFigure subject
Taking
the subject moves
9. Curvaturevehicle
two
vehicleat ais
lanes (0
in m,
constant
moves
of the −1.75
both m)
directions
speed
at aplanned
trajectory and
constant of the
as
10 antarget
m/s,
by speed example,
the
the improved positionin
initial
of 10 m/s, theis (100
initial
positions m,
the initial
APF method; 1.75
state,
of m).
the
staticAssuming
position
positions
(a) trajectory of
of sta
curvatu
that
the the obstacle
subject subject
vehicle vehicle (0 moves
is (b) m, −1.75 atm)acurvature
2 constant
and(15 thefor speed
target of 10obstacles
position m/s,isthe initial positions of static
obstacle vehiclefor
obstacle1single
andvehicle static 1 obstacle
case;
and static vehicle
trajectory
obstacle are
vehicle m, 1.75
2 multiple
are (15 m)m,and1.75(100
(50 m) m,
case.m, 1.75
2.5m).
and−(50 m),Assuming
m, −2.5 m),
obstacle
thethat
respectively,spectively, vehiclevehicle
the subject
location 1 and staticmoves obstacle invehicle
at a constant 2 are of
speed (1510m, 1.75the
m/s, m)initial
and (50 m, −2.5ofm),
positions re
static
therelationships
location relationships are shown are Figure
shown10. in Figure 10.
spectively,
obstacle
5.2.2. Simulation the location
vehicle 1 and
of relationships
Pathstatic obstacle
Planning are shown
vehicle
in Static 2 are
Obstacles in Figure m, 10.
(15Situations
1.75 m) and (50 m, −2.5 m), re-
4
2
4 spectively, the location relationships are shown in Figure 10. Obstacle 1
Obstacle
Obstacle 1 2
0
2 Taking two lanes in both directions as an example, in the initial state, the position Obstacle 2
Obstacle
Obstacle 3
3
40 Target
Obstacle 1
-2
2
-2 the subject vehicle is (0 m, −1.75 m) and the target position is (100 m, 1.75 m). Assumin Target
Starting
Obstacle
Starting
Path
Path
Obstacle
2
Point
3
Point

-4 0
0 -4
-2 that the subject vehicle moves at a constant speed of 10 m/s, the initial positions of stat
0
20
20
40
40
60
60
80
80
100
100
Target
Starting Point
Path
-4

Figure 10. Static obstacle

Figure vehicle
10. Static
Figure 10. 1environment.
obstacles
Static anddriving
obstaclesstatic obstacle vehicle 2 are (15 m, 1.75 m) and (50 m, −2.5 m), r
environment.
driving environment.
0 20 40 60 80 100
obstacles driving
spectively, the location relationships are
Figure 10. Static obstacles driving environment. shown in Figure 10.
4 The The simulation
simulation results
results of
The simulation results of trajectory planned by theof trajectory
trajectory planned
planned by the and
by
traditional traditional
traditionaland
improvedandimproved APF
improved
APF A Obstacle 1
2 Obstacle 2
0 methods aremethods methods
shown in The
are are shown
simulation
shown5.2.2
Section in Figures
results
in Figures of 11
11 and
and Figure and 12.
trajectory
12.12. planned by the traditional and improved APF Obstacle 3
Target
Starting Point
4
-2
-4
4 methods are shown in Figures 11 and 12. Path1
Obstacle
2 0 20 40 60 80 100 Obstacle 2 1
Obstacle
2 Obstacle
Obstacle 3 2
40
Target
Obstacle
Obstacle 1 3
0 Starting Point
2
-2 Target
Obstacle 2
-2 0
-4
Figure 10. Static obstacles driving environment. Path
Obstacle
Target
3 Point
Starting
Path 120
0 20 40 60 80 100
-4 -2 Starting Point
0 20 40 60 80 100 Path 120
-4
Figure 11.simulation
Trajectory planned
0
byof
thetrajectory
traditional APF method inthe
static obstacles driving
and environment.
20 40 60 80 100 120

Figure 11.The
Trajectory planned results
by the traditional APFplanned
methodby in statictraditional
obstacles drivingimproved AP
environment
Figure
methods
Figure 11. Trajectory 11. Trajectory
are by
planned shown planned by the
in FiguresAPF
the traditional traditional
11 and APF method in static obstacles driving
12. in static obstacles driving environment.
method environment.
4
Obstacle 1
2
0
Although the traditional APF method also successfully planned the trajectory, it can Obstacle 2
Obstacle 3
Target
-2
-4
be seen that the improved APF method planned a smoother trajectory than the traditional Starting Point
Path
0 20 40 60 80 100 120
APF method by comparing Section 5.2.2 and Figure 12, which is also confirmed by the
Figure 11.in
curvature comparison Trajectory
Figure planned
13. by the traditional APF method in static obstacles driving environment.
 4
Obstacle 1
World Electr.
2 Veh. J. 2022, 13, x FOR PEER REVIEW Obstacle 2
Obstacle 3
13 of 21
0
Target
-2 Starting Point
Path
-4
0 20 40 60 80 100
4
Obstacle 1
4
World Electr. Veh. J. 2022, 13, 97
World
2
Electr. Veh.2 J. 2022, 13, x FOR PEER REVIEW 13 of 21
Obstacle
Obstacle
Obstacle
1 2
2 3
Obstacle 13 of 21
0 Obstacle 3
0
Target
-2 Target
Starting Point
-2 Starting
PathPoint
-4 Path
0 -4 20 40 60 80 100
0 20 40 60 80 100
4
4 Obstacle 1
Obstacle 1
2 24 Obstacle
Obstacle 2 2
Obstacle 1
Obstacle
0 Obstacle 3 3
02 Obstacle 2
Target
Target
Obstacle 3
Starting
-2 -20 Starting PointPoint
Target
PathPath
-4 -2
-4 Starting Point
0 0 20 20 40 40 60 60 8080 100
100 Path
-4
4 4
0 20 40 60 80 100 Obstacle 1 1
Obstacle
2 24 Obstacle 2 2
Obstacle
Obstacle 1
Obstacle 3
02 Obstacle
Obstacle 2 3
0 Target
Target
Obstacle 3
-20 Starting Point
Target
-2 PathStarting Point
-2
-4 Starting
PathPoint
-4 0 20 40 60 80 100 Path
0 -4 20 40 60 80 100
4
0 20 40 60 80 100 Obstacle 1
24 Obstacle 2
Obstacle 1
Obstacle 3
02
-20
Figure 12. Trajectory planned by the improved APF method in static obstacles driving environme Obstacle 2
Target
Obstacle 3
Starting Point
Target
Path
-2
-4 Starting Point
0 20 40 60 80 100 Path
-4

4
0
Although the traditional APF method also successfully planned the trajectory, it c
20 40

Figure 12. Trajectory planned by the improved APF method in static obstacles driving environment
60 80 100

Obstacle 1
2 Obstacle 2

0 be seen that the improved APF method planned a smoother trajectory than the traditio Obstacle 3
Target
-2 Starting Point

-4 APF method by comparing

Although Figures
the traditional APF 11 and 12,
method alsowhich is alsoplanned
successfully confirmed by the curvatu
the trajectory, it can Path

0 20 40 60 80 100

comparison in Figure 13.

be seen that the improved APF method planned a smoother trajectory than the traditional
Figure 12. Trajectory
Figure 12. Trajectory planned byAPF
the improved APF method in static obstacles driving environment
APFplanned
methodbybythe improved
comparing Figures method in
11 and static obstacles
12, which driving
is also environment.
confirmed by the curvature
comparison in Figure 13.
2 Although the traditional APF method also successfully planned the trajectory, it can
0.6

1.5 be seen that the improved APF method planned a smoother trajectory than the traditiona
0.4
2 0.6
1 APF method by comparing Figures 11 and 12, which is also confirmed by the curvature
1.5
0.5 comparison in Figure 13. 0.2 0.4
Curvature/m-1

Curvature/m-1
1
0
0.5 0 0.2
-0.5
Curvature/m-1

Curvature/m-1
2 0.6
0
-1 1.5 -0.2 0
-0.5 0.4
1
-1.5
-1 -0.2
-0.4
0.5 0.2
-2
-1.5
Curvature/m-1

Curvature/m-1
0 -0.4
-2.5 -2 -0.6 0
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
-0.5
-2.5 Longitudinal displacement / m -0.6 Longitudinal displacement / m
-1 0 10 20 30 40 50 60 70 80 90 100 -0.20 10 20 30 40 50 60 70 80 90 100

-1.5
(a)
Longitudinal displacement / m (b)
Longitudinal displacement / m

(a) -0.4 (b)

Figure 13. Comparison
Figure 13.-2 Comparison of curvature of of curvature of plannedin
planned trajectories trajectories in static
static obstacles obstacles
driving driving environme
environment;
(a) Figure 13.ofComparison of curvature of
-0.6planned trajectories in static obstacles driving environment
of curvature planned trajectory byAPF;
traditional APF; (b)
ofcurvature of90 planned bytrajectory by
-2.5
(a) curvature 0 10 planned
20 30 40 trajectory
50 60 70 by
80 traditional
90 100 0 (b)
10 curvature
20 30 40 50 planned
60 70 80 trajectory
100 im-
proved (a)Longitudinal
curvature
APF. of/ mplanned trajectory by traditionalLongitudinal
displacement APF; displacement
(b) curvature
/m of planned trajectory by im-
proved APF. proved APF. (a) (b)
5.2.3.Figure
5.2.3. Simulation Simulation
of Path of PathinofPlanning
Planning
13. Comparison curvature in
Dynamic Dynamic
Obstacle
of planned Obstacle
Situation
trajectories in Situation
static obstacles driving environment
5.2.3. Simulation of Path Planning in Dynamic Obstacle Situation
(a)
Still take two curvature
lanes
Still in
taketake
twoof
bothplanned trajectory
directions
lanes in both as byantraditional
directionsexample, APF; (b)
thecurvature
inexample,
initial of planned
instate, trajectory
the initial by init
im
proved Still
APF. two lanes in both directionsasasan an example, in the
the initial
initial state,thethe
state, initial
position of position
the subject of vehicle
position the subject
of the subject −1.75
is (0vehicle
m,vehicle m)
is is
(0 m,and
(0 them)
m,−1.75
−1.75 target
m)andandposition
the targetisposition
the target (100 m, is1.75
position is(100m).
(100m,m, 1.75
1.75 m)m
Assuming that the
Assuming subject
Assuming that vehicle
the
that thesubjectis moving
vehicle at
is a constant
moving at a speed
constant of 10 m/s, the initial
speed of 10 m/s, the initial po-p
speed of 10 m/s, the initial
5.2.3. Simulation ofsubject vehicle is
Path Planning in moving
DynamicatObstacle
a constant Situation
positions ofsitions
dynamic
sitions obstacle
of dynamic
of dynamic vehicle
obstacle 1 vehicle
obstacleand dynamic
vehicle andobstacle
1 1and dynamicvehicle
dynamic obstacle
obstacle2 are (15 m,
vehicle
vehicle 22are1.75
are(15 m)
(15
m,m, 1.75
1.75 m)
and (50 m, − 2.5(50
m) m, Still take twothe
respectively, lanes in bothrelationships
location directions asare an shown
example, in in the initial state, the initia
and and (50−2.5 m) respectively,
m, −2.5 m) respectively, the thelocation
locationrelationships
relationships areFigure
are shown14.
shown ininFigure
Figure 14.14.
position of the subject vehicle is (0 m, −1.75 m) and the target position is (100 m, 1.75 m)
4 4 Assuming that the subject vehicle is moving at a constant speed of 10 m/s, the initial po- Obstacle vehicle
Obstacle 1
vehicle 1
2 2 Obstacle vehicle 2

0
0 sitions of dynamic obstacle vehicle 1 and dynamic obstacle vehicle 2 are (15 m, 1.75 m) Obstacle vehicle 2
Target
Target
Starting Point
-2 Starting Point
-2
-4
and (50 m, −2.5 m) respectively, the location relationships are shown in Figure 14. Path
Path
-4 0 20 40 60 80 100
0 20 40 60 80 100

4
Figure
Figure 14. Dynamic 14. Dynamic
obstacles obstacles
driving driving environment.
environment. Obstacle vehicle 1
2
0
Figure 14. Dynamic obstacles driving environment. Obstacle vehicle 2
Target
Starting Point
-2
Path
-4 Assume that
Assume that dynamic dynamic
obstacle obstacle
vehicle vehicle
1 and 1 and dynamic
dynamic obstacleobstacle
vehiclevehicle
2 are at2 are
a at a uni-
Assume that dynamic obstacle vehicle 1 and
form
0

speedand
of 5 3m/s and respectively.
3 m/s, respectively.
The
dynamic
simulation
obstacle
20

results
vehicle 2 are at a u
oftrajectory
40

trajectory planned
60 80 100

uniform
World Electr. Veh. J. 2022, 13, speed of 5REVIEW
x FOR PEER
Figure 14.m/s m/s, The simulation results of 14 of 21
formby speed ofDynamic
the traditional
obstacles
5 m/s and
and 3 m/s,driving
improved
environment.
respectively. The simulation results 15
of and
trajectory plann
planned by the traditional and improved APF APF methods
methods are are shown
shown in Figures
in Figures 15 and 16.
16.
by the traditional and improved APF methods are shown in Figures 15 and 16.
Assume that dynamic obstacle vehicle 1 and dynamic obstacle vehicle 2 are at a uni-
4
2 form speed of 5 m/s and 3 m/s, respectively. The simulation results of trajectory planned Obstacle vehicle 1
Obstacle vehicle 2
0 Target

-2
by the traditional and improved APF methods are shown in Figures 15 and 16. Starting Point
Path
-4
0 20 40 60 80 100

Figure
Figure 15. Trajectory 15. Trajectory
planned planned byAPF
by the traditional the traditional
method inAPF methodobstacles
dynamic in dynamic obstacles
driving driving environment
environment.
4
Obstacle vehicle 1
2 It can be seen from the figure that the trajectories are successfully planned by the APF Obstacle vehicle 2
Target
0

method before and after the improvement. However, as can be seen from Figure 17, the
-2
Starting Point
Path
-4
peak curvature of the trajectory planned by the improved APF method is approximately
4
0 20 40 60 80 100

Obstacle vehicle 1

half of the curvature of the trajectory planned by the traditional APF method. This means
2
0
Obstacle vehicle 2
Target
Starting Point
-2
Path
-4
0 20 40 60 80 100
4
Obstacle vehicle 1
2 Obstacle vehicle 2
0 Target
Starting Point
-2
Path
-4
0 20 40 60 80 100
 Path
-4
0 20 40 60 80 100

Figure 15. Trajectory planned by the traditional APF method in dynamic obstacles driving environme
World Electr.
4 Veh. J. 2022, 13, x FOR PEER REVIEW 14 of 21
Obstacle vehicle 1
2 Obstacle vehicle 2
0 Target
World Electr. Veh. J. 2022, 13, 97
-2
14 of 21
Starting Point
Path
-4
0 4 20 40 60 80 100
Obstacle vehicle 1
4
2 Obstacle
Obstacle vehicle
vehicle 2 1
2 0 Target
Obstacle vehicle 2
Starting
TargetPoint
0 -2
Path
Starting Point
-2
-4
that the improved APF method planned the trajectory more smoothly and the vehicle is
-4
0 20 40 60 80 100
Path

0 20 40 60 80 100
4 easier to maneuver when
Figure avoiding
15. Trajectory obstacles.
planned by the traditional APF method in dynamic obstacles driving environment Obstacle vehicle 1
2 Obstacle vehicle 2
0 Target
4 Starting Point
-2 Obstacle
Path vehicle 1
2 Obstacle vehicle 2
-4
0 0 20 40 60 80 100 Target
Starting Point
-2
Path
-4
0 Figure 16. Trajectory planned by the improved APF method in dynamic obstacles driving environme
20 40 60 80 100
4
Obstacle vehicle 1
2 Obstacle vehicle 2
0 Target

-2 It can be seen from the figure that the trajectories are successfully planned by the AP Starting Point
Path
-4

4
method before and after the improvement. However, as can be seen from Figure 17, t
0 20 40 60 80 100

Obstacle vehicle 1
2
0
peak curvature of the trajectory planned by the improved APF method is approximate Obstacle vehicle 2
Target
Starting Point
-2
-4
half of the curvature of the trajectory planned by the traditional APF method. This mea Path

0 20 40 60 80 100
that the improved APF method planned the trajectory more smoothly and the vehicle
easierFigure
Figure 16. Trajectory 16. Trajectory
toplanned
maneuverby the planned byAPF
improved
when the improved
avoiding method inAPF
obstacles. method
dynamic in dynamic
obstacles obstacles
driving driving environment
environment.

0.7
It can be seen from the figure
0.35
that the trajectories are successfully planned by the APF
0.6
method before and after the improvement.
0.3
However, as can be seen from Figure 17, the
peak curvature of the trajectory planned by the improved APF method is approximately
0.5 0.25
half of the curvature of the trajectory planned by the traditional APF method. This means
0.4 0.2
that the improved APF method planned the trajectory more smoothly and the vehicle is
Curvature/m-1

Curvature/m-1
0.3 0.15
easier to maneuver when avoiding obstacles.
0.2 0.1

0.1 0.7 0.05 0.35

0 0.6 0 0.3

0.5
-0.1 -0.05 0.25

0.4 0.2
-0.2 -0.1
Curvature/m-1

Curvature/m-1

0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
0.3 0.15
Longitudinal displacement / m Longitudinal displacement / m
0.2 0.1
(a) (b)
0.1 0.05

Figure 17. Comparison

Figure 17.0 Comparison of curvatureofofcurvature of planned
planned trajectories
0 trajectories
in dynamic in dynamic
obstacles obstacles
driving driving enviro
environ-
ment. (a) Curvature of planned trajectory
-0.05by traditional APF; (b) curvature of planned trajectory
ment. (a) Curvature of planned trajectory by traditional APF; (b) curvature of planned trajectory by
-0.1

improved-0.2 improved
0APF.
10 20 30
APF.
40 50 60 70 80 90 100
-0.1
0 10 20 30 40 50 60 70 80 90 100
Longitudinal displacement / m Longitudinal displacement / m

5.3. Simulation of Tracking

5.3. Simulation theTracking
(a) of Planned the Trajectory
Planned Trajectory (b)
In order toInfurther
Figure 17. verify
order to the reliability
further
Comparison verify the of
of curvature ofthe pathof
reliability
planned planning
the path
trajectories inof the improved
planning
dynamic APF
of thedriving
obstacles improved AP
environ-
method, we havement.
method, we(a) have
tracked Curvature
thetrackedof planned
planned the trajectorytrajectories
trajectories
planned by
in traditional
static and inAPF; (b) curvature
dynamic
static andobstacle of planned
dynamic driving trajectory
obstacle driviby
environments onimproved
the Carsim
environments APF. simulation platform and analyzed the simulation results.
on the Carsim simulation platform and analyzed the simulation results.
In order toInexhibit
order to theexhibit
amount theof steering
amount of wheel
steering angle
wheel variation of the intelligent
angle variation of the intelligent v
5.3. Simulation of Tracking the Planned Trajectory
vehicle under different speeds, we set two groups of speed
hicle under different speeds, we set two groups of speed commonly commonly used in urbanusedroadin urban ro
environment, 10 m/s Inand
order
15 to further
m/s. These verify
two the reliability
groups of of the
speed willpath
be planning
used as of expected
the the improved APF
environment, 10 m/s and 15 m/s. These two groups of speed will be used as the expect
speed of the method,
subject we have
vehicle to tracktracked the planned
thetotrajectory trajectories in previous
static andsection.
dynamicSince obstacle
it driving
speed of the subject
environments onvehicle
the Carsim track theplanned
simulationtrajectory in the
platformplanned in thethe
and analyzed previous
simulationsection. Since
results.
is only necessary
is only to obtain the
necessary to variation
obtain the rules of steering
variation rules wheel
of angle
steering before
wheel andbefore
angle after theand after t
In order to exhibit the amount of steering wheel angle variation of the intelligent ve-
improved APF method,
improved the article
APF method, only takes
the article the default
only takestracking model
theofdefault in Carsim.
tracking model In static
hicle under different speeds, we set two groups speed commonly used ininurbanCarsim.
road
obstacle environment, Figure
staticenvironment,
obstacle 18 compared the 18
environment, planned trajectory and the actual trajectory
10 m/s and Figure
15 m/s. Thesecompared
two groups the planned
of speed trajectory
will be used and
as the actual tr
the expected
for expected vehicle
jectory forspeed
speed of the of
expected 10 m/s.
vehicle
subject ThetoEuclidean
speed
vehicle of 10the
track m/s. errors ofplanned
the planned
The Euclidean
trajectory in the trajectory
errors of the planned
previous and trajecto
section. Since it
the actual and
trajectory
the
is onlyfor
actual expected
trajectory
necessary vehicle speed
for expected
to obtain of 10
vehicle
the variation m/s
rulesspeed are shown
of 10 m/s
of steering wheelin Figure
areangle
shown 19.in The
before Figure
and after19.the
T
variation curves of steering
improved
variation curves ofwheel
APF method,
steeringanglethe for
wheel expected
article
angleonly vehicle
fortakes speed
the default
expected vehicle of 10 m/s
tracking
speed in
10 static
model
of in Carsim.
m/s in staticIno
obstacles driving environment
static obstacle are shown
environment, in Figure
Figure
stacles driving environment are shown in Figure 20. 18 20.
compared the planned trajectory and the actual tra-
Figure 21 compared the planned
jectory for expected vehicle trajectory
speed of 10 andm/s.theTheactual
Euclideantrajectory
errors of for
theexpected
planned trajectory
vehicle speed ofand 15 the
m/s. Thetrajectory
actual Euclidean for errors
expected of vehicle
the planned
speed of trajectory
10 m/s areand shownthe inactual
Figure 19. The
trajectory for expected
variationvehicle
curves speed of 15 m/s
of steering wheelareangle
shown for in Figurevehicle
expected 22. Thespeedvariation
of 10curves
m/s in static ob-
of steering wheel stacles
angle driving environment
for expected vehicle are shown
speed of in Figure
15 m/s20.in static obstacles driving
environment are shown in Figure 23.
 World Electr. Veh. J. 2022, 13, x FOR PEER REVIEW 15 of 21
World Electr. Veh. J. 2022, 13, 97 15 of 21
World Electr. Veh. J. 2022, 13, x FOR PEER REVIEW 15 of 21

World Electr. Veh. J. 2022, 13, x FOR PEER REVIEW 15 of 21

(a) (a)

(a)

(b)
Figure 18.(b)
Comparison of planned trajectory and actual trajectory for expected vehicle speed of 10
Figure 18. Comparison of planned trajectory and actual trajectory for expected vehicle speed of
m/s in static obstacles driving environment. (a) Planned trajectory by traditional APF and actual
10 m/s18.
Figure inComparison
static obstacles driving
of planned
trajectory environment.
trajectory
of vehicle; (a)
and actual
(b) planned Planned trajectory
trajectory
trajectory byand
for expected
by improved APF traditional
vehicle APFofof
speed
actual trajectory and actual
10vehicle.
m/s in staticofobstacles
trajectory driving
vehicle; (b) environment.
planned trajectory(a) (b)
byPlanned trajectory
improved APF and by actual
traditional APF and
trajectory actual
of vehicle.
trajectory of vehicle; Figure
(b) planned trajectory by improved APF and
18. Comparison of planned trajectory andactual
actualtrajectory
trajectoryof
forvehicle.
expected vehicle speed of 10
m/s in static obstacles driving environment. (a) Planned trajectory by traditional APF and actual
trajectory of vehicle; (b) planned trajectory by improved APF and actual trajectory of vehicle.
Euclidean error / m
Euclidean error / m

Euclidean error / m

Figure 19. Euclidean errors of trajectories planned for expected vehicle speed of 10 m/s in static
obstacles driving environment.

Figure
Figure19. Euclidean
19.Euclidean errors
errors of of trajectories
trajectories planned
planned for expected
for expected vehicle
vehicle speedspeed of 10
of 10 m/s inm/s
staticin static
obstacles driving environment.
obstacles driving environment.
Figure 19. Euclidean errors of trajectories planned for expected vehicle speed of 10 m/s in static
obstacles driving environment.

Figure 20. Steering wheel angles for expected vehicle speed of 10 m/s in static obstacles driving
environment.
 Figure 20. Steering wheel angles for expected vehicle speed of 10 m/s in static obstacles driving
environment.
environment.
Figure 21 compared the planned trajectory and the actual trajectory for expected ve-
Figure 21 compared the planned trajectory and the actual trajectory for expected ve-
hicle speed of 15 m/s. The Euclidean errors of the planned trajectory and the actual trajec-
hicle speed of 15 m/s. The Euclidean errors of the planned trajectory and the actual trajec-
tory for expected vehicle speed of 15 m/s are shown in Figure 22. The variation curves of
World Electr. Veh. J. 2022, 13, 97 tory for expected vehicle speed of 15 m/s are shown in Figure 22. The variation curves16 ofof
21
steering wheel angle for expected vehicle speed of 15 m/s in static obstacles driving envi-
steering wheel angle for expected vehicle speed of 15 m/s in static obstacles driving envi-
ronment are shown in Figure 23.
ronment are shown in Figure 23.

(a)
(a)

(b)
(b)
Figure Figure
21. Comparison of planned
21. Comparison trajectory
of planned and actual
trajectory andtrajectory for expected
actual trajectory vehicle speed
for expected of 15
vehicle speed of
m/s in Figure
static 21. Comparison
obstacles driving ofenvironment.
planned trajectory
(a) and actual
Planned trajectory
trajectory by for expected
traditional APF vehicle
and speed of 15
actual
15 m/s in static obstacles driving environment. (a) Planned trajectory by traditional APF and actual
m/sofinvehicle;
trajectory static obstacles driving
(b) planned environment.
trajectory (a) Planned
by improved APF and trajectory
actual by traditional APF and actual
trajectory of vehicle; (b) planned trajectory by improved APF and trajectory of vehicle.
actual trajectory of vehicle.
trajectory of vehicle; (b) planned trajectory by improved APF and actual trajectory of vehicle.
1.4
1.4 Traditional APF
Improved APFTraditional APF
1.2 Improved APF
1.2

1
1

0.8
0.8

0.6
0.6

0.4
0.4

0.2
0.2

0
0 0 1.25 2.5 3.75 5 6.25 7.5
0 1.25 2.5
Time / s 3.75 5 6.25 7.5
Time / s

Figure 22. Euclidean errors of trajectories planned for expected vehicle speed of 15 m/s in static
obstacles driving environment.

In dynamic obstacles environment, Figure 24 compared the planned trajectory and

the actual trajectory for expected vehicle speed of 10 m/s. The Euclidean errors of the
planned trajectory and the actual trajectory for expected vehicle speed of 10 m/s are shown
in Figure 25. The variation curves of steering wheel angle for expected vehicle speed of
10 m/s in dynamic obstacles driving environment are shown in Figure 26.
 World Electr. Veh. J. 2022, 13, x FOR PEER REVIEW 17 of 21

World Electr. Veh. J. 2022, 13, x FOR PEER Figure

REVIEW 22. Euclidean errors of trajectories planned for expected vehicle speed of 15 m/s in static17 of 21
obstacles driving environment.

World Electr. Veh. J. 2022, 13, 97 17 of 21

Figure 22. Euclidean errors of trajectories planned for expected vehicle speed of 15 m/s in static
obstacles driving environment.

Figure 23. Steering wheel angles for expected vehicle speed of 15 m/s in static obstacles driving
environment.

In dynamic obstacles environment, Figure 24 compared the planned trajectory and

the actual trajectory for expected vehicle speed of 10 m/s. The Euclidean errors of the
planned trajectory and the actual trajectory for expected vehicle speed of 10 m/s are shown
Figure 23. SteeringFigure
wheel 23. Steering
angles
in Figure forwheel
25. The angles
expected
variation for of
vehicle
curves expected
speed vehicle
steering of speed
15 m/s
wheel angle forof
in 15 m/s
static in staticdriving
obstacles
expected vehicle obstacles
speed of 10driving
environment.
m/s in dynamic obstacles driving environment are shown in Figure 26.
environment.
In dynamic obstacles environment, Figure 24 compared the planned trajectory and
the actual trajectory for expected vehicle speed of 10 m/s. The Euclidean errors of the
planned trajectory and the actual trajectory for expected vehicle speed of 10 m/s are shown
in Figure 25. The variation curves of steering wheel angle for expected vehicle speed of 10
m/s in dynamic obstacles driving environment are shown in Figure 26.

(a)

(a)

(b)

Figure 24. Comparison of planned

Figure trajectory
24. Comparison andtrajectory
of planned actual trajectory for expected
and actual trajectory vehicle
for expected speed
vehicle speedof
of 10
World Electr. Veh. J. 2022, 13, x FOR PEER REVIEW m/s in dynamic
10 m/s in dynamic obstacles drivingobstacles driving environment.
environment. (a) Planned
(a) Planned trajectory
18 of 21 by traditional APF and actual
trajectory by traditional APF and
trajectory of vehicle; (b) planned trajectory by improved APF and actual trajectory of vehicle.
actual trajectory of vehicle; (b) planned trajectory by improved APF and actual trajectory of vehicle.

(b)

Figure 24. Comparison of planned trajectory and actual trajectory for expected vehicle speed of 10
m/s in dynamic obstacles driving environment. (a) Planned trajectory by traditional APF and actual
trajectory of vehicle; (b) planned trajectory by improved APF and actual trajectory of vehicle.

Figure 25. Euclidean errors of trajectories planned for expected vehicle speed of 10 m/s in dynamic
Figure 25. Euclidean errors of trajectories planned for expected vehicle speed of 10 m/s in dynamic
obstacles driving environment.
obstacles driving environment.
 Figure 25. Euclidean errors of trajectories planned for expected vehicle speed of 10 m/s in dynamic
World Electr. Veh. J. 2022, 13, 97 obstacles driving environment. 18 of 21
Figure 25. Euclidean errors of trajectories planned for expected vehicle speed of 10 m/s in dynamic
obstacles driving environment.

Figure 26. Steering wheel angles for expected vehicle speed of 10 m/s in dynamic obstacles driving
Figure 26. Steering Figure
environment. wheel26. Steering
angles wheel angles
for expected for expected
vehicle vehicle
speed of 10 m/sspeed of 10 m/s obstacles
in dynamic in dynamicdriving
obstacles driving
environment.
environment.
Figure 27 compared the27planned
Figure compared trajectory andtrajectory
the planned the actualandtrajectory
the actualfor expected
trajectory ve-
for expected ve-
Figure 27 compared
hicle speed of 15 m/s. the planned trajectory and the actual trajectory for expected
hicle The
speedEuclidean
of 15 m/s. errors of the planned
The Euclidean errors of trajectory
the plannedand the actual
trajectory trajec-
and the actual trajec-
vehicle
tory for speed of tory
expected 15 m/s.
vehicle The Euclidean
speed
for expected ofvehicle
15 m/sspeederrors
are of m/s
shown
of 15 the planned
inare
Figure
shown28.trajectory
inThe andThethe
variation
Figure 28. actual
curves ofcurves of
variation
trajectory for expected
steering wheel angle vehicle
for wheel
steering speed
expectedangle of 15
vehicle m/s are shown
speedvehicle
for expected of 15 m/sin Figure
speed 28.
inofdynamic The variation
15 m/s inobstacles curves
dynamic driving
obstacles driving
of steering wheel
environment angle for
in expected
are environment
shown 29.vehicle
are shown
Figure speed
in Figure 29.of 15 m/s in dynamic obstacles driving
environment are shown in Figure 29.

World Electr. Veh. J. 2022, 13, x FOR PEER REVIEW 19 of 21

(a)

(a)

(b)

Figure
Figure 27. Comparisonof
27. Comparison ofplanned
plannedtrajectory
trajectoryand
andactual
actualtrajectory
trajectory
forfor expected
expected vehicle
vehicle speed
speed of
of 15
15 m/s in dynamic obstacles driving environment. (a) Planned trajectory by traditional
m/s in dynamic obstacles driving environment. (a) Planned trajectory by traditional APF and actual APF and
actual trajectory
trajectory of vehicle;
of vehicle; (b) planned
(b) planned trajectory
trajectory by improved
by improved APFactual
APF and and actual trajectory
trajectory of vehicle.
of vehicle.

As can be seen in figure about Euclidean errors of trajectories planned, by using the
same Carsim tracking model, the tracking Euclidean error of the trajectory planned by the
improved APF is smaller, which indicates that the trajectory planned by the improved APF
method is easier to be tracked. Due to the better tracked trajectory obtained, the steering
wheel angle of the vehicle is smaller when tracking the trajectory planned by the improved
APF. It indicates that the intelligent vehicle is easier to be maneuvered.
 (b)

World Electr. Veh. J. 2022, 13, 97

Figure 27. Comparison of planned trajectory and actual trajectory for expected vehicle speed19of 15
of 21
m/s in dynamic obstacles driving environment. (b) (a) Planned trajectory by traditional APF and actual
trajectory of vehicle; (b) planned trajectory by improved APF and actual trajectory of vehicle.
Figure 27. Comparison of planned trajectory and actual trajectory for expected vehicle speed of 15
m/s in dynamic obstacles driving environment. (a) Planned trajectory by traditional APF and actual
trajectory of vehicle; (b) planned trajectory by improved APF and actual trajectory of vehicle.

Figure 28. Euclidean errors of trajectories planned for expected

expected vehicle
vehicle speed
speed of
of 15
15 m/s
m/s in dynamic
Figure 28. Euclidean errors of trajectories planned for expected vehicle speed of 15 m/s in dynamic
obstacles driving environment.
obstacles driving environment.

Figure 29. SteeringFigure

wheel29. Steering
angles wheel angles
for expected for expected
vehicle vehicle
speed of 15 m/s speed of 15 m/s in
in dynamic dynamicdriving
obstacles obstacles driving
Figure 29. Steering environment.
wheel angles for expected vehicle speed of 15 m/s in dynamic obstacles driving
environment.
environment.
6. Conclusions As can be seen in figure about Euclidean errors of trajectories planned, by using the
As can be seen same
inCarsim
figure tracking model, theerrors
about Euclidean trackingofEuclidean
trajectorieserror of the trajectory
planned, by usingplanned
the by the
(1) This paper introduces the principle ofwhich
the traditional APFthe method andplanned
its advantages
same Carsim tracking model, the tracking Euclidean error of the trajectory planned by theimproved
improved APF is smaller, indicates that trajectory by the
and shortcomings, solving the problem of excessive
Due to theinitial
betterattractive force and intelli-the steer-
improved APF isAPF method
smaller, is easier
which to be tracked.
indicates that the trajectory tracked trajectory
planned obtained,
by the improved
gent vehicleingcannot
wheel reach
anglethe target
of the by improving
vehicle the potential
is smaller when field
tracking the functions.
trajectory At the
planned by the im-
APF method is easier
same time, to be tracked.
establishing the Dueboundary
road to the better tracked
potential trajectory
field combined obtained,
with theactual
the steer-
proved APF. It indicates that the intelligent vehicle is easier to be maneuvered.
ing wheel angle of
application the vehicle is smaller when tracking the trajectory planned by the im-
scenario.
(2) A strategy of jump that
proved APF. It indicates thelocal
out of intelligent
minima vehicle
basedison
easier to besteering
smaller maneuvered.
angles has been
proposed, solving the problem of local minima that the traditional APF method tends
to fall into by finding smaller steering angles and determining the appropriate jump
out step length in the steering angle range of the vehicle.
(3) The improved APF method can not only jump out local minima but also plan smooth
trajectories by simulation in Matlab. By comparing the magnitude of curvature and
tracking the planned trajectories in Carsim platform, the reduction of Euclidean error
and steering wheel angle proved that the trajectories planned by the improved APF
method are easier to be tracked.
However, there are two drawbacks in the research of this paper:
1. The driving environment designed in this paper is urban road and the general vehicle
speed limit range is 30–60 km/h in the city. 10 m/s (36 km/h) and 15 m/s (54 km/h)
are the common speeds in the speed limit range, so they are used as the simulation
speed in this paper. Furthermore, obstacle avoidance of high-speed vehicles is a
complex motion planning involving braking and steering. If the speed of the vehicle is
too high, the actual trajectory will have a large deviation from the planned trajectory
World Electr. Veh. J. 2022, 13, 97 20 of 21

due to the inertia of the vehicle, which may directly cause the vehicle to have a lateral
collision with the obstacle and lead to the failure of the path planning. In this paper,
under the assumption of uniform vehicle motion, the vehicle trajectory planning
under high-speed motion is not considered.
2. The APF model in this paper does not consider the differences of obstacle avoidance
trajectories of different vehicle types in the actual road environment, and only consid-
ers the obstacle avoidance scenarios of flat and straight roads, which is a relatively
single scene.
In view of the above deficiencies, the subsequent research will keep improve the APF
model and enhance the adaptability of the model. Hence, future work should be devoted to
establishing the APF model under high-speed motion. In addition, an avenue for our future
work would be to research the interaction between the subject vehicle and the obstacle
vehicle with different parameters in the obstacle avoidance process.

Author Contributions: Conceptualization, J.T.; methodology, J.T.; software, J.T.; validation, J.T.;
formal analysis, B.L. and S.B.; investigation, J.T.; resources, J.T., B.L. and S.B.; data curation, J.T.;
writing—original draft preparation, J.T., H.H. and Z.Q.; writing—review and editing, J.T., S.B., B.L.,
D.Z., X.Z. and H.T.; visualization, J.T.; supervision, B.L. and S.B.; project administration, B.L. and S.B.;
funding acquisition, not applicable. All authors have read and agreed to the published version of
the manuscript.
Funding: This research was funded by the Natural Science Foundation of the Jiangsu Higher Educa-
tion of China under grant number 21KJA580001, the National Natural Science Foundation of China
under grant number 52172367, and the National Natural Science Foundation of China under grant
number 52105260. The APC was funded by 52172367.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Zhang, Y.; Xu, K.; Zheng, C.; Feng, W.; Xu, G. Advanced research on information perception technologies of intelligent electric
vehicles. Chin. J. Sci. Instrum. 2017, 38, 794–805.
2. LaValle, S.M. Planning Algorithms; Cambridge University Press: Cambridge, UK, 2006; pp. 64–76.
3. Bao, Q.; Li, S.; Shen, Z.; Men, X. Survey of local path planning of autonomous mobile robot. Transducer Microsyst. Technol. 2009,
28, 1–4.
4. Ma, R.; Guan, Z. Summarization for Present Situation and Future Development of Path Planning Technology. Mod. Mach. 2008, 3,
22–24.
5. González, D.; Pérez, J.; Milanés, V.; Nashashibi, F. A review of motion planning techniques for automated vehicles. IEEE Trans.
Intell. Transp. Syst. 2015, 17, 1135–1145. [CrossRef]
6. Yu, H.; Li, X. Fast Path Planning Based on Grid Model of Robot. Microelectron. Comput. 2005, 6, 98–100.
7. Ouyang, X.; Yang, S. Obstacle Avoidance Path Planning of Mobile Robots Based on Potential Grid Method. Control Eng. China
2014, 21, 134–137.
8. Dijkstra, E.W. A note on two problems in connexion with graphs. Numer. Math. 1959, 1, 269–271. [CrossRef]
9. Wang, D. Indoor mobile-robot path planning based on an improved A*algorithm. J. Tsinghua Univ. (Sci. Technol.) 2012, 52,
1085–1089.
10. Zhao, X.; Wang, Z.; Huang, C.; Zhao, Y. Mobile Robot Path Planning Based on an Improved A* Algorithm. Robot 2018, 40, 903–910.
11. Chen, Q.; Jiang, H.; Zheng, Y. Summary of Rapidly-Exploring Random Tree Algorithm in Robot Path Planning. Comput. Eng.
Appl. 2019, 55, 10–17.
12. Lindqvist, B.; Agha-Mohammadi, A.A.; Nikolakopoulos, G. Exploration-rrt: A multi-objective path planning and exploration
framework for unknown and unstructured environments. arXiv 2021, arXiv:2104.03724.
13. LaValle, S.M.; Kuffner, J.J., Jr. Randomized kinodynamic planning. Int. J. Robot. Res. 2001, 20, 378–400. [CrossRef]
14. Vega-Brown, W.; Roy, N. Asymptotically optimal planning under piecewise-analytic constraints. In Algorithmic Foundations of
Robotics XII; Springer: Berlin/Heidelberg, Germany, 2020; pp. 528–543.
15. Nakrani, N.M.; Joshi, M.M. A human-like decision intelligence for obstacle avoidance in autonomous vehicle parking. Appl. Intell.
2022, 52, 3728–3747. [CrossRef]
World Electr. Veh. J. 2022, 13, 97 21 of 21

16. Chen, Y.; Liang, J.; Wang, Y.; Pan, Q.; Tan, J.; Mao, J. Autonomous mobile robot path planning in unknown dynamic environments
using neural dynamics. Soft Comput. 2020, 1420, 289–299. [CrossRef]
17. Jafari, M.; Xu, H.; Carrillo, L.R.G. Brain emotional learning-based path planning and intelligent control co-design for unmanned
aerial vehicle in presence of system uncertainties and dynamic environment. In Proceedings of the 2018 IEEE Symposium Series
on Computational Intelligence (SSCI), Bangalore, India, 18–21 November 2018; pp. 1435–1440.
18. Li, L.; Gan, J.; Qu, X.; Mao, P.; Ran, B. Car-following Model Based on Safety Potential Field Theory Under Connected and
Automated Vehicle Environment. China J. Highw. Transp. 2019, 32, 76–87.
19. Khatib, O. Real-time obstacle avoidance for manipulators and mobile robots. In Autonomous Robot Vehicles; Springer: New York,
NY, USA, 1986; pp. 396–404.
20. Choe, T.S.; Jin, W.H.; Chae, J.S.; Park, Y.W. Real-time collision avoidance method for unmanned ground vehicle. In Proceedings of
the International Conference on Control, Honolulu, HI, USA, 8–11 July 2008.
21. Huang, Y.; Ding, H.; Zhang, Y.; Wang, H.; Cao, D.; Xu, N.; Hu, C. A motion planning and tracking framework for autonomous
vehicles based on artificial potential field-elaborated resistance network (apfe-rn) approach. IEEE Trans. Ind. Electron. 2019, 67,
1376–1386. [CrossRef]
22. Fan, S.; Zhang, H. Research on Local Path Planning of Vehicles Active Obstacle Avoidance Based on Improved Artificial Potential
Field Method. J. Qingdao Univ. (Eng. Technol. Ed.) 2022, 37, 50–57.
23. Li, E.; Wang, Y. Research on Obstacle Avoidance Trajectory of Mobile Robot Based on lmproved Artificial Potential Field. Comput.
Eng. Appl. 2022, 58, 296–304.
24. Bounini, F.; Gingras, D.; Pollart, H.; Gruyer, D. Modified Artificial Potential Field Method for Online Path Planning Applications.
In Proceedings of the Intelligent Vehicles Symposium, Los Angeles, CA, USA, 11–14 June 2017.
25. Cummings, M.L.M.; Marquez, J.J.; Roy, N.G.D. Human-automated path planning optimization and decision support. Int. J.
Hum.-Comput. Stud. 2012, 70, 116–128. [CrossRef]
26. Luo, D.; Wu, S. Ant colony optimization with potential field heuristic for robot path planning. Syst. Eng. Electron. 2010, 32,
1277–1280.
27. Cheng, Z. Robot Path Planning Based on Artificial Potential Field Method. Master Thesis, Yanshan University, Hebei, China, 2016.
28. Zhu, W. Research on Path Planning and Obstacle Avoidance Method for Vehicle Based on Improved Artificial Potential Field.
Master’s Thesis, Jiangsu University, Jiangsu, China, 2017.
29. Osman, K.; Ghommam, J.; Mehrjerdi, H.; Saad, M. Vision-based curved lane keeping control for intelligent vehicle highway
system. Proceedings of the Institution of Mechanical Engineers Part I. J. Syst. Control Eng. 2018, 233, 961–979.