Accident Analysis and Prevention 184 (2023) 106999

Contents lists available at ScienceDirect

Accident Analysis and Prevention

journal homepage: www.elsevier.com/locate/aap

Collision-avoidance lane change control method for enhancing safety for

connected vehicle platoon in mixed traffic environment
Yitao Ma a, Qiang Liu a, b, *, Jie Fu a, c, Kangmin Liufu a, Qing Li d
a
School of Intelligent Systems Engineering, Sun Yat-Sen University, Shenzhen 518000, China
b
Guangdong Marshell Electric Technology Company, Zhaoqing 526238, China
c
Sun Yat-Sen University-Guangzhou Automobile Research Institute Joint Laboratory of Intelligent Transportation and Artificial Intelligence, Guangzhou 516000, China
d
School of Aerospace, Mechanical and Mechatronic Engineering, University of Sydney, Sydney, NSW 2006, Australia

A R T I C L E I N F O A B S T R A C T

Keywords: In a mixed traffic environment, the connected vehicle platoon cannot communicate and collaborate with the
Collision-avoidance lane change surrounding vehicles. In this case, there is a high risk of collision in large vehicle platoon’s lane change scenario
Connected vehicle platoon where the non-connected surrounding vehicle occupies the target lane-changing space of the platoon. This study
Mixed traffic
proposes a collision-avoidance lane change control method for a connected bus platoon to elude the non-
Control method
Finite State Machine
connected vehicle in the target lane for completing lane change in the mixed traffic environment safely. A
platoon vehicle sensor system with low-cost and low data processing complexity is designed, which equips with
multiple sensors in longitudinal and lateral directions. Under control of the proposed platoon controller on the
basis of vehicle-to-vehicle (V2V) communication, the platoon following vehicles are fully autonomous in both
longitudinal and lateral directions. The safe lane change decision-maker is designed based on the Finite State
Machine (FSM). The decision-maker fuses multiple sensor data and determines the lane change operation of the
following vehicles. To verify the effectiveness of the proposed method, a three-vehicle platoon is carried out the
lane change experiments in a high-fidelity mixed traffic scenario built by the PreScan-Simulink joint simulation
platform. Exposure-to-Risk Index (ERI) of the platoon vehicles is adopted to evaluate the collision risk of the
platoon during lane changing process. Three typical case scenarios are tested, including unimpeded lane change,
passive waiting lane change, and active accelerating lane change. The simulation results show that all platoon
vehicles have an excellent success rate in lane change without collision with the non-connected surrounding
vehicle in these scenarios. The proposed method exhibits compelling benefits on improving the safety of platoon
vehicles in the mixed traffic environment.

1. Introduction benefits, such as increasing road capacity, improving road safety, and
reducing energy consumption (Axelsson, 2017, Sun and Yin, 2021, Liu
Connected and automated vehicle (CAV) integrates two advanced et al., 2022). In particular, the large vehicle platoon, such as bus platoon
technologies, namely vehicle-to-everything (V2X) communication and and truck platoon, can bring more outstanding socioeconomic benefits
vehicle automation, to capacitate cooperative planning and control for (Zhang et al., 2022b). Stimulated by these benefits, many research in­
vehicular operation (Wang et al., 2022b, Li et al., 2023). The previous stitutes and manufacturers have devoted to the study on platoon tech­
studies have shown that CAVs are effective in improving operational nology, such as PATH (Milanés and Shladover, 2014), SARTRE (Larburu
efficiency as well as traffic safety (Bian et al., 2019, Xiao et al., 2021). et al., 2012), Energy ITS (Sugimachi et al., 2013, Tsugawa, 2013),
There have also been expanding studies on exploring the applications of Scania (Alam et al., 2014), Volvo (Axelsson, 2017), etc., to realize the
CAV, such as platooning, cooperative lane change and collaborative practical applications of platoon technology recently. Before large-scale
perception (Yu et al., 2022, Zhang et al., 2022a, Zhao et al., 2022). deployment, it is necessary to thoroughly investigate the safety issues of
Platooning is considered one of the most promising CAV applications, the connected vehicle platoon (Li et al., 2018, Rahman and Abdel-Aty,
which has been recognized with its significant social and environmental 2018, Yang et al., 2022).

* Corresponding author at: School of Intelligent Systems Engineering, Sun Yat-Sen University, Shenzhen 518000, China.
E-mail address: [email protected] (Q. Liu).

https://doi.org/10.1016/j.aap.2023.106999
Received 16 August 2022; Received in revised form 10 January 2023; Accepted 6 February 2023
Available online 11 February 2023
0001-4575/© 2023 Published by Elsevier Ltd.
Y. Ma et al. Accident Analysis and Prevention 184 (2023) 106999

Lane change signifies an essential case scenario in practical appli­ some lane change management models and platoon organization stra­
cations of vehicle platoon. However, lane change is a main cause for a tegies for various mixed traffic environments supposing that the CAVs
significant number of accidents (Falsone et al., 2020, Peng et al., 2020), can obtain SV’s information via onboard sensing. However, the detailed
and vehicular crashes due to unsafe lane changes can lead to serious layout of the vehicle sensor system for the SV detection was not detailed
consequences (Adanu et al., 2021, Khattak et al., 2021). For the large in these articles. If a large vehicle transplants such existing sensor sys­
vehicle platoon which is nearly 50–100 m in size, lane-changing is rather tems implemented in an automated car (Khalifa et al., 2021, Wang et al.,
demanding. For this reason, many researchers have explored the safe 2022a), more radars, cameras and Lidars will need to be installed to
platoon lane change method. For example, Firoozi et al., proposed an cover the area around the large vehicle. That will increase the cost of the
optimal collision-avoidance motion planner and a Model Predictive sensor system and the difficulty of sensor system design and sensor data
Control (MPC) based path-following controller to navigate the platoon fusion. Therefore, it is worth exploring design of a platoon sensor system
lane changing smoothly (Firoozi et al., 2021). Zhang et al., developed a with low-cost and low data processing complexity for large vehicles.
hybrid MPC-based platoon-centered controller to address multiple lane Although the platoon gap is fairly short in general, the large vehicle
change requests from adjacent lanes next to the CAV platoon (Zhang platoon can be still nearly 50–100 m in size. Finding enough space for all
et al., 2022a). Xu et al., considered the scenario of a lane change control platoon vehicles to change lanes simultaneously might be rather difficult
maneuver for a car merging into the platoon from the adjacent lane and in increasingly-congested traffic. It is common that a platoon’s leading
proposed a safe lateral operation management strategy using dynamic vehicle starts changing lanes, whereas the platoon’s following vehicles
programming (Xu et al., 2021). By assuming that the vehicles could cannot follow immediately. This is due to the fact that the target lane-
communicate with each other locally, Gao et al., proposed a multi-lane changing space for the following vehicles could have been occupied
platoon controller based on the distributed graph and the potential field by the non-connected SV driving parallel to the platoon (Skottke et al.,
approach. The controller improved the obstacle avoidance ability of the 2014, Razmi Rad et al., 2020, 2021), as shown in Fig. 1. In this situation,
platoon by changing lanes and reshaping platoon formation before the risk of vehicular side-collision between the platoon and the SV in­
encountering obstacles (Gao et al., 2019). Note that some efficient creases (Hyun et al., 2021). In the mixed traffic environment, the pla­
methods for platoon and surrounding vehicles (SVs) cooperating in safe toon and the non-connected SV cannot coordinate lane change through
lane-changing were proposed in the previous studies which took ad­ vehicle-to-vehicle (V2V) communication (Ding et al., 2022). In this case,
vantages of CAV’s capability of communication and cooperation. the interactions between human-driven vehicles and CAV platoons are
However, a pure CAV traffic environment has not been realized to date inevitable (Mahdinia et al., 2021). The previous studies had proposed
(Rahman and Abdel-Aty, 2018, Chu and Kalabic, 2019, Zhu et al., 2021, some control methods to improve platoon safety in different scenarios
Hamad and Alozi, 2022, Khattak et al., 2022). (Axelsson, 2017). However, the case scenario where the non-connected
In the transition period, known as a mixed traffic environment, to a SV runs parallel to the platoon during platoon lane changing has yet to
pure CAV environment, how the CAV platoon eludes the non-connected be considered. To increase the safety of the large vehicle platoon in the
SV and completes lane change safely without inter-vehicular commu­ mixed traffic environment, it is necessary to thoroughly explore the
nication and collaboration remains a significant issue to be addressed control method for platoon safe lane change under the target lane-
thoroughly prior to practical implementation (Virdi et al., 2019, Ma changing space occupied condition.
et al., 2022). To date, only limited research has concerned platoon lane- This paper proposes a collision-avoidance lane change control
changing in a mixed traffic environment. In this regard, Khattak et al., method for a large vehicle platoon in a mixed traffic environment. We
developed an enhanced prototype CAV application for lane management choose a 12 m bus as the representative large vehicles to study the safe
in a mixed traffic environment of human-driven vehicles (HDVs) and lane-changing method for the large vehicle platoon. First, we design an
CAVs (Khattak et al., 2022). Woo and Skabardonis constructed a enhanced platoon vehicle sensor system with multiple sensors in lon­
microscopic traffic model to evaluate the flow disruption of platoon lane gitudinal and lateral directions, which are millimeter-wave radars and
change under some mixed traffic conditions (Woo and Skabardonis, ultrasonic radars, to help the platoon vehicles detect the non-connected
2021). Liu et al., proposed a deep reinforcement learning-based proac­ surrounding vehicle in the target lane. According to the leader–follower
tive longitudinal controller for CAVs platoon to counteract disruptive policy, a V2V communication-based platoon vehicle controller is
HDVs lane change behaviors that can induce disturbances (Liu et al., devised. Then, a collision-avoidance lane change decision-maker is
2021). Note that accurately detecting the non-connected SV is consid­ developed based on a Finite State Machine (FSM) to fuse multiple sen­
ered a precondition to elude it successfully. These above works proposed sors’ data and instruct the safe lane change maneuvers for the platoon in

Fig. 1. The accident between platoon vehicle and non-connected vehicle during platoon lane changing.

2
Y. Ma et al. Accident Analysis and Prevention 184 (2023) 106999

dangerous scenarios. We validate the effectiveness of the proposed achieved fully autonomous yet (Li et al., 2021). Therefore, the leading
method by simulation tests conducted in the PreScan-Simulink joint vehicle in the platoon still needs an experienced driver to manipulate at
simulation platform. The platoon equipped with the enhanced sensor present (Wei et al., 2019, Lee et al., 2020, Khalifa et al., 2021, Zhang
system and the devised controller is tested in the unimpeded lane change et al., 2022b). The SV is a non-connected human-driven vehicle. The SV
scenario as well as the lane change space occupied scenario. Exposure- and the LV maneuver along a reference trajectory consisting of a preset
to-Risk Index (ERI) is used as an indicator to evaluate the collision path and speed profile (Wang et al., 2022a, Zhang et al., 2022b). In the
risk of the platoon vehicles during lane changing. The results demon­ tests, the SV travels along the center line of the left lane on the straight
strate that the proposed method can help platoon vehicles elude the non- road. The LV travels along the center line of the center lane, and then
connected surrounding vehicle in the target lane and complete lane changes to the left lane on the straight road. The lane change trajectory
change without collision. In the course of lane changing, the collision of the LV is approximated with a quintic polynomial (Peng et al., 2020).
risk of the platoon remains at a low level. The FV1 and the FV2 are controlled automatically by the platoon
controller proposed in Section 3. In the beginning of the simulation, the
2. Simulation modeling platoon and SV travel at the same velocity in their current lanes. A few
seconds later, the LV shifts to the left in the lane where the SV travels, as
2.1. Problem statement shown in Fig. 2.

From the previous studies, it has been recognized that the in­ 2.3. Vehicle kinematic modeling
teractions between the lane-changing vehicle and its surrounding ve­
hicles in the target and original lane significantly influence collision risk In the platoon controller design, the behavior of the platoon vehicle
of vehicular lane-changing (Chen et al., 2019, Chen et al., 2020b). In the needs to be modeled for determining correct motions (Polack et al.,
study concerning the effects of CAVs platoon on the behavior of manual 2017). The nonlinear behavior of each following vehicle is modeled by
vehicle drivers, it was found that HDVs intended to emulate the inter- the kinematic model, which is a common modeling approach in platoon
vehicle distance and speed to run parallel to the platoon (Martínez- control (Wang et al., 2015, Gao et al., 2019, Firoozi et al., 2021, Xu et al.,
Díaz et al., 2021, Razmi Rad et al., 2021). As shown in Fig. 1(a), a 2021). The kinematic model of the platoon following vehicles is shown
leading CAV (signed as LV) driven by an experienced driver and two in Fig. 3. Let (X, Y) denote the generalized coordinates of the vehicular
following CAVs (signed as FV1 and FV2) formed a human-lead-platoon moving, where the positive direction of the X-axis is due east and the Y-
via V2V communication. The surrounding vehicle (SV) is a non- axis positive direction is due north. P1(Xr, Yr) is the midpoint of the
connected vehicle driven by a human driver. While much work has vehicle’s rear axle, and P2(Xf, Yf) is the midpoint of the vehicle’s front
been conducted for safe platoon lane changing in the pure CAV envi­ axle. Define the vehicle-body local coordinate (X’, Y’) with the rear axle
ronment (Gao et al., 2019, Firoozi et al., 2021, Xu et al., 2021, Zhang midpoint P1(Xr, Yr) as the origin of the coordinate, where the x-axis
et al., 2022a), development of a feasible method is necessary for platoon positive direction is the forward direction of the vehicle. φ is defined as
vehicles to detect and avoid surrounding vehicles during platoon lane the heading angle of the vehicle, in which the positive direction is
changing in the mixed traffic environment. In some dangerous case anticlockwise. The dynamics of the vehicle can be determined by the
scenarios, if the LV changes to the left lane and the FV1 follows the state vector [Ẋr , Ẏr , φ̇]T. Thus, the kinematic model of an autonomous
moving of the LV simultaneously, it is easy to cause a severe side-impact following vehicle can be described in the following equation:
accident between the SV and the FV1, as shown in Fig. 1(b). To reduce ⎡ ⎤ ⎡ ⎤
the danger in this scenario, we develop a new sensor system for Ẋr cosφ
⎣ Ẏr ⎦ = ⎣ sinφ ⎦vr (1)
detecting surrounding vehicles in the adjacent lane and a control
φ̇ tanδf /l
method for collision-avoidance lane change with the large vehicle pla­
toon in a mixed-flow traffic environment.

2.2. Test scenario construction

Considering the safety and economy of the real-life tests, it is

demanding to establish the actual experiment for platoon collision-
avoidance lane change. For this reason, simulation experiments are
conducted to verify the effectiveness of the proposed method in a mixed
traffic environment. First, a high-fidelity platoon lane change scenario in
a mixed traffic environment is constructed in PreScan, which is an
Advanced Driving Assistance System (ADAS) and active safety system
simulation program applied widely (Peng et al., 2020, Liufu et al.,
2021).
A 1500 m long three-lane straight road with a lane width of 3.5 m is
considered in the simulation. On the road, there is a non-connected
surrounding vehicle (a truck 10 m long and 2.1 m wide) and a platoon
consisting of three CAVs, signed as LV, FV1 and FV2, as shown in Fig. 2.
Fig. 3. Vehicle nonlinear kinematic model.
Due to existing technological limitations, the leading CAV has not

Fig. 2. Platoon lane change scenario in a mixed traffic environment built by PreScan.

3
Y. Ma et al. Accident Analysis and Prevention 184 (2023) 106999

where l is the length of the wheelbase, vr is the velocity of the vehicle, surrounding vehicles in the adjacent lanes. The mounting height of all
and δf is the steering angle of the front wheel. In this study, the ultrasonic sensors is 0.8 m from the ground. The horizontal distance
wheelbase of the platoon vehicles is 5.88 m, and the length and width of between the two adjacent ultrasonic sensors is 2.0 m, much smaller than
the platoon vehicles are 12.00 m and 2.54 m, respectively. the length of the shortest road vehicles. Hence, it can guarantee that no
surrounding vehicle can hide in the blind zone between two adjacent
ultrasonic sensors. The vehicle’s rear end is equipped with a rearward
2.4. Design and modeling of the vehicle sensor system
radar (FOV: ±75◦ , MDR: 5 m) to detect surrounding vehicles behind and
in adjacent lanes. The system only uses radar-type sensors, whose cost is
Sensor perception is a primary way for CAVs to detect and track non-
relatively lower. Since the sensors’ perception data format is consistent,
connected vehicles as CAVs cannot communicate with non-connected
the data processing and fusion are relatively simple. Compared with the
HDVs to acquire information in a mixed traffic environment (Ding
existing sensory systems in automated vehicles, the proposed sensor
et al., 2022). In literature, some researchers have developed multiple
system exhibits the advantage of low-cost and low data processing
sensor systems for platoon vehicles (Tsugawa, 2013, Lu and Shladover,
complexity.
2014, Song et al., 2016). These sensor systems play a key role in
maintaining platoon gap, speed difference and string stability. In the
3. Platoon controller modeling
recent studies, the sensor system of CAVs consisted of Radars, Lidars and
video cameras, which were mounted on the front of the vehicle body
The platoon controller described in this section allows automatically
(Schmidt, 2017, Cui et al., 2018, Lefeber et al., 2020, Meyer et al., 2021,
controlling the acceleration, deceleration and steering of the following
Balador et al., 2022). It is noted that if the existing sensor system is
vehicles in the platoon. The platoon controller is designed based on the
transplanted into a large vehicle, more radars, cameras and Lidars will
leader–follower strategy (Park and Yoo, 2021). The principle of the
need to be installed to cover the area around the vehicle to avoid blind
leader–follower strategy is shown in Fig. 5, in which the central radar
spots of perception. This will increase the cost of the sensor system and
coordinate (x, y) with the sensor’s mounting point is considered as the
the difficulty of system design and sensor data fusion. Therefore, how to
origin of the coordinate. The position deviation between two consecu­
design the platoon vehicle sensor system with low-cost and low data
tive vehicles can be defined as a vector [Range, θ], where Range and θ
processing complexity remains a significant issue worth investigating.
represent the distance and azimuthal angle in the radar coordinate at
To ensure safety in the course of platoon lane changing in a mixed traffic
which the predecessor is detected. The longitudinal gap xreal and the
environment, an enhanced platoon sensor system is proposed here by
lateral distance yreal between the two consecutive vehicles can be
considering longitudinal and lateral direction perception.
calculated by the following equations, respectively:
Fig. 4 presents the enhanced platoon vehicle sensor system, which is
{
constructed of the physics-based sensor model provided by PreScan. xreal = Range⋅cosθ
These sensors are mounted on the front, sides, and rear of the 3D vehicle (2)
yreal = Range⋅sinθ
body. In the front of the vehicle, three millimeter-wave radars are used
to detect the position data of the forward vehicle or obstacle, including a Vehicles in the platoon are expected to travel stably with a specific
central radar (FOV (Field of View): ±30◦ , MDR (Maximum Detection longitudinal gap xreal and a specific lateral distance yreal (usually set to be
Range): 100 m), and two short-range radars (FOV: ±60◦ , MDR: 20 m) on zero). In this study, the longitudinal gap xreal is 6 m and the lateral
the both sides. The central radar measures the distance and the distance yreal is 0. Hence, the goal of the platoon controller is to reduce
azimuthal angle between the forward vehicle and itself. Two short-range errors δx and δy to zero, in which the δx and δy are calculated as follows:
radars measure and output differential azimuthal angle signals to reduce {
δx = xreal − xexp
the lateral error between the forward vehicle and itself. (3)
δy = yreal − yexp
Different from the conventional cooperative adaptive cruise control
(CACC) platoon vehicle sensor system, our system adds multiple ultra­ The platoon motion controller consists of a longitudinal controller
sonic radars on both sides of the vehicle whose beam center line is not and a lateral controller, which control the speed and front wheel steering
parallel to the vehicle’s driving direction. Two kinds of ultrasonic radars angle of the vehicle, respectively. The safe lane change decision-maker
with different MDR are applied here. The beam center line of the short- estimates whether it is safe enough to change lanes and follow the
range ultrasonic sensor is perpendicular to the driving direction of anterior vehicle based on the fusion perception data from the sensors.
vehicle. The angle between the beam center line of the long-range ul­
trasonic sensor and the horizontal center line of the vehicle is 45◦ . Two
different kinds of ultrasonic radars are used to detect the presence of the

Fig. 4. The schematic of enhanced platoon vehicle sensor system.

4
Y. Ma et al. Accident Analysis and Prevention 184 (2023) 106999

Fig. 5. Illustration of the position deviation between the two consecutive platoon vehicles.

3.1. Longitudinal controller

vout (i) = vraw (i) (9)
By controlling the velocity of the host vehicle (FV1 and FV2), the
longitudinal controller maintains its speed to be the same as the leading 3.2. Lateral controller
vehicle, thereby guaranteeing a stable longitudinal gap between the two
consecutive vehicles involved. The expected speed can be calculated as: The lateral controller is designed to reduce the lateral distance yreal
between the two consecutive vehicles to zero. The lateral distance yreal ,
vraw = αδx* + βvlv + γ(alv − ahv ) (4) calculated by forward-looking radars’ detection data, goes through a
first-order filter to remove the noise for obtaining δy∗ . Let dp be the
where vraw is the roughly expected velocity of the host vehicle; δx∗ is the steering preview distance, which is equal to the expected longitudinal
value that the longitudinal gap difference δx between the actual distance →
gapxexp = 6 m. In the xoy coordinate plane, the steering vector S (dp ,
xreal and the expected platoon gap xexp converted by a first-order filter;
δy ) is constructed with the coordinate origin as the starting point. Let λ
∗
vlv is the velocity of the leading vehicle in the platoon; alv is the accel­ →
eration of the leading vehicle; ahv is the acceleration of the host vehicle be the argument of vector S (dp , δy∗ ). With the above transformation,
(FV1 and FV2); α, β, and γ are the weighting coefficients for each the lateral error, calculated by the forward-looking radars’ fusion data,
parameter in Eq. (4). Note that the vlv and alv are transmitted between can be converted to the angle in between the traveling directions of two
the leading vehicle and the host vehicle (FV1 and FV2) via the V2V consecutive vehicles in the platoon. The PI (Proportional-Integral)
communication. The communication delay in transmitting the speed lateral controller, whose input is λ and output is front wheel steering
and acceleration messages from the LV to the following vehicles is angle δf , is designed as:
considered. Due to the constant longitudinal gap between two platoon
∑
i
members, the delay time between the two contiguous vehicles is usually δf (i) = KP λ(i) + KI λ(j)δT (10)
configured as a constant value (Cui et al., 2018). In this study, the j=0

communication delay between the LV and FV1 is set as 5 ms; and the
communication delay between the LV and FV2 is set to be 10 ms (Song where KP is the proportional gain; KI is the integral gain. The parameters
et al., 2016). of the PI controller, KP and KI, are obtained by manual adjustment for
vraw is not the final output expected velocity of the longitudinal the best possible values.
controller. To avoid excessive vehicle speed fluctuation or speed varia­ It is necessary to introduce a dead zone block to avoid repeated jitter
tion rate that is out of the vehicle’s physical limit, we design a velocity in the δf output from the PI controller when λ tends to zero, which may
variation limiter to constrain the output value of the expected speed cause vehicle’s unstable lateral moving. Let -ξ be the start value of the
model defined in Eq. (4). The limiter will calculate the first derivative dead zone and ξ be the end value of the dead zone. The dead zone block
v̇raw of the vraw and determine whether or not the v̇raw meets the generates a zero output when the δf output from the PI controller is in
condition: the specified region [ξ, -ξ]. Finally, a saturation block is connected in
serial behind the dead zone block to avoid the vehicle steering angle
amin ⩽v̇raw ⩽ amax (5) exceeding the limit of vehicle dynamics, causing vehicle rollover. The
upper limit of the saturation block is δmax and the lower limit is -δmax . In
where amin and amax are the maximum deceleration and maximum ac­
other words, when δf is within the region [-δmax , δmax ], the lateral con­
celeration that the vehicle can achieve at the current speed, respectively.
troller’s final output is δf ; when δf is less than -δmax , the lateral con­
The derivative v̇ raw is calculated as follows:
troller’s final output is -δmax ; and when δf is greater than δmax , the lateral
vraw (i) − vout (i-1) controller’s final output is δmax .
v̇raw = (6)
δT
3.3. Longitudinal controller and lateral controller performance
where vraw (i) is the current input of the velocity variation limiter; verification
vout (i-1) is the output of the velocity variation limiter at the previous
step; δT is the simulation step size. The output of the velocity variation A straight cruising simulation test is conducted to verify the perfor­
limiter is determined by comparing the v̇raw to the maximum decelera­ mance of the longitudinal and the lateral controller. A platoon with
tion amin and the maximum acceleration amax parameters: three vehicles is applied for the verification here. The desired platoon
(1) If v̇raw is greater than the maximum acceleration amax , the output gap is 6 m and the desired lateral error between the leading vehicle and
of limiter is calculated as: the following vehicle is 0 m. The initial speed of the LV is 0 km/h. In the
vout (i) = δT • amax + vout (i-1) (7) acceleration stage, the LV accelerates smoothly until reaching the
desired speed of 60 km/h. The LV cruises at 60 km/h for about 10 s and
(2) If v̇raw is less than the maximum deceleration amin , the output of then decelerates at a fixed rate of 3 m/s2 until it entirely stops. During
limiter is calculated as: the test, the following vehicles (FV1 and FV2) are automatically
vout (i) = δT • amin + vout (i-1) (8) controlled by the proposed controller. The critical parameters of the
platoon controller are shown in Table 1.
(3) If v̇raw is between the bounds of amin and amax , the output of limiter Fig. 6 presents the velocity, longitudinal error, and lateral error of
is equal to the input: each vehicle (LV, FV1 and FV2). It can be seen from the figure that the

5
Y. Ma et al. Accident Analysis and Prevention 184 (2023) 106999

Table 1 sensors on the left side of the vehicle body monitor the surrounding
The controller parameters used in the simulation. vehicles in the left lane in real-time. When all the ultrasonic sensors do
α β γ KP KI ξ (◦ ) δmax (◦ ) not detect any target, it is confirmed that the left lane is clear. Rearward
radar scans the space on the left-rear of the ego vehicle in real-time.
4 1 0.5 0.6 0.01 0.1 1
When the distance between the ego vehicle and the detected target is
greater than the preset safety threshold, it is confirmed that the left-rear
longitudinal controller shows good performance in speed tracking and space is clear. Finally, after the clearance maintains for more than 1 s,
platoon gap control. The maximum longitudinal error is − 0.46 m and the following vehicle changes from KCL state to CLL state. In the CLL
the maximum longitudinal error rate is − 7.6 %. The initial lateral error state, the following vehicle commanded by the controller mentioned in
between LV and FV1 as well as between LV and FV2 is set as 0.5 m and Sections 3.1 and 3.2 steers automatically to change lanes and trace the
− 0.5 m (here deviation from the center axis of the LV is positive on the leading vehicle. After the center point of the vehicle crosses the center
left side and negative on the right side), respectively. Evidently, the line of the left lane (C7), the vehicle stops changing lanes and resumes
initial lateral error is reduced to nearly zero after 10 s. The measurement keeping the current lane state.
error of the radars causes the tracking errors of the platoon vehicle. Note
that the radar detection point of the same target is not fixed. It implies
that the longitudinal and lateral distances between the center points of
the ego vehicle’s front bumper and the preceding vehicle’s rear bumper
measured by radars are not 100 % precisely.
According to the requirements of the “Test Procedures for Automatic
Driving Function of Intelligent Connected Vehicles of China” (NTCAS,
2018), the actual longitudinal gap error between platoon vehicles
should be kept within ± 25 % of the expected longitudinal gap during
platooning, and the lateral error relative to the leading vehicle should be
kept within ±0.5 m when the platoon is cruising. From this perspective,
the proposed platoon controller satisfies the above performance
requirements.

3.4. Safe lane change decision-maker

The safe lane change decision-maker is designed based on Finite

State Machine (FSM) (Chen et al., 2020a, Cui et al., 2021, He et al.,
2021). The lateral motion control of platoon vehicles on the straight
road can be divided into three states, which are keeping the current lane Fig. 7. The transition diagram of Finite State Machine (FSM) and the spatial
(KCL), changing lanes to the left (CLL), and changing lanes to the right division surrounding the vehicle.
(CLR). Fig. 7(a) provides a schematic of the FSM-based decision-maker.
If the sensor system detects the surrounding vehicle in the target lane-
Table 2
changing space, the following vehicles enter the KCL state. Mean­
The transition conditions of Finite State Machine (FSM) for safe platoon lane
while, the decision-maker updates the expected lateral distance from change.
0 to yreal . The following vehicle can change from KCL state to CLL state
Transition Explanation
or CLR state only when the vehicle gets enough safe lane-changing
condition
space. When the safe lane-change condition is satisfied, the expected
C1 The space on the left side of the vehicle is clear.
lateral distance is updated from yreal to 0 and the following vehicles
C2 The space on the right side of the vehicle is clear.
changes lanes. C3 The space on the left-rear of the vehicle is clear.
The transition conditions are summarized in Table 2. For example, C4 The space on the left-rear of the vehicle is clear.
the following vehicle follows the leading vehicle to change lanes to the C5 The clearance on the left and the left-rear of the vehicle
left lane for analyzing the state transition process of the following maintains more than 1 s.
C6 The clearance on the right and the right-rear of the vehicle
vehicle. The KCL state is the initial state of the following vehicle. To maintains more than 1 s.
switch from the KCL state to the CLL state, the following vehicle must C7 Left lane change is complete
meet three conditions, including left side clear (C1), left-rear clear (C3), C8 Right lane change is complete
and clearance maintaining more than 1 s (C5). Multiple ultrasonic

Fig. 6. The results of platoon straight cruising test.

6
Y. Ma et al. Accident Analysis and Prevention 184 (2023) 106999

4. Simulation experiments and results FV2′ s geometric center the from LV’s geometric center in the lateral
direction in meters.
4.1. Experimental setup
4.1.3. Surrogate safety measure
4.1.1. Scenario design and settings Exposure-to-Risk Index (ERI) is adopted here to evaluate the safety of
Three types of platoon lane change scenarios are considered here, the following vehicles during lane changing. ERI is the probability of
namely unimpeded lane change scenario, platoon passive waiting sce­ exposure to a dangerous situation in the course of lane change, to reflect
nario, and platoon active accelerating scenario. All the experiments are the likelihood of crash occurrence between the following vehicles and
conducted in the PreScan-Simulink joint simulation platform. Those the SV (Zhang et al., 2023). ERI is calculated by the following equation:
scenarios aim to test whether the platoon can avoid the surrounding
TGTTC< GTTCt
vehicles and complete lane change safely; and how the proposed platoon ERI = (11)
TLC
longitudinal and lateral controller would affect lane change perfor­
mance. Note that the architecture of the safe lane change decision-maker where, GTTC represents the generalized time-to-collision, GTTCt is the
is symmetric based on FSM, and the performance of the platoon is threshold of GTTC, TGTTC< GTTCt is the duration when GTTC is smaller
consistent in the case of lane change to the left and lane change to the than the threshold GTTCt, and TLC is the duration of lane change. The
right. Therefore, to reduce duplication of the experiments, the platoon is threshold value of GTTC selected can vary from 1.5 s to 4 s in literature
considered to change lanes to the left in all the simulations. (Zhang et al., 2023), so this study selected six thresholds of GTTC (4, 3.5,
In this section, fifteen simulation tests are implemented, whose ve­ 3, 2.5, 2, and 1.5 s).
locity parameters of the vehicles are summarized in Table 3. In the GTTC is proposed by Zhang et al. (Zhang et al., 2023) to evaluate the
different tests, the LV and SV are configured to drive at speeds of 40–60 risk in the lane change. The outer contour of vehicles is imaged as a
km/h, which is a realistic speed range for a three-lane urban road in rectangle and some points on the rectangle’s edges are created to sub­
China (Zheng et al., 2022). To standardize each test, the driving path of stitute the rectangle, as shown in Fig. 8. The distance Dij of the two
the LV is set to be uniform in all the tests that are performed in the same vehicles (denoted as Vi and Vj) is the shortest distance among these point
scene constructed in Section 2.2. All simulation tests with a sampling
pairs. Ḋij is the change rate of the distance Dij . GTTC can be computed as
time of 5 ms are performed on a computer with i7-10700F CPU.
follows:
4.1.2. Measurement of effectiveness − Dij
GTTC = (12)
The success rate of collision avoidance of the following vehicles is Ḋij
considered to be an indicator to evaluate the effectiveness of the pro­
posed method (Li et al., 2020a, Cui et al., 2021, Hang et al., 2022). The
longitudinal error (m) and the lateral error (m) are adopted to assess the 4.2. Unimpeded lane change scenarios
performance of the safe platoon lane change control method. The lon­
gitudinal error measures the difference between desired platoon gap and In these scenarios, one leading vehicle and two following vehicles
the actual distance gap. The lateral error is the deviation of the ego form a single-lane platoon cruising with an expected velocity and an
vehicle’s geometric center from the LV’s geometric center in the lateral expected longitudinal gap of 6 m and then change lanes from the middle
direction. To be more specific, there are four variables selected to to the left lane. The simulation results of the unimpeded lane change
measure the performance, which are listed as follows: scenarios are shown in Fig. 9 and Fig. 10. The maximum longitudinal
Elong.FV1-LV: The longitudinal error of FV1 and LV. It is the difference error is 0.09 m. Moreover, the maximum lateral error is 0.07 m after the
between the desired platoon gap and the actual gap between FV1 and LV entire platoon completes the lane change. As shown in Fig. 9, the lon­
in meters. gitudinal error stays almost the same as the cruising speed increases,
Elong.FV2-FV1: The longitudinal error of FV2 and FV1. It is the differ­ while the lateral error increases from 0.04 m to 0.07 m. The average
ence between the desired platoon gap and the actual gap between FV2 longitudinal error is 0.05–0.07 m in these tests. Note that a slight speed
and FV1 in meters. oscillation is observed one second after the simulation starts. To save
Elat.FV1-LV: The lateral error of FV1and LV. It is the deviation of the time for the platoon to accelerate from zero to target cruising speed in
FV1′ s geometric center the from LV’s geometric center in the lateral every simulation, the following vehicles are controlled at a constant
direction in meters. expected speed and gap by the program in the first second. After the first
Elat.FV2-LV: The lateral error of FV2and LV. It is the deviation of the second, the following vehicles are controlled by the platoon controller
proposed in Section 3. When the control of the following vehicles is

Table 3
The velocity parameters of the vehicles in the simulation tests.
Scenario type Test number LV initial velocity (VLI) LV final velocity (VLF) SV initial velocity (VSI) SV final velocity (VSF) Velocity difference (δ V)
(km/h) (km/h) (km/h) (km/h) (km/h)

Unimpeded lane change 1 40 40 / / /

2 50 50 / / /
3 60 60 / / /
Platoon passive waiting 4 40 40 40 35 5
5 40 40 40 30 10
6 50 50 50 45 5
7 50 50 50 40 10
8 60 60 60 55 5
9 60 60 60 50 10
Platoon active accelerating 10 40 45 40 40 5
11 40 50 40 40 10
12 50 55 50 50 5
13 50 60 50 50 10
14 60 65 60 60 5
15 60 70 60 60 10

7
Y. Ma et al. Accident Analysis and Prevention 184 (2023) 106999

Fig. 8. Points on vehicles and the shortest distance between them.

Fig. 9. Simulation results of unimpeded lane change scenarios.

rate is 1.5 %. After the platoon finished lane change and regained
cruising, the maximum value of Elat.FV1-LV and Elat.FV2-LV is 0.04 m and
0.07 m, respectively. It can be seen that the platoon cruises as expected,
as both longitudinal error and lateral error stay stable nearly zero during
the simulation. More importantly, the ability of the proposed platoon
controller that it can control the following vehicles to change lanes
safely by following the leading vehicle is confirmed.

4.3. Lane change space occupied scenarios

In the case shown in Fig. 2, the platoon can take two different ma­
neuvers to increase the speed difference with the SV so as to obtain safe
lane-changing space. The first maneuver is passive waiting, where the
following vehicles wait for the non-connected surrounding vehicle to
slow down to make room for a proper lane change prior to making the
change. The second maneuver is active accelerating, where the leading
vehicle accelerates after changing lanes while the following vehicles
trace and accelerate, waiting to gain enough space before changing
lanes. Two different maneuvers for lane change and collision avoidance
of the following vehicles are assessed in the mixed traffic conditions. The
Fig. 10. Performances statistics during unimpeded lane changing. simulation results of these two scenarios are described in Sections 4.3.1
and 4.3.2, respectively.
switched from program to platoon controller, the platoon controller
4.3.1. Passive waiting maneuver scenario tests
adjusts the vehicle speed based on the radar output distance with
The simulation processes are shown in Fig. 11. In the beginning, the
measurement error and brings a marginal speed oscillation. The same
three vehicles platoon is traveling in the middle lane with an expected
phenomenon also occurs in the other simulated scenarios.
speed (40, 50, 60 km/h) and a platoon gap of 6 m. The SV, whose ve­
Fig. 10 provides summary statistics of the platoon controller per­
locity is the same as the platoon, drives side-by-side with FV1 in the left
formance. As the cruising speed increases, longitudinal error stays
lane. After traveling the same distance, the LV starts changing to the left
relatively the same. In all tests of this scenario, the maximum longitu­
lane from the middle lane. After the LV finishes lane change, the SV
dinal error occurs during platoon lane changing and the maximum error
decelerates with a constant deceleration of 1 m/s2 until the speed

8
Y. Ma et al. Accident Analysis and Prevention 184 (2023) 106999

Fig. 11. Major behaviors of the platoon at six moments during lane changing.

difference with respect to the platoon is 5 km/h or 10 km/h; and then 0.04 m after the entire platoon completes the lane change. When the
the SV keeps a new constant cruising speed. following vehicle’s sensor system detects a surrounding vehicle in the
In all the tests of this scenario, the following vehicles have a 100 % target lane, it is demonstrated from Fig. 11 and Fig. 12 that the following
success rate of collision avoidance with the SV, complete lane change, vehicle does not immediately follow the preceding vehicle to change
and recover platoon configuration cruising in the left lane. The perfor­ lanes but stays in the current lane under the instruction of the safe lane
mances during simulation are presented in Fig. 12 and Fig. 13. The lane- change decision-maker. In this case, the following vehicles successfully
changing durations are marked as “lane change” in Fig. 12. The avoid collision with the SV. While waiting for the SV to decelerate, the
maximum longitudinal error is 0.11 m and the maximum lateral error is lateral errors between the following vehicles and LV are about 3.5 m,

Fig. 12. Simulation results of platoon passive waiting scenarios.

9
Y. Ma et al. Accident Analysis and Prevention 184 (2023) 106999

Fig. 13. Performances statistics during passive waiting lane changing.

Fig. 14. Simulation results of platoon active accelerating scenarios.

10
Y. Ma et al. Accident Analysis and Prevention 184 (2023) 106999

equaling the lane width. After the entire platoon finishes lane change vehicles change lanes to the left.
and regains cruising, the maximum value of Elat.FV1-LV and Elat.FV2-LV are The magnitude of the lateral error does not increase noticeably with
0.03 m and 0.04 m, respectively. It indicates that the platoon can recover the increase in platoon cruising speed. After the entire platoon finishes
the expected zero lateral distance after avoiding the SV and finishing lane change and regains cruising, the maximum values of Elat.FV1-LV and
lane-changing. Elat.FV2-LV are 0.03 m and 0.07 m, respectively, which is similar to the
As shown in Fig. 12, the velocity of the following vehicles is basically unimpeded lane change scenario. It demonstrates that the proposed
the same as the LV during lane changing, even though the following controller and safe lane change decision-maker can make platoon lane-
vehicles are not in the same lane as LV sometimes. The average longi­ changing safely under various speed conditions.
tudinal error is 0.05–0.07 m, which is the same as the results of unim­ For the longitudinal error, the averages of Elong.FV1-LV and Elong.FV2-FV1
peded lane change scenario. Whenever a vehicle in the platoon starts remain mostly the same with increasing speed. However, the maximum
changing lanes, there is a slight speed oscillation with a magnitude of error is greater than those presented in the unimpeded lane change
about ±1 km/h, which is caused by the change of the radar detecting scenario and the passive waiting maneuver scenario. When the LV ac­
points. When the anterior vehicle starts to change lanes, the azimuthal celerates, the longitudinal error increases until the LV stops acceler­
angle θ between two contiguous vehicles changes, and the radar ating. The reason is that due to the communication latency, the
detecting points of the following vehicle move horizontally so that the followers’ response would always be slower than the LV. Based on the
radar measure distance presents marginal saltation, leading to the speed facts that Elong.FV2-FV1 is less than Elong.FV1-LV, the gap oscillation is not
oscillations. amplified along with the platoon when the LV accelerates. The longi­
In all the tests of this scenario, the maximum longitudinal error of the tudinal error then converges to the average value when the LV starts
following vehicles is less than 0.12 m, which is slightly greater than that travelling at a constant speed. Hence, the platoon string stability still
of the unimpeded lane change scenario, as shown in Fig. 13. However, it holds in this scenario.
indicates that even if the following vehicle is not in the same lane as the
preceding vehicle, the longitudinal performance of the proposed platoon 4.4. Exposure-to-risk index of the following vehicles
controller does not degrade. To sum up, the proposed control method for
collision-avoidance lane change is effective and does not influence the Fig. 16 presented the ERI of the following vehicles (FV1 and FV2)
longitudinal and lateral performance of the platoon controller. during platoon lane changing. As the threshold of GTTC decreases, ERI
of the following vehicles also decreases. When the threshold of GTTC is
4.3.2. Active accelerating maneuver scenario tests great, ERI of the following vehicles remains at a low level (less than
In this scenario, the initial motion state of the platoon and the SV is 0.27). It indicates that the proposed collision-avoidance lane change
consistent with the passive waiting maneuver scenario. The SV, whose control method can help the platoon change lanes safely maintaining a
velocity is the same as the platoon, drives side-by-side with FV1 in the low collision risk with the SV.
left lane. Then, the LV starts changing from the middle lane to the left From Fig. 16, ERI of the following vehicles in the platoon active
lane, as illustrated in the T1 and T2 screenshots in Fig. 11. Unlike the accelerating scenario is lower than the platoon passive waiting scenario
passive waiting maneuver scenario, the SV constantly maintains its under the same speed condition. It suggests that when the SV is taking
initial speed (40, 50 or 60 km/h) during the tests. To make room for the platoon’s target lane-changing space, the leading vehicle actively
lane-changing of the following vehicles, the LV accelerates until the accelerating and leading the platoon to overtake the SV is a safer ma­
speed difference relative to the SV is 5 km/h or 10 km/h and then neuver with lower collision risk, compared to passive waiting for the SV
maintains the final speed. to decelerate. In all the scenario involving occupied lane change space
It is found that in all the tests of the active accelerating maneuver (Test 4–15), ERI of the FV2 is higher than the FV1. This is due to the fact
scenario, the FV1 and the FV2 have a 100 % success rate of collision that the FV2 and the SV drive side by side for a longer time. Therefore, to
avoidance with the SV during lane changing. The simulation results are ensure the platoon’s lane change safety, the number of vehicles in the
presented in Fig. 14 and Fig. 15. The average longitudinal error is platoon should not be excessive.
0.05–0.07 m, which is the same as the previous two scenarios. The
maximum longitudinal error is 0.25 m; and the maximum lateral error is 4.5. Comparative analysis of total lane change time of the platoon
0.07 m after the entire platoon completes the lane change. In Fig. 14, it
can be seen that due to impedance by the SV, which does not decelerate, Fig. 17 presents the total lane change time of the platoon in each test,
the following vehicles stay in the middle lane and follow the accelera­ which shows that as the platoon cruising speed increases, it takes less
tion of the LV. When safe lane change conditions are met, the following time for the entire platoon to finish lane change. Comparing the total

Fig. 15. Performances statistics during active accelerating lane changing.

11
Y. Ma et al. Accident Analysis and Prevention 184 (2023) 106999

Fig. 16. ERI of the following vehicles with different thresholds of GTTC.

and safe lane change control method can help the following vehicles
elude the non-connected surrounding vehicle, which occupies the target
lane, and complete lane change under the mixed traffic condition.
Meanwhile, the platoon’s string stability and lateral error control are not
degraded during platoon lane changing. The proposed method is suit­
able for platoon lane-changing in the mixed traffic environment.
Although the proposed method aimed to improve the safety of pla­
toon lane changing in the mixed traffic environment, whether the
method is applicable to other vehicles and how to enable the method to
increase the efficiency of platoon lane change are worth further
studying.
The proposed method has the potential to be used as an auxiliary and
backup safety system for the CAV. Note that CAVs mainly obtain the
status data of the cooperative vehicles through V2V communication in
the coordinate operation. When CAV loses connection or suffers from
Fig. 17. Statistics of the total lane change time for the entire platoon in cyberattacks, it cannot obtain correct surrounding information to make
each test. proper path planning and vehicle control. The proposed sensor system
can scan around the CAV and provide location data of the detected
lane change time of the platoon passive waiting scenario and the platoon targets. The data can be fused with V2V communication information,
active accelerating scenario, the total lane change time of the platoon is providing more precise environmental sensing information to the CAV
almost the same when the initial cruising speeds and the speed differ­ path planner. When the V2V communication system fails or is unreli­
ence between the LV and SV remain unchanged. The speed difference able, the sensor system acts as a backup safety system and provides
between the LV and SV is a decisive factor in determining the time for location data of the obstacles to CAV. Based on the sensor data, the FSM-
the platoon to avoid the SV and complete lane change safely. Due to based decision-maker can plan collision-free motion. Therefore, the
impedance by the SV, the total time for platoon lane change is 2–4 times proposed method can help CAV reduce the collision risk even if the V2V
more than that of the unimpeded lane change scenario. The longer the communication system fails.
following vehicles side-by-side with the surrounding vehicles, the higher The proposed method prevents the following vehicle from following
the risk of accidents. Therefore, even if equipped with the platoon the predecessor blindly when the obstacle presents in the target lane
collision-avoidance lane change system, the driver of the LV should in­ during platoon lane changing, which improves the safety of the platoon.
crease the speed difference with the surrounding vehicles as soon as Nevertheless, it cannot provide some speed suggestions to the driver of
possible under the condition that the target lane is occupied by the non- the leading vehicle or the surrounding vehicle to complete platoon lane
connected vehicle. change efficiently. The simulation results show that the speed difference
is a critical factor in lane change time. If the platoon and the surrounding
5. Discussion and future work vehicle keep the same speed, completing the platoon lane change is
impossible. Whether there are more efficient methods for the platoon to
The simulation results demonstrate that the proposed sensor system elude the non-connected vehicles and complete lane change in a mixed

12
Y. Ma et al. Accident Analysis and Prevention 184 (2023) 106999

traffic environment is indeed worth further investigation. For example, will be enhanced and the state switch conditions will be adjusted. These
using the visual or auditory techniques to broadcast the platoon lane enhancements will further improve the platoon lane change safety.
change intention and speed suggestions in advance to the non-connected
vehicles in the target lane. Moreover, there are several existing models CRediT authorship contribution statement
for lane-changing trajectories (Ziegler and Stiller, 2009, Zeng et al.,
2019, Li et al., 2020b). It is worth investigating and comparing the Yitao Ma: Methodology, Data curation, Formal analysis, Software,
performance and safety of the proposed method and the lane-changing Writing – original draft. Qiang Liu: Conceptualization, Writing – review
trajectory planning and following method in the same scenario in the & editing, Supervision, Project administration, Funding acquisition. Jie
future. Fu: Investigation, Writing – review & editing. Kangmin Liufu:
Conceptualization, Writing – review & editing. Qing Li: Writing – re­
6. Conclusion view & editing.

This study proposed a collision-avoidance lane change control Declaration of Competing Interest
method for the human-lead-platoon to improve the platoon lane change
safety in the mixed traffic environment. Within its limitation, it can draw The authors declare that they have no known competing financial
the following conclusion: interests or personal relationships that could have appeared to influence
the work reported in this paper.
(1) The proposed collision-avoidance lane change control method
consists of an enhanced platoon vehicle sensor system, a platoon Data availability
vehicle controller based on the V2V communication and a safe
lane change decision-maker based on the Finite State Machine No data was used for the research described in the article.
(FSM). The simulation results demonstrate that the proposed
method can eliminate the side-impact accidents between the
Acknowledgments
connected vehicle platoon and the non-connected surrounding
vehicles in the course of platoon lane changing.
This work was supported by the National Natural Science Foundation
(2) The enhanced platoon vehicle sensor system has the advantages
of China (51675540); Fourth Xijiang Innovation and Entrepreneurship
of low-cost and low data processing complexity. The vehicle
Team and Leader Project of Zhaoqing, China; and Guangdong Basic and
sensor system effectively detects and monitors the non-connected
Applied Basic Research Foundation, China (2022A1515010692,
surrounding vehicles in the target lane during platoon lane
2020A1515110160).
changing. The design of the presented sensor system has the po­
tential to be used as an auxiliary and backup safety system for
References
CAV.
(3) Using the enhanced platoon vehicle sensor system and the safe Adanu, E.K., Lidbe, A., Tedla, E., Jones, S., 2021. Injury-severity analysis of lane change
lane change decision-maker, all following vehicles have a 100 % crashes involving commercial motor vehicles on interstate highways. J. Saf. Res. 76,
success rate of collision avoidance with the non-connected sur­ 30–35.
Alam, A., Gattami, A., Johansson, K.H., Tomlin, C.J., 2014. Guaranteeing safety for
rounding vehicle in the platoon lane change space occupied heavy duty vehicle platooning: Safe set computations and experimental evaluations.
scenarios. This study suggests that in the mixed traffic environ­ Control Eng. Pract. 24, 33–41.
ment, it is necessary to add the sensor in the lateral direction of Axelsson, J., 2017. Safety in vehicle platooning: A systematic literature review. IEEE
Trans. Intell. Transp. Syst. 18 (5), 1033–1045.
the large platoon vehicle to prevent accidents caused by the Balador, A., Bazzi, A., Hernandez-Jayo, U., De La Iglesia, I., Ahmadvand, H., 2022.
platoon vehicle following the predecessor blindly when the A survey on vehicular communication for cooperative truck platooning application.
obstacle presents in the target lane during platoon lane changing. Veh. Commun. 35, 100460.
Bian, Y.G., Zheng, Y., Ren, W., Li, S., Wang, J.Q., Li, K.Q., 2019. Reducing time headway
(4) The proposed collision-avoidance lane change control method for platooning of connected vehicles via v2v communication. Transport. Res. Part C-
can help the platoon change lanes safely and maintain a low Emerg. Technol. 102, 87–105.
collision risk with the surrounding vehicles. Exposure-to-Risk Chen, K., Pei, X., Okuda, H., Zhu, M., Guo, X., Guo, K., Suzuki, T., 2020a. A hierarchical
hybrid system of integrated longitudinal and lateral control for intelligent vehicles.
Index (ERI) of the following vehicles in the platoon active
ISA Trans. 106, 200–212.
accelerating scenarios is lower than that of the platoon passive Chen, T., Shi, X., Wong, Y.D., 2019. Key feature selection and risk prediction for lane-
waiting scenarios under the same speed condition. When the SV changing behaviors based on vehicles’ trajectory data. Accid. Anal. Prev. 129,
occupies the platoon’s target lane-changing space, the leading 156–169.
Chen, T., Shi, X., Wong, Y.D., Yu, X., 2020b. Predicting lane-changing risk level based on
vehicle actively accelerating and leading the platoon to overtake vehicles’ space-series features: A pre-emptive learning approach. Transport. Res.
the SV is a safer maneuver with lower collision risk, compared to Part C: Emerg. Technol. 116, 102646.
passive waiting for the SV to decelerate. Chu, T.S., Kalabic, U., 2019. Model-based deep reinforcement learning for cacc in mixed-
autonomy vehicle platoons. 2019 IEEE 58th Conference on Decision and Control
(5) Under the control of the proposed platoon vehicle controller, the (Cdc), 4079–4084.
platoon’s string stability and lateral stability are not degraded Cui, Q., Ding, R., Wei, C., Zhou, B., 2021. A hierarchical framework of emergency
during platoon lane changing. However, impeded by the non- collision avoidance amid surrounding vehicles in highway driving. Control Eng.
Pract. 109, 104751.
connected surrounding vehicle, the total time for platoon lane Cui, L., Hu, J., Park, B.B., Bujanovic, P., 2018. Development of a simulation platform for
change is 2–4 times more than that of the baseline scenario safety impact analysis considering vehicle dynamics, sensor errors, and
(unimpeded lane change scenario). It is found that the time of the communication latencies: Assessing cooperative adaptive cruise control under cyber
attack. Transport. Res. Part C-Emerg. Technol. 97, 1–22.
total platoon lane change is sensitive to the platoon cruising Ding, H., Pan, H., Bai, H., Zheng, X., Chen, J., Zhang, W., 2022. Driving strategy of
speed and the speed difference between the platoon and the connected and autonomous vehicles based on multiple preceding vehicles state
surrounding vehicle. estimation in mixed vehicular traffic. Physica A 596, 127154.
Falsone, A., Sakcak, B., Prandini, M., 2020. Coordinated lane change in autonomous
driving: A computationally aware solution. IFAC-PapersOnLine 53 (2),
The proposed method is tested in scenarios where only one non- 15211–15216.
connected surrounding vehicle is in the target lane. Future studies will Firoozi, R., Zhang, X., Borrelli, F., 2021. Formation and reconfiguration of tight multi-
construct more complicated scenarios involving more surrounding ve­ lane platoons. Control Eng. Pract. 108, 104714.
Gao, L., Chu, D., Cao, Y., Lu, L., Wu, C., Year. Multi-lane convoy control for autonomous
hicles in the original and target lane to verify and optimize the method. vehicles based on distributed graph and potential field. In: Proceedings of the 2019
Based on the simulation results, the decision-maker state architecture IEEE Intelligent Transportation Systems Conference (ITSC), pp. 2463–2469.

13
Y. Ma et al. Accident Analysis and Prevention 184 (2023) 106999

Hamad, K., Alozi, A.R., 2022. Shared vs. Dedicated lanes for automated vehicle vehicles? In: Proceedings of the 2017 IEEE Intelligent Vehicles Symposium (IV), pp.
deployment: A simulation-based assessment. Int. J. Transp. Sci. Technol. 812-818.
Hang, J., Yan, X., Li, X., Duan, K., Yang, J., Xue, Q., 2022. An improved automated Rahman, M.S., Abdel-Aty, M., 2018. Longitudinal safety evaluation of connected
braking system for rear-end collisions: A study based on a driving simulator vehicles’ platooning on expressways. Accid. Anal. Prev. 117, 381–391.
experiment. J. Saf. Res. 80, 416–427. Razmi Rad, S., Farah, H., Taale, H., Van Arem, B., Hoogendoorn, S.P., 2020. Design and
He, S., Zeng, J., Zhang, B., Sreenath, K., 2021. Rule-based safety-critical control design operation of dedicated lanes for connected and automated vehicles on motorways: A
using control barrier functions with application to autonomous lane change. conceptual framework and research agenda. Transport. Res. Part C: Emerg. Technol.
Hyun, K., Mitra, S.K., Jeong, K., Tok, A., 2021. Understanding the effects of vehicle 117, 102664.
platoons on crash type and severity. Accid. Anal. Prev. 149, 105858. Razmi Rad, S., Farah, H., Taale, H., Van Arem, B., Hoogendoorn, S.P., 2021. The impact
Khalifa, A., Kermorgant, O., Dominguez, S., Martinet, P., 2021. Platooning of car-like of a dedicated lane for connected and automated vehicles on the behaviour of drivers
vehicles in urban environments: An observer-based approach considering actuator of manual vehicles. Transport. Res. F: Traffic Psychol. Behav. 82, 141–153.
dynamics and time delays. IEEE Trans. Intell. Transp. Syst. 22 (9), 5684–5696. Schmidt, K.W., 2017. Cooperative adaptive cruise control for vehicle following during
Khattak, Z.H., Smith, B.L., Fontaine, M.D., 2021. Impact of cyberattacks on safety and lane changes. IFAC-PapersOnLine 50 (1), 12582–12587.
stability of connected and automated vehicle platoons under lane changes. Accid. Skottke, E.-M., Debus, G., Wang, L., Huestegge, L., 2014. Carryover effects of highly
Anal. Prev. 150, 105861. automated convoy driving on subsequent manual driving performance. Hum. Factors
Khattak, Z.H., Smith, B.L., Fontaine, M.D., Ma, J., Khattak, A.J., 2022. Active lane 56 (7), 1272–1283.
management and control using connected and automated vehicles in a mixed traffic Song, G., Lim, A., Bevly, D., 2016. An empirical study of dsrc v2v performance in truck
environment. Transport. Res. Part C: Emerg. Technol. 139, 103648. platooning scenarios. Digital Commun. Networks 2 (4), 233–244.
Larburu, M., Urquiza, A., Sanchez, J., 2012. Safe road trains for the environment (sartre): Sugimachi, T., Fukao, T., Suzuki, Y., Kawashima, H., 2013. Development of autonomous
Validation of sartre platoon service and the sartre hmi. In: Proceedings of the Its platooning system for heavy-duty trucks. IFAC Proc. Volumes 46 (21), 52–57.
World Congress. Sun, X., Yin, Y., 2021. Decentralized game-theoretical approaches for behaviorally-stable
Lee, Y., Ahn, T., Lee, C., Kim, S., Park, K., 2020. A novel path planning algorithm for and efficient vehicle platooning. Transp. Res. B Methodol. 153, 45–69.
truck platooning using v2v communication. Sensors. Tsugawa, S., 2013. An overview on an automated truck platoon within the energy its
Lefeber, E., Ploeg, J., Nijmeijer, H., 2020. Cooperative adaptive cruise control of project. IFAC Proceed. Volumes 46 (21), 41–46.
heterogeneous vehicle platoons. IFAC-PapersOnLine 53 (2), 15217–15222. Virdi, N., Grzybowska, H., Waller, S.T., Dixit, V., 2019. A safety assessment of mixed
Li, L., Gan, J., Zhou, K., Qu, X., Ran, B., 2020b. A novel lane-changing model of fleets with connected and autonomous vehicles using the surrogate safety assessment
connected and automated vehicles: Using the safety potential field theory. Physica A module. Accid. Anal. Prev. 131, 95–111.
559, 125039. Wang, X., Fu, M., Ma, H., Yang, Y., 2015. Lateral control of autonomous vehicles based
Li, Y., Tu, Y., Fan, Q., Dong, C., Wang, W., 2018. Influence of cyber-attacks on on fuzzy logic. Control Eng. Pract. 34, 1–17.
longitudinal safety of connected and automated vehicles. Accid. Anal. Prev. 121, Wang, S., Ma, Y., Liu, J., Yu, B., Zhu, F., 2022a. Readiness of as-built horizontal curved
148–156. roads for lidar-based automated vehicles: A virtual simulation analysis. Accid. Anal.
Li, J., Yao, L., Xu, X., Cheng, B., Ren, J., 2020a. Deep reinforcement learning for Prev. 174, 106762.
pedestrian collision avoidance and human-machine cooperative driving. Inf. Sci. Wang, Y., Wang, L., Guo, J., Papamichail, I., Papageorgiou, M., Wang, F.-Y., Bertini, R.,
532, 110–124. Hua, W., Yang, Q., 2022b. Ego-efficient lane changes of connected and automated
Li, K., Zhang, R., Wang, H., Yu, F., 2021. Multi-intelligent connected vehicle longitudinal vehicles with impacts on traffic flow. Transport. Res. Part C: Emerg. Technol. 138,
collision avoidance control and exhaust emission evaluation based on parallel 103478.
theory. Process Saf. Environ. Prot. 150, 259–268. Wei, S., Zou, Y., Zhang, X., Zhang, T., Li, X., 2019. An integrated longitudinal and lateral
Li, D., Zhu, F., Chen, T., Wong, Y.D., Zhu, C., Wu, J., 2023. Coor-plt: A hierarchical vehicle following control system with radar and vehicle-to-vehicle communication.
control model for coordinating adaptive platoons of connected and autonomous IEEE Trans. Veh. Technol. 68 (2), 1116–1127.
vehicles at signal-free intersections based on deep reinforcement learning. Transport. Woo, S., Skabardonis, A., 2021. Flow-aware platoon formation of connected automated
Res. Part C: Emerg. Technol. 146, 103933. vehicles in a mixed traffic with human-driven vehicles. Transport. Res. Part C:
Liu, R., Ren, Y., Yu, H., Li, Z., Jiang, H., 2022. Connected and automated vehicle platoon Emerg. Technol. 133, 103442.
maintenance under communication failures. Veh. Commun. 35, 100467. Xiao, G., Lee, J., Jiang, Q., Huang, H., Abdel-Aty, M., Wang, L., 2021. Safety
Liu, Y., Zhou, A., Wang, Y., Peeta, S., 2021. Proactive longitudinal control to manage improvements by intelligent connected vehicle technologies: A meta-analysis
disruptive lane changes of human-driven vehicles in mixed-flow traffic. IFAC- considering market penetration rates. Accid. Anal. Prev. 159, 106234.
PapersOnLine 54 (2), 321–326. Xu, J., Zhang, J., Zhang, R., Shen, T., 2021. String stability guaranteed lane change
Liufu, K., Liu, Q., Lu, Y., Chen, Z., Zhang, Z., Li, Q., 2021. Multiobjective optimization on maneuver for automated vehicles with vehicle-to-vehicle communication. IFAC-
cooperative control of autonomous emergency steering and occupant restraint PapersOnLine 54 (10), 330–335.
system for enhancing occupant safety. Accid. Anal. Prev. 159, 106302. Yang, X., Zou, Y., Chen, L., 2022. Operation analysis of freeway mixed traffic flow based
Lu, X.Y., Shladover, S.E., 2014. Automated truck platoon control and field test. Road on catch-up coordination platoon. Accid. Anal. Prev. 175, 106780.
Vehicle Automat. 247–261. Yu, G., Li, H., Wang, Y., Chen, P., Zhou, B., 2022. A review on cooperative perception
Ma, X., Tu, Z., Chen, F., 2022. A reliability-based approach to evaluate the lateral safety and control supported infrastructure-vehicle system. Green Energy Intelligent
of truck platoon under extreme weather conditions. Accid. Anal. Prev. 174, 106775. Transport. 1 (3), 100023.
Mahdinia, I., Mohammadnazar, A., Arvin, R., Khattak, A.J., 2021. Integration of Zeng, D., Yu, Z., Xiong, L., Zhao, J., Zhang, P., Li, Z., Fu, Z., Yao, J., Zhou, Y., Year. A
automated vehicles in mixed traffic: Evaluating changes in performance of following novel robust lane change trajectory planning method for autonomous vehicle. In:
human-driven vehicles. Accid. Anal. Prev. 152, 106006. Proceedings of the 2019 IEEE Intelligent Vehicles Symposium (IV), pp. 486-493.
Martínez-Díaz, M., Al-Haddad, C., Soriguera, F., Antoniou, C., 2021. Platooning of Zhang, H., Du, L., Shen, J., 2022a. Hybrid mpc system for platoon based cooperative lane
connected automated vehicles on freeways: A bird’s eye view. Transp. Res. Procedia change control using machine learning aided distributed optimization. Transp. Res.
58, 479–486. B Methodol. 159, 104–142.
Meyer, M.-A., Granrath, C., Feyerl, G., Richenhagen, J., Kaths, J., Andert, J., 2021. Zhang, Y.M., Wu, Z.Z., Zhang, Y., Shang, Z.Y., Wang, P., Zou, Q.Q., Zhang, X.H., Hu, J.,
Closed-loop platoon simulation with cooperative intelligent transportation systems 2022b. Human-lead-platooning cooperative adaptive cruise control. IEEE
based on vehicle-to-x communication. Simul. Model. Pract. Theory 106, 102173. Transactions On Intelligent Transportation Systems.
Milanés, V., Shladover, S.E., 2014. Modeling cooperative and autonomous adaptive Zhang, J., Lee, J., Abdel-Aty, M., Zheng, O., Xiao, G., 2023. Enhanced index of risk
cruise control dynamic responses using experimental data. Transport. Res. Part C: assessment of lane change on expressway weaving segments: A case study of an
Emerg. Technol. 48, 285–300. expressway in china. Accid. Anal. Prev. 180, 106909.
Ntcas, 2018. Test procedures for automatic driving function of intelligent connected Zhao, H., Sun, D., Zhao, M., Pu, Q., Tang, C., 2022. Combined longitudinal and lateral
vehicles of china. control for heterogeneous nodes in mixed vehicle platoon under v2i communication.
Park, B.S., Yoo, S.J., 2021. Connectivity-maintaining obstacle avoidance approach for IEEE Trans. Intell. Transp. Syst. 23 (7), 6751–6765.
leader-follower formation tracking of uncertain multiple nonholonomic mobile Zheng, J., Ma, L., Zhang, W., 2022. Promotion of cooperative lane changes by use of
robots. Expert Syst. Appl. 171, 114589. emotional vehicle-to-vehicle communication. Appl. Ergon. 102, 103742.
Peng, T., Su, L., Zhang, R., Guan, Z., Zhao, H., Qiu, Z., Zong, C., Xu, H., 2020. A new safe Zhu, L., Tang, Y., Yang, D., 2021. Cellular automata-based modeling and simulation of
lane-change trajectory model and collision avoidance control method for automatic the mixed traffic flow of vehicle platoon and normal vehicles. Physica A 584,
driving vehicles. Expert Syst. Appl. 141, 112953. 126368.
Polack, P., Altché, F., D’andréa-Novel, B., Fortelle, A.D.L., Year. The kinematic bicycle Ziegler, J., Stiller, C., Year. Spatiotemporal state lattices for fast trajectory planning in
model: A consistent model for planning feasible trajectories for autonomous dynamic on-road driving scenarios. In: Proceedings of the 2009 IEEE/RSJ
International Conference on Intelligent Robots and Systems, pp. 1879–1884.

14