1 Adaptive Leading Cruise Control in Mixed Trafﬁc Considering Human Behavioral Diversity

2025-04-28 0 0 2.59MB 10 页 10玖币

侵权投诉

Adaptive Leading Cruise Control in Mixed Trafﬁc

Considering Human Behavioral Diversity

Qun Wang, Haoxuan Dong, Fei Ju, Weichao Zhuang, Chen Lv, Liangmo Wang, and Ziyou Song

Abstract—This paper presents an adaptive leading cruise

control strategy for the connected and automated vehicle (CAV)

and ﬁrst considers its impact on the following human-driven

vehicle (HDV) with diverse driving characteristics in the uniﬁed

optimization framework for improved holistic energy efﬁciency.

The car-following behaviors of HDV are statistically calibrated

using the Next Generation Simulation dataset. In a typical single-

lane car-following scenario where CAVs and HDVs share the

road, the longitudinal speed control of CAVs can substantially

reduce the energy consumption of the following HDV by avoiding

unnecessary acceleration and braking. Moreover, apart from the

objectives including car-following safety and trafﬁc efﬁciency,

the energy efﬁciencies of both CAV and HDV are incorporated

into the reward function of reinforcement learning. The speciﬁc

driving pattern of the following HDV is learned in real-time

from historical speed information to predict its acceleration and

power consumption in the optimization horizon. A comprehensive

simulation is conducted to statistically verify the positive impacts

of CAV on the holistic energy efﬁciency of the mixed trafﬁc ﬂow

with uncertain and diverse human driving behaviors. Simulation

results indicate that the holistic energy efﬁciency is improved by

4.38 % on average.

I. INTRODUCTION

THE transportation sector, which consumes 25% of global

energy resources, is one of the main sources of green-

house gas emissions and air pollution [1]. Extensive efforts

have been made to improve vehicle efﬁciency and lower

emissions of on-road vehicles in response to the increasingly

stringent emission standards [2]–[4]. As a crucial technology

in saving energy consumed by vehicles, eco-driving has been

extensively discussed [5]–[9], with the core idea of adjusting

vehicle speed and maintaining an energy-efﬁcient driving

style [10]. More recently, the development of connectivity

and automation technologies has provided another promising

opportunity to further cut down energy consumption through

the deployment of connected and automated vehicles (CAVs).

Qun Wang is with the Department of Mechanical Engineering, National

University of Singapore, Singapore 117575, Singapore, and also with the

School of Mechanical Engineering, Nanjing University of Science and Tech-

nology, Nanjing 210094, China (e-mail: wangqun@u.nus.edu).

Haoxuan Dong and Weichao Zhuang are with the School of Mechanical

Engineering, Southeast University, Nanjing 211189 , China (e-mail: dong-

haox@foxmail.com; wezhuang@seu.edu.cn).

Fei Ju and Liangmo Wang are with the School of Mechanical Engineering,

Nanjing University of Science and Technology, Nanjing 210094, China (e-

mail:jufei@njust.edu.cn; liangmo@njust.edu.cn).

Chen Lv is with the School of Mechanical and Aerospace Engi-

neering, Nanyang Technological University, Singapore 639798 (e-mail:

lyuchen@ntu.edu.sg).

Ziyou Song is with the Department of Mechanical Engineering, Na-

tional University of Singapore, Singapore 117575, Singapore (e-mail:

ziyou@nus.edu.sg).

Corresponding authors: Weichao Zhuang; Ziyou Song.

With the assistance of vehicle-to-anything (V2X) communica-

tions [11] and sensor fusion techniques [12], CAVs can take

advantage of the rich information to optimize their operations,

such as vehicle acceleration [13], motor torque regulation [14],

path planning [15], etc.

Despite some promising results indicating that CAVs can

save energy, the impacts of connectivity and autonomy on

the trafﬁc efﬁciency and energy performance of neighboring

vehicles have not been extensively studied, while this type

of study can provide insights for policymakers and incentives

to further promote CAVs. For example, Joshua et al. [16]

analyzed the mobility and energetic impacts introduced by

CAVs’ deployment. Results demonstrate that the trafﬁc ﬂow

is improved with the increasing travel demand and decreasing

travel time. Fakhrmoosavi et al. [17] explored the inﬂuences

of a mixed trafﬁc ﬂeet on several aspects from a network

level, indicating that CAVs can enhance trafﬁc safety, mobility,

and emission reduction of the trafﬁc system. Zhao et al. [18]

assessed the impacts of CAVs under eight different testing

scenarios with a travel demand model and simulation results

indicate that the travel demand in Austin, Texas can increase

by at least 20%.

Moreover, it is anticipated that CAVs and human-driven ve-

hicles (HDVs) will co-exist on the same road in the near future

[19], [20]. Human drivers will still remain to be the majority

who take charge of vehicle operations for a long period. Hence,

it is imperative to develop eco-driving strategies for CAVs in

the mixed trafﬁc ﬂow in which CAVs frequently interact with

HDVs. Lu et al. [21] proposed an energy-efﬁcient adaptive

cruise control model for electric CAVs in a mixed trafﬁc ﬂow.

Simulations are performed in a mixed single-lane trafﬁc ﬂow

with different market penetration rates of CAVs, indicating

that the proposed method exhibits a superior performance

in energy saving compared to other existing adaptive cruise

control and cooperative adaptive cruise control methods. Zhu

et al. [22] designed a novel model predictive control (MPC)

method to enhance energy efﬁciency and keep driving safety

for the CAV in a mixed trafﬁc ﬂow. An integrated data-driven

model of car-following is used in the MPC framework to

predict the behaviors of HDVs. Simulation results validate its

effectiveness in energy efﬁciency improvement and robustness.

Li et al. [23] developed a cooperative controller for CAVs

in a mixed trafﬁc platoon based on multi-agent reinforcement

learning. Compared with MPC, the proposed strategy performs

better in dampening trafﬁc oscillations and reducing energy

consumption. Ma et al. [24] investigated the energy-saving

potentials of the following human-driven platoon enabled by

eco-driving control of CAVs ahead. Especially, the inﬂuences

arXiv:2210.02147v1 [eess.SY] 5 Oct 2022

of diverse characteristics of human behaviors are evaluated

through extensive numerical analyses, which statistically show

a positive inﬂuence of the proposed strategy on the subsequent

platoon.

However, all the aforementioned studies only focus on the

optimization of the CAV, while neglecting its impact on the

following HDV. To achieve a higher holistic energy efﬁciency

of the mixed trafﬁc ﬂow, this study ﬁrst incorporates the HDV

energy consumption in the optimization framework. In most

existing studies, car-following models (e.g., optimal velocity

model [25] and intelligent driver model (IDM) [26]) are

usually utilized to describe the behaviors of HDVs. In [27], ﬁve

representative microscopic car-following models were used

to calibrate the behaviors of drivers in Shanghai, and the

IDM outperformed the other models from the perspectives

of accuracy and stability. The parameters of these models

indicating different driving styles are either set as constant

[28], or assumed to follow some predeﬁned uniform distri-

bution [24]. However, in a dynamic trafﬁc environment, the

behaviors of HDVs are quite stochastic and do not follow

deterministic patterns. It remains challenging to accurately

predict the behaviors of HDVs, which are the necessary

previews of most predictive control schemes.

More recently, model-free reinforcement learning (RL) al-

gorithms has been widely applied in many areas such as

autonomous driving [29], [30], battery management [31], and

eco-driving for electriﬁed vehicles [32], [33]. One of the main

advantages of model-free RL is that the agent can interact

with the stochastic environment and try to maximize the ac-

cumulated reward in a learning manner. Instead of attempting

to model the complicated environment with high stochasticity

(e.g., uncertain human driving behaviors in this study), model-

free methods directly improves system performance based on

the explored samples [34].

Motivated by the discussion above, this study aims to

design an adaptive leading cruise control strategy to reduce the

holistic energy consumption of both CAV and HDV by consid-

ering the diverse human driving behaviors in a reinforcement

learning framework. The contributions and novelties of this

study are summarized as follows:

1) In addition to car-following safety and trafﬁc efﬁciency,

the dynamics of both CAV and the following HDV are

considered in the optimization framework for improved

holistic energy efﬁciency.

2) HDVs are calibrated into a joint distribution using the

IDM based on the ﬁeld-collected Next Generation Simu-

lation (NGSIM) dataset to cover a wide range of stochas-

tic and realistic driving behaviors.

3) The inﬂuences of diverse driving behaviors on the im-

provement of energy efﬁciency using the proposed control

algorithm are quantitatively analyzed.

The rest of this paper is organized as follows. Section

II presents the problem formulation including scenario de-

scription, vehicle dynamics, energy consumption model, the

intelligent driver model as well as control objectives of this

paper. In Section III, the stochastic behaviors of human drivers

are developed and the detailed design process of reinforcement

learning is given. Simulation results and performance analysis

Preceding Vehicle

(PV)

Connected and Automated

Vehicle (CAV)

Human-driven

Vehicle (HDV)

Traffi c flow direction

Fig. 1. Schematic of the 3-vehicle ﬂow consisting of PV, CAV, and HDV.

TABLE I

PARAMETERS OF VEHICLE LONGITUDINAL DYNAMIC MODEL

Model Parameter Value

Vehicle mass m1005 kg

Aerodynamic drag coefﬁcient CD0.3

Vehicle frontal area Af2.02 m2

Rolling resistance coefﬁcient f0.015

Air density ρ1.206 kg/m3

Gravitational acceleration g9.81 m/s2

Wheel radius r0.28 m

Rotational inertia coefﬁcient δ1.02

are provided in Section IV. Conclusions of this paper are

presented in Section V.

II. PROBLEM FORMULATION

A. Scenario Description

Similar to existing studies in [24] and [35], a common

scenario of mixed trafﬁc stream is investigated in this study,

where there is a human-driven preceding vehicle (PV), a CAV,

and a following HDV, as shown in Fig. 1. Assume the CAV

can obtain velocity and gap distance information of both the

PV and the following HDV through onboard sensors (e.g.,

millimeter-wave radars). There may exist more vehicles (CAVs

or HDVs) before the PV, which means this scenario is just a

fraction of a long mixed trafﬁc ﬂow, while the trafﬁc before

the PV is not the focus of this study.

B. Vehicle Longitudinal Dynamics

Since this study emphasizes energy-efﬁcient driving in a

car-following scenario, only vehicle longitudinal dynamics is

considered here, as described in Eq. (1).

˙v=Fd−mgf cos α−mg sin α−0.5CDAfρv2

δm (1)

where vis the vehicle velocity. Fddenotes the driving force of

the vehicle. mis vehicle mass and gdenotes the gravitational

acceleration. CD,Af, and ρare aerodynamic drag coefﬁcient,

vehicle frontal area, and air density, respectively. δis the

rotational inertia coefﬁcient. For simplicity, road slope is not

considered here (α= 0). Table I lists the critical parameters

of the longitudinal dynamics model of the vehicle.

C. Energy Consumption Model

Generally, an approximated and differentiable energy con-

sumption model in a polynomial expression is sufﬁcient to

develop the eco-driving algorithm. According to the experi-

mental data in [32], the demand power of the motor Pmot

can be written as a nonlinear function of the vehicle speed

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

6howmuchtimeremainsbeforecollision,istreatedasthemetricforsafety,asgivenby:TTC(t)=d01(t)v01(t)(10)where0istheindexforprecedingvehicleand1istheindexforCAV;d01denotestheirgapdistanceandv01representstheirrelativespeed(v01=v0v1).AsmallerTTCvalueisassociatedwithhighercrashrisk,andviceversa[39].Thelo...

展开>> 收起<<

1 Adaptive Leading Cruise Control in Mixed Trafﬁc Considering Human Behavioral Diversity.pdf

共10页,预览2页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

1 Adaptive Leading Cruise Control in Mixed Trafﬁc Considering Human Behavioral Diversity

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: