没有合适的资源?快使用搜索试试~ 我知道了~
基于隐马尔可夫模型和模型预测控制的驾驶员跟车行为建模:一种网络物理系统方法
需积分: 0 6 下载量 194 浏览量
2021-05-19
06:17:05
上传
评论
收藏 283KB PDF 举报
温馨提示
试读
6页
本文的目的是基于隐马尔可夫模型开发一种新的驾驶员追随行为的运动视野优化建模方法,可以有效地模拟驾驶员的追随过程和驾驶特性。 首先,对驾驶员的驾驶行为与马尔可夫随机过程之间的关系进行了分析,证明了驾驶员期望的驾驶行为具有马尔可夫性质。 然后,提出了具有移动视野优化特征的建模框架,包括预览和感知模块,预测模块,优化模块。 在此框架下,以纵向加速度为隐藏状态,以时空为输出状态,给出了驾驶员跟车行为的隐马尔可夫模型。 为了获得纵向加速度命令,通过使后验概率最大化来使用优化算法。 最后,基于NGSIM数据集,通过Baum-Welch算法识别隐马尔可夫模型的参数,并从某些典型驱动器的闭环响应中讨论了所提出的基于HMM的建模方法的有效性和准确性。
资源推荐
资源详情
资源评论
Modeling Driver’s Car-Following Behavior Based on Hidden Markov
Model and Model Predictive Control: A Cyber-Physical System
Approach
Ting Qu
1
, Shuyou Yu
2
, Zhuqing Shi
1
and
∗
Hong Chen
1,2
1. State Key Laboratory of Automotive Simulation and Control, Jilin University (Campus NanLing)
Renmin Str. 5988, Changchun 130025, China
2. Department of Control Science and Engineering, Jilin University (Campus NanLing)
Renmin Str. 5988, Changchun 130025, China
Emails: quting@jlu.edu.cn; shuyou@jlu.edu.cn; shizq16@mails.jlu.edu.cn;
∗
Corresponding author: Email: chenh@jlu.edu.cn
Abstract— The goal of this paper is to develop a novel
moving horizon optimization modeling method of driver’s car-
following behavior based on hidden Markov model, which could
effectively mimic driver’s car-following process and driving
characteristic. First, the analysis of relation between the driver’s
driving behavior and Markov random process is proposed,
and the result of driver’s desired driving behavior has the
Markov property is proven. Then, a modeling framework with
moving horizon optimization characteristic is presented, includ-
ing the preview and perception module, prediction module,
optimization module. In this framework, the hidden Markov
model of driver’s car-following behavior is given by taking the
longitudinal acceleration as the hidden state and time headway
as the output state. To obtain the longitudinal acceleration
command, the optimization algorithm is used by maximizing
the posterior probability. Finally, based on the NGSIM data
set, the parameters of hidden Markov model are identified by
the Baum-Welch algorithm, and the effectiveness and accuracy
of proposed HMM-based modeling method are also discussed
from the closed-loop responses of certain typical drivers.
I. INTRODUCTION
Cyber-Physical Systems are multi-dimensional complex
feedback systems, which expand the interaction between the
cyber and physical worlds through the communication, com-
puting and control technologies, and possibly with humans
in the loop
[1]
. Intelligent transportation systems (ITS) as the
special Cyber-Physical Systems have earned the sustained at-
tention and interests in driver-oriented intelligent vehicles
[2]
,
which are motivated by their potentials for enhancements
of driving safety, comfortable and efficiency. Understanding
and modeling of human driving behavior under the complex
driving scenarios became important issues to be studied, and
received continuous interest in recent years
[3,4]
.
The car-following characteristic is one of the main be-
haviors of human drivers in vehicle manipulation. The de-
velopment of accurate driver’s car-following behavior model
could effectively understand the driver’s driving process and
mimic the driver’s control action. Many longitudinal driver
models have been presented in the previous researches. In
[5], a “follow the leader model” was addressed, and the aim
of the driver was to maintain a following distance from the
leading vehicle. Based on the assumption that the driver’s
desired braking and acceleration rates were constrained, in
[6], a new switching vehicle speed model was constructed for
the following vehicle, and the driver’s characteristics were
represented by the parameters of model. In [7], to evalu-
ate the performance of adaptive cruise control systems, an
accurate longitudinal human driving model was developed,
and six driver models were evaluated based on two selected
databases. The research results illustrate that the Gipps
model
[6]
was the most promising one, and a modified version
of the model was also suggested and implemented in a
microscopic traffic simulator, whose behavior was consistent
with the macroscopic traffic one very well.
In the actual driving process, human drivers do not behave
deterministically, some stochastic uncertainties were also
exhibited
[8]
. This property attracts many researchers’ atten-
tions. In [9], An errorable car-following driver model was
presented and modeled as a random process. To analyze and
mimic the driver’s driving behavior, the stochastic process
method was applied, and the Road-Departure Crash-Warning
System Field Operational Test data has been used to identify
the model parameters and validate its effectiveness. From the
results of a car-following experiment, it is shown that in [10],
the driver’s driving behavior can be represented by a simple
scheme: the acceleration a(t) is held approximately constant
for a certain time interval, followed by a jump to a new
acceleration. It also illustrates that the driver’s behavior has
deterministic and stochastic components
[11]
.
In the process of driver’s car-following, usually the vehicle
movements can be observed and recorded, while the driver’s
behavior within vehicles can not be obtained directly. To
derive unobservable driver behaviors, many modeling meth-
ods based on hidden Markov model (HMM) were developed.
In [12], a HMM-based driver behavior model with layered
structure was proposed. Meanwhile, the identification and e-
valuation algorithms of bad driving behavior were presented.
In [13], HMM was also used to mimic the driver’s various
behaviors when driving at the crossing. To predict and
simulate the driver’s driving characteristics, the Intersection
2017 11th Asian Control Conference (ASCC)
Gold Coast Convention Centre, Australia
December 17-20, 2017
978-1-5090-1572-6/17/$31.00 ©2017 IEEE 114
资源评论
weixin_38716423
- 粉丝: 4
- 资源: 929
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
最新资源
- HCIP-Datacom笔记 (1).pdf
- yolov5,SSD 可能使用到的一些代码
- bbbbbbbbbbbbbbbbbb
- 安卓逆向学习笔记之Frida Stalker 还原OLLVM AES.docx
- 安卓逆向学习笔记之unicorn来trace还原OLLVM Base64.docx
- 最新版本私钥助记词碰撞器大富豪使用python进行制作通过接口的方式进行验证支持多币种多链多网络一分钟万次验证高出货率
- 介绍离散性制造行业的MES系统流程
- Arduino IDE压缩包版本,2024年4月26日,最新版本
- 基于IDEA-CCNL/Randeng-Pegasus-238M-Summary-Chines微调的中文文本摘要任务源码+数据集
- 自动驾驶-状态估计和定位之直方图滤波(Histogram+Filter)定位应用和源码.pdf
资源上传下载、课程学习等过程中有任何疑问或建议,欢迎提出宝贵意见哦~我们会及时处理!
点击此处反馈
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功