节点文献
高速汽车弯道前方碰撞预警算法的研究
Study of Forward Collision Warning Algorithm for Curved Highway
【作者】 张立存;
【导师】 管欣;
【作者基本信息】 吉林大学 , 车辆工程, 2004, 硕士
【摘要】 汽车的出现无疑给人们的生活带来巨大的便利,但频繁的交通事故、惨重的人员伤亡和巨大的财产损失使汽车交通安全问题日益成为人们关注的焦点。与以安全带和安全气囊为主要设施的汽车被动安全性研究相比,汽车主动安全性的研究对解决汽车交通安全问题起着更大作用。汽车主动安全性研究不但是指汽车的操纵性能、制动性能等汽车本身性能品质的研究;也指借助智能控制技术、信息通信技术、计算机技术等建立智能辅助驾驶系统帮助驾驶员驾驶汽车,以提高汽车行驶的安全性。汽车碰撞预警系统更是智能辅助驾驶系统的研究热点之一。 汽车碰撞预警系统是一种驾驶员警示系统,它能感知道路交通环境中潜在的危及本车安全的信息并通过声光信号等给予驾驶员警示作用,从而弥补驾驶员感官能力的不足,帮助驾驶员避免可能产生的交通事故,提高行车的安全性。汽车碰撞预警系统是先进车辆控制与安全系统(Advanced Vehicle Controland Safety System,AVCSS)的重要研究方向之一。汽车碰撞预警系统一般包括道路交通环境感知、安全逻辑判断和驾驶员安全信息感知三个部分。其中道路交通环境信息感知部分包括本车运动状态参数如汽车速度、加速度信息的获得、前方行驶汽车运动参数如前车速度、加速度的获得、与前方其它车辆之间相对位置关系如相对速度、相对距离的获得。安全逻辑算法部分主要根据道路交通环境感知部分获得的信息对本车行驶的安全性进行判断,以便给出预警与否的信息并送给预警信息的驾驶员感知部分,预警信息的驾驶员感知部分则通 71<WP=77>吉林大学硕士学位论文过声光等信号警示驾驶员,这样驾驶员就可以通过预警信息提前做出判断,避免交通事故的发生。 汽车的行驶受到驾驶员的指令动作、汽车本身的响应和道路对驾驶员的影响三方面因素的限制。无论是对汽车本身性能品质的优化设计,还是对由“人—车—路” 所组成的闭环系统的研究,都离不开一个准确而实用的驾驶员模型。驾驶员模型的发展由来已久,自五十年代起,人们就开始对驾驶员的行为特点进行研究,并建立了多种驾驶员模型,总体上分为:方向控制模型、速度控制模型和方向与速度综合控制模型。上述模型中有单一对汽车方向或速度进行控制的模型。也有对驾驶员驾驶过程中,如何改变方向盘转角与油门踏板/制动踏板,实现方向和速度综合控制的模型,但其结构十分复杂,实时计算十分不方便。考虑以上各方面的原因管欣教授和郭孔辉院士提出了“驾驶员方向与速度综合控制模型的研究” 这一课题,并建立了驾驶员最优预瞄加速度模型。该模型详细分析了驾驶员操纵行为特性及驾驶员、汽车和道路三者之间的相互关系。在驾驶员操纵行为特性和汽车动力学响应特性分析的基础上,综合预瞄跟随理论、广义预测控制理论、系统模糊决策理论、非线性系统控制理论中的描述函数法和系统辨识理论,提出了驾驶员稳态预测动态校正假说,并以此建立了一个全面描述驾驶员如何控制汽车油门踏板、制动踏板和方向盘转角的驾驶员方向与速度综合控制模型,收到了良好的效果。 本文首先详细分析了驾驶员最优预瞄加速度决策模型,针对模型中存在的问题提出了改进的方法,重新建立了驾驶员最优预瞄加速度模型模糊决策环节的四个基本评价指标—驾驶安全性评价指标、驾驶合法性评价指标、驾驶轻便性评价指标和驾驶工效性评价指标的隶属函数。仿真表明,改进后的模型消除了平台效应。在对驾驶员最优预瞄加速度模型研究的基础上,综合考虑了驾驶员的行为特性和汽车的运动特性,并将之引入到高速汽车前方碰撞预警系统的研究中,开发了基于雷达获取前方道路交通环境信息的高速汽车弯道前方碰撞预警算法,不但适用于高速公路的直道预警,而且适用于高速公路弯道预警。通过全文的理论分析和数字仿真,总结有以下几个结论: 1、通常的高速汽车前方碰撞预警系统,都是由安装在汽车上的大量的传感器来获取道路交通环境信息,成本昂贵。也有单纯依靠雷达为获取前方道路72<WP=78>摘 要交通环境信息传感器的预警系统,但是这些系统很少综合考虑汽车的运动特性和驾驶员的行为特性,因而只能适用于高速公路直道的预警,对高速公路弯道无能为力。针对这种情况,本文在对驾驶员最优预瞄加速度模型充分研究的基础上,综合考虑了驾驶员的行为特性和汽车的运动特性,以雷达作为前方道路交通环境信息传感器,利用驾驶员最优预瞄加速度模型中“稳态预测”假说研制了高速汽车弯道前方碰撞预警算法,由于成本低而更易于实用化。2、传统的基于雷达的高速汽车前方碰撞预警算法由于无法分辨前方道路上不同汽车所在的车道,因而无法对高速公路的弯道进行预警。本文在综合考虑驾驶员行为特性的基础上,提出了横向与纵向综合预测算法预测汽车行驶的轨迹,并由此对汽车行驶的道路边界进行拟合计算。然后基于雷达信息提出了虚拟弯道检测算法,对雷达获取的前方不同车道的障碍物进行识别,得出位于本车道内会对本车构成危险的车辆的相对位置和相对速?
【Abstract】 It is doubtless that vehicle have brought us great convenience in thelife, but frequent traffic accident, miserable casualty and tremendousloss of property have made traffic safety to be a focus among peoples.Different from the passive safety technology such as safety belt, safetygasbag and so on, active safety technology has more important functionin improving driving safety. One aspect of vehicle positive safety isthe research of vehicle handling and stability performance, brakingperformance etc. Another aspect is to improve driving safety by usingdriving assistant system that is based on technology of intelligentcontrol technology, information communication technology and computertechnology. Vehicle collision warning system is a hotspot of drivingassistant research.The function of collision warning system is to give driver warningthrough sound or light signal by apperceive latency danger of trafficcircumstance, thus can make up the shortage of driver and improve thesafety of driving by avoiding traffic accident. Vehicle collisionwarning system is an important research direction of Advanced VehicleControl and Safety System. Vehicle collision warning system alwaysincluding three sections : Apperception of traffic circumstance,74<WP=80>Judgment of safety logical and driver apperception of warning signal.The first part includes the information obtaining of velocity andacceleration of frontal vehicle and body vehicle, the relative velocityand distance between the two vehicles. The function of the second partis to judge whether it is safety and provide information for the thirdpart. The third part is mainly to give signals to driver in order toavoid traffic accident.Vehicle driving is influenced by operation of driver, response ofvehicle and road circumstance. No matter what the design of driverperformance, the research of close-loop that make up ofdriver-vehicle-road, it is necessary to establish an applied drivermodel. Since 1950’s scientists had began to study thedriver’s behavior property and established several drivermodels. As a whole they are divided to direct control model,velocity control model and integrated control model. Amongthe above models narrated whether it is direct control model,velocity control model or integrated model, they always havea complicated structure. Considering the reasons above Prof.Guan Hsin and Academician of Chinese Academy of Engineering-K.H.Guo brought forward researching of integrated controlbehavior model and established driver optimal previewacceleration model. Driver characters and the interrelationamong driver, vehicle and road were studied in this model.Integrated preview-follow theory, description functionmethod in non-linear system control theory and systemidentification theory, the static preview and dynamiccorrection assumption was put forward, based on theoriesabove driver optimal preview acceleration model wasestablished which on the whole described driver’s behaviorproperty of how to control acceleration pedal, braking pedal 75<WP=81>and steering wheel, and expecting result was got.Firstly driver optimal preview acceleration model was studiedin this paper; improving method according to the bug of themodel was put forward. Four new evaluation indexes in thefuzzy decision section for drive safety, drive handiness,driver legality and drive efficiency were established. Basedon the model, considering driver behavior and characters ofvehicle, we established forward collision warning algorithmfor curved highway based on radar getting forward roadcircumstance. According to the whole theory and simulationof this paper, we have several conclusions below:1、The traditional forward collision warning system usually dependson plenty of sensors fixed on automobile to get road circumstance ,sohas a high cost. Some warning system based on radar can not be used tocurved highway for not c
【Key words】 Driver Model; Warning of Curved Highway; Check of Virtual Curved Road; Flat Effect;
- 【网络出版投稿人】 吉林大学 【网络出版年期】2004年 04期
- 【分类号】U462
- 【被引频次】58
- 【下载频次】2155