【作者】 柴毅；

【导师】 黄席樾；

【作者基本信息】 重庆大学 ， 控制理论与控制工程， 2001， 博士

【摘要】 本文针对汽车行驶中的安全问题，以汽车主动安全系统为研究对象，采用信息处理技术和智能化技术对信息的表示、行驶安全的动态模型、系统的构成和建模、信息的实时获取、行驶防撞等进行了较系统的研究。 从汽车和安全的角度，概括叙述了汽车及其安全性的有关问题，讨论了智能汽车系统(ITS)以及汽车防撞安全系统，对汽车主动安全系统的研究内容进行了的分析。 针对人-车-道路环境构成的闭环系统，驾驶过程中道路-环境状态的感知这个首要因素，本文研究了驾驶过程状态信息的表达方法，建立了汽车-环境状态特征模型，描述了汽车在行驶过程中时变动态的驾驶状态。 以汽车驾驶中行驶速度和与车距这两个重要安全因素为基础，分析了汽车行驶中感知、判断决策、操作和汽车响应与速度的时间关系。提出了以安全行驶为核心的安全行驶动态模型P~2OT，并从时间量上，提出了基于汽车安全行驶模型P~2OT的时间表达方法，并将此结论应用于系统分析与设计的时间约束，以及汽车安全行驶防撞专家系统的安全(时间)模型。 本文提出了基于安全行驶行为的汽车主动安全系统的原理结构。利用多智能体计算行为的并发性和异构性，提出了基于多智能体的汽车主动安全系统，通过相互协同完成从动态数据的实时获取到专家系统推理决策报警的全过程。 针对环境信息获取的实时性和并发性，提出了一种支持计时实时动态流程的计时受控有色petri网(TC~2PN)，建立汽车行驶复杂动态环境的多智能体系统流程的协作模型，描述实时分布式多智能体系统中的多任务协调过程，以及基于时间安全模型的多智能体系统协调求解机制。 本文利用计算机视觉理论和技术，讨论了一种道路和目标(车辆)的视觉图像检测和提取的方法。对已检测出的前方目标，给出了在结构化道路上进行相对径向距离的计算方法。 本文研究了汽车行驶的时间安全模型和汽车行驶规则描述，提出了“目标-规则基-特征状态体”的领域知识表示模型，给出一种汽车安全报警防撞专家系统，采用主-从两级推理机制完成对安全的推理求解策略，实现安全报警。 最后，为了使驾驶行为得到有效的监督，规范安全驾驶，本文提出了一种利用KDD技术发现驾驶行为和习惯的方法，从汽车行驶数据中发现驾驶规律和习惯，从而对驾驶的安全性进行审计，提高驾驶员安全驾驶的意识。更多还原

【Abstract】 Aiming at the safety problem of vehicle driving, this thesis focuses on the following topics in a vehicle initiative safety system using information processing technologies and smart agent technologies: the representation of information, the dynamic model of safety during steering, the components and models, the information acquisition in real-time and avoiding collision in steering.At the aspect of vehicle and safety, the vehicles and its safety related problems are explored briefly. The intelligent traffic system (ITS) and the safety system to preventing collision are discussed also. In addition, the research content of vehicle initiative safety system is analyzed.In the closed system composed of human, vehicles and road surrounding, the perceiving of the state of road surrounding is the first key factor to assure safety when steering. This paper studies the representation approaches of state information during driving and established vehicle-surrounding state feature model to illustrate the variable and dynamic state of running vehicles.Based on velocity and interval between vehicles during driving which are the two most important factors related to safety, the time relationship between surrounding perceiving, decision making, operation, vehicle response and driving speed is analyzed. A dynamic 20T is presented, beyond this, the time representation method of this model under the perspective of time is presented, which is also employed, in system analysis and design, as time restriction, and the safety model of expert system of collision preventing.This paper presented the theoretical structure of vehicle initiative safety system based on steering safety behavior. Using the concurrence and isomerism characteristic of multi-agent computation, the initiative safety system based on multi-agent is stated, which can perform the process from dynamic data acquiring in real-time to decision making of expert system by the mutual cooperation between agents.In allusion to the real-time and concurrence of acquiring surrounding information, a timing controlled colored petri-net supporting timing dynamic process is presented and a cooperative model of multi-agent system of the complex dynamic surrounding of steering is established too. Describing process of multi task coordination of real-time distributed multi-agent system and mechanism of multi-agent coordinating problem solving based on time-safety model.- II -ABSTRACTBased on computer vision theory and technology, a detection and retrieval approach of road and target barrier images is discussed and to the detected target, a method computing the relative radial distance in structured road is presented.Studies the time-safety model of steering vehicle and the representation of steering rules, this thesis presents a domain knowledge representation of target-rule radix-feature state body. Besides these, a vehicle safety expert system consisting of master-slave reasoning mechanism is designed which can alarm while encountering emergency.Finally in order to supervise the driving behavior effectively and regulate safety steering, this paper presents a knowledge discovering method to discover the driver’s steering habits and regular pattern from the driving record using KDD, which can examine steering safety and improve the driver’ safety consciousness greatly.更多还原