随着城市化的进展和汽车的普及，交通运输问题日益严重，道路拥挤，交通事故频发，交通环境不断恶化。在这种背景下，运用各种高新技术系统的解决交通问题的智能交通系统(Intelligent Transport System)也就应运而生。其中，高级汽车安全控制系统和智能车辆的研究作为ITS的重要领域之一，吸引了大量学者。 本课题的研究目标是设计出能够在各种不同的交通环境下控制车辆自动运行的智能代理系统。这种代理系统能够实现驾驶任务中决策层和执行层的各种功能，可以接收驾驶员或者导航层代理的命令，根据周围的交通环境做出实时的驾驶决策，并根据驾驶决策做出各种驾驶操作。这种系统对于AHS的研究有着重要的意义，同时也可以用于人工交通系统中交通子系统建模的研究。 为了实现上述研究目标，本文首先研究了汽车动力学模型，提出一种包含重力计算模型、空气阻力计算模型、悬挂系统模型、滚动阻力模型、轮胎模型、车身平移模型和旋转模型在内的汽车动力学模型的设计框架。并基于LuGre轮胎/摩擦力模型提出了一种新的汽车动力学模型。该模型可以采用模块化模型和集总式模型两种类型，可以用于各种汽车控制问题的研究和汽车仿真程序的设计。 接下来，研究了各种汽车纵向和横向控制问题，对横向控制问题中的车道保持控制器的设计进行了综述，并提出一种基于模糊逻辑的车道保持控制器。 在对汽车的各种控制系统有了直观的认识之后，研究了代理和多代理系统的基本理论，学习了面向代理的软件工程方法学，掌握了设计基于多代理的智能驾驶系统所必需的代理相关的知识。 接下来，基于多代理系统理论设计了智能驾驶代理控制系统，提出驾驶任务的四层结构代理模型：导航层代理、协调层代理、决策层代理和执行层代理，同时对每一层代理的功能和实现结构进行了详细的描述。为了测试智能驾驶代理系统，我们利用交通仿真器，在仿真系统中实现了各种交通对象和智能驾驶代理系统，并给出了各种对象和代理系统的UML实现类图以及相互作用的关系图。通过几个场景的实验发现，该代理系统能够完成决策和驾驶车辆的任务。
With the development of urbanization and generalization of automobiles, more considerations have been put on the serious traffic problems, such as road congestion, air pollution, vehicle accidents and decreasing driving safety. To address these issues, research has been conducted under the framework of Intelligent Transportation Systems (ITS). Advanced Vehicle Control Safety Systems and Intelligent Vehicles, as two important research fields of ITS, have attract a lot of researchers. The research objective of this thesis is to design a multi-agent based intelligent driving system, which can control an autonomous vehicle in different environments. With implementing all kinds of functionalities of both the tactical level and the operational level of the driving task, this system can makes its decisions based on the received input from its sensors and its instructions and translates to control operations that are sent to the vehicle it controls. The implementation of this system is of considerable significance to the research on Automated Highway System (AHS) and can also be used in the research of Artificial Intelligent System (AIS). To reach the goal, the project first has a research on the dynamic modeling of vehicles and presents a design architecture of a dynamic vehicle model, which includes the modeling of gravitation, air dynamics, suspension system, rolling resistance, tire/ground friction, chassis translation and chassis rotational dynamics. Using this design architecture we present a novel dynamic vehicle model based on the LuGre tire/road friction model. This model can be used in the design of advanced vehicle control systems and simulation of dynamic vehicle modeling. In the following part, we do some research on issues of the longitudinal and lateral control of vehicles and have a overview on the design of lane keeping controller. A lane following controller based on fuzzy logic is then described and the simulation result is also presented. The theory of agent and multi-agent system is a basis for the design of multi-agent based intelligent driving system, and the agent oriented software engineering methodology is also a necessity. So we do some research on these contents in the following part. After we have the necessary knowledge of vehicle control and multi-agent system, we begin the design of the multi-agent based intelligent driving system. We first bring forward a four-level agent model of the driving task, including navigation agent, coordination agent, tactical agents and operational agent. The detailed functionality and the implantation architecture of all these agents are then presented. We must have a traffic simulator to verify the intelligent driving agents system, and the design of the traffic simulator is our last work.