【作者】 马晓凤；

【导师】 严新平；

【作者基本信息】 武汉理工大学 ， 计算机应用技术， 2008， 硕士

【摘要】 现有研究中,Multi-Agent思想多用于宏观角度以解决交通诱导等交通问题。文中从微观角度建立了面向驾驶员特性的多智能体交通流模型框架。该模型实现了车辆Agent的行为建模,改进了以往模型驾驶员特性单一的缺陷;设定了车辆Agent间的车道位置协商、换道冲突协商等策略;借鉴了元胞自动机的高速演化策略,便于数据分析。在此框架模型的基础上,重点研究了以下三个非常态交通场景的交通流演化规律。在每个场景中,针对不同的交通环境因素,细化了车辆Agent的驾驶员特性函数、决策模型,使其反映特定的交通行为。并提出了交通参数的设置方案,可作为交通拥堵疏散、交通管理的参考。(1)直路段上的车辆违规过街模型。将驾驶员的不同特性引入车辆Agent,给出了视线遮挡情景下的决策模型。结果表明:穿越时垂直行驶方向的事故率较高,而追尾事故多发生在流量较大的条件下;穿越概率较小时,对流量的影响并不明显,但是事故率随穿越概率增大而增加,在穿越概率大于0.015时考虑设定交通设施;另外,研究了不同驾驶员比例对事故数的影响,当反应迟缓的驾驶员比例较高时,事故率相对较高,流量明显下降。(2)大型活动对周边道路交通流的影响研究。建立了停车场、管制道路和正常道路三个子模型,研究了大型活动散场后,私家车对周围交通流的影响。并且以沌口体育馆的调查数据作为对比,模拟了在停车场、管制道路和正常道路上的车辆行驶,反映了驾驶员偏好等特性。模型可以反映出交通流拥堵、交通流高峰期传播的规律,可用该模型对大型活动后的交通流扩散进行预测,为后续大型活动疏散模型的建立提供参考。(3)高速公路车道维修对车流拥堵的影响。分别对正常道路、行车道维修和超车道维修建立了决策模型。着重分析了车辆类型、驾驶员特性和道路维修标志位置对交通拥堵的影响。结果表明:驾驶员特性在低密度下对拥堵的影响比较明显,但整体上影响不大。主要影响因素为车辆类型和提示标志的放置位置。若行车道维修,应该避免在车流密度0.1～0.2、大型车辆比例大约为0.4时进行维修:若超车道维修,应该避免在车流密度0.25、大型车比例较小时进行维修;维修标志宜放置于维修点前7.5 m左右。更多还原

【Abstract】 Multi-Agent system has been used to solve the problem of traffic guidance from macroscopy. The traffic flow model based on Multi-Agent system and driver attributes is set up in this paper, in which the car-agent model is the most importance part. This study focuses on some significant parameters: driver attributes and car-type, that make this model distinct with other studies. Besides, the negotiations between car agents reflect the car-agents’ motives and beliefs including the parking negotiation, lane negotiation and changing-lane negotiation. Based on this frame model, three traffic scenes are simulated in the following chapters in which the driver attributes functions and the decision-making models are instantiated according to the traffic environment. These scenes have typical character of abnormal traffic, so these simulations are benefit for the study of traffic collision. Three scenes are as follows:(1) Cars’ Crossing Willfully to Traffic Flow based on Multi-agent System.Multi-agent system is used to describe the behavior of cars’ crossing willfully when they are not on the crossroad. Moreover, the driver attributes are considered as important parameters to each agent, and some models are proposed. To get the effect of the cars’ crossing willfully, we set simulations that depict cars cross road with some probability. According to the results, there are some conclusions as follows: first, car tracing cauda is more frequent when the flux is larger. Second, when the probability of crossing is small, it rarely influences flux. Once the probability of crossing exceed 0.015, it is necessary to set some traffic signal to warning the crossing drivers. Third, as far as driver attributes, the number of accident is increased with the ratio of slowness driver increasing; besides, the flux decreases.(2) A Multi-Agent Model for Evacuation System under Large-scale Events.The model of Chapter 2 is instantiated and used to simulate the large-scale activity. Car’s variability is introduced by superimposing agents, with different driver’s activity agenda, motives and beliefs. In the large-scale events, there are three sub-models which based on the multi-agent model: the park model, the evacuation- road model and the normal-road model. After above three phases, traffic flow time space dissipation models under large-scale activity are analyzed, moreover, investigated data of traffic flow under large-scale activity dissipation are used to test the accuracy and applied example of these models, such as the flux of some investigation site, the evacuation time of audience. The results show that these models have small errors about forecast and can reflect the changing of traffic flow after large-scale events, which might be helpful to traffic management scheme.(3) The effects of road maintenance in the freeway system.The normal model, lane maintenance model and overtaking lane maintenance model are proposed. After simulating, we analyze how the driver attributes, car types and the position of maintenance signal influence on the traffic flow and collision. The results show that: the driver attributes have relationship with traffic collision under low flux, but it has little influence on the form of collision as a whole. Besides, the effect of car types and the position of signal to collision are obvious. It had better not maintain the lane if the flux is between 0.1 and 0.2, and the big-vehicle proportion is 0.4. When it comes to overtaking lane, it had better not maintain when the flux is 0.25 and the big-vehicle proportion is low. Moreover, the maintenance signal should have a reasonable position according to the flux.Based on three models’ simulations and analyses, some advice and traffic parameters are proposed, and they might helpful to the traffic management and traffic evacuation.更多还原