郑玲钰
随着我国经济的快速发展和城市人口的急剧膨胀，大城市机动车保有量迅速攀升。有限的城市道路系统的供给能力越发难以满足不断增长的交通需求。如何整治城市交通拥堵问题已经成为中国，乃至全球各个国家所面临的重大课题。随着城市的发展，可用于建设交通基础设施的空间日益减小，单纯通过扩大路网规模已经无法缓解交通拥堵问题。鉴于此，高效利用已有的交通基础设施，提高其运行效率，便成了缓解交通拥堵问题的主要手段，而城市交通信号控制系统即为缓解交通拥堵问题的重要手段之一。
通过对已有的交通控制系统及相关理论研究深入探讨，城市道路交通控制根据其控制思想可分为两类：一类是根据实时的交通流量进行调整和优化的“被动响应式控制”；一类是根据检测数据得到的交通数据，预测所控制交叉口的交通需求，并基于此优化生成控制方案的“半主动式控制”。基于此，本文借鉴主动交通管理的概念，归纳了“主动式控制”的概念，并认为“主动式控制”应从道路使用者的实际交通需求预测和道路使用者在不同控制策略影响下的反应出发，在维持畅通运行的基础上，进一步避免拥堵、事故等现象发生。为实现“主动式控制”策略的设计及实施，本文从以下四方面进行了深入的研究：
（1） 基于交通控制理论的研究现状，梳理城市道路主动式控制理论体系，阐述城市道路主动式控制的内涵和外延，提出控制策略的逻辑框架，并对其实施所需的基础理论和关键技术进行分析和探讨。
（2） 构建基于用户感知的道路交通服务水平评价体系，基于服务水平的概念深入分析道路使用者对道路状况、交通条件、道路环境等方面的感知情况，并借鉴美国HCM2010和上海市的划分标准，分别构建基于灰类白化权函数的城市道路路段和交叉口的服务水平评价模型，以期更好的描述道路使用者理解的道路交通状态。其中，路段的服务水平评价模型运用路段平均行程速度与自由流车速之比，路段流量与通行能力之比和速度标准差三个参数作为评价依据，交叉口的服务水平评价模型运用交叉口的流量与通行能力之比和每车信控延误两个参数作为评价依据。
（3） 提出城市道路控制优先级判断方法。分别从微观和宏观两个角度对城市道路短时交通流量进行预测：在微观层面上从路段角度出发，通过路段的交通流量对未来的交通流量变化情况进行预测；在宏观层面上从变化的OD需求出发，兼顾道路的供给情况及道路使用者的需求情况，运用动态交通分配模型对研究范围内的路段流量进行分配。根据微观和宏观两个角度的预测结果，结合基于用户感知的道路交通服务水平评价体系对城市道路单点信号控制的优先级进行判断。
（4） 从主动式控制策略的主要任务和目标分析入手，提出了总体的主动式控制策略的总体优化逻辑。在此基础上，针对MFD控制小区和边界控制小区两个不同小区的控制目标进行了基于强化学习的主动式控制策略的设计和控制效果验证。
本文通过对城市道路交通控制策略的演化发展的挖掘，基于系统科学理念，构筑包括基于用户感知的道路交通服务水平评价体系、城市道路控制优先级判断方法和基于强化学习的主动式控制策略的城市道路主动式控制理论体系，丰富和完善了城市道路控制策略的理论与方法。限于数据、现场实验等条件和主动式控制策略本身及其实施的复杂性，研究尚有诸多不足和缺陷，特别是未能对策略效果进行现场实验检验和分析，这是后续研究的重要内容。同时，城市道路主动式控制系统平台的开发也是未来的重要研究方向。
关键词：主动式控制，动态交通分配，服务水平评价，流量预测，强化学习
ABSTRACT
With the rapid development of China's economy and expansion of urban population, the number of motor vehicles in big cities is rising rapidly. The limited supply capacity of urban road system is increasingly difficult to meet the growing traffic demand. How to rectify urban traffic congestion has become a major issue faced with China and even all countries around the world. With the development of cities, the space that can be used to construct traffic infrastructure is becoming smaller and smaller. It is impossible to solve traffic congestion simply by expanding the scale of road network. In view of this, the efficient use of existing transport infrastructure, improve its operational efficiency, has become the main means to solve the problem of traffic congestion, and urban traffic signal control system is one of the important means to alleviate traffic congestion.
Through the thorough discussion of the existing traffic control system and related theoretical research, urban road traffic control can be divided into two categories according to its control idea: one is the "passive control" which adjusts and optimizes the real-time traffic flow; the other is the "passive control" which predicts the traffic demand of the controlled intersection based on the traffic data obtained from the detection data, and generates the control scheme based on this optimization. Semi-active control. Based on this, this paper puts forward the concept of "pro-active control", and holds that "pro-active control" should proceed from the actual traffic demand prediction of road users and the response of road users under the influence of different control strategies, and on the basis of maintaining smooth operation, further avoid congestion, accidents and other phenomena. In order to realize the design and implementation of the "pro-active control" strategy, this paper makes a thorough study from the following four aspects:
(1) Construct the theoretical system of urban road pro-active control, elaborate the connotation and extension of urban road pro-active control, put forward the logical framework of control strategy, and analyze and discuss the basic theory and key technology needed for its implementation.
(2) Construct the evaluation system of road traffic service level based on user perception. Based on the concept of service level, the perception of road users on road condition, traffic condition and road environment is analyzed in depth. Referring to the dividing standards of HCM2010 and Shanghai, the ratio of average travel speed to free flow speed, traffic volume and traffic volume are used respectively. Two evaluation models of urban road service level based on Grey whitening weight function are constructed to better describe the road traffic state understood by road users.
(3) A method for judging the priority of single-point control of urban roads is proposed. The short-term traffic flow of urban roads is forecasted from the micro and macro perspectives respectively: from the micro level, the future traffic flow changes are forecasted through the traffic flow of the sections; from the macro level, from the changing OD demand, considering both the supply of roads and the demand of road users, the dynamic traffic assignment model is used. Distribution of section flow within the scope of study. According to the prediction results from both micro and macro perspectives, the priority of urban road single-point signal control is judged by combining the evaluation system of road traffic service level based on user perception.
(4) Starting from the analysis of the main tasks and objectives of the pro-active control strategy, the overall optimization logic of the overall pro-active control strategy is proposed. On this basis, the pro-active control strategy based on reinforcement learning is designed and validated for two different control objectives of MFD control cell and boundary control cell.
Through excavating the evolution and development of urban road traffic control strategy, based on the concept of system science, this paper constructs the evaluation system of road traffic service level based on user perception, the priority judgment method of single-point control of urban road and the theory system of pro-active control of urban road based on reinforcement learning, which enriches and improves the control strategy of urban road. Theories and methods. Limited to the conditions of data, field experiments and the complexity of pro-active control strategy itself and its implementation, there are still many shortcomings and shortcomings in the study, especially the failure to test and analyze the effect of the strategy, which is an important part of the follow-up study. At the same time, the development of urban road pro-active control system platform is also an important research direction in the future.
Key Words: pro-active traffic control, dynamic traffic assignment, level of service, traffic flow prediction, reinforcement learning
