英文文献单片机
Abstract . Traffic in urban areas is generally regulated by traffic lights, which
represent a traffic flow bottleneck if they are not effectively configured. The dynamic nature of traffic flows makes reliable prediction possible only over a limited time horizon, and hence necessitates continual re-computation solutions. As intersections network size increases, the exponential growth in joint signal timing and traffic states exhibit a computational barrier to determining effective traffic lights timing. So, real-time traffic signal control optimization becomes a challenging problem. The present work investigates the issue of adaptive traffic control using real-time traffic data in multiple intersections. Each network intersection is controlled by an
autonomous agent operating with a view of incoming traffic. The generated model presents ability to identifying dynamic changes in traffic flow conditions as well as adjusting green lights sequences and offering a good coordination between each
intersections neighboring. The main contribution of this model is a real-time adaptive control of the traffic lights, according to discontinuity and heterogeneity of traffic flow. Simulation results confirm the system effectiveness compared to fixed time control, actuated time control and green wave method. And that by mitigating the congestion in terms of maximum throughput traffic flow, minimum waiting time and stops while maintaining fairness among all the network traffic lights. Keywords: Traffic lights control, multiple intersections, adaptive, self-organizing, coordination
1. Introduction
Traffic congestion has become a crucial problem especially in urban locations. This is due to the incapacity of roads to satisfy the expectations of an exponential vehicles number. The consequences are significant costs in terms of man hours lost, excessive fuel consumption, and also an adverse environmental effect through carbon emissions. Thus, congestion reduces the quality of life for many people and represents a major handicap for the road transport system. Nevertheless, it remains the displacement means most used of citizens.
Traffic lights are the key element of traffic control. The combination of several
constraints makes harder to design an effective traffic light control strategy to achieve an optimal real time schedule.
This is especially the case of multiple intersections scenarios since traffic conditions from neighboring intersection have a dynamic impact on the local traffic lights.
Indeed, these flows should become part of the local traffic flow after a certain period. Therefore, the coordination between adjacent intersections should also be taken into account.
Existing methods for traffic lights management can be grouped into three categories: fixed time control,actuated control and adaptive control.
Fixed time control is the most commonly implemented control strategy in the real word. The timing signal is previously calculated based on traffic conditions studied
which is provided by day of week and time of day. A large amount of historical traffic data needs to be collected, recorded and then used to derive some traffic patterns. This is based on different geographic location of intersections and different time periods (e.g.,peak hours).However, these systems cannot be effective because the traffic conditions may change dynamically and significantly from time to time throughout the day. It is only applicable in stable traffic conditions.
The other scheme, called a traffic actuated controller, works by adapting to the volume of vehicles on the road. It does not set a timer for the cycle of signals, but instead measures the volume of vehicles on the road at any given moment and then sets the frequency and interval of traffic lights accordingly. Actuated traffic signal provides a minimum length of green time to each approach during a cycle. This length may be incremented based on vehicle arrivals. The length of every green interval is also constrained by a maximum green time specification.
Compared to fixed time control, actuated control is based on online data which is collected via traffic sensors, as induction loops or infrared sensors that report measurements to a central entity. These traffic volumes data are transferred to the signal control system to decide. However, the variability and unpredictability of
traffic demand on arterial systems often outpace the ability to update signal timings so that signalized intersections operate efficiently.
In the other hand, adaptive traffic control covers a set of techniques which provide a systematic approach for automatic adjustment of traffic lights states in real time. The system aim is to achieve and to maintain a desired level of control system
performance when traffic parameters change in time.
A large variety of adaptive traffic tools have been developed. Thus, the conventional green wave mechanism has been widely used in many cities. It is a traffic control strategy which assigns a fixed offset to a sequence of traffic lights such that if vehicles travel at a certain velocity from specified junction to others, they will not stop at any red lights. The offset is the time needed for vehicles to travel from an
intersection to another one at an expected constant speed. The vehicles are grouped in platoons of varied sizes, determined by signal timings, which progress through the green wave at uniform speed.
Thus, effective adaptive models aim to coordinate signals of controlled intersections so as to improve performances and reduce congestion in urban networks areas.
Generally, adaptive traffic control systems are defined as the application of computing, information, and communications technologies to the real-time vehicles management.
The major researches are focused on Artificial Intelligence (AI) techniques. IA
provides to be a highly promising field for solving the traffic control problems. Indeed, fuzzy systems, artificial neural networks, multiagent systems, evolutionary computing and swarm intelligence are effective computing control tools in dealing with
complexity and dynamics of traffic situation. Implementation of these traffic lights controls is supposed respond to traffic network demand, adapt timing plans in time, and implement real-time control.
In this paper, an adaptive traffic lights control scheme in multiple intersections is proposed based on multi-agents framework. Multi-Agents Systems (MAS) provide an intelligent approach for addressing the realtime traffic signal control problem, given the distributed nature of traffic flow information. The data provided by the sensor detection equipment are assumed to be available to the proposed traffic control system.
First, prior researches in intelligent traffic signal control are reviewed in this paper. Adaptive traffic signal control problem is addressed. Our basic approach to modeling traffic flows and making traffic signal control decisions within each agent is described in Sections III. In Sections IV, experimental results that indicate the performance characteristics of the proposed approach are presented. Finally, in Section V, summarize different results of the present work and indicate guidelines for future research work.
2. Related works
The traffic congestion problem has motivated researchers to study new control strategies to efficiently manage the traffic movement in urban area. Artificial
intelligence techniques have been widely investigated urban traffic control in different areas. In this section, related works of control traffic lights and principally associated intelligent agent’s applications are presented.
2.1. Knowledge based traffic lights models
Initially, Fuzzy logic (FL) field is adopted to model expert’s knowledge by adjusting traffic signal control parameters such cycle length and phase sequence. FL signal controllers [1–5] use a set of rules to determine the preferred action based on a number of inputs. Although results have shown that these methods arecapable of effective control within small networks, the use of a static rule-base implies that further updates to the system may be required over time. Hence, it is difficult to generate an efficient model, especially for complex intersections.
Decision support systems (DSS) consists of computer technology solutions that can be used to support complex decision making and aims to assist human operator in selecting effective measures. Many DSS traffic control systems were developed such Hoogendoorn and Almejalli models [6, 7]. While it has been demonstrated that these
systems are capable of suggesting effective actions, they rely on large databases of historical traffic data and expert knowledge, which may be hard to acquire and maintain. In addition, the data-bases must be constantly updated with new traffic scenarios and control action.
2.2. Neural network traffic lights models
Artificial neural networks (ANNs) have been extensively explored as approaches for decision making. Thus, the authors in [8, 9] have successfully proposed ANN to train a traffic signal controller. ANN Models compute decisions by learning from
successfully solved examples. However, as the size of the road network increases, neural networks suffer significant performance degradation. Moreover, when traffic volumes change, they have to relearn an effective method of control.
2.3. Queue theory traffic lights models
The theory of queues is also a widely used approach when it comes to model the tails present at intersections [10, 11]. The queuing theory algorithm uses the law of “Little” in order to determine the lengths of queues. Using queues information„ the model is able to develop a dynamic plan for traffic lights in order to reduce the queue average length as well as the average waiting time. However, this theory requires a number of measuring points to manage the intersection and cannot deal with minor conflicts.
2.4. Agent based traffic lights models
Multiagent Systems (MAS) is an IA tool that provides principles for modeling of complex systems involving multiple agents with centralized and/or decentralized coordination mechanisms. There are many researches in traffic lights modeling using MAS. Thus, Dresner and Stone [12] have tackled the problem using a reservation system for collision avoidance at an intersection. In this system, vehicles make a request to a central agent. If the request is accepted, the vehicle needs to follow the prescribed path, to be safe while crossing the intersection. Preliminary
experimentations with this type of system are promising, with minimum delay of vehicles at intersections. However, it lacks the coordination aspect between multiple traffic lights, which is one of the most important problems that affects the design of such systems. Roozemond [13] presents agent model acting autonomously while
sharing gathered data. Then, intersections use this information to make both short and long term traffic predictions with accordingly adjustment. Bazzan [14] has used a decentralized approach combining MAS and evolutionary game theory. This approach models each intersection as an individually motivated agent which must focuses not only on getting vehicles through the intersection, but also on reducing travel times for all vehicles.
2.4.1. Traffic lights models using reinforcement learning
Several applications of reinforcement learning (RL) exist, such as in the works
[15–19]. These models attempt to infer a reward model for various pairs of
state/action by allowing interaction between the controlling agents and traffic
environment. In a large network, the problem of the large number of these pairs restricts the use of RL method in this domain, (convergence problem).
2.4.2. Traffic lights models using swarm optimization methods
The swarm optimization methods are another field of artificial intelligence which was applied to control the traffic lights. The works of Oliveira [20] use the principle of optimization distribution of signal based on the amount of pheromone emitted by waiting vehicles. In [21], authors propose autonomous intersection management
(AIM) based on Ant Colony System (ACS) algorithm that evacuates vehicles for each sequence of vehicles arrivals and for large number of lanes. In addition, numerous works has used evolutionary computing to optimize signal plans within a network
[22–24]. These approaches typically aim to represent the signal plans for the entire network as a set of chromosomes and then an evolutionary algorithm is applied. However, this type of solution requires an extremely high puissance of computing with the network size. In complex networks, real-time control may be impossible because the computation may not complete in a suitable time.
2.4.3. Adaptive traffic lights models
Adjustment of traffic lights configuration is the key aspect in optimization of traffic flow model presented in major previous works. However, in the real world, the
configuration parameters of the traffic lights are based on data and designer’s model which is updated frequently.
The multi-agents model can be used in order to deal with these real situations. But, to provide more accurate data, it must be more adaptive with change. The idea is to
endow the system with its own control system, so it can face its dynamic environment and the multiple situations it will encounter. Thereby, the self-organized system is a particularly appropriate approach to apply within traffic control.
Gershenson [25, 26] presented one of the initial applications of self-organized systems to traffic signal control. The author shows that an agent-based solution is capable of controlling traffic signals effectively. Lammer and Helbing extend this work [27] and present another self-organized control system which is shown to effectively handling several traffic scenarios within hypothetical grid-like networks.
The approach proposed by Xaio-Feng Xie [28], presents a model of aggregates incoming vehicles into clusters which enables the real-time computation of signal timing policies. These Self-scheduling approaches give rise to a self-organizing system.
Zhou and al. [29] provides an adaptive traffic lights algorithm which selects the sequence of phases in a cycle, according to the following criteria: the presence of priority vehicles, totalwaiting time and queue length. This algorithm, however, requires the same type of vehicle with same speed.
Thework presented in this paper is based on the principles of self-organized and
adaptive systems, including large-scale autonomous decentralized traffic management. Each intersection is controlled by an intelligent traffic lights agent, operating for real-time with view of incoming traffic. Independently, intersections traffic controls follow the same behavior, based only on the local traffic information.
Thus, the model provides a decentralized management between adjacent intersections based on detected traffic information and using a Wireless Sensor Network (WSN).
With a hierarchical architecture designed to be flexible and adaptable to
heterogeneous flows, the situation is re-evaluated at every phase based on selective traffic parameters.
3. Proposed model: Adaptive traffic signal control
In this paper, an adaptive traffic lights model in multiple intersections is proposed. Each traffic light adjusts its states depending on the nature of traffic flow and the degree of saturation of all connected lanes. Coordination between traffic lights allows the presence of a “green wave", progressing through consecutive intersections. This permits the greatest flexibility at intersections most in demand and so improves network throughput.
The model assumes two main steps, detection of primary traffic data then green light assignment decision. The first step detects and calculates traffic information in real-time. Green light assignment decision uses traffic information to determine the next traffic movementand duration required to serve it. Before describing the proposed model, the problem formulation is presented.
3.1. Problem formulation
In urban networks, an intersection is a shared area of a road network where vehicles with conflicting paths cross. Vehicles enter an intersection via entry and exit lanes. Each lane has a set of allowed turning movements. Each period of time that allows one group of turning movements is called a phase. All the phases of a traffic signal combined together constitute the cycle of the signal. Traffic lights work by cycling through their light phases, with green permitting vehicles to travel through the
intersection, yellow requiring them to slow down and prepare to stop, and red forcing vehicles to stop at the intersection. Traffic signals manage the intersection by alternating between different traffic flows of allowed turning movements.
Four directions are present (north, south, east, and west), and each of which has two lanes to go, one for going forward and the other for turning left (Fig. 1).
A wide range of sensors can be used as data detectors in the traffic networks including
inductive loops, cameras, magnetometers, wireless sensors, and radars. We define a sensor as a node that has a detection unit and generally limited capacities in terms of energy, memory, computation power and communication.
In most works, wireless sensors wireless sensors usage is motivated by various characteristics such: low manufacturing cost, small size, ease and rapid installation, reactivity, communication and memory units. Wireless sensors may be located on various parts of lane. Based on the results from the literature, the distance between sensors should be sufficient in order to have a correct sampling [30].
The proposed model involves three different placements of sensors at a given
intersection (Fig. 2.). Thus, the traffic on every incoming lane is monitored by these sensors nodes:
- The Exit Sensor node (ES) is located close to the traffic light, counting the vehicles departures from the intersection when the light is green;
- The Advanced Sensor node (AS) is placed at an appropriate distance before the light, counting vehicles arrivals continuously. It can be located at a fixed distance of the light, to consider as soon as possible the entrances on the intersection.
- The Movement Sensor node (MS) is located at the entry of the outgoing lanes. Such deployment better differentiates vehicles exiting the intersection and allows informing neighbor intersections about the coming movement flow. Movement Sensors nodes allow collecting and aggregating traffic of all nodes lanes emitting in its direction.
Submitted to traffic safety rules and for each intersection, there are twelve authorized movements shown in Fig. 3. Thereby, the decision step is transformed to a decision setting the duration time of the green light to each movement and the displaying sequence between them.
3.2. System architecture
The proposed multi-agent framework is described in Fig. 4. Three main components are included in this framework:
1. Sensor component: Collecting real-time traffic flow information;
2. Traffic control agent: It is in charge of autonomous control of traffic lights in intersection;
3. Traffic Lights: Device which defines a road intersection behavior by controlling when each traffic light becomes red or green and for how long.
Urban traffic depends strongly on rapid response of the control strategy of the traffic lights system in each intersection. This control strategy mechanism can be described as follows:
• First, the detection component (WSN sensor) is used to monitor the traffic volume and forward the traffic volume data to decision component (Agent);
• After collecting all of the required information, decision component (Agent) began to generate an appropriate traffic signal control strategy;
• Execution Component (traffic lights) is used to regulate traffic flow.
The system involves double objectives: The local objective is to improve traffic in each intersection. The global objective is to place the traffic signals on a coordinated system so that vehicles encounter green lights successively and therefore platoons of vehicles can proceed through a continuous series of green lights.
4.Conclusion
The problem of traffic congestion is a serious problem in urban life causing social disorders such as delays,economic losses and environmental pollution. In this work, an adaptive approach for traffic lights control in multiple intersections is proposed. Based on the local real-time traffic data and traffic condition of neighbor intersections, the sequences of green lights are self-adjusted.
The self-organized model contains two steps: primary real-traffic data detection and assignment decision of green light. First twelve permitted movements are defined. Based on the traffic flow features, assignment decision of green light uses computed movement rank to determine the next movement in each intersection most in need and sufficient duration required to serve it as well as coordinate between neighboring. So, the proposed architecture aims to combine green wave and local adaptation in a dynamic way in the
network. Indeed, in multiple intersections, adoption of the coordinated traffic control could reduce vehicle delays and stops.
It should be noted that the efficiency of the road network is not based solely on the effectiveness of
each traffic lights but also on their collaboration. Thus, an optimum traffic flow can be achieved through the road network by the proposed traffic lights coordination mechanism.
Traffic simulator (SUMO) is used to test and validate the system. Various simulations are conducted
to examine the performance against a three strategies: adaptive green wave method, actuated time control and fixed time control. The results confirmed that the model outperforms the previous approaches in terms of throughput, average waiting time and average number of stops.
The improvement in performance is attributed to the ability of the model to create green waves and to act locally. The simulations results encourage to extend the model in future work to introduce new elements, like public transport and pedestrian crossing which would be closer with the reality. 摘要。城市交通的交通灯通常是受交通信号灯的,如果没有有效的配置,就代表了交通流的瓶颈。交通流的动态特性使得可靠的预测可能只在有限的时间范围内,因此,必须不断重新计算解决方案。由于交叉口的网络规模的增加,在联合信号的时间和交通状态的指数增长表现出计算障碍,以确定有效的交通灯时间。因此,实时交通信号控制的优化成为一个具有挑战性的问题。本工作研究的问题,在多
个交叉口的实时交通数据的自适应流量控制。每个网络的交叉点是由一个独立的代理操作控制 有一个视图的传入流量。所产生的模型提出的能力,以确定在交通流条件下的动态变化,以及调整绿灯序列,并提供了一个很好的协调之间的每个交叉点。该模型的主要贡献是一个实时的交通灯的自适应控制,根据交通流的不连续性和异质性。仿真结果证实了系统的有效性相比,固定时间控制,驱动时间控制和绿波法。这是通过减轻拥塞的最大吞吐量流量,最小等待时间和停止,同时保持所有的网络交通灯之间的公平性。
关键词:交通灯控制,多个交叉口,自适应,自组织,协调
1、景区简介
交通拥堵已成为一个重要的问题,特别是在城市的位置。这是由于道路不能满足车辆数量的预期指数。后果是显着的成本,在工时损失,过量的燃料消耗,以及通过碳排放量的不利的环境影响。因此,拥堵降低了生活质量,许多人,代表一个主要 公路运输系统的障碍。尽管如此,它仍然是最常用的公民的位移。 交通灯是交通控制的关键因素。几个约束的组合使得难以设计一个有效的交通灯控制策略,以实现最佳的实时时间表。
这是特别的情况下,多个交叉口的情况下,由于交通条件从邻近的路口有一个动态的本地交通灯的影响。事实上,这些流量在一定时期内成为局部交通流的一部分。因此,相邻的交叉点之间的协调也应考虑在内。
现有的交通灯管理方法可分为三大类:固定时间控制、驱动控制和自适应控制。 定时控制是最常用的控制策略,在实际的话。的定时信号是先前计算的基础上研究的交通条件,这是一天一天的时间和时间。大量的历史交通数据需要收集,记录,然后用来推导出一些流量模式。这是基于不同的地理位置的交叉点和不同的时间段(例如,峰值时间),但是,这些系统不能是有效的,因为交通条件可能会动态变化,并显着的时间在整个白天。它只适用于稳定的交通条件。
另一种称为交通致动控制器的方案是在道路上的车辆数量进行调整的方法。它没有设置信号周期的计时器,而是在任何特定的时刻测量道路车辆的数量,并据此设置交通灯的频率和时间间隔。驱动的交通信号提供了一个最小长度的绿灯时间,在一个周期中的每一个方法。这个长度可以增加基于车辆到达。每一个绿灯间隔的长度也被一个最大的绿色时间规范约束。
相对于固定时间控制,感应控制是基于通过流量传感器收集的在线数据,作为感应回路或红外传感器,报告测量到一个中央实体。这些交通量数据被传送到信号控制系统来决定。然而,在动脉系统的变化和交通需求的不可预测性往往超过更新信号的能力,信号交叉口的有效运作。
另一方面,自适应交通控制涵盖了一套技术,提供了一个系统的方法,实时交通灯状态的自动调整。该系统的目标是实现和保持一个理想的控制系统的性能,当流量参数变化的时间。
已经开发了大量的自适应交通工具。因此,传统的绿波机制在许多城市得到了广泛的应用。它是一个分配一个固定偏移量的交通控制策略交通信号灯的序列,如果车辆行驶在一定的速度从指定的路口向他人,他们不会停止在任何红色的灯。该偏移量是车辆从一个交叉点到另一个在期望的恒定速度行驶所需的时间。车辆 分成大小不同的排,由信号决定的,它通过匀速绿色的波浪。
因此,有效的自适应模型的目的是协调信号的控制交叉点,以提高性能,减少拥堵的城市网络区。
一般而言,自适应交通控制系统被定义为应用程序的计算,信息和通信技术的实时车辆管理。
人工智能技术的研究主要集中在人工智能(人工智能)技术。为解决交通问题提供了一个非常有前途的领域。事实上,模糊系统、人工神经网络、多智能体系统、进化计算和群智能是处理交通情况的复杂性和动态有效的计算控制工具。实施这些交通灯控制应响应交通网络的需求,及时调整时间计划,并实施实时控制。 本文提出了一种基于多智能体的自适应交通灯控制方案。多智能体系统提供了一个智能的方法来解决实时交通信号控制问题,给出了分布式的流量信息的性质。所提供的传感器检测设备的数据被假定为提供给拟议的交通控制系统。
首先,本文综述了智能交通信号控制的研究进展。自适应交通信号控制问题得到解决。我们的基本方法来模拟交通流量和交通信号控制决策的每个代理中的第四节,实验结果表明所提出的方法的性能特点。最后,在第五部分,总结了不同的结果,目前的工作,并指出未来研究工作的指导方针。
2。相关作品
交通拥堵问题促使研究人员研究新的控制策略,有效地管理城市地区的交通运动。人工智能技术在不同地区的城市交通控制中得到了广泛的研究。在这一节中,控制交通灯和主要相关联的智能代理的应用程序的相关工作。
2.1。基于知识的交通灯模型
最初,模糊逻辑(佛罗里达州)领域是通过调整交通信号控制参数,这样的周期长度和相位序列模型专家的知识。佛罗里达州的信号控制器[ 1,5 ]使用一组规则,以确定优先行动的基础上的数量的投入。虽然结果表明,这些方法能够有效的控制在较小的网络中,一个静态的规则库的使用意味着进一步的系统更新可能需要时间。因此,它是很难产生一个有效的模型,特别是对于复杂的交叉点。 决策支持系统(决策支持系统)由计算机技术解决方案,可用于支持复杂的决策和目标,以协助人力运营商在选择有效的措施。许多决策支持系统,交通控制系统开发等hoogendoorn 和almejalli 模型[ 6,7 ]。虽然它已经表明,这些系统是能够提出有效的行动,他们依靠大量的历史数据和专家知识,这可能是很难获得和维护的数据库。此外,数据基地必须不断更新,新的流量方案和控制作用。
2.2。神经网络交通灯模型
人工神经网络(ANN )已得到了广泛的探讨,为决策方法。因此,作者在[ 8,9 ]已经成功地提出了人工神经网络训练交通信号控制器。人工神经网络模型计算的决定,通过学习成功解决的例子。然而,随着道路网络规模的增加,神经网络受到明显的性能下降。此外,当交通量的变化,他们必须重新学习控制的有效方法。
2.3。排队论交通灯模型
排队论也是一种广泛使用的方法,当它涉及到模型的尾巴在交叉点[ 10,11 ]。排队论算法采用“小”的规律,以确定队列的长度。使用队列信息„模型是能够发展为交通灯为了减少队列平均长度和平均等待时间的动态规划。然而,这个理论需要一个测量点来管理的交叉点,不能处理小冲突。
2.4。基于代理的交通灯模型
多Agent 系统(MAS )是一个IA 的工具,为复杂系统的集中与分散协调机制的多Agent 建模提供了原则。多Agent 系统的交通灯建模研究。因此,Dresner 和石[ 12 ]已经解决了这个问题,使用防撞在路口预订系统。在该系统中,车辆向中央代理提出请求。如果请求被接受,车辆需要按照规定的路径,在交叉路口时要安全。这类系统的初步实验是有希望的,在路口的车辆延误最小。然而,它缺乏多个交通灯之间的协调方面,这是一个最重要的问题,影响这样的系统的设计。roozemond [ 13 ]提出了代理模型的自主行动而共享数据。然后,交叉点使用此信息,使短期和长期的交通预测,相应的调整。bazzan [ 14 ]使用了一种结合MAS 和进化博弈理论的分散方法。这种方法模型的每个路口作为一个单独的动机的代理,它必须关注的不仅是获得车辆通过路口,但也减少旅行时间为所有车辆。
2.4.1。基于强化学习的交通灯模型
强化学习(RL )存在的几个应用,如在15–作品[ 19 ]。这些模型试图通过允许控制剂和交通环境之间的相互作用,推断出一个奖励模型的各种状态/行动。在大型网络中,这些对问题的大量 制约强化学习方法在这一领域的使用,(收敛)。
2.4.2。基于群优化算法的交通灯模型
群优化算法是人工智能的又一个领域,它被应用于交通灯的控制。奥利维拉的作品[ 20 ]利用等待车辆发射的信息素信息量的优化分配原理。在[ 21 ]中,作者提出了自主的交叉管理 (目的)基于蚁群系统(ACS )算法将车辆各车辆到达顺序和大量的车道。此外,许多作品已经使用进化计算,以优化网络中的信号计划
[ 22,24 ]。这些方法通常的目标是代表整个网络的信号计划,作为一组的染色体,然后应用进化算法。然而,这种解决方案需要与网络规模计算极高的权势。在复杂的网络中,实时控制可能是不可能的,因为在一个合适的时间,计算可能不完整。
2.4.3。自适应交通灯模型
交通灯配置的调整是主要的前期工作中提出的交通流模型优化的关键环节。然而,在现实世界中,交通灯的配置参数是基于数据和设计师的模型,这是经常更新。
可以使用多代理模型,以处理这些实际情况。但是,提供更准确的数据,它必须更适应变化。其思想是赋予系统以其自身的控制系统,使其能面对它的动态环境和遇到的多种情况。因此,自组织系统是一种特别适用于交通控制的方法。
格申森[ 25,26 ]提出了一个自组织系统的交通信号控制中的初步应用。作者表明,基于代理的解决方案是能够有效地控制交通信号。Lammer 和Helbing 扩大这项工作[ 27 ]提出了一种自组织控制系统,能有效地处理多种交通情景假设的网格状网络内。
由校风邪[ 28 ]的方法,提出了一种集料进入的车辆为簇使信号的实时计算模型 定时政策。这些自调度方法产生自组织系统。
周和铝。[ 29 ]提供一种在一个周期中选择一个周期序列的自适应交通灯算法,根据以下标准:存在 优先权的车辆,totalwaiting 时间和排队长度。然而,该算法需要相同的类型的车辆以相同的速度。
这项工作提出了基于自组织、自适应的系统原理,包括大规模自主分散的交通管理。每个路口都是由一个智能交通灯控制的代理,操作的实时与输入的流量。独立地,交叉口的交通控制遵循相同的行为,仅根据当地交通信息。
因此,该模型提供了一个分散的管理之间的基础上检测到的交通信息和使用无线传感器网络(无线传感器网络)的相邻的交叉点之间的管理。
在一个分层的体系结构,设计灵活,适应异构流量,在每一个阶段的情况下,重新评估的基础上选择性的流量参数。
3。模型:自适应交通信号控制
在本文中,提出了一种自适应交通灯模型在多个路口。每一个交通灯都根据交通流的性质和所有连接车道的饱和程度来调整它的状态。交通灯之间的协调,允许存在的“绿灯”,通过连续交叉的进展。这允许最大的灵活性,在交叉点最需要的,所以提高网络吞吐量。
该模型假定2个主要步骤,检测的主要交通数据,然后绿灯分配决定。第一步检测并实时计算交通信息。绿色的光分配决定利用交通信息来确定所需的服务下一个交通运动持续时间。在描述所提出的模型之前,提出的问题制剂。
3.1。问题的提法
在城市网络中,一个路口是一个共享的道路网络,车辆与冲突的路径交叉。车辆通过进入和退出车道进入一个交叉口。每个车道都有一套允许转弯的动作。每一段时间,允许一组旋转运动被称为相位。一个交通信号的所有阶段组成的信号的循环。交通灯的工作,通过骑自行车通过他们的光阶段,绿色允许车辆行驶通过路口,黄要求他们放慢和准备停止,和红色迫使车辆停止在路口。交通信号管理交叉口之间的交替变化,允许的转向运动的流量。
四个方向是(北,南,东,西),每一个都有2条通道,一个向前和向左转(图
1)。
范围广泛的传感器可以作为交通网络包括感应线圈、摄像机、磁强计、无线传感器数据探测器和雷达。我们定义了一个传感器节点,具有检测单元和一般能力有限的能量,内存,计算能力和通信方面的能力。
在大多数工作中,无线传感器的无线传感器的使用是出于各种特性,如:制造成本低,体积小,易于安装和快速安装反应,通信和内存单元。无线传感器可以位于不同的车道上。从文献的结果的基础上,传感器之间的距离应该是足够的,以便有一个正确的采样[ 30 ]。
所提出的模型包括三个不同位置的传感器在一个给定的交叉点(图2)。因此,每一个输入通道的流量监测这些 传感器节点:
•出口传感器节点位于靠近交通灯处,计算在绿灯时,车辆偏离路口;
•先进的传感器节点(如)放置在一个适当的距离前的光,计数车辆到达不断。它可以在一个固定的距离的光,考虑尽快在路口的入口。
•运动传感器节点(多通道)位于输出车道的入口。这样的部署,更好地区分车辆退出的交叉口,并允许通知邻居的交点,未来的运动流。移动传感器节点可以收集和汇总各节点的流量,在它的方向发射。
提交交通安全规则和每个路口,有十二个授权的动作如图3所示。因此,决策步骤被转换为一个决定,设置绿灯的持续时间,每一个运动和它们之间的显示序列。
3.2。系统的体系结构
建议多代理框架,如图4所示。三个主要组件包括在这个框架中:
1。传感器组件:采集实时交通流信息;
2。交通控制智能代理:路口交通灯的自主控制;
3。交通灯:装置通过控制交通灯,当每一个交通灯变成红色或绿色,并用多长时间来定义一个路口的行为。
城市交通强烈依赖于每个路口的交通灯系统的控制策略的快速反应。该控制策略机制可以描述如下:
首先,检测元件(无线传感器网络传感器)是用来监测交通量和转发的交通量数据的决策组件(代理);
在收集所有所需信息后,决定组件(代理)开始产生一种合适的交通信号控制策略;
执行组件(交通灯)是用来调节流量的。
4结论
该系统涉及双重目标:本地的目标是提高每个路口的交通量。全球目标是把交通信号放在一个协调的地方系统使车辆遇到绿灯先后因此排的车辆可以继续通过一系列连续的绿灯。
交通拥堵问题是城市生活中的一个严重问题,如延误、经济损失、环境污染等。在这项工作中,在多个路口的交通灯控制的自适应方法。基于本地实时交通数据和邻居的交叉口交通状况,绿色照明的序列是自我调整的。
自组织模型包含两个步骤:一次真正的交通数据检测和绿色照明的分配决定。前十二个允许的动作被定义。根据交通流的特点,绿色照明的分配决定使用计算的运动等级,以确定在每个路口的下一个动作,最需要的和足够的时间,以服务于它以及相邻的坐标。
因此,拟议的架构的目的是结合的动态方式在网络中的绿色浪潮和本地的适应。事实上,在多个路口,通过协调的交通控制,可以减少车辆的延误和停止。 应该指出的是,道路网络的效率不是完全基于每个交通灯的有效性,但也对他们的合作。因此,一个最佳的交通流量,可以通过所提出的交通灯协调机制的道路网络实现。
交通仿真(SUMO )是用来测试和验证系统。进行各种模拟研究一三种策略的性能:自适应的绿波法、驱动时间控制和定时控制。结果证实,该模型优于以前的方法在吞吐量方面,平均等待时间和平均次数的停止。
性能的改善是由于模型的能力,创建绿色的波,并在当地采取行动。模拟结果鼓励延长模型在未来的工作中引入新的元素,如公共交通和行人过路,将更接近现实。