CN104157150A

CN104157150A - Novel single-intersection traffic signal lamp control method

Info

Publication number: CN104157150A
Application number: CN201410390398.4A
Authority: CN
Inventors: 蒋昌俊; 闫春钢; 陈闳中; 叶晨; 杨忠程
Original assignee: Tongji University
Current assignee: Tongji University
Priority date: 2014-08-08
Filing date: 2014-08-08
Publication date: 2014-11-19
Anticipated expiration: 2034-08-08
Also published as: CN104157150B

Abstract

A novel single-intersection traffic signal control method, including: (1), separating the decision-making model of the signal light from the control algorithm: encapsulating the mathematical model of decision-making into a function; (2), no traffic congestion or congestion will not worsen Based on the premise that the waiting time of all vehicles at the red light is the shortest, a control model is proposed. The characteristics of the control model are as follows: (2.1), using the shortest vehicle length model; (2.2), using the optimal vehicle waiting time model; (2.3 ), two thresholds are proposed, which are the phase vehicle length threshold and the total vehicle length threshold; (2.4), a hybrid control model is proposed by combining the above three elements; (3), for the established model, a decision-making algorithm will be designed, Realize traffic light planning for several traffic light cycles in the future, and obtain a feasible plan that has the minimum value obtained by the above model and is real-time. The present invention can well dispatch common emergencies, real-time control, and good control effect.

Description

A new control method for single intersection traffic lights

技术领域technical field

本发明涉及实时单路口交通信号灯的控制决策的数学模型建立及求解该模型的算法。The invention relates to the establishment of a mathematical model for the control decision-making of a real-time single intersection traffic signal lamp and an algorithm for solving the model.

背景技术Background technique

近年来随着经济和科学技术的发展，城市的汽车数量不断增多，这也带来了城市交通拥挤问题。如今这些问题越来越受到人们的重视，各个城市也逐渐建立一些智能交通的项目，摄像头、传感器、地磁线圈等信息采集设备收集到了大量复杂的交通信息，这些数据怎么作用于交通指挥中提出了新的更高的要求。针对这些路口交通信息，怎么有效控制红绿灯使之能有效的缓解交通压力成为了今年来的一个研究热点。In recent years, with the development of economy and science and technology, the number of cars in cities has been increasing, which has also brought about urban traffic congestion. Nowadays, these problems have been paid more and more attention by people, and various cities have gradually established some intelligent transportation projects. Information collection equipment such as cameras, sensors, and geomagnetic coils have collected a large amount of complex traffic information. How these data can be used in traffic command is proposed. New and higher requirements. According to the traffic information of these intersections, how to effectively control the traffic lights so that they can effectively relieve the traffic pressure has become a research hotspot this year.

现在，绝大多数交通信号灯的控制模型和算法在设计之初都是把模型建立与算法设计两块混合完成，即首先确立算法模型，然后根据该模型特点设计出求解该模型的算法。因此存在许多问题。其中主要存在以下两种问题At present, most of the control models and algorithms of traffic lights are designed by combining model establishment and algorithm design, that is, first establish the algorithm model, and then design an algorithm to solve the model according to the characteristics of the model. Therefore there are many problems. There are mainly two problems

1、在模型设计方面因为模型过于简化，不能很好的调度常见的突发状况如车辆稀少、早晚高峰等。1. In terms of model design, because the model is too simplified, it cannot well dispatch common emergencies such as scarce vehicles, morning and evening peaks, etc.

2、为求效果算法过于复杂不能实时控制，或者算法过于简单控制效果不佳。2. The algorithm is too complex for real-time control, or the algorithm is too simple and the control effect is not good.

发明内容Contents of the invention

本发明目的在于克服现有技术不足，公开一种新型的单路口交通信号灯控制决策模型及算法，能很好的调度常见的突发状况，实时控制，控制效果佳。The purpose of the present invention is to overcome the deficiencies of the prior art, and disclose a new single intersection traffic signal light control decision-making model and algorithm, which can well dispatch common emergencies, real-time control, and good control effect.

本发明方法思路：将交通信号灯的控制决策模型和算法进行分离，设计出一个更全面的信号灯决策模型，再设计一个更有效的实时控制算法。The idea of the method of the present invention is to separate the control decision-making model and the algorithm of the traffic signal light, design a more comprehensive signal light decision-making model, and then design a more effective real-time control algorithm.

原理：本发明将实时信号灯控制问题的控制决策模型和算法进行了分离。首先设计了一种混合控制决策模型，使的信号灯控制决策在不出现交通情况恶化的前提下尽可能人性化。其次，设计了动态规划算法将复杂的信号灯决策变成可以在多项式时间内求得。最终，在动态规划的状态转移中调用混合控制模型，进而达到模型和算法的结合，使得能快速得到一个具有一定人性化的信号灯控制决策。Principle: The present invention separates the control decision-making model and algorithm of the real-time signal light control problem. Firstly, a hybrid control decision-making model is designed to make the control decision-making of signal lights as humanized as possible without deteriorating traffic conditions. Secondly, a dynamic programming algorithm is designed to make the complex signal light decision can be obtained in polynomial time. In the end, the hybrid control model is called in the state transition of dynamic programming, and then the combination of the model and the algorithm is achieved, so that a certain humanized signal light control decision can be quickly obtained.

为解决以上技术问题，本发明采用如下的技术方案：In order to solve the above technical problems, the present invention adopts the following technical solutions:

一种新型的单路口交通信号灯控制决策模型及算法，其特征在于，该方法包括如下步骤：A novel single intersection traffic light control decision-making model and algorithm, characterized in that the method comprises the following steps:

(1)、对于将信号灯的决策模型和控制算法分离，本发明将决策的数学模型封装成一个函数，即将每一时刻的路口信息提交给该数学模型函数，则可以得到一个函数值表示该路口的交通指数，数值越低则交通越好，反之越差。如此，决策的数学模型对于后续的算法设计相当于黑盒操作，即改变模型不影响以后的算法决策。(1), for separating the decision-making model of signal lights from the control algorithm, the present invention encapsulates the mathematical model of decision-making into a function, is about to submit the crossing information at each moment to the mathematical model function, and then a function value can be obtained to represent the crossing The traffic index of , the lower the value, the better the traffic, and vice versa. In this way, the mathematical model of decision-making is equivalent to a black-box operation for subsequent algorithm design, that is, changing the model does not affect future algorithm decisions.

(2)、在模型设计方面，本发明以交通不出现拥堵或拥堵不会恶化为前提让所有车辆的红灯等待时间最短为目的提出了控制模型。模型如下：(2), in terms of model design, the present invention proposes a control model for the purpose of making the red light waiting time of all vehicles the shortest on the premise that traffic jams do not occur or congestion does not worsen. The model is as follows:

(2.1)、为了防止或者减轻拥堵，本发明采用最短车辆长度模型。对某一时刻计算出各个路口的排队车辆长度，对它们的长度求和就是该模型函数的函数值。(2.1), in order to prevent or alleviate congestion, the present invention adopts the shortest vehicle length model. Calculate the length of queued vehicles at each intersection at a certain moment, and the sum of their lengths is the function value of the model function.

(2.2)、为了让所有车等待时间越少，本发明采用了最优车辆等待时间模型。对每一辆车当它进入了车辆等待队列就对它进行计时，计时时间为该车辆的等待时间，对应某一时刻，对所有的车辆的等待时间求和就是该模型函数的函数值。(2.2), in order to allow the waiting time of all cars to be less, the present invention adopts an optimal vehicle waiting time model. When each vehicle enters the vehicle waiting queue, it is timed. The timing time is the waiting time of the vehicle. Corresponding to a certain moment, the sum of the waiting time of all vehicles is the function value of the model function.

(2.3)、为了融合上述两个模型，本发明提出了两阈值，即相位车辆长度阈值和总车辆长度阈值。相位车辆长度阈值为，对一个路口的某一个相位的车辆长度提出阈值，阈值一般为该相位的车辆长度在其最长绿灯时间的2到5倍(具体系数视具体情况待定)时长内刚够将所有车驶出。总车辆长度阈值为，对各个相位车辆长度总和提出阈值，阈值一般为对该路口所有车辆长度在标准红绿灯周期的2到5倍时间(具体系数视具体情况待定)内刚够将所有车驶出。(2.3), in order to fuse the above two models, the present invention proposes two thresholds, namely phase vehicle length threshold and total vehicle length threshold. The phase vehicle length threshold is a threshold value for the vehicle length of a certain phase at an intersection. The threshold is generally just enough for the vehicle length of this phase to be 2 to 5 times the longest green light time (the specific coefficient is to be determined according to the specific situation). Get all cars out. The total vehicle length threshold is a threshold value for the sum of the vehicle lengths of each phase. The threshold value is generally that the length of all vehicles at the intersection is within 2 to 5 times of the standard traffic light cycle (the specific coefficient is to be determined according to the specific situation) and it is just enough to drive all vehicles out. .

(2.4)、将上述三要素融合提出了本发明的混合控制模型。当前路口各个相位的车辆长度小于相位车辆长度阈值并且总车辆长度小于总车辆长度阈值时采用最优车辆时间等待模型，否则采用最短车辆等待时间模型。注：作为比较的车辆长度为当前时刻的车辆长度，和调用该函数所传入的路口信息无关。(2.4), combining the above three elements to propose a hybrid control model of the present invention. When the vehicle length of each phase at the current intersection is less than the phase vehicle length threshold and the total vehicle length is less than the total vehicle length threshold, the optimal vehicle time waiting model is adopted, otherwise the shortest vehicle waiting time model is adopted. Note: The vehicle length to be compared is the vehicle length at the current moment, and has nothing to do with the intersection information passed in by calling this function.

(3)、对于已经建立的模型，本发明将设计出一种决策算法，算法将实现对未来几个红绿灯周期的红绿灯规划，得到一种可行的规划对上述模型求得的值最小。且在满足该特性的同时保证具有较高的实时性。算法步骤如下：(3), for the established model, the present invention will design a kind of decision-making algorithm, and algorithm will realize the traffic light planning to several traffic light cycles in the future, obtains a kind of feasible planning and obtains the minimum value to above-mentioned model. And while satisfying this characteristic, high real-time performance is guaranteed. The algorithm steps are as follows:

(3.1)、算法的预处理。算法预处理包括对现有得到的路口信息数据进行二次处理，对处理后的数据分析对未来车流量的预测和绿灯期开走车辆的预测。(3.1), Algorithm preprocessing. Algorithm preprocessing includes secondary processing of the existing intersection information data, analysis of the processed data, prediction of future traffic flow and prediction of vehicles driving away during the green light period.

(3.2)、对算法模型设计出一个整数规划模型。本发明本根据红绿灯显示时间的最小单位(秒)作为基本单位，整个模型就可以化简为，第i时刻第j个相位红绿灯情况为P_ij，P_ij为一个0/1变量，0表示红灯，1表示绿灯。那么整个问题就可以化简成有许许多多个P_ij求一种最优的规划，使得上文所建立的数学模型的结果最小。当然P_ij还需要满足一些红绿灯特有的限制。(3.2) Design an integer programming model for the algorithm model. According to the present invention, the minimum unit (second) of traffic light display time is used as the basic unit, and the whole model can be simplified as, the situation of the jth phase traffic light at the i moment is P _ij , and P _ij is a 0/1 variable, and 0 represents red Light, 1 means green light. Then the whole problem can be simplified to find an optimal plan with many P _ij , so that the result of the mathematical model established above is the smallest. Of course, P _ij also needs to meet some traffic light-specific restrictions.

(3.3)根据红绿灯的特殊性质设计出动态规划方法求解整数规划模型。如果直接求解线性规划问题，那么状态总量为2^n*m，n表示决策的时间长度，m表示相位数量，可以看出这是一个NP问题，而在实际红绿灯控制中，人们希望他具有很强的精度同时必须要在短时间内求解出来。但是这个问题本身并不是一个纯正的毫无约束的0/1规划问题，其本身存在一些很强的约束条件，通过对该0/1规划问题的研究和红绿灯特有的性质，本课题对0/1规划模型做出了修改为一个可以使用动态规划求解的整数规划问题，最后在多项式时间内(实际运行中时间复杂度小于1秒)求解出来。(3.3) According to the special properties of traffic lights, a dynamic programming method is designed to solve the integer programming model. If the linear programming problem is directly solved, then the total state is 2 ^n*m , n represents the time length of decision-making, m represents the number of phases, it can be seen that this is an NP problem, and in the actual traffic light control, people hope that it has a lot Strong precision must be solved in a short time at the same time. However, this problem itself is not a pure unconstrained 0/1 programming problem. It has some strong constraints. Through the research on the 0/1 programming problem and the unique properties of traffic lights, this topic is 0/1 1 The planning model is modified to an integer programming problem that can be solved by dynamic programming, and finally solved in polynomial time (the time complexity in actual operation is less than 1 second).

本发明的创新点体现在：The innovation of the present invention is reflected in:

1)将控制决策模型和算法分离。1) Separate the control decision-making model from the algorithm.

2)设计出一种混合控制模型。2) Design a hybrid control model.

3)设计出一种新型决策算法：基于整数规划的动态规划算法。3) Design a new decision-making algorithm: dynamic programming algorithm based on integer programming.

附图说明Description of drawings

图1本发明的总流程图The general flowchart of Fig. 1 the present invention

图2控制模型流程图Figure 2 Control Model Flowchart

图3决策算法流程图Figure 3 Decision Algorithm Flowchart

公式说明Formula description

公式1最短车辆长度模型Formula 1 Minimum Vehicle Length Model

公式2最优车辆等待时间模型Formula 2 optimal vehicle waiting time model

公式3模型转换条件Formula 3 Model Conversion Conditions

公式4混合控制模型Formula 4 Hybrid Control Model

公式5算法状态转移方程Formula 5 Algorithm State Transition Equation

具体实施方式Detailed ways

以下结合附图和公式对本发明技术方案作进一步说明。The technical solutions of the present invention will be further described below in conjunction with the accompanying drawings and formulas.

原理：本发明提出对信号灯控制决策模型和算法的分离，并在此基础上分别对模型建立和算法设计提出优化。Principle: The present invention proposes the separation of signal light control decision-making model and algorithm, and on this basis, optimizes model establishment and algorithm design respectively.

如图1是本发明的总流程图，首先通过对模型和算法分离，将控制模型封装成一个黑盒的函数，然后通过随后的决策算法不断调用模型函数，以达到决策目的。Figure 1 is the general flowchart of the present invention. Firstly, by separating the model from the algorithm, the control model is encapsulated into a black box function, and then the model function is continuously called through the subsequent decision-making algorithm to achieve the decision-making purpose.

图2是本发明的所建立的信号灯数学模型，他基于了公式(1)的最短车辆长度模型和公式(2)的最优车辆等待时间模型加入了公式(3)的决策阈值，组合成一个混合决策模型，模型的数学表达为公式(4)。Fig. 2 is the mathematical model of signal lights of the present invention, he has added the decision-making threshold of formula (3) based on the shortest vehicle length model of formula (1) and the optimal vehicle waiting time model of formula (2), is combined into a Mixed decision model, the mathematical expression of the model is formula (4).

公式(1)：L_m为总车辆长度，L_i为第i号相位的车辆长度。Formula (1): L _m is the total vehicle length, L _i is the vehicle length of the i-th phase.

公式(2)：T_n为总车辆等待时间，t_i为第i号车辆的等待时间。Formula (2): T _n is the total vehicle waiting time, t _i is the waiting time of the i-th vehicle.

公式(3)：L_max1为相位车辆长度阈值，L_max1为总车辆长度阈值。Formula (3): L _max1 is the phase vehicle length threshold, and L _max1 is the total vehicle length threshold.

公式(4)：G(t)为模型函数，t对应t时刻。Formula (4): G(t) is a model function, and t corresponds to time t.

${L L}_{m m} = = {Σ Σ}_{i i = = 11}^{m m} {L L}_{i i}$

公式1最短车辆长度模型Formula 1 Minimum Vehicle Length Model

${T T}_{n no} = = {Σ Σ}_{i i = = 11}^{n no} {t t}_{i i}$

L_i≤L_max1(1≤i≤m)∩L_m≤L_max2 L _i ≤L _max1 (1≤i≤m)∩L _m ≤L _max2

公式3模型转换条件Formula 3 Model Conversion Conditions

$G G ((t t)) = = \{\begin{matrix} {Σ Σ}_{i i = = 11}^{m m} {L L}_{i i} (({L L}_{i i} \leq \leq {L L}_{max max 11} \cap \cap {L L}_{m m} \leq \leq {L L}_{max max 22})) \\ {Σ Σ}_{i i = = 11}^{n no} {t t}_{i i} (({L L}_{i i} > > {L L}_{max max 11} \cup \cup {L L}_{m m} > > {L L}_{max max 22})) \end{matrix}$

公式4混合控制模型Formula 4 Hybrid Control Model

$F f ((t t,, i i)) = = \{\begin{matrix} \underset{{g g}_{min min} \leq \leq g g \leq \leq {g g}_{max max}}{min min} ((F f ((t t - - g g,, i i - - 11)) + + {Σ Σ}_{k k = = t t - - g g}^{t t} G G ((k k)))) ((i i &NotEqual; &NotEqual; m m)) \\ \underset{{g g}_{min min} \leq \leq g g \leq \leq {g g}_{max max}}{min min} ((F f ((t t - - g g,, 11)) + + {Σ Σ}_{k k = = t t - - g g}^{t t} G G ((k k)))) ((i i = = m m)) \end{matrix}$

公式5算法状态转移方程Formula 5 Algorithm State Transition Equation

如图3为本发明的决策算法流程，如图所示算法首相对数据进行预处理和对未来车流量的预测，然后通过对信号灯模型特点建立整数规划模型，最后通过模型特点使用动态规划算法求解。Figure 3 is the decision-making algorithm process of the present invention, as shown in the figure, the algorithm prime minister performs preprocessing on the data and forecasts the future traffic flow, then establishes an integer programming model through the characteristics of the signal lamp model, and finally uses the dynamic programming algorithm to solve the problem through the characteristics of the model .

公式(5)是动态规划算法的状态转移方程，其中t表示第t时刻，i表示当前第i路口为绿灯，g为i路口的绿灯时间。Formula (5) is the state transition equation of the dynamic programming algorithm, where t represents the t-th moment, i represents the current green light at the i-th intersection, and g is the green light time at the i-th intersection.

综上所述，整个实现流程概括为：In summary, the entire implementation process is summarized as follows:

1)将信号灯的控制模型和决策算法分离。1) Separate the control model of the signal light from the decision-making algorithm.

2)对控制模型设计出混合控制模型。2) Design a hybrid control model for the control model.

3)对决策算法设计出一个整数规划模型并用动态规划算法进行求解。3) Design an integer programming model for the decision-making algorithm and use the dynamic programming algorithm to solve it.

Claims

1. a novel single intersection method for controlling traffic signal lights, is characterized in that, the method comprises the steps:

(1), the decision model of signal lamp is separated with control algolithm:

The mathematical model of decision-making is packaged into a function, the intersection information in each moment is submitted to this mathematical model function, obtain the traffic index that a functional value represents this crossing, the lower traffic of numerical value is better, otherwise poorer;

(2), can not to worsen as prerequisite makes the red light stand-by period of all vehicles the shortest in object has proposed control model do not appear blocking up or blocking up in traffic, this control model feature is characterized by:

(2.1), adopt the shortest Vehicle length model: being calculated to the queuing Vehicle length at each crossing a certain moment, is exactly the functional value of this pattern function to their length summation;

(2.2), adopted optimum vehicle stand-by period model: each car is just carried out to timing to it when it has entered vehicle waiting list, timing time is the stand-by period of this vehicle, the corresponding a certain moment is exactly the functional value of this pattern function to the stand-by period summation of all vehicles;

(2.3), two threshold values are proposed, be phase place Vehicle length threshold value and total Vehicle length threshold value:

Phase place Vehicle length threshold value is, the Vehicle length of the some phase places to a crossing proposes threshold value, and the Vehicle length that threshold value is generally this phase place just enough rolls all cars away from 2 to 5 times of durations of its maximum green perild;

Total Vehicle length threshold value is that, to each phase place Vehicle length summation proposition threshold value, threshold value is that all Vehicle lengths in this crossing are just enough rolled away from all cars within 2 to 5 times of times in standard traffic lights cycle;

(2.4), above-mentioned three elements are merged and have proposed Hybrid Control Model: the Vehicle length of each phase place of current crossing is less than when phase place Vehicle length threshold value and total Vehicle length are less than total Vehicle length threshold value and adopts optimum vehicle time Holding Model, otherwise adopts the shortest vehicle stand-by period model;

(3), for the model of having set up, will design a kind of decision making algorithm, realize the traffic lights planning to several traffic lights cycle in future, obtain the value minimum that a kind of feasible planning is tried to achieve above-mentioned model, and real-time, step is as follows:

(3.1), the pre-service of algorithm: algorithm pre-service comprises carries out secondary treating to the existing intersection information data that obtain, the drive away prediction of vehicle of the prediction to data analysis after treatment to future car flow and green light phase;

(3.2), algorithm model is designed to an integer programming model: as base unit, whole model can abbreviation be just that the i moment, j phase place traffic lights situation was P according to the least unit of traffic lights displaying time (second) _ij, P _ijbe 0/1 variable, 0 represents red light, and 1 represents green light, and whole problem reduction becomes to have various P _ijask a kind of planning of optimum, make the result minimum of the mathematical model set up above;

(3.3) design dynamic programming method and solve integer programming model.