CN115995152A - Iterative learning boundary control method for dynamically balancing traffic loads of multiple sub-areas of city - Google Patents

Abstract

The invention relates to the field of urban traffic signal control, in particular to an iterative learning boundary control method for dynamically balancing urban multi-sub-area traffic loads. Including the following steps, S100: Select the urban road network area to be studied, collect road network data and analyze it; S200: Divide traffic sub-areas within the selected urban road network, and establish each control sub-area according to the collected road network data S300: Analyze the traffic flow in and out of the sub-area boundary, establish a sub-area traffic flow model; S400: Implement boundary control for each sub-area, build an iterative learning control model according to the repetitive characteristics of urban traffic flow, and obtain an iterative control plan And set the signal timing strategy. The invention comprehensively considers the inflow and outflow of vehicles between multiple sub-areas and the traffic flow between the sub-areas and the road network boundary, improves the efficiency of iterative learning control, improves the robustness of the system, and makes the boundary control show real-time and optimal characteristics. .

Description

An iterative learning boundary control method for dynamically balancing traffic loads in multiple urban sub-areas

Technical Field

The present invention relates to the field of urban traffic signal control, and in particular to an iterative learning boundary control method for dynamically balancing the traffic loads of multiple sub-areas in an urban area.

Background Art

As the income of urban residents continues to rise, the number of cars owned by families is also increasing, which has led to more serious urban congestion problems, and it is more likely to cause traffic accidents and traffic safety hazards. In order to alleviate the pressure of urban road traffic and reduce the growing travel difficulties year by year, traffic signal control has become a research direction for many scholars. In the early days, due to insufficient traffic flow, road congestion problems often occurred at a few intersections in the core area of the city and rarely spread to other areas. However, today, traffic congestion has evolved into a large-scale, multi-intersection complex regional problem. The road network macro basic diagram (MFD) is a basic attribute of the road network. It can analyze the traffic conditions in the road network area from a macro level, determine the saturated traffic flow of the road network, and help researchers better solve the congestion problem.

Traffic control sub-areas can help traffic management departments divide the road network into small areas with similar characteristics when regulating a large-scale road network, and solve the congestion problem by adjusting the traffic signals in each area. At present, there is a lot of research on sub-area boundary control based on MFD. In the early days, boundary control based on a single sub-area was mainly to limit the flow of vehicles into the sub-area at the boundary intersection of the controlled area. Now, boundary control for multiple sub-areas can evenly distribute traffic flow among the sub-areas and reduce the oversaturation of road sections. Due to the repetitive nature of road network traffic flow, iterative learning control can use historical traffic flow data to construct new control input variables, fully track the expected trajectory within a limited time, and improve control quality.

There are still many shortcomings in the existing research on the boundary control of multiple sub-areas in urban road networks. Unreasonable sub-area division will affect the drawing of the sub-area macro basic map, resulting in errors in the saturated traffic flow in each sub-area and affecting boundary regulation. In addition, most studies tend to consider the inflow and outflow of vehicles between sub-areas and ignore the traffic flow between sub-areas and boundaries. Uniformly distributing traffic flow within sub-areas may lead to aggravated vehicle queuing at boundary intersections and traffic overflow, thus leading to uneven vehicle distribution and diffusion of congestion.

Summary of the invention

In order to solve the above problems, the present invention provides an iterative learning boundary control method for dynamically balancing the traffic loads of multiple sub-areas in a city.

The present invention adopts the following technical scheme: an iterative learning boundary control method for dynamically balancing the traffic load of multiple sub-areas in a city, comprising the following steps:

S100: Select the urban road network area to be studied, collect road network data and analyze it;

S200: Divide the traffic sub-areas within the selected urban road network, and establish a macro basic map of each control sub-area based on the collected road network data;

S300: Analyze the traffic flow in and out of the sub-area boundary and establish a sub-area traffic flow model;

S400: Implement boundary control on each sub-area, build an iterative learning control model according to the repetitive characteristics of urban traffic flow, obtain an iterative control scheme and set a signal timing strategy.

The specific steps of S100 include:

S101: Obtain the road topology structure of the urban traffic road network to be studied, and set up data collectors at the entrance and exit locations of each signalized intersection;

S102: The actual traffic flow data, road section length and number of lanes, red light conditions, pedestrian crossings and parking lots of each intersection within the road network are obtained through the set data collector and Baidu map, and the traffic flow characteristics, path flow, average travel time, lane branching and turning ratio of the road network are analyzed.

The specific steps of dividing the traffic sub-areas in S200 include:

S201: Calculate the path association between adjacent intersections in the road network:

为相邻交叉口i与交叉口j之间的交通关联度；T为车辆在交叉口i，j间的平均行程时间，单位为min；

为路段流量不均衡系数；m为来自上游交叉口的关联流向数；

为到达下游交叉口的车流量总和；

为上游交叉口的最大交通量，即

的最大值；in,

is the traffic correlation between adjacent intersections i and j; T is the average travel time of vehicles between intersections i and j, in min;

is the unbalanced coefficient of the road section flow; m is the number of associated flow directions from the upstream intersection;

is the total volume of vehicles arriving at the downstream intersection;

is the maximum traffic volume at the upstream intersection, that is

The maximum value of

S202: Calculate the path correlation of all adjacent intersections within the road network to be studied through the road network data collected by S100, calculate the road network density Laplace matrix and the eigenvalues corresponding to the eigenvectors, list the eigenvectors as a matrix as the algorithm input for fast global K-means clustering, select the sample point that minimizes the square error criterion function as the best clustering center of the cluster, save the clustering results, divide the intersections in a cluster into a sub-area, and merge the non-clustered intersections into adjacent sub-areas through boundary adjustment, so as to complete the urban road network sub-area division.

The step of establishing a macro basic map of each control sub-area in S200 includes collecting the cumulative number of vehicles in each sub-area within a unit time interval and the number of vehicles leaving the road network through a data collector, and performing data processing to remove data with obvious errors and redundant data, draw an MFD scatter plot, perform curve fitting based on the scatter data, and obtain the macro basic map of each traffic sub-area and its corresponding critical cumulative number of vehicles.

The traffic flow model of the sub-area in S300 is:

其中，

为第k个采样时刻子区i的内部车辆数，单位为veh；

为控制周期，单位为s；

为第k个采样时刻相邻子区m向子区i输入的流量，单位为veh/h；

为第k个采样时刻路网边界向子区i输入的流量，单位为veh/h；

为第k个采样时刻子区i向子区m输出的流量，单位为veh/h；

为第k个采样时刻子区i向路网边界输出的流量，单位为veh/h；

为子区i内部交通发生源或消散源产生或减少的流量，单位为veh/h。

in,

is the number of vehicles in sub-area i at the kth sampling moment, in veh;

is the control period, in seconds;

is the flow rate input from the adjacent sub-area m to the sub-area i at the k-th sampling moment, in veh/h;

is the flow rate input from the road network boundary to sub-area i at the kth sampling moment, in veh/h;

is the flow rate output from sub-area i to sub-area m at the kth sampling moment, in veh/h;

is the flow rate output from sub-area i to the road network boundary at the kth sampling time, in veh/h;

It is the flow rate generated or reduced by the traffic generation source or dissipation source within sub-area i, in veh/h.

The specific process of S400 is:

S401: According to the principle of conservation of vehicle number, a state space expression of the road network is established, and iterative learning control is introduced into the sub-area traffic flow model to obtain an iterative learning control model. In the iterative learning control model, the state equation of the sub-area at the nth iteration is:

表示系统的状态变量；

表示系统的输入变量；

表示系统的输出变量；

为单位矩阵；

为输入矩阵；

为输出矩阵；

为状态扰动矩阵；in,

Represents the state variables of the system;

Represents the input variables of the system;

Represents the output variable of the system;

is the identity matrix;

is the input matrix;

is the output matrix;

is the state perturbation matrix;

The tracking error of the system is:

表示系统的期望输出；

表示系统的输出扰动；in,

Represents the expected output of the system;

represents the output disturbance of the system;

S402: setting an iterative learning control rate for the divided sub-areas. The sub-areas and the outer boundary are open-closed loop PD type, denoted as 1, and the inner boundary of the sub-area is P type, denoted as 2:

为第n次迭代的控制输入饱和函数；

为微分学习增益；

，

为比例反馈学习增益；

为第n次迭代时的系统跟踪误差；in,

The control input saturation function for the nth iteration;

is the differential learning gain;

,

is proportional feedback learning gain;

is the system tracking error at the nth iteration;

的控制输入

，从而保证受控子区处于稳定运行状态，各个子区交通负荷动态平衡于整体路网。S403: Iterative calculation is performed on the iterative learning control model. In the first iteration, the original signal timing scheme of the road network sub-area is used. From the second iteration onwards, the input variable of the current system is compared with the cumulative number of vehicles collected in the previous time. The iterative learning control rate obtained above is used. After multiple iterations of learning control, the cumulative number of vehicles in each sub-area is calculated to follow the expected output.

Control input

, thereby ensuring that the controlled sub-areas are in a stable operating state and the traffic load of each sub-area is dynamically balanced with the overall road network.

Compared with the prior art, the present invention provides an iterative learning boundary control method for dynamically balancing the traffic load of multiple sub-areas in a city. Based on the characteristics of the current urban traffic road network being complex and large and difficult to regulate globally, the road network traffic flow data is collected, homogeneous traffic sub-areas are divided, and the signals are regulated by points instead of faces. According to the repeatability characteristics of macroscopic traffic flow, the iterative learning control method is used to stabilize the number of vehicles in each sub-area road network to the desired state, and the boundary control method is used to achieve dynamic balance of the overall road network traffic load. The present invention comprehensively considers the inflow and outflow of vehicles between multiple sub-areas and the vehicle flow between the sub-areas and the road network boundaries, and combines the use of PD-type and P-type iterative learning control to perform macroscopic regulation of the boundaries of each sub-area, and simultaneously utilizes the errors generated by this learning and the previous learning, further improving the efficiency of iterative learning control, improving the robustness of the system, and making the boundary control show real-time and optimal characteristics.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic flow chart of an iterative learning boundary control method for dynamically balancing traffic loads in multiple sub-districts of a city provided by the present invention;

FIG2 is a schematic diagram of the flow direction of traffic at adjacent intersections of an iterative learning boundary control method for dynamically balancing the traffic loads of multiple sub-areas in an urban area provided by the present invention;

FIG3 is a schematic diagram of multi-sub-area traffic flow according to an iterative learning boundary control method for dynamically balancing urban multi-sub-area traffic loads provided by the present invention.

DETAILED DESCRIPTION

In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described in addition are only used to explain the present invention, but are not used to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without making creative work will fall within the scope of protection of the present invention.

Referring to FIG. 1 , the present invention provides an iterative learning boundary control method for dynamically balancing the traffic load of multiple sub-areas in a city, comprising:

S100: Select a suitable urban road network area to be studied, collect road network data and analyze it;

S200: Divide the traffic sub-areas within the selected urban road network and establish a macro basic diagram (MFD) of each control sub-area based on the collected traffic data;

S300: Analyze the traffic flow in and out of the sub-area boundary and establish a sub-area traffic flow model;

S400: Implement boundary control on each sub-area, build an iterative learning control model based on the repetitive characteristics of urban traffic flow, obtain an iterative control scheme and set a signal timing strategy to ensure the dynamic balance of vehicle load within the research road network and improve urban traffic congestion.

The steps of collecting and analyzing road network data include:

Select the urban traffic road network to be studied, obtain the road topology information, and design data collectors at the entrance and exit locations of each signal intersection;

Through the set data collector and Baidu map, the actual traffic flow data, road section length and number of lanes, red light conditions, pedestrian crossings and parking lots at each intersection within the road network are obtained, and the traffic flow characteristics, path flow, average travel time, lane branching and turning ratio of the road network are analyzed.

The specific steps of dividing the traffic sub-areas include:

First, the path correlation between adjacent intersections in the road network is calculated:

为相邻交叉口i与交叉口j之间的交通关联度；T为车辆在交叉口i，j间的平均行程时间，单位为min；

为路段流量不均衡系数；m为来自上游交叉口的关联流向数；

为到达下游交叉口的车流量总和；

为上游交叉口的最大交通量，即

的最大值；图2展示了相邻交叉口的流量流向示意图来帮助说明上述关联度计算公式。in,

is the traffic correlation between adjacent intersections i and j; T is the average travel time of vehicles between intersections i and j, in min;

is the unbalanced coefficient of the road section flow; m is the number of associated flow directions from the upstream intersection;

is the total volume of vehicles arriving at the downstream intersection;

is the maximum traffic volume at the upstream intersection, that is

The maximum value of; Figure 2 shows a schematic diagram of the flow direction of adjacent intersections to help illustrate the above correlation calculation formula.

The path correlation of all adjacent intersections in the road network to be studied is calculated through the collected data, the road network density Laplace matrix and the eigenvalues corresponding to the eigenvectors are calculated, the eigenvectors are listed as matrices as algorithm inputs for fast global K-means clustering, and the sample points that minimize the square error criterion function are selected as the best clustering center of the cluster. The clustering results are saved, and the intersections in a cluster are divided into a sub-area. The intersections that are not clustered are merged into adjacent sub-areas through boundary adjustment, thereby completing the sub-area division of the urban road network.

The steps of establishing the macro basic diagram of each control sub-area include:

The data collector collects the cumulative number of vehicles in each sub-area within a unit time interval, as well as the number of vehicles leaving the road network, and processes the data to remove obviously erroneous and redundant data, draws an MFD scatter plot, and performs curve fitting based on the scattered data to obtain a macro basic map of each traffic sub-area and its corresponding critical cumulative number of vehicles.

The steps of establishing the sub-area traffic flow model include:

The macro basic diagram of multiple sub-areas divided by the urban road network is analyzed. Here, it is assumed that the traffic flow in each sub-area is relatively stable, as shown in Figure 3, which is defined as sub-areas i, j, and m. One of the sub-areas i is taken as a reference object for research, and the number of vehicles in the sub-area at the kth sampling time is determined to be:

为第k个采样时刻子区i的内部车辆数，单位为veh；

为控制周期，单位为s；

为第k个采样时刻相邻子区m向子区i输入的流量，单位为veh/h；

为第k个采样时刻路网边界向子区i输入的流量，单位为veh/h；

为第k个采样时刻子区i向子区m输出的流量，单位为veh/h；

为第k个采样时刻子区i向路网边界输出的流量，单位为veh/h；

为子区i内部交通发生源或消散源（如居民区，停车场）产生或减少的流量，单位为veh/h。in,

is the number of vehicles in sub-area i at the kth sampling moment, in veh;

is the control period, in seconds;

is the flow rate input from the adjacent sub-area m to the sub-area i at the k-th sampling moment, in veh/h;

is the flow rate input from the road network boundary to sub-area i at the kth sampling moment, in veh/h;

is the flow rate output from sub-area i to sub-area m at the kth sampling moment, in veh/h;

is the flow rate output from sub-area i to the road network boundary at the kth sampling time, in veh/h;

It is the flow rate generated or reduced by the internal traffic generation or dissipation sources (such as residential areas and parking lots) in sub-area i, in veh/h.

Among them, the boundary control of each sub-area is carried out, the state space expression of the road network is established according to the principle of conservation of vehicle number, and iterative learning control is introduced into the traffic flow model. The iterative learning control model is constructed by using the repetitive characteristics of urban traffic flow, including:

Defined in the iterative learning control model, the state equation of the sub-zone at the nth iteration is:

表示系统的状态变量；

表示系统的输入变量；

表示系统的输出变量；

为单位矩阵；

为输入矩阵；

为输出矩阵；

为状态扰动矩阵；in,

Represents the state variables of the system;

Represents the input variables of the system;

Represents the output variable of the system;

is the identity matrix;

is the input matrix;

is the output matrix;

is the state perturbation matrix;

The tracking error of the system is:

表示系统的期望输出；

表示系统的输出扰动；in,

Represents the expected output of the system;

represents the output disturbance of the system;

The iterative learning control rate is designed for the divided sub-areas. The sub-areas and the outer boundary are open-closed loop PD type, denoted as 1, and the inner boundary of the sub-area is P type, denoted as 2:

为第n次迭代的控制输入饱和函数；

为微分学习增益；

，

为比例反馈学习增益；

为第n次迭代时的系统跟踪误差。in,

The control input saturation function for the nth iteration;

is the differential learning gain;

,

is proportional feedback learning gain;

is the system tracking error at the nth iteration.

The present invention considers balancing the traffic load of multiple sub-areas, that is, considering the relationship between the number of vehicles in the current road network and the optimal cumulative number of vehicles in the road network. If the number of vehicles in the current road network is equal to the optimal cumulative number of vehicles in the road network, it means that the traffic network is in a saturated state and can carry the maximum traffic flow without congestion, that is:

为路网最优累计车辆数，单位为veh；

为第k个采样时刻路网的内部车辆数，单位为veh。in,

is the optimal cumulative number of vehicles in the road network, in veh;

is the number of vehicles in the road network at the kth sampling moment, in veh.

The boundary control idea of this patent is to dynamically balance the traffic load of each sub-area with the overall road network, namely:

为子区i的最优累计车辆数。Where m is the number of sub-areas;

is the optimal cumulative number of vehicles in sub-area i.

的控制输入

，从而保证受控子区处于稳定运行状态，各个子区交通负荷动态平衡于整体路网。S403: Iterative calculation is performed on the iterative learning control model. In the first iteration, the original signal timing scheme of the road network sub-area is used. From the second iteration onwards, the input variable of the current system is compared with the cumulative number of vehicles collected in the previous time. The iterative learning control rate obtained above is used. After multiple iterations of learning control, the cumulative number of vehicles in each sub-area is calculated to follow the expected output.

Control input

, thereby ensuring that the controlled sub-areas are in a stable operating state and the traffic load of each sub-area is dynamically balanced with the overall road network.

The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims (6)

1. An iterative learning boundary control method for dynamically balancing the traffic load of multiple sub-areas in an urban area, characterized by comprising the following steps: S100: Select the urban road network area to be studied, collect road network data and analyze it; S200: Divide the traffic sub-areas within the selected urban road network, and establish a macro basic map of each control sub-area based on the collected road network data; S300: Analyze the traffic flow in and out of the sub-area boundary and establish a sub-area traffic flow model; S400: Implement boundary control on each sub-area, build an iterative learning control model according to the repetitive characteristics of urban traffic flow, obtain an iterative control scheme and set a signal timing strategy. 2. The iterative learning boundary control method for dynamically balancing urban multi-sub-area traffic loads according to claim 1 is characterized in that: the specific steps of S100 include: S101: Obtain the road topology structure of the urban traffic road network to be studied, and set up data collectors at the entrance and exit locations of each signalized intersection; S102: The actual traffic flow data, road section length and number of lanes, red light conditions, pedestrian crossings and parking lots of each intersection within the road network are obtained through the set data collector and Baidu map, and the traffic flow characteristics, path flow, average travel time, lane branching and turning ratio of the road network are analyzed. 3. The iterative learning boundary control method for dynamically balancing urban multi-sub-area traffic loads according to claim 2 is characterized in that: the specific step of dividing the traffic sub-areas in S200 includes: S201: Calculate the path association between adjacent intersections in the road network:

in,

is the traffic correlation between adjacent intersections i and j; T is the average travel time of vehicles between intersections i and j, in min;

is the unbalanced coefficient of the road section flow; m is the number of associated flow directions from the upstream intersection;

is the total volume of vehicles arriving at the downstream intersection;

is the maximum traffic volume at the upstream intersection, that is

The maximum value of S202: Calculate the path correlation of all adjacent intersections within the road network to be studied through the road network data collected by S100, calculate the road network density Laplace matrix and the eigenvalues corresponding to the eigenvectors, list the eigenvectors as a matrix as the algorithm input for fast global K-means clustering, select the sample point that minimizes the square error criterion function as the best clustering center of the cluster, save the clustering results, divide the intersections in a cluster into a sub-area, and merge the non-clustered intersections into adjacent sub-areas through boundary adjustment, so as to complete the urban road network sub-area division. 4. The iterative learning boundary control method for dynamically balancing urban multi-sub-area traffic loads according to claim 2 is characterized in that: the step of establishing a macro basic map of each control sub-area in S200 includes collecting the cumulative number of vehicles in each sub-area within a unit time interval and the number of vehicles leaving the road network through a data collector, and performing data processing to remove data with obvious errors and redundant data, draw an MFD scatter plot, perform curve fitting based on the scatter data, and obtain a macro basic map of each traffic sub-area and its corresponding critical cumulative number of vehicles. 5. The iterative learning boundary control method for dynamically balancing urban multi-sub-district traffic loads according to claim 1 is characterized in that: the sub-district traffic flow model in S300 is:

in,

is the number of vehicles in sub-area i at the kth sampling moment, in veh;

is the control period, in seconds;

is the flow rate input from the adjacent sub-area m to the sub-area i at the k-th sampling moment, in veh/h;

is the flow rate input from the road network boundary to sub-area i at the kth sampling moment, in veh/h;

is the flow rate output from sub-area i to sub-area m at the kth sampling moment, in veh/h;

is the flow rate output from sub-area i to the road network boundary at the kth sampling time, in veh/h;

It is the flow rate generated or reduced by the traffic generation source or dissipation source within sub-area i, in veh/h. 6. The iterative learning boundary control method for dynamically balancing urban multi-sub-area traffic loads according to claim 1 is characterized in that: the specific process of S400 is: S401: According to the principle of conservation of vehicle number, a state space expression of the road network is established, and iterative learning control is introduced into the sub-area traffic flow model to obtain an iterative learning control model. In the iterative learning control model, the state equation of the sub-area at the nth iteration is:

in,

Represents the state variables of the system;

Represents the input variables of the system;

Represents the output variable of the system;

is the identity matrix;

is the input matrix;

is the output matrix;

is the state perturbation matrix; The tracking error of the system is:

in,

Represents the expected output of the system;

represents the output disturbance of the system; S402: setting an iterative learning control rate for the divided sub-areas. The sub-areas and the outer boundary are open-closed loop PD type, denoted as 1, and the inner boundary of the sub-area is P type, denoted as 2:

in,

The control input saturation function for the nth iteration;

is the differential learning gain;

,

is proportional feedback learning gain;

is the system tracking error at the nth iteration; S403: Iterative calculation is performed on the iterative learning control model. In the first iteration, the original signal timing scheme of the road network sub-area is used. From the second iteration onwards, the input variable of the current system is compared with the cumulative number of vehicles collected in the previous time. The iterative learning control rate obtained above is used. After multiple iterations of learning control, the cumulative number of vehicles in each sub-area is calculated to follow the expected output.

Control input

, thereby ensuring that the controlled sub-areas are in a stable operating state and the traffic load of each sub-area is dynamically balanced with the overall road network.

Images