A new dynamic sub-area division method for urban road network

Abstract

The application relates to a city traffic road network division method, in particular to a novel city road network dynamic sub-area division method. The method comprises the following steps: S100, selecting a traffic road network to be researched, collecting and analyzing road network data; S200, calculating the correlation degrees of all adjacent intersections in the selected researched road network according to the collected road network data and a road network topological structure, and obtaining a distance matrix between the intersections according to the correlation degrees of all the adjacent intersections; S300, dividing the intersections in the road network, determining core intersections, adjacent intersections and noise intersections, and clustering the intersections to form clusters; and S400, repeatedly merging the clusters separated from S300 based on similarity, and completing city road network dynamic sub-area division. The application solves the problem that a traditional clustering method is difficult to remove outliers or has requirements for a data set shape and influences clustering precision.

Description

Technical Field

[0001] This invention relates to a method for dividing urban traffic networks, specifically a novel method for dynamic sub-region division of urban road networks. Background Technology

[0002] With the continuous improvement of people's living standards, the proportion of private cars in daily travel is rapidly increasing, and more and more cities are facing increasingly serious traffic congestion problems. Limited by city size and road boundaries, traffic management departments find it difficult to meet the ever-growing traffic demand by increasing road facilities or expanding the road network area. Improving the operational efficiency of the urban road network based on existing traffic facilities through traffic management and control methods is a good option. Traffic control sub-zones can divide the entire road network into multiple independent small areas when traffic management departments are implementing traffic control on a large-scale urban road network. In each small area, the optimal control strategy is determined based on its own traffic flow characteristics, thus optimizing signal timing.

[0003] Current research on sub-region division is abundant. Early studies primarily focused on statically dividing the road network for a given time period based on traffic organization channelization. However, with the increase in vehicle ownership, research has shifted from static to dynamic approaches. By tracking the control sub-region division results of the road network at the previous moment, algorithms are used at the next moment to identify congested road segments within the sub-regions and perform re-division. In this new approach, the division target is only a few highly heterogeneous road segments, resulting in lower computational complexity compared to the previous method of statically re-dividing the road network at every moment.

[0004] Existing methods for sub-zone division still have many shortcomings. Due to the complexity and diversity of road network environments and varying road conditions, outlier data frequently occurs. Some methods fail to account for outliers, potentially leading to inaccurate sub-zone division. Other methods rely solely on data during sub-zone division, ignoring the actual situation and assigning distant intersections to the same sub-zone, resulting in significant errors. Ignoring these issues can lead to unreasonable sub-zone division of the road network, thereby affecting the coordinated control of the urban road network and exacerbating congestion. Summary of the Invention

[0005] To address the problems existing in current sub-region division methods, this invention provides a novel dynamic sub-region division method for urban road networks.

[0006] The present invention adopts the following technical solution: a novel method for dynamic sub-region division of urban road networks, comprising the following steps: S100: selecting the traffic road network to be studied and collecting and analyzing road network data; S200: calculating the correlation degree of all adjacent intersections within the selected road network based on the collected road network data and road network topology, and obtaining the distance matrix between intersections based on the correlation degree of all adjacent intersections; S300: dividing the intersections within the road network, determining core intersections, adjacent intersections, and noisy intersections, and clustering them to form clusters; S400: repeatedly merging the clusters separated in S300 based on similarity to complete the dynamic sub-region division of the urban road network.

[0007] In step S100, the road network data collection includes: S101: obtaining the road topology of the urban traffic network to be studied; S102: obtaining the actual traffic flow data of each intersection in the road network, the length of the road segment and the number of lanes, and analyzing the traffic flow characteristics, path flow, average travel time and turning ratio of the road network.

[0008] In step S200, the traffic connectivity between adjacent intersections is calculated as follows:

[0009]

[0010]

[0011] in, denoted as the traffic correlation degree between adjacent intersections i and j; T represents the average travel time of vehicles between intersections i and j, in minutes. is the traffic flow imbalance coefficient for the road segment; m is the number of associated flow directions from the upstream intersection; This represents the total traffic flow reaching the downstream intersection. The maximum traffic volume at the upstream intersection, i.e. The maximum value; calculate the correlation degree between each adjacent intersection, and denot its value as . Then the correlation matrix of each adjacent intersection is: .

[0012] In step S300, the intersections within the studied road network are defined as... Through a set of neighborhoods ( , This can be used to describe the intensiveness of traffic data collection. Describes the selected intersection The neighborhood distance threshold of an element The distance between sample points is described as The threshold for the number of samples in the neighborhood, i.e., the number of samples at the intersection. With center at and radius at, The number of other intersections within the area;

[0013] Core intersection: When the intersection of The neighborhood contains at least Sample, i.e. ,but As the core intersection;

[0014] Adjacent to intersection: When the intersection of The neighborhood contains fewer than However, within its neighborhood at other core intersections, It is an adjacent intersection;

[0015] Noisy intersections: Points at non-core intersections and adjacent intersections.

[0016] The clustering process in step S300 includes,

[0017] S301: Use the correlation matrix of adjacent intersections as input to the clustering algorithm, in order to determine... Then, find all the core intersections that meet the conditions, and use them to determine the neighboring intersections in the neighborhood of the core intersections and the noisy intersections that are not in the neighborhood;

[0018] S302: Classify all adjacent intersections, if the reachable distance kd from an adjacent intersection to its core intersection is less than or equal to the distance kd from the intersection to the core intersection. If these data points are classified into the same category, then in the urban traffic network, the intersections represented by these data points are roughly divided into the same sub-cluster.

[0019] S303: If the reachable distance kd> Then determine the core distance hd: hd > This is a noisy intersection; hd < This creates new clusters, ultimately dividing the data into suitable subclusters to complete the intersection data processing.

[0020] Among them, for intersections within the road network In the given and After that, The minimum neighborhood radius that becomes the core intersection is called Core distance hd; core intersections within the road network and its adjacent intersection The larger of the Euclidean distance d and the core distance hd is called the intersection. to the intersection Let kd be the reachable distance.

[0021] In step S400, the step of dividing the entire road network into appropriate traffic sub-zones includes:

[0022] S401: Calculate the interconnectivity and similarity of all adjacent clusters obtained in step S300, and calculate the similarity of adjacent clusters using the metric formula;

[0023] Interconnectivity: Considering the distance between two clusters and the distance between their internal elements, it is quantified using relative interconnectivity RI.

[0024]

[0025] in, This indicates that cluster C is divided into two clusters. and The weight of the cut edge; Indicates will The weight of the cut edge divided into two approximately equal parts;

[0026] Approximation: Considering the similarity of features between two clusters, quantification is achieved using the relative similarity (RC).

[0027]

[0028] in, This indicates that cluster C is divided into two sub-clusters. and Average weight of the cut edge; Indicates will The average weight of the cut edge divided into two approximately equal parts;

[0029] Measurement formula: *

[0030] Here, α is a parameter used to adjust the weight of the two parameters. α>1 emphasizes relative approximation, while α<1 emphasizes relative interconnectivity.

[0031] S402: Compare with the merging threshold to determine whether to merge two clusters. Select the two clusters that maximize the above formula and merge them. Repeat this process until there are no clusters that meet the conditions, thereby obtaining high-quality clusters of arbitrary shapes, which is the final division result of the dynamic sub-regions of the road network.

[0032] Compared with existing technologies, this invention collects road network data and calculates the correlation between adjacent intersections. It then uses a density-based clustering algorithm to classify each intersection and provide simple aggregation, providing a minimal cluster set of data points for the Chameleon algorithm. Furthermore, it iteratively merges these sub-clusters based on similarity through agglomerative hierarchical clustering, ultimately obtaining high-quality traffic sub-regions of arbitrary form. This invention considers the complex and variable traffic conditions in real urban road networks, resulting in abnormal vehicle movement. These abnormalities are merged or removed during sub-region division. It also combines the Chameleon algorithm's ability to discover high-quality, arbitrary-form clusters with the OPTICS algorithm's ability to detect noise points and merge core points, solving the problems of traditional clustering methods struggling to remove outliers or having requirements on dataset shape that affect clustering accuracy. Attached Figure Description

[0033] Figure 1 A flowchart illustrating a novel dynamic sub-region division method for urban road networks provided by this invention;

[0034] Figure 2 A schematic diagram of traffic flow direction at adjacent intersections in a novel dynamic sub-region division method for urban road networks provided by the present invention;

[0035] Figure 3 This is a schematic diagram illustrating the division process of a novel urban road network dynamic sub-region division method provided by the present invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described are only for explaining the invention and are not intended to limit the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort will fall within the scope of protection of this invention.

[0037] See attached document Figure 1 A novel method for dynamic sub-region division of urban road networks includes the following steps:

[0038] S100: Select the traffic network to be studied and collect and analyze the network data;

[0039] S200: Calculate the correlation degree of all adjacent intersections within the selected study road network by collecting road network data and road network topology, and obtain the distance matrix between intersections based on the correlation degree of all adjacent intersections;

[0040] S300: Divide the intersections within the road network, identify core intersections, adjacent intersections, and noisy intersections, and cluster them into clusters;

[0041] S400: The sub-clusters separated from S300 are repeatedly merged based on similarity to complete the dynamic sub-region division of the urban road network.

[0042] In step S100, road network data collection includes,

[0043] S101: Obtain the road topology of the urban traffic network to be studied, and design detectors at the entrance and exit points of each signalized intersection.

[0044] S102: Obtain actual traffic flow data, road segment length and number of lanes at each intersection within the road network, collect Baidu traffic index data for relevant road networks, and analyze the traffic flow characteristics, path flow, average travel time and turning ratio of the road network.

[0045] In step S200, the traffic connectivity between adjacent intersections is calculated as follows:

[0046]

[0047]

[0048] in, denoted as the traffic correlation degree between adjacent intersections i and j; T represents the average travel time of vehicles between intersections i and j, in minutes. is the traffic flow imbalance coefficient for the road segment; m is the number of associated flow directions from the upstream intersection; This represents the total traffic flow reaching the downstream intersection. The maximum traffic volume at the upstream intersection, i.e. The maximum value.

[0049] The traffic correlation degree of all adjacent intersections within the traffic network is calculated and plotted based on the collected data, resulting in a distance matrix between the intersections.

[0050] Calculate the degree of correlation between adjacent intersections, and denote its value as . Then the correlation matrix of each adjacent intersection is: .

[0051] In step S300, the intersections within the studied road network are defined as... Through a set of neighborhoods ( , This can be used to describe the level of detail in the collection of traffic data. Describes the selected intersection The neighborhood distance threshold of an element The distance between sample points is described as The threshold for the number of samples in the neighborhood, i.e., the number of samples at the intersection. With center at and radius at, The number of other intersections within the range.

[0052] Core intersection: When the intersection of The neighborhood contains at least One sample, i.e. ,but It is the core intersection.

[0053] Adjacent to intersection: When the intersection of The neighborhood contains fewer than However, within its neighborhood at other core intersections, It is an adjacent intersection.

[0054] Noisy intersections: Points at non-core intersections and adjacent intersections.

[0055] The clustering process in step S300 includes,

[0056] S301: Use the correlation matrix of adjacent intersections as input to the clustering algorithm, in order to determine... Then, all core intersections that meet the criteria are found, and neighboring intersections within the core intersection's neighborhood and noisy intersections not within the neighborhood are identified.

[0057] S302: Classify all adjacent intersections, if the reachable distance kd from an adjacent intersection to its core intersection is less than or equal to the distance kd from the intersection to the core intersection. If these data points are classified into the same category, then in the context of urban traffic networks, the intersections represented by these data points are roughly divided into the same sub-cluster.

[0058] S303: If the reachable distance kd> Then determine the core distance hd: hd > This is a noisy intersection; hd < This creates new clusters, ultimately dividing the data into suitable subclusters to complete the intersection data processing.

[0059] Among them, for intersections within the road network In the given and After that, The minimum neighborhood radius that becomes the core intersection is called Core distance hd; core intersections within the road network and its adjacent intersection The larger of the Euclidean distance d and the core distance hd is called the intersection. to the intersection Let kd be the reachable distance. A brief analysis shows that the reachable distance from adjacent intersections within the core distance to the core intersection is a fixed value hd, while the reachable distance from adjacent intersections outside the core distance to the core intersection is the Euclidean distance d.

[0060] In step S400, the Chameleon algorithm is used to divide the entire road network into appropriate traffic sub-regions. The steps are as follows:

[0061] First, define the interconnectivity and approximation between clusters:

[0062] Interconnectivity: Considering the distance between two subclusters and the distance between their internal elements, it is quantified using relative interconnectivity RI:

[0063]

[0064] in, This indicates that cluster C is divided into two sub-clusters. and The weight of the cut edge; Indicates will The weight of the cut edge divided into two roughly equal parts.

[0065] Approximation: Considering the similarity of features between two subclusters, it is quantified using relative approximation (RC):

[0066]

[0067] in, This indicates that cluster C is divided into two sub-clusters. and Average weight of the cut edge; Indicates will The average weight of the cut edge divided into two roughly equal parts.

[0068] Define the measurement formula: *

[0069] Here, α is a parameter used to adjust the weight of the two parameters. α>1 emphasizes relative approximation, while α<1 emphasizes relative interconnectivity.

[0070] S401: Calculate the interconnectivity and similarity of all adjacent clusters obtained in step S300, and calculate the similarity of adjacent clusters using the metric formula.

[0071] S402: Compare with the merging threshold to determine whether to merge two clusters. Select the two clusters that maximize the above formula and merge them. Repeat this process until there are no clusters that meet the conditions, thereby obtaining high-quality clusters of arbitrary shapes, which is the final division result of the dynamic sub-regions of the road network.

[0072] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A novel method for dynamic sub-region division of urban road networks, characterized in that: Includes the following steps, S100: Select the traffic network to be studied, and collect and analyze the network data; S200: Calculate the correlation degree of all adjacent intersections within the selected study road network by collecting road network data and road network topology, and obtain the distance matrix between intersections based on the correlation degree of all adjacent intersections; S300: Divide the intersections within the road network, identify core intersections, adjacent intersections, and noisy intersections, and cluster them into clusters; In step S300, the intersections within the studied road network are defined as... Through a set of neighborhoods ( , This can be used to describe the intensiveness of traffic data collection. Describes the selected intersection The neighborhood distance threshold of an element The distance between sample points is described as The threshold for the number of samples in the neighborhood, i.e., the number of samples at the intersection. With center at and radius at, The number of other intersections within the area; Core intersection: When the intersection of The neighborhood contains at least Sample, i.e. ,but As the core intersection; Adjacent to intersection: When the intersection of The neighborhood contains fewer than However, within its neighborhood at other core intersections, It is an adjacent intersection; Noisy intersections: Points at non-core intersections and adjacent intersections; The clustering process in step S300 includes, S301: Use the correlation matrix of adjacent intersections as input to the clustering algorithm, in order to determine... Then, find all the core intersections that meet the conditions, and use them to determine the neighboring intersections in the neighborhood of the core intersections and the noisy intersections that are not in the neighborhood; S302: Classify all adjacent intersections, if the reachable distance kd from an adjacent intersection to its core intersection is less than or equal to the maximum distance kd. If these data points are classified into the same category, then in the urban traffic network, the intersections represented by these data points are roughly divided into the same sub-cluster. S303: If the reachable distance kd> Then determine the core distance hd: hd > This is a noisy intersection; hd < This creates new clusters, ultimately dividing the data into suitable subclusters to complete the intersection data processing. Among them, for intersections within the road network In the given and After that, The minimum neighborhood radius that becomes the core intersection is called The core distance hd; the larger of the Euclidean distance d between a core intersection and its adjacent intersections within the road network and the core distance hd is called the intersection. to the intersection The reachable distance is kd; S400: The clusters separated from S300 are repeatedly merged based on similarity to complete the dynamic sub-region division of the urban road network.

2. The novel urban road network dynamic sub-region division method according to claim 1, characterized in that: In step S100, road network data collection includes, S101: Obtain the road topology of the urban traffic network to be studied; S102: Obtain actual traffic flow data, road segment length and number of lanes at each intersection within the road network, and analyze the traffic flow characteristics, path flow, average travel time and turning ratio of the road network.

3. The novel urban road network dynamic sub-region division method according to claim 1, characterized in that: In step S200, the traffic correlation degree between adjacent intersections is calculated as follows: in, denoted as the traffic correlation degree between adjacent intersections i and j; T represents the average travel time of vehicles between intersections i and j, in minutes. is the traffic flow imbalance coefficient for the road segment; m is the number of associated flow directions from the upstream intersection; This represents the total traffic flow reaching the downstream intersection. The maximum traffic volume at the upstream intersection, i.e. The maximum value; calculate the correlation degree between each adjacent intersection, and denot its value as . Then the correlation matrix of each adjacent intersection is: .

4. The novel urban road network dynamic sub-region division method according to claim 1, characterized in that: Step S400 includes: S401: Calculate the interconnectivity and similarity of all adjacent clusters obtained in step S300, and calculate the similarity of adjacent clusters using the metric formula; Interconnectivity: Considering the distance between two clusters and the distance between their internal elements, it is quantified using relative interconnectivity RI: in, This indicates that cluster C is divided into two clusters. and The weight of the cut edge; Indicates will The weight of the cut edge that divides into two equal parts; Approximation: Considering the similarity of features between two clusters, quantification is achieved using the relative similarity (RC). in, This indicates that cluster C is divided into two sub-clusters. and Average weight of the cut edge; Indicates will The average weight of the cut edge divided into two equal parts; Measurement formula: Here, α is a parameter used to adjust the weight of the two parameters. α>1 emphasizes relative approximation, while α<1 emphasizes relative interconnectivity. S402: Compare with the merging threshold to determine whether to merge two clusters. Select the two clusters that maximize the above formula and merge them. Repeat this process until there are no clusters that meet the conditions, thereby obtaining high-quality clusters of arbitrary shapes, which is the final division result of the dynamic sub-regions of the road network.

Images