Expressway emergency rescue point site selection method based on data driving

Abstract

The invention relates to the field of expressway network emergency rescue, in particular to a data-driven expressway emergency rescue site selection method, which comprises the following steps of: rasterizing a target range; matching the vehicle toll station OD data of various vehicle types into the target range grid to generate an expressway network; carrying out community division on the expressway network according to different vehicle types; calculating the community centroid of each community of each vehicle type; calculating the vehicle model weight of each vehicle model for the highway traffic accident; dividing the community centroids corresponding to the various vehicle types into corresponding centroid groups, and performing weighted calculation on all community centroids in each centroid group according to the vehicle type weights of the various vehicle types; and matching the nearest toll station for the weighted centroid of each community as an emergency rescue point. According to the method, reasonable emergency rescue points can be selected according to the operation space-time distribution characteristics of different vehicle types in the expressway and the influence weights of the different vehicle types on the emergency rescue node layout.

Description

Technical Field

[0001] The invention relates to the field of highway network emergency rescue, and in particular to a method for selecting a highway emergency rescue point based on data-driven. Background Art

[0002] As the scale of highways continues to grow, the frequency and impact of vehicle breakdowns and traffic accidents on highways are also increasing. Road traffic accidents often cause property losses and even endanger the lives of road users. Effectively preventing and controlling road traffic operation risks and ensuring road traffic safety have always been complex issues that plague many countries. Traffic accidents threaten people's safety, and improper emergency response will increase rescue time, cause secondary damage to the accident site, waste resources, and hinder sustainable development. Therefore, the demand for emergency rescue points on highways is also increasing, so the location of emergency rescue points on highways has become the key to the safe and efficient operation of highways. Highways are designed to have high speeds, large traffic volumes, and complex traffic flow compositions, with high probability of traffic accidents and high accident severity. Due to the sudden nature of highway traffic accidents, emergency rescue on highways is difficult.

[0003] Highway emergency rescue points refer to emergency service places that are set up to reduce casualties, property losses and restore traffic order when a traffic accident or vehicle breakdown occurs on a highway. They are usually located at the entrances and exits of highways and are equipped with corresponding personnel, vehicles and facilities and equipment. When a traffic accident or vehicle breakdown occurs, the incident needs to be handled quickly and the traffic order on the scene needs to be restored as soon as possible, so emergency rescue teams are required to arrive at the scene immediately. The location of highway emergency points is a key factor in ensuring the efficiency of highway emergency rescue, so the location and layout of highway emergency rescue centers are of great significance.

[0004] With the development of the highway network, the density of the high-speed network is increasing day by day. Relying solely on the emergency needs of highways to establish a site selection model can no longer meet the increasingly complex highway network. The problem of site selection for highway emergency rescue points belongs to the problem of facility site selection. Many domestic and foreign scholars have done a lot of relevant research on the problem of facility site selection. In terms of data-driven facility site selection, many scholars obtain the distribution of facility demand through data-driven methods and conduct facility site selection based on the demand distribution. Lage, MD et al. calculated the spatial distribution of potential travel demand through the Geographic Information System (GIS) and used two hierarchical facility location models to determine the ideal site location. Shen obtained the demand points for the recycling of construction waste in the urban area based on the characteristics of the generation end and the demand end in the new urban area and studied the optimal site selection of centralized and decentralized treatment facilities for construction waste. Li used data-driven behavior analysis to determine functional areas, such as work and living areas, providing a basis for the government to carry out urban design and construction and the site selection of service facilities. Lan used land use information and public transport network datasets to construct a geospatial graph convolutional network based on the public transport system to predict the attractiveness of different store locations in the community, thus selecting the store location. Castillo-Neyra, Ricardo used the Poisson regression model to quantify the participation probability based on the walking distance to the nearest vaccination site, regression-fitted the survey data, and then used the computational recursive exchange technique to solve the facility location problem to find the best location of a set of fixed vaccination sites, extracting people's travel patterns and activities, referring to a large number of high-resolution, anonymous GPS traces and visited points of interest (e.g., restaurants and grocery stores).

[0005] However, there are still the following problems in the current research on the problem of site selection for highway emergency rescue points: 1) In terms of the site selection method for emergency rescue facilities, it is generally carried out based on optimization models, and there is less research on the application of data-driven in site selection. 2) The converted traffic volume is generally used as the decision variable. In fact, the accident probability and accident intensity caused by different vehicle types have different degrees of influence on the layout of emergency rescue facilities from the degree of vehicle type conversion in traffic volume, and there is a lack of refined characterization of the influence mechanism of vehicle types on the layout of emergency rescue facilities. Summary of the Invention

[0006] Aiming at the deficiencies of the above-mentioned existing technologies, the technical problem to be solved by the present invention is: how to provide a data-driven site selection method for highway emergency rescue points, which can select reasonable emergency rescue points according to the operation spatio-temporal distribution characteristics of different vehicle types in highways and the influence weights of different vehicle types on the layout of emergency rescue nodes, so as to ensure the practicability and accuracy of the site selection of highway emergency rescue points.

[0007] To solve the above technical problems, the present invention adopts the following technical solutions:

[0008] A data-driven method for locating highway emergency rescue points, including:

[0009] S1: Grid the target range to generate target range grids;

[0010] S2: Match the vehicle toll station OD data of various vehicle types to the target range grids to generate a highway network;

[0011] S3: Conduct community division on the highway network for different vehicle types to generate community division results corresponding to various vehicle types;

[0012] S4: Calculate the community centroid of each community in the community division results corresponding to each vehicle type to generate community centroids corresponding to various vehicle types;

[0013] S5: Calculate the vehicle type weights of each vehicle type for highway traffic accidents;

[0014] S6: Divide the community centroids corresponding to various vehicle types into corresponding centroid groups, and then perform weighted calculation on all community centroids in each centroid group according to the vehicle type weights of various vehicle types to generate the community weighted centroid of the centroid group;

[0015] S7: Match the nearest toll station to each community weighted centroid as its corresponding emergency rescue point.

[0016] Preferably, the target range grids are generated through the following steps:

[0017] S101: Calculate the central coordinates (lon, lat) of the target range according to the longitude and latitude (lon1, lat1) at the lower left corner and the longitude and latitude (lon2, lat2) at the upper right corner of the target range;

[0018] The formula description is:

[0019]

[0020] S102: Calculate the longitude and latitude (Δlon, Δlat) of each grid in the target range according to the earth radius R at the target range and the set grid length a;

[0021] The formula description is:

[0022]

[0023]

[0024] S103: Use the lower left corner of the target range as the center point of the first grid to fill the entire target range with grids, and number each grid with column numbers and row numbers (loncol, latcol);

[0025] S104: Calculate the central coordinates (hblon, hblon) of each grid;

[0026] The formula description is as follows:

[0027] hblon = loncol·Δlon + lon1;

[0028] hblat = latcol·Δlat + lat1;

[0029] S105: Calculate the coordinates of the four vertices of each grid: One vertex is deduced for each grid, and the other three vertices are deduced from the surrounding three grids;

[0030] For any grid, starting from the bottom - left vertex, the coordinates of the four vertices in the counter - clockwise direction are:

[0031]

[0032]

[0033]

[0034]

[0035] In the formula: The subscripts i and j represent the row number and column number of the grid respectively.

[0036] Preferably, use longitude and latitude to match the OD data of vehicle toll stations of various vehicle types to the target - range grids;

[0037] The formula for obtaining the corresponding grid is as follows:

[0038]

[0039]

[0040] In the formula: x and y represent the longitude and latitude of the toll stations in the OD data of vehicle toll stations respectively.

[0041] Preferably, perform community division on the highway network through the Fastunfolding algorithm;

[0042] The specific steps are as follows:

[0043] S301: Initialize, divide each node in the highway network into different communities;

[0044] S302: For each node, it is planned to divide it into the community where its adjacent nodes are located, and calculate the change value of modularity before and after the division: If the change value of modularity is zero, then abandon this division; otherwise, accept this division;

[0045] 1) The calculation formula for modularity is as follows:

[0046]

[0047] Where: M represents the modularity of the highway network; w ij represents the weight of the edge between v i and v j ; k i represents the sum of the weights of the edges connected to the node v i ; c i represents the community assigned to the node i;

[0048] When μ = v, δ(μ, v) = 1; in other cases, δ(μ, v) = 0;

[0049] 2) The calculation formula for the change value of modularity is as follows:

[0050] ΔM = M f - M b ;

[0051] Where: ΔM represents the change in modularity M before and after the partition, that is, the change value of modularity; M f , M b respectively represent the modularity M before and after the change;

[0052] S303: Repeat step S302 until all nodes have been visited and no update occurs, and generate the corresponding community structure;

[0053] S304: Reconstruct the highway network according to the community structure generated in step S303, aggregate all nodes within the same community together to form a new node;

[0054] S305: Repeat steps S302 to S304 until the structure of the highway network no longer changes, and generate the corresponding community partition result.

[0055] Preferably, calculate the centroid of each community as its community centroid;

[0056] The formula description is as follows:

[0057]

[0058]

[0059] Where: C x , C y respectively represent the abscissa and ordinate of the community centroid; X i , Y i respectively represent the abscissa and ordinate of the nodes within the community; n represents the number of nodes within the community.

[0060] Preferably, calculate the corresponding weights of each vehicle type for the probability and intensity of highway traffic accidents as the vehicle type weights;

[0061] The specific steps are as follows:

[0062] S501: Construct an initial sample matrix R = (r ij ) 2×4 , r ij ≠0 (i = 1, 2, j = 1, 2, 3, 4), where i represents two evaluation indicators of traffic accident probability or accident intensity, and j represents four vehicle types;

[0063] S502: Standardize the initial sample matrix R to generate a standard sample matrix D = (d ij ) 2×4 , where d ij is the standardized sample value;

[0064] The formula description is:

[0065]

[0066] S503: Calculate the proportion h ij of each vehicle type in each group of samples in the sample;

[0067] The formula description is:

[0068]

[0069] S504: Calculate the entropy E j of the j-th vehicle type;

[0070] The formula description is:

[0071]

[0072] S505: Calculate its entropy weight w j according to the entropy E j of the j-th vehicle type, that is, the vehicle type weight;

[0073] The formula description is:

[0074]

[0075] In the formula: w j represents the entropy weight of the j-th vehicle type, that is, the vehicle type weight; E j represents the entropy of the j-th vehicle type.

[0076] Preferably, generate a community weighted centroid through the following steps:

[0077] S601: Define the community division results of four vehicle types as A, B, C, and D;

[0078] The formula description is:

[0079] A = {A1, A2,... A k};

[0080] B = {B1, B2,... B l};

[0081] C = {C1, C2,... C m};

[0082] D = {D1, D2,... D n};

[0083] In the formula: k, l, m, n respectively represent the number of communities in the corresponding community division results;

[0084] S602: Select the community division result F ∈ (A, B, C, D) with the least number of communities, where f = min(k, l, m, n);

[0085] S603: Define the set of community centroids of F as W, and the sets of community centroids of the community division results of the other three vehicle types as X, Y, and Z respectively:

[0086] The formula description is:

[0087] W = {W1, W2, W3... W k};

[0088] X = {X1, X2, X3... X l};

[0089] Y = {Y1, Y2, Y3... Y m};

[0090] Z = {Z1, Z2, Z3... Z n};

[0091] where k < l / m / n;

[0092] In the formula: W i , X i , Y i , Z i respectively represent the community centroids in W, X, Y, and Z; k, l, m, n respectively represent the number of community centroids;

[0093] S604: Construct k centroid groups with each community centroid W i in W as the benchmark;

[0094] S605: Calculate the centroid distances between the community centroids in X, Y, and Z and the community centroid in W respectively through the KNN algorithm, and correspondingly divide the community centroids in X, Y, and Z into the centroid groups corresponding to the community centroids with the closest centroid distances in W. Finally, all the community centroids in X, Y, and Z are correspondingly divided into k centroid groups;

[0095] The calculation formula for the centroid distance is:

[0096]

[0097] In the formula: (x i , y i ) represent the coordinates of the community centroids in X, Y, and Z, represents the coordinate of the community centroid in W;

[0098] S606: Calculate the community weighted centroid of each centroid group according to the position coordinates of all the community centroids in the centroid group and the vehicle type weights, and finally obtain k community weighted centroids;

[0099] The calculation formula for the community weighted centroid is:

[0100]

[0101]

[0102] In the formula: C wx , C wy respectively represent the abscissa and ordinate of the community weighted centroid; ω i represents the vehicle type weight corresponding to the community centroid with the coordinate of ; n represents the number of community centroids in the centroid group.

[0103] Preferably, match the nearest toll station for the community weighted centroid through the KDTree algorithm;

[0104] The specific steps are as follows:

[0105] S701: Determine whether the community weighted centroid P(k) is less than the set threshold m: If so, visit the left subtree; otherwise, visit the right subtree;

[0106] where m represents the middle distance value in the distance set between the toll stations and the community weighted centroid in the same dimension;

[0107] If the distance between a certain toll station and the community weighted centroid is less than m, divide this station into the left subtree;

[0108] If the distance between a certain toll station and the community weighted centroid is greater than or equal to m, divide this station into the right subtree;

[0109] S702: Starting from the root node, visit each node along the search path in sequence until reaching the leaf node Q. At this time, the leaf node Q is the nearest node to P(k), and the distance between Q and P(k) is d;

[0110] S703: Starting from the nearest node, backtrack along the search path to the root node. At the same time, calculate the distance d' between each node on the search path and P(k). If there exists a node Q' with a distance d' less than d, then take the node Q' as the new nearest node;

[0111] S704: Include the children nodes of the new nearest node in the search scope and return to step S702;

[0112] S705: Repeat steps S702 to S704 until all search paths are empty. Take the latest nearest node as the toll station nearest to the community weighted centroid P(k), that is, the emergency rescue point.

[0113] Preferably, after determining the emergency rescue points R i (i = 1, 2... m), first take each toll station as the rescue demand point H i (i = 1, 2... n); then calculate the average rescue time of all rescue demand points and determine whether the average rescue time is greater than the set maximum emergency rescue time: if so, return to step S3, further divide the communities in the community division result with the least number of communities to obtain a new community division result for various vehicle types, and then continue to execute steps S4 to S7 and add the newly generated emergency rescue points to R i (i = 1, 2... m); otherwise, output the emergency rescue point R i (i = 1, 2... m).

[0114] Preferably, calculate the average rescue time through the following steps:

[0115] 1) Calculate the distance S between each rescue demand point H i (i = 1, 2... n) and the nearest emergency rescue point R i (i = 1, 2... m); i (i = 1, 2... n);

[0116] 2) Calculate the rescue time of each rescue demand point

[0117]

[0118] In the formula: v represents the driving speed of the emergency rescue vehicle;

[0119] 3) Calculate the average rescue time of all rescue demand points

[0120]

[0121] In the formula: n represents the number of rescue demand points.

[0122] Compared with the prior art, the method for selecting the location of highway emergency rescue points based on data driving in the present invention has the following beneficial effects:

[0123] Based on the vehicle detection data of highway toll stations, the present invention deeply mines the traffic operation data of highways and proposes a method for selecting the location of highway emergency rescue points driven by data. First, the OD data of vehicle toll stations on the highway is divided by vehicle type and matched to the map grid. On this basis, a complex network is constructed and the data community discovery results are obtained respectively, the centroids of different vehicle type communities are obtained, and the KNN algorithm is used to merge the centroid points of the communities into the community division result with the fewest community groups. After determining the different weights of various vehicle types, the centroid points of the same partition are merged into a community centroid by using the weights, and the highway toll station closest to the community centroid point is matched as the location selection of the emergency rescue point. On the one hand, when selecting the location of highway emergency rescue points, it is necessary to obtain the vehicle operation distribution characteristics, because the vehicle operation distribution characteristics reflect the highway emergency rescue demand to a certain extent. Therefore, the present invention selects reasonable emergency rescue points based on the operation spatio-temporal distribution characteristics of different vehicle types on the highway, which is beneficial to ensuring the efficiency of highway emergency rescue, reducing the cost of highway emergency rescue, and thus ensuring the practicability of the location selection of highway emergency rescue points. On the other hand, the accident probabilities and accident intensities of different types of vehicles are different, and the demands for highway emergency rescue facilities are also different. Therefore, the present invention distinguishes different vehicle types and obtains the accident characteristics of different vehicle types, and then determines the influence weights of different vehicle types on the layout of emergency rescue nodes. Finally, after obtaining the operation spatio-temporal distribution characteristics of different vehicle types and the influence weights of different vehicle types on the layout of emergency rescue points, the operation spatio-temporal distribution characteristics of highway vehicles are merged and processed according to the weights of different vehicle types to determine the location selection of highway emergency rescue points, realizing the selection of reasonable emergency rescue points according to the operation spatio-temporal distribution characteristics of different vehicle types on the highway and the influence weights of different vehicle types on the layout of emergency rescue nodes, thus ensuring the accuracy of the location selection of highway emergency rescue points.

[0124] By rasterizing the target range, the present invention can utilize the more convenient advantages of raster data overlap and combination to better and more conveniently analyze the operation spatio-temporal distribution characteristics of different vehicle types in the OD data of vehicle toll stations, thereby assisting in improving the efficiency of subsequent location selection of highway emergency rescue points.

[0125] The present invention divides the highway network into communities for different vehicle types. Through community division, the locations within the research scope can be divided into different communities or groups according to their similarities and connections, which helps to enhance the closeness of communication among different regions, enables the optimal layout of rescue points within and between different communities, and further improves the accuracy of the location selection of highway emergency rescue points.

[0126] Considering that different vehicle types have different degrees of influence on traffic accidents, the present invention assigns corresponding weights to the accident probabilities and accident intensities of different vehicle types respectively, so as to determine the influence weights of different vehicle types on the layout of emergency rescue nodes, thereby ensuring the accuracy of the location selection of highway emergency rescue points.

[0127] The present invention divides the centroid of the community corresponding to each vehicle type into the corresponding centroid group, and then performs weighted calculation on all community centroids in each centroid group according to the vehicle type weight of each vehicle type to generate the community weighted centroid of the centroid group. Finally, based on the community weighted centroid, the emergency rescue point is matched, which can better realize the selection of reasonable emergency rescue points according to the operation spatio-temporal distribution characteristics of different vehicle types on the highway and the influence weights of different vehicle types on the layout of emergency rescue nodes.

[0128] The present invention matches the nearest toll station for each community weighted centroid through the KDTree algorithm, which can ensure the efficiency and accuracy of the nearest toll station matching.

[0129] After the present invention selects the nearest toll station matching each community weighted centroid as the emergency rescue point, it further judges whether the number and location of the emergency rescue points are reasonable according to the average rescue time of all rescue demand points (for example, when the community scope is large, selecting only one highway emergency rescue point in each community cannot meet the highway emergency demand), thereby further improving the practicality of the location selection of highway emergency rescue points. BRIEF DESCRIPTION OF THE DRAWINGS

[0130] In order to make the objectives, technical solutions and advantages of the invention more clear, the present invention will be further described in detail below with reference to the drawings, where:

[0131] Figure 1 is the overall framework diagram of the method for locating highway emergency rescue points based on data driving;

[0132] Figure 2 is the process framework diagram of highway emergency rescue zoning;

[0133] Figure 3 is the process framework diagram of the location selection of highway emergency rescue points;

[0134] Figure 4 is an example of a Voronoi diagram;

[0135] Figure 5 It is a process framework diagram for charging stations matched with the community weighted centroid. Specific implementation manners

[0136] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. The components of the embodiments of the present invention usually described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0137] It should be noted that like reference numerals and letters denote like items in the following drawings, and thus, once an item is defined in one drawing, it does not require further definition and explanation in subsequent drawings. In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the inventive product is customarily placed during use. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be construed as a limitation of the present invention. In addition, the terms "first", "second", "third", etc. are only used for distinguishing descriptions and cannot be construed as indicating or implying relative importance. In addition, terms such as "horizontal" and "vertical" do not mean that the components are required to be absolutely horizontal or hanging, but can be slightly inclined. For example, "horizontal" only means that its direction is more horizontal relative to "vertical", and does not mean that the structure must be completely horizontal, but can be slightly inclined. In the description of the present invention, it should also be noted that unless otherwise clearly specified and limited, the terms "set", "installed", "connected", "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, and can be the internal communication of two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0138] The following is a further detailed description through specific implementation manners:

[0139] Embodiment:

[0140] To establish the emergency rescue site selection covering the highway network, the following assumptions are made in this example:

[0141] 1) There will be no significant changes in the highway network in a short period of time.

[0142] 2) The OD distribution characteristics of vehicle toll stations on highways show stability characteristics.

[0143] 3) Under the condition of path accessibility, it is considered that the rescue vehicle can run smoothly during the rescue process without being interfered by the outside world.

[0144] 4) Due to its own characteristics, the community algorithm will consider that some highway toll stations with less traffic data cannot form a large community with other data, and this part of the data will not affect the site selection modeling.

[0145] 5) The accident probability and accident intensity distribution of different vehicle types on highways remain stable in a long period of time.

[0146] Based on the above assumptions, this embodiment discloses a data-driven method for selecting emergency rescue points on highways.

[0147] As Figure 1 shown, the data-driven method for selecting emergency rescue points on highways includes:

[0148] S1: rasterize the target range to generate target range grids;

[0149] S2: match the OD data of vehicle toll stations of various vehicle types to the target range grids to generate a highway network;

[0150] S3: conduct community division on the highway network for different vehicle types to generate community division results corresponding to various vehicle types; combined with Figure 2 shown, the highway emergency rescue zoning based on community discovery is the basis for establishing the emergency rescue point site selection model of the present invention. First, rasterize the research (target) range and map the highway toll data to the grids; then establish a highway traffic network and conduct community division on the highway toll data separately.

[0151] S4: calculate the community centroid of each community in the community division result corresponding to each vehicle type to generate community centroids corresponding to various vehicle types; combined with Figure 3As shown in the figure, based on the community division for each vehicle type, first determine the vehicle type weights of different vehicle types for highway traffic accident probability and accident severity. Then, use the centroid method to determine the centroid coordinates of each community, use the weights to merge the centroids in the same partition into a weighted centroid, and finally use the KDTree algorithm to obtain the toll station closest to the weighted centroid, and use this toll station as the location selection for the emergency rescue point.

[0152] S5: Calculate the vehicle type weights of each vehicle type for highway traffic accidents;

[0153] S6: Divide the community centroids corresponding to various vehicle types into corresponding centroid groups, and then perform weighted calculations on all community centroids in each centroid group according to the vehicle type weights of various vehicle types to generate the community weighted centroid of this centroid group;

[0154] S7: Match the closest toll station for each community weighted centroid as its corresponding emergency rescue point.

[0155] Based on the vehicle detection data of highway toll stations, this invention deeply mines the traffic operation data of highways and proposes a method for locating emergency rescue points on highways driven by data. First, divide the OD data of vehicle toll stations on the highway by vehicle type and match it to the map grid. On this basis, construct a complex network and obtain the data community discovery results respectively, obtain the centroid of different vehicle type communities, and use the KNN algorithm to merge the centroid points of the community into the community division result with the fewest community groups. After determining the different weights of various vehicle types, use the weights to merge the centroid points in the same partition into a community centroid, and match the highway toll station closest to the community centroid point as the location selection for the emergency rescue point. On the one hand, the location selection of highway emergency rescue points needs to obtain the vehicle operation distribution characteristics, because the vehicle operation distribution characteristics reflect the highway emergency rescue needs to a certain extent. Therefore, this invention selects a reasonable emergency rescue point based on the operation spatio-temporal distribution characteristics of different vehicle types on the highway, which is beneficial to ensuring the efficiency of highway emergency rescue, reducing the cost of highway emergency rescue, and thus ensuring the practicality of the location selection of highway emergency rescue points. On the other hand, the accident probability and accident severity of different types of vehicles are different, and the requirements for highway emergency rescue facilities are also different. Therefore, this invention distinguishes different vehicle types and obtains the accident characteristics of different vehicle types, and then determines the influence weights of different vehicle types on the layout of emergency rescue nodes. Finally, after obtaining the operation spatio-temporal distribution characteristics of different vehicle types and the influence weights of different vehicle types on the layout of emergency rescue points, merge and process the highway vehicle operation spatio-temporal distribution characteristics according to the weights of different vehicle types to determine the location selection of highway emergency rescue points, realizing the selection of a reasonable emergency rescue point based on the operation spatio-temporal distribution characteristics of different vehicle types on the highway and the influence weights of different vehicle types on the layout of emergency rescue nodes, thus ensuring the accuracy of the location selection of highway emergency rescue points.

[0156] In the specific implementation process, community discovery belongs to spatial statistical analysis. Raster data has incomparable advantages over vector data in terms of spatial analysis: the overlap and combination of raster data are more convenient, and using raster data makes various spatial analyses more convenient. The process of creating a raster for the research scope is called rasterization, that is, dividing a target scope (a region) into a series of discrete pixels or grid cells, and each cell has a specific attribute value. The rasterization of the research scope can be achieved through existing mature algorithms. The present invention generates a raster for the target scope through the following steps:

[0157] S101: Calculate the central coordinates (lon, lat) of the target scope according to the longitude and latitude (lon1, lat １ ) and the longitude and latitude (lon2, lat2) of the upper right corner of the target scope;

[0158] The formula description is:

[0159]

[0160] S102: Calculate the longitude and latitude (Δlon, Δlat) of each raster in the target scope according to the radius R of the earth at the target scope and the set raster length a; the circumference of the meridian is 2πR, and the circumference of the parallel is 2πR cos(lat)

[0161] The formula description is:

[0162]

[0163]

[0164] S103: Use the lower left corner of the target scope as the center point of the first raster to fill the entire target scope with rasters, and number each raster with column numbers and row numbers (loncol, latcol);

[0165] S104: Calculate the central coordinates (hblon, hblon) of each raster;

[0166] The formula description is:

[0167] hblon = loncol·Δlon + lon1;

[0168] hblon = latcol·Δlat + lat1;

[0169] S105: Calculate the coordinates of the four vertices of each grid: deduce one vertex for each grid, and deduce the other three vertices with the surrounding three grids;

[0170] For any grid, starting from the bottom-left vertex, the coordinates of the four vertices in the counterclockwise direction are as follows:

[0171]

[0172]

[0173]

[0174]

[0175] where the subscripts i and j represent the row number and column number of the grid, respectively.

[0176] In the specific implementation process, the OD data of vehicle toll stations of various vehicle types are matched to the grid in the target range by using longitude and latitude; the corresponding grid formula is as follows:

[0177]

[0178]

[0179] where x and y represent the longitude and latitude of the toll station in the OD data of the vehicle toll station, respectively.

[0180] By rasterizing the target range, the present invention can take advantage of the more convenient overlapping and combination of raster data to better and more conveniently analyze the operation spatio-temporal distribution characteristics of different vehicle types in the OD data of vehicle toll stations, thereby assisting in improving the efficiency of subsequent highway emergency rescue point location.

[0181] In the specific implementation process, a complex network describes the structure of a group of objects, and there is a certain "connection" between some objects in the network. A network consists of nodes (Nodes) and edges (Edges). The objects in the network are nodes, and the connections between nodes are the edges of the network. The highway network is a complex network. Specifically: the highway network is represented as G = {N, E}, where N = {N i |i ∈ {1, 2, 3... n}} is the node set of the highway network, and the nodes represent toll stations; E = {E ij |i, j ∈ {1, 2, 3... n}} is the edge set of the highway network, and the edges represent the paths between toll stations;

[0182] The weight set of E is ω = {ω ij , i ≠ j, i, j ∈ {1, 2, 3... n}}, where w ij represents the weight between edge v i and v j ;

[0183] In the weight set ω, the weight value ωij The calculation formula is as follows:

[0184]

[0185]

[0186] In the formula: q represents the traffic volume between node N i and N j ; M represents the number of passing vehicles between node N i and node N j within time T.

[0187] In the specific implementation process, the present invention uses community discovery to explore the travel characteristics of the complex expressway traffic network, providing a basis for the location selection of expressway emergency rescue points. Any complex network is composed of nodes and the connections between nodes. The cohesive subgraph formed by nodes in a complex network is called a sub-region, and any complex network is the union of several sub-regions. The traffic network is a complex network. The nodes of a complex network have the property of aggregation. The nodes that gather together form a community. The connection edge density between nodes within the same community is relatively high, while the connection edge density between nodes in different communities is relatively low. The process of revealing such community structure in a complex network is called community discovery.

[0188] Based on the fact that nodes and edges in the same community usually have similar characteristics, community discovery reflects the structural characteristics of the complex network well. The method for revealing such community structure of the complex network is called a community discovery algorithm. Community discovery is considered an effective solution for extracting useful information from complex networks and has been widely applied in a series of fields such as biological networks, social sciences, and regional geography.

[0189] Commonly used community discovery algorithms include algorithms based on modularity optimization, algorithms based on spectral analysis, and algorithms based on information theory, etc. The present invention uses a community discovery algorithm based on modularity optimization. This method uses modularity as a metric to measure the quality of community partitioning results. Modularity is a metric value between -1 and 1, which is used to compare the internal connection tightness of community partitioning results and the connection tightness between partitioned communities. Among them, the Fastunfolding algorithm is an algorithm based on modularity optimization. The main goal of this algorithm is to continuously partition communities so that the modularity of the entire network after partitioning continuously increases. It mainly includes two stages: The first stage is called Modularity Optimization, mainly to partition each node into the community where its adjacent nodes are located to make the value of modularity continuously increase; the second stage is called Community Aggregation, mainly to aggregate the communities partitioned in the first step into a single point, that is, to reconstruct the network according to the community structure generated in the previous step.

[0190] Specifically, the processing steps of the Fastunfolding algorithm are as follows:

[0191] S301: Initialization. Each node in the highway network is divided into different communities; at this time, the number of communities is the same as the number of nodes;

[0192] S302: For each node, it is proposed to divide it into the community where its adjacent nodes are located, and calculate the change value of modularity before and after the division: if the change value of modularity is zero, then this division is abandoned; otherwise, this division is accepted;

[0193] 1) The calculation formula of modularity is:

[0194]

[0195] In the formula: M represents the modularity of the highway network; w ij represents the weight between edges v i and v j ; k i represents the sum of the weights of the edges connected by node v i ; c i represents the community assigned to node i;

[0196] When μ = v, δ(μ, v) = 1; in other cases, δ(μ, v) = 0;

[0197] 2) The calculation formula of the change value of modularity is:

[0198] ΔM = M f - M b ;

[0199] In the formula: ΔM represents the change amount of modularity M before and after the division, that is, the change value of modularity; M f , M b respectively represent the modularity M before and after the change;

[0200] S303: Repeat step S302 until all nodes have been visited once and no update occurs. At this time, the optimal community division under greed is reached, and the corresponding community structure is generated;

[0201] S304: Reconstruct the highway network according to the community structure generated in step S303, and aggregate all nodes in the same community together to form new nodes;

[0202] S305: Repeat steps S302 to S304 until the structure of the highway network no longer changes, and the corresponding community division result is generated.

[0203] The present invention divides the highway network into communities for different vehicle models. Through community division, the locations within the research scope can be divided into different communities or groups according to their similarities and connections, which helps to enhance the closeness of communication among different regions, enables the optimal layout of rescue points within and between communities, and further improves the accuracy of the location selection of highway emergency rescue points.

[0204] Through gridification and community division, the present invention can better understand and optimize the structure of the rescue network.

[0205] In the specific implementation process, to reflect the integrity of each community and determine the service scope of the emergency rescue points, the present invention uses the Voronoi diagram to divide different communities. The Voronoi diagram consists of a set of continuous polygons formed by the perpendicular bisectors of the lines connecting adjacent points. As Figure 4 shown, each Voronoi diagram contains only one discrete point data; the points within the Voronoi diagram are the closest to the corresponding discrete point; the points located on the edge of the Voronoi diagram are equidistant from the discrete points on both sides.

[0206] The present invention uses the Voronoi diagram to display the service scope of the emergency rescue points for various communities;

[0207] The discrete point set of the target scope is the point set N = {N1, N2, N3... N n}, where N i ≠N j (i≠j, i∈{1, 2, 3... n}, j∈{1, 2, 3... n}), that is, any two points do not coincide and any four points are not concyclic;

[0208] For any discrete point N i , the definition of its Voronoi diagram is:

[0209] V (i) = {x: D(x, N i ) < D(x, N j )|N i , N j ∈N, N i ≠N j , D is the Euclidean distance}.

[0210] The present invention determines the service scope of the community through the Voronoi diagram, which can provide a basis and foundation for subsequent centroid grouping and merging of communities, thus assisting in realizing the location selection of highway emergency rescue points.

[0211] In the specific implementation process, after the community division is completed, it is first necessary to determine the centroid of each partition, so that the location of the emergency rescue point is the shortest distance from each demand point within the partition, reducing the driving time of emergency rescue vehicles and improving the efficiency of emergency rescue. The centroid is the center of the geometric shape, and the center of mass is the center in terms of mass. If the density is uniform, the centroid and the center of mass should overlap. The present invention believes that the regional density of the community is uniform after the community division, so the center of mass can be obtained to represent the centroid of the community to find the center of each community.

[0212] The present invention calculates the center of mass of each community as its community centroid;

[0213] The formula description is as follows:

[0214]

[0215]

[0216] In the formula: C x , C y respectively represent the abscissa and ordinate of the community centroid; X i , Y i respectively represent the abscissa and ordinate of the nodes within the community; n represents the number of nodes within the community.

[0217] Since the regional density of the community is uniform after the community division, the present invention calculates the center of mass of each community as the community centroid, which can better improve the accuracy of the location selection of highway emergency rescue points.

[0218] In the specific implementation process, considering that different vehicle types have different degrees of influence on traffic accidents, it is necessary to assign corresponding weights to the accident probabilities and accident intensities of different vehicle types respectively. Based on the highway traffic accident data, the present invention uses the entropy weight method to assign the weights of the accident probabilities and accident intensities of four vehicle types respectively. The entropy weight method is to calculate the information entropy of the index and measure the effective information and index weights contained in the known data by the degree of difference of the index. Entropy weight refers to the relative importance coefficient of each index in the competition when making decisions or evaluating schemes under the conditions of given evaluation objects and evaluation indexes, but it does not represent the important coefficient of the index in the actual sense.

[0219] The present invention calculates the corresponding weights of each vehicle type for the highway traffic accident probability and accident intensity as the vehicle type weights;

[0220] The specific steps are as follows:

[0221] S501: Construct an initial sample matrix R=(r ij ), r 2×4 , r ij ≠0 (i = 1, 2, j = 1, 2, 3, 4), where r ijRepresents each element of the matrix, i represents two evaluation indicators, namely the probability of traffic accidents or the intensity of accidents, and j represents four vehicle types, including cars, buses, light trucks, and heavy trucks;

[0222] S502: Standardize the initial sample matrix R to generate a standard sample matrix D=(d ij ) 2×4 , where d ij is the standardized sample value;

[0223] The formula is described as:

[0224]

[0225] S503: Calculate the proportion h ij of each vehicle type in each group of samples in the sample;

[0226] The formula is described as:

[0227]

[0228] S504: Calculate the entropy E j of the j-th vehicle type;

[0229] The formula is described as:

[0230]

[0231] S505: Calculate the entropy weight w j of the j-th vehicle type according to its entropy E j , that is, the vehicle type weight;

[0232] The formula is described as:

[0233]

[0234] In the formula: w j represents the entropy weight of the j-th vehicle type, that is, the vehicle type weight; E j represents the entropy of the j-th vehicle type;

[0235] Finally, obtain the weight matrix W=(w j ) 2×4 .

[0236] The present invention takes into account that the influence degrees of different vehicle types on traffic accidents are different, so corresponding weights are respectively assigned to the accident probability and accident intensity of different vehicle types, so that the influence weights of different vehicle types on the layout of emergency rescue nodes can be determined, thereby ensuring the accuracy of the location selection of highway emergency rescue points.

[0237] In the specific implementation process, after the network is constructed, community detection is carried out for four types of vehicles respectively, and four different community detection results can be obtained, which can intuitively show the OD distribution characteristics of the four types of vehicles and provide a basis for subsequent data integration. The community division results of different vehicle types are different, and the number and distribution of their centroid points are also different. Therefore, the KNN algorithm needs to be used to merge the community centroids. The present invention generates a community weighted centroid through the following steps:

[0238] S601: Define the community division results of the four vehicle types as A, B, C, and D;

[0239] The formula description is:

[0240] A = {A1, A2,... A k};

[0241] B = {B1, B2,... B l};

[0242] C = {C1, C2,... C m};

[0243] D = {D1, D2,... D n};

[0244] In the formula: k, l, m, n respectively represent the number of communities in the corresponding community division results;

[0245] S602: Select the community division result F ∈ (A, B, C, D) with the smallest number of communities, where f = min(k, l, m, n);

[0246] S603: Define the set of community centroids of F as W, and the sets of community centroids of the community division results of the other three vehicle types as X, Y, and Z respectively;

[0247] The formula description is:

[0248] W = {W1, W2, W3... W k};

[0249] X = {X1, X2, X3... X l};

[0250] Y = {Y1, Y2, Y3... Y m};

[0251] Z = {Z1, Z2, Z3... Z n};

[0252] where k < l / m / n;

[0253] In the formula: W i 、X i 、Yi , Z i respectively represent the centroids of the communities in W, X, Y, and Z; k, l, m, and n respectively represent the number of community centroids, which is equal to the number of communities;

[0254] S604: Construct k centroid groups based on each community centroid W in W i as a benchmark;

[0255] S605: Calculate the centroid distances between the community centroids in X, Y, and Z and the community centroids in W respectively through the KNN algorithm, and divide the community centroids in X, Y, and Z into the centroid groups corresponding to the community centroids with the closest centroid distance to the community centroids in W. Finally, all the community centroids in X, Y, and Z are correspondingly divided into k centroid groups;

[0256] In this embodiment, if the community centroids in X, Y, and Z that are closest to the community centroids in W do not belong to the same community (based on the community division result of F), then divide the community centroids in X, Y, and Z into the centroid groups corresponding to the community centroids in W that belong to the same community.

[0257] The calculation formula for the centroid distance is:

[0258]

[0259] where: (x i , y i ) represent the coordinates of the community centroids in X, Y, and Z, represents the coordinate of the community centroid in W;

[0260] S606: Calculate the community weighted centroid of each centroid group based on the position coordinates and vehicle type weights of all community centroids in the centroid group, and finally obtain k community weighted centroids;

[0261] The calculation formula for the community weighted centroid is:

[0262]

[0263]

[0264] where: C wx , C wy respectively represent the abscissa and ordinate of the community weighted centroid; ω i represents the vehicle type weight corresponding to the community centroid with the coordinate ; n represents the number of community centroids in the centroid group.

[0265] The present invention divides the centroids of communities corresponding to various vehicle types into corresponding centroid groups, and then calculates the weighted sum of all community centroids in each centroid group based on the vehicle type weights of various vehicle types to generate the community weighted centroid of the centroid group. Finally, the emergency rescue points are matched based on the community weighted centroid, which can better achieve the selection of reasonable emergency rescue points according to the operation spatio-temporal distribution characteristics of different vehicle types on the highway and the influence weights of different vehicle types on the layout of emergency rescue nodes.

[0266] In the specific implementation process, the method for matching the nearest toll station in the present invention is the KDTree (K-Dimensional Tree) algorithm. The KDTree algorithm is a binary tree data structure used to store data points in a k-dimensional space and perform fast retrieval on them. The KDTree algorithm divides the data points into two subsets, each subset is stored in a node of the tree, and the division is based on selecting a dimension and the value on that dimension. The construction process of the KDTree algorithm is to cyclically select each dimension of the data points as the splitting dimension, take the median value of the data points on this dimension as the splitting hyperplane, hang the data points on the left side of the median value on its left subtree, and hang the data points on the right side of the median value on its right subtree, and recursively process its subtrees until all data points are mounted.

[0267] In the KDTree, each node represents a hyperplane that is perpendicular to the coordinate axis of the current division dimension and divides the space into two parts in this dimension, one part is in its left subtree and the other part is in its right subtree. This makes the KDTree efficient in processing multi-dimensional space data.

[0268] The characteristics of the binary search tree are as follows:

[0269]

[0270] Where T l_sub is the left subtree, T l_root is the root node of the left subtree T l_sub is the right subtree, T r_sub is the right subtree T r_root is the root node of the right subtree T r_sub of.

[0271] The construction process of the KDTree is as follows:

[0272] 1) Calculate the variances of each dimension and obtain the dimension K with the largest variance from them.

[0273]

[0274] 2) Arrange the data on dimension K in ascending order to obtain the data set N Kis the number of data in the K dimension. Calculate the median m in the K dimension.

[0275]

[0276] 3) Set the threshold to the median m, and two sets K sub_high and K sub_low can be obtained, and a tree node for storage is created:

[0277]

[0278] 4) Repeat the above operations for the two obtained subsets until all subsets can no longer be divided; when a subset can no longer be divided, save the data in the subset to the leaf node without child nodes.

[0279] During the search process, by dividing the plane, it is possible to avoid comparing the search point with each point one by one, reducing the search difficulty and search time. The distance calculated by the KDTree is the Euclidean distance, and the distance calculation formula is as follows:

[0280]

[0281] where d is the Euclidean distance between the point (x i , y i ) and the community centroid.

[0282] Specifically, the process of the KDTree finding the nearest highway toll station is as Figure 5 shown. For the given community centroid, it is necessary to start comparing from the root node of the KDTree. Here, P(K) is the value corresponding to the data point P on the division dimension K of the current root node, and m is the threshold for the current node division. When P(K) > m, visit the right subtree; otherwise, visit the left subtree until reaching the leaf node Q. At this time, Q is the current nearest toll station, and the distance d between P and Q is the current minimum distance. Then, backtrack up the original search path to the root node. If a point with a distance less than d is found during this process, its child nodes need to be included in the search range, and the nearest toll station is updated in a timely manner until all search paths are empty, and the process of using the KDTree algorithm to match the centroid with the nearest toll station ends. The KDTree algorithm respectively finds the toll station with the shortest distance to the nearest community centroid, and selects this toll station as the location for the emergency rescue point in this area.

[0283] The specific steps are as follows:

[0284] S701: Determine whether the community weighted centroid P(k) is less than the set threshold m: if so, visit the left subtree; otherwise, visit the right subtree;

[0285] Where m represents the middle distance value among the distance set between the toll station and the community weighted centroid in the same dimension;

[0286] If the distance between a toll station and the community weighted centroid is less than m, then divide this station into the left subtree;

[0287] If the distance between a toll station and the community weighted centroid is greater than or equal to m, then divide this station into the right subtree;

[0288] S702: Starting from the root node, visit each node along the search path in turn until reaching the leaf node Q. At this time, the leaf node Q is the nearest node to P(k), and the distance between Q and P(k) is d;

[0289] S703: Starting from the nearest node, backtrack along the search path to the root node. At the same time, calculate the distance d' between each node on the search path and P(k). If there exists a node Q' with a distance d' less than d, then take the node Q' as the new nearest node;

[0290] S704: Incorporate the children nodes of the new nearest node into the search scope, and return to step S702;

[0291] S705: Repeat steps S702 to S704 until all search paths are empty, and take the latest nearest node as the toll station closest to the community weighted centroid P(k), that is, the emergency rescue point.

[0292] The present invention matches the nearest toll station for each community weighted centroid through the KDTree algorithm, which can ensure the efficiency and accuracy of the nearest toll station matching.

[0293] In the specific implementation process, when the community scope is large, selecting only one highway emergency rescue point for each community cannot meet the highway emergency needs. The location model based on community discovery can be iterated repeatedly, and the above location process is repeated for each community after community division.

[0294] In the present invention, after determining the emergency rescue point R i (i = 1, 2... m), first take each toll station as the rescue demand point H i (i = 1, 2... n); then calculate the average rescue time of all rescue demand points, and determine whether the average rescue time is greater than the set maximum emergency rescue time (studies have shown that when the emergency rescue team arrives at the scene within 30 minutes after the accident, it can effectively reduce the casualties and property losses of the accident, that is, the maximum emergency rescue time T max(i)It can be set to 30 min): If so, return to step S3, further divide the communities in the community division result with the least number of communities to obtain new community division results for various vehicle types, and then continue to execute steps S4 to S7, and add the generated new emergency rescue points to R i (i = 1, 2...m); Otherwise, output the emergency rescue point R i (i = 1, 2...m).

[0295] In actual application, the new emergency rescue point serves as the rescue point of the next level. The first community division is the first-level emergency rescue point, the second is the second-level... and so on. It is set that the configuration levels of rescue points at each level are different, and the traffic accidents that can be served are also different. The higher-level rescue points can serve more traffic accidents.

[0296] Specifically, the average rescue time is calculated through the following steps:

[0297] 1) Calculate the distance S between each rescue demand point H i (i = 1, 2...n) and the emergency rescue point R i (i = 1, 2...m) closest to it; i (i = 1, 2...n);

[0298] 2) Calculate the rescue time of each rescue demand point

[0299]

[0300] In the formula: v represents the driving speed of the emergency rescue vehicle;

[0301] 3) Calculate the average rescue time of all rescue demand points

[0302]

[0303] In the formula: n represents the number of rescue demand points.

[0304] After the present invention selects the toll station closest to the weighted centroid of each community as the emergency rescue point, it further judges whether the number and location of the emergency rescue points are reasonable according to the average rescue time of all rescue demand points (for example, when the community scope is large, selecting only one highway emergency rescue point in each community cannot meet the highway emergency demand), thereby further improving the practicability of the location selection of highway emergency rescue points.

[0305] To characterize the stability of emergency rescue time, and due to the significant differences in rescue times between first- and second-level traffic accidents and third- and fourth-level traffic accidents, the present invention uses the coefficient of variation CV to measure the stability of highway emergency time. The coefficient of variation can eliminate the influence of different measurement scales and dimensions. To calculate the coefficient of variation, it is first necessary to calculate the average rescue time and its standard deviation. The formula for the standard deviation is as follows:

[0306]

[0307] The formula for the coefficient of variation is as follows:

[0308]

[0309] Based on the vehicle toll data of a certain city's highway in China for method instantiation application, the results show that for the zoning site selection of 8 first-level emergency rescue points and 23 second-level emergency rescue points divided in this area, compared with the P-center site selection model, the average rescue time for first- and second-level traffic accidents is reduced by approximately 22.02%, and the average rescue time for third- and fourth-level traffic accidents is reduced by 21.33%. The coefficients of variation of the emergency rescue times for the two site selection models are reduced by 37.37% and 16.14% respectively, indicating that the method of the present invention is more in line with the characteristics of highway emergency rescue needs, and significantly improves the rationality of the layout of rescue center sites.

[0310] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit the technical solutions. Those of ordinary skill in the art should understand that any modifications or equivalent replacements made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions should be covered by the scope of the claims of the present invention.

Claims

1. A data-driven method for locating highway emergency rescue points, characterized in that, Including: S1: Rasterize the target range to generate a target range raster; S2: Match the OD data of vehicle toll stations of various vehicle types to the target range raster to generate a highway network; S3: Conduct community division on the highway network for different vehicle types to generate community division results corresponding to various vehicle types; S4: Calculate the community centroid of each community in the community division result corresponding to each vehicle type to generate community centroids corresponding to various vehicle types; S5: Calculate the vehicle type weights of each vehicle type for highway traffic accidents; S6: Divide the community centroids corresponding to various vehicle types into corresponding centroid groups, and then perform weighted calculation on all community centroids in each centroid group according to the vehicle type weights of various vehicle types to generate the community weighted centroid of the centroid group; S7: Match the nearest toll station to each community weighted centroid as its corresponding emergency rescue point.

2. The method for selecting an emergency rescue point on a highway based on data-driven as claimed in claim 1, wherein: In step S1, the target range raster is generated through the following steps: S101: Calculate the central coordinates (lon, lat) of the target range according to the longitude and latitude (lon1, lat1) of the lower left corner and the longitude and latitude (lon2, lat2) of the upper right corner of the target range; The formula description is: S102: Calculate the longitude and latitude (Δlon, Δlat) of each raster in the target range according to the earth radius R at the target range and the set raster length a; The formula description is: S103: Use the lower left corner of the target range as the center point of the first raster to fill the entire target range with rasters, and number each raster with column numbers and row numbers (loncol, latcol); S104: Calculate the central coordinates (hblon, hblon) of each raster; The formula description is: hblon = loncol·Δlon + lon1; hblon = latcol·Δlat + lat1; S105: Calculate the coordinates of the four vertices of each grid: infer one vertex for each grid, and infer the other three vertices from the surrounding three grids; For any grid, starting from the lower left vertex, the coordinates of the four vertices in the counterclockwise direction are: Where: the subscripts i and j represent the row number and column number of the grid respectively.

3. The method for selecting an emergency rescue point on a highway based on data driving according to claim 2, wherein: In step S2, the OD data of vehicle toll stations of various vehicle types are matched to the target range raster using longitude and latitude; The formula for obtaining the corresponding raster is as follows: Where: x and y represent the longitude and latitude of the toll station in the OD data of the vehicle toll station.

4. The method for selecting the location of highway emergency rescue points based on data driving according to claim 1, wherein: In step S3, the community division of the highway network is performed through the Fastunfolding algorithm; The specific steps are as follows: S301: Initialize and divide each node in the highway network into different communities; S302: For each node, it is planned to divide it into the community where its adjacent node is located, and calculate the change value of modularity before and after the division: if the change value of modularity is zero, this division is abandoned; otherwise, this division is accepted; 1) The formula for modularity is: Where: M represents the modularity of the highway network; w ij represents the weight between edges v i and v j ; k i represents the sum of the weights of the edges connected to node v i ; c i represents the community assigned to node i; When μ = v, δ(μ, v) = 1; in other cases δ(μ, v) = 0; 2) The formula for the change value of modularity is: ΔM = M f -M b ; Where: ΔM represents the change in modularity M before and after division, that is, the modularity change value; M f and M b represent the modularity M before and after the change, respectively; S303: Repeat step S302 until all nodes have been visited and no updates occur, generating the corresponding community structure; S304: Reconstruct the highway network according to the community structure generated in step S303, aggregating all nodes within the same community together to form new nodes; S305: Repeat steps S302 to S304 until the structure of the highway network no longer changes, generating the corresponding community division result.

5. The method for selecting an emergency rescue point on an expressway based on data driving according to claim 1, characterized in that: In step S4, calculate the centroid of each community as its community centroid; The formula description is: Where: C x and C y represent the abscissa and ordinate of the community centroid respectively; X i , Y i represent the abscissa and ordinate of the nodes within the community respectively; n represents the number of nodes within the community.

6. The method for selecting emergency rescue points on expressways based on data driving according to claim 1, wherein: In step S5, calculate the corresponding weights of each vehicle type for the probability and intensity of highway traffic accidents as the vehicle type weights; The specific steps are as follows: S501: Construct an initial sample matrix \(R=(r ij ) 2×4 ,r ij \neq0 (i = 1, 2, j = 1, 2, 3, 4)\), where \(i\) represents two evaluation indicators of traffic accident probability or accident intensity, and \(j\) represents four vehicle types; S502: Standardize the initial sample matrix R to generate a standard sample matrix D = (d ij ) 2×4 , where d ij is the standardized sample value; The formula description is: S503: Calculate the proportion h of each vehicle model in each group of samples in the samples ij ; The formula description is: S504: Calculate the entropy E of the j-th vehicle model j ; The formula description is: S505: Calculate its entropy weight w j according to the entropy E of the j-th vehicle model j , that is, the vehicle model weight; The formula description is: where: w j represents the entropy weight of the j-th vehicle model, i.e., the vehicle model weight; E j represents the entropy of the j-th vehicle model.

7. The method for selecting an emergency rescue point on an expressway based on data driving according to claim 1, wherein: In step S6, generate the community weighted centroid through the following steps: S601: Define the community division results of four vehicle types as A, B, C, and D; The formula description is: A = {A1, A2,... A k}; B = {B1, B2,... B l}; C = {C1, C2,... C m}; D = {D1, D2,... D n}; In the formula: k, l, m, n respectively represent the number of communities in the corresponding community division results; S602: Select the community division result F ∈ (A, B, C, D) with the least number of communities, where f = min(k, l, m, n); S603: Define the set of community centroids of F as W, and the sets of community centroids of the community division results of the other three vehicle types as X, Y, and Z respectively; The formula description is: W = {W1, W2, W3... W k}; X = {X1, X2, X3... X l}; Y = {Y1, Y2, Y3... Y m}; Z = {Z1, Z2, Z3... Z n}; where k < l / m / n; Where: W i , X i , Y i , Z i respectively represent the community centroids of W, X, Y, and Z; k, l, m, and n respectively represent the number of community centroids; S604: Construct k centroid groups with each community centroid W in W i as the benchmark; S605: Calculate the centroid distances between the community centroids in X, Y, and Z and the community centroids in W respectively through the KNN algorithm, and divide the community centroids in X, Y, and Z into the centroid groups corresponding to the community centroids with the closest distance to the centroids in W. Finally, all community centroids in X, Y, and Z are divided into k centroid groups; The calculation formula for the centroid distance is: where: (x i , y i ) represents the coordinates of the centroid of the community in X, Y, Z, represents the coordinates of the centroid of the community in W; S606: Calculate the community weighted centroid of each centroid group according to the position coordinates and vehicle type weights of all community centroids in the centroid group, and finally obtain k community weighted centroids; The calculation formula for the community weighted centroid is: Where: C wx , C wy respectively represent the abscissa and ordinate of the community weighted centroid; ω i represents the vehicle type weight corresponding to the community centroid with coordinates ; n represents the number of community centroids in the centroid group.

8. The method for selecting highway emergency rescue site based on data driving according to claim 1, characterized in that: In step S7, match the closest toll station for the community weighted centroid through the KDTree algorithm; The specific steps are as follows: S701: Determine whether the community weighted centroid P(k) is less than the set threshold m: If so, visit the left subtree; otherwise, visit the right subtree; where m represents the middle distance value in the set of distances between toll stations and community weighted centroids in the same dimension; If the distance between a certain toll station and the community weighted centroid is less than m, then divide this station into the left subtree; If the distance between a certain toll station and the community weighted centroid is greater than or equal to m, then divide this station into the right subtree; S702: Starting from the root node, visit each node along the search path in turn until reaching the leaf node Q. At this time, the leaf node Q is the closest node to P(k), and the distance between Q and P(k) is d; S703: Starting from the closest node, backtrack along the search path to the root node, and at the same time calculate the distances d' between each node on the search path and P(k). If there exists a node Q' with a distance d' less than d, then take the node Q' as the new closest node; S704: Incorporate the children nodes of the new nearest node into the search scope and return to step S702; S705: Repeat steps S702 to S704 until all search paths are empty, and take the latest nearest node as the toll station closest to the community weighted centroid P(k), that is, the emergency rescue point.

9. The method for selecting a highway emergency rescue point based on data driving according to claim 1, wherein: In step S7, determine the emergency rescue point R i (i = 1, 2... m), first take each toll station as the rescue demand point H i (i = 1, 2... n); then calculate the average rescue time of all rescue demand points, and determine whether the average rescue time is greater than the set maximum emergency rescue time: if so, return to step S3, further divide the communities in the community division result with the least number of communities to obtain a new community division result for various vehicle types, and then continue to execute steps S4 to S7, and add the generated new emergency rescue points to R i (i = 1, 2... m); otherwise, output the emergency rescue point R i (i = 1, 2... m).

10. The method for selecting the location of highway emergency rescue points based on data-driven as claimed in claim 9, wherein: Calculate the average rescue time through the following steps: 1) Calculate each rescue demand point H i (i = 1, 2... n) and the distance to the nearest emergency rescue point R i (i = 1, 2... m) is S i (i = 1, 2... n); 2) Calculate the rescue time for each rescue demand point In the formula: v represents the driving speed of the emergency rescue vehicle; 3) Calculate the average rescue time for all rescue demand points In the formula: n represents the number of rescue demand points.

Images