A VANET Clustering Algorithm Based on Support Vector Machine

Abstract

The present invention provides a VANET clustering algorithm based on a support vector machine, which belongs to the technical field of vehicle network clustering. It solves the problems of instability of network topology and restricted movement trajectory caused by the high mobility of vehicle nodes in the vehicle network VANET. Its technical solution is as follows: it includes the following steps: step 1, extraction of vehicle node attribute characteristic parameters and construction of support vector machine; step 2, use of AP clustering algorithm to select cluster head node at the initial moment; step 3, use of support vector machine to complete clustering of road vehicle nodes; step 4, cluster maintenance operation during vehicle driving; step 5, after the cluster head vehicle node finishes driving, the support vector machine parameters of the cluster head node are adjusted and optimized based on the statistical survival time of the nodes in the cluster. The beneficial effects of the present invention are: the present invention improves the stability of the cluster; and can complete the rapid clustering judgment by itself.

Description

Technical Field

[0001] The present invention relates to the field of vehicle networking and machine learning technology, and in particular to a VANET clustering algorithm based on a support vector machine. Background Art

[0002] VANET is a type of mobile ad-hoc network (MANET) whose network nodes are vehicles traveling on roads. Communication in VANETs typically involves vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I). Due to the high mobility of vehicle nodes, the constraints of road geometry and driving regulations, and the physical connectivity limitations of urban environments, vehicle movement trajectories are semi-organized. Consequently, ensuring stable communication between vehicle nodes in VANETs presents a unique set of challenges.

[0003] Clustered communication helps improve the stability and reliability of communication in VANETs. Clustering involves grouping road vehicles based on their spatial distribution and relative speed, thereby forming a hierarchical communication network structure. Clustering can serve as the foundation for routing algorithms, accident and congestion detection, information dissemination, and entertainment applications. Although VANETs are derived from MANETs, ​​the high mobility and road topology mean that vehicles in a cluster will inevitably disconnect from the VANET. Therefore, clustering algorithms developed for MANETs cannot be directly applied to VANETs.

[0004] How to solve the above technical problems is the subject faced by the present invention. Summary of the Invention

[0005] The purpose of the present invention is to provide a VANET clustering algorithm based on a support vector machine. The present invention mainly targets the characteristics of the high mobility of vehicle nodes in the vehicle network (VANET), which leads to the instability of the network topology and the restricted movement trajectory. By using a support vector machine algorithm to cluster the vehicle nodes, a relatively stable communication environment and reliable management means are provided for the vehicle nodes in the cluster, thereby improving the stability of the cluster; an improved AP clustering algorithm is used to select cluster head nodes, making cluster head election more efficient. Compared with existing clustering algorithms, in the present invention, vehicle nodes do not need to use third-party equipment to enter the cluster, and the clustering efficiency is higher.

[0006] In order to achieve the above-mentioned object of the invention, the present invention adopts a technical solution specifically as follows: a VANET clustering algorithm based on support vector machine, specifically comprising the following steps:

[0007] Step 1: Extract feature vectors from each vehicle node that is newly on the road, and construct a support vector machine model for each vehicle node based on the feature vectors;

[0008] In the step 1:

[0009] S101. Create a data structure for the vehicle as shown in Table 1. This structure includes the current time (time), which is the vehicle's travel time on the target road section; the vehicle type (vehicle_type), which is a value based on the vehicle type (bus, car, truck, etc.), and requires dimensionality conversion to 0, 1, or 2, respectively; the vehicle's x-coordinate or longitude (vehicle_x); the vehicle's y-coordinate or latitude (vehicle_y); the vehicle's speed (vehicle_speed); and the vehicle ID (vehicle_id), which is generally an English identifier consisting of the vehicle's direction of travel and the lane it is in. Initially, the vehicle's entry time in_time and exit time out_time are empty 0. When subsequent vehicles join or leave the cluster while driving, the current time is recorded; the vehicle identification flag has an initial value of false. When the vehicle enters the cluster while driving, the flag switches to True; the vehicle state state has an initial value of UN. When the vehicle becomes a cluster head node during driving, the state switches to CH. If it becomes a cluster member node, the state switches to CM; the vehicle neighbor table neighbor has an initial value of empty. Subsequently, the corresponding data in the node data structure is added according to the distance; the vehicle cluster member table member has an initial value of empty. Later, when the node becomes a cluster head node, the corresponding data of the cluster member node is added to member; the vehicle's cluster head ID head_id has an initial value of empty. When the node becomes a cluster member node, it is updated to the ID number of the cluster head node to which it belongs; the vehicle support vector machine parameters C and gamma, C is the penalty parameter set to 0.25, gamma is the kernel parameter, and the initial value is randomly set from 1 to 500. C is the penalty coefficient, which can be understood as the tolerance to error. The higher C is, the easier it is to overfit, and the smaller C is, the easier it is to underfit.

[0010] S102: Extract the vehicle node's own attributes into a 4-dimensional feature vector. The attributes include the vehicle type, vehicle x-coordinate, vehicle y-coordinate, and vehicle speed in the above data structure. After data normalization, the normalization process of feature t is as follows:

[0011]

[0012] Among them, t' represents the standardized t, mean is the mean of all data of feature t, and std is the standard deviation of all data of feature t.

[0013] Get the vehicle node feature vector x_vehicle i as follows:

[0014] x_vehicle i =(type i ,xpos i ,ypos i , s i )

[0015] Where type i For standardized vehicle types, xpos i is the standardized vehicle x coordinate, ypos i is the standardized vehicle y coordinate, s i is the normalized vehicle speed.

[0016] S103: Build a support vector machine model for each vehicle node, which is used to extract C and gamma from the data structure to make judgments when the node becomes a cluster head or enters a cluster. The decision function is as follows:

[0017]

[0018] Where sign() is the sign function. When n is positive, sign(n) is 1, when n is negative, sign(n) is -1, and when n is zero, sign(n) is 0. Gamma is a parameter that comes with the Gaussian kernel function after it is selected as the kernel function. The larger the gamma value, the fewer support vectors, and the smaller the gamma value, the more support vectors. The number of support vectors affects the speed of the algorithm. i and α i * is the Lagrange multiplier, which is positive, y i Indicates the category of cluster head node, y j is the category feature of the input node.

[0019] Step 2: In the clustering stage, the UN vehicle node that first enters the road uses the improved AP clustering algorithm to select the cluster head. Specifically:

[0020] S201: All vehicles that have just entered the road are sequentially similarized using the improved AP clustering algorithm DAPC, as follows:

[0021]

[0022] Among them, D i is the average displacement of node i at all times, that is, the mathematical expectation of the displacement value of node i. MR is the maximum communication distance of the vehicle, that is, the one-hop distance. Speed i is the mathematical expectation of the instantaneous speed of the node, speed m is the maximum speed limit.

[0023] At the initial moment, the vehicle nodes on the road are selected as cluster heads according to the following criteria: If r(i, i) + a(i, i) > 0, node i is selected as a cluster head node.

[0024] Where r(i, k) is the attraction degree, which is the message sent from node k to node i, indicating the suitability of node i to select node j as the cluster head, and r(i, i) is the self-attraction degree. a(i, k) is the attribution degree, which is the message sent from node i to node k, indicating the suitability of node k to be the cluster head of node i, and a(i, i) is the self-attribution degree. It is calculated as follows:

[0025] r(i,k)=s(i,k)-max{a(i,k′)+s(i,k′)}

[0026] a(i,k)=min{0,r(k,k)+∑max(0,r(i′,k))}

[0027] a(k,k)=∑max(0,r(i′,k))

[0028] Where k≠k', k' represents a node other than k. a(i,k') represents the ownership of point i by points other than k, initially 0; s(i,k') represents the attraction of points other than k to i, meaning that all points other than i are vying for ownership of i; r(i',k) represents the similarity of point k as the cluster center of points other than i. This is the cumulative evidence that point i chooses k as its cluster center, given the support of data points with attraction values ​​greater than 0.

[0029] S202. In order to formulate the optimal cluster head selection criteria in the neighbor table, the following composite evaluation function is constructed:

[0030] (1) The position scaling function is defined as follows: the node un needs to be added to the cluster where the cluster head node t is located, and its position needs to be transformed:

[0031]

[0032] where num t is the number of nodes in the current cluster t. t is the x coordinate of the cluster head node, y t is the y coordinate of the cluster head node. un is the x coordinate of the UN vehicle node, y un is the y coordinate of the vehicle node.

[0033] (2) Sum of belongingness and attraction:

[0034]

[0035] Among them, CH i The cluster head of node i is node k. Its value is the sum of the belonging degree and the attraction degree.

[0036] S203, construct the optimal cluster head selection rule in the DAPC algorithm. The definition is as follows:

[0037] CH m =w1arg min DF+w2argmin(-CH old )

[0038] Where w1+w2=1. The w1 value corresponding to this function should be greater than w2. The values ​​of w1 and w2 depend on the performance of AP clustering.

[0039] Step 3: After the optimal cluster head selection is completed, the vehicle node will enter the cluster.

[0040] Specifically, a vehicle node uses the support vector machine model of the cluster head node to join the cluster. At this time, the feature vector of the UN vehicle node is input into the support vector machine classifier carried by the cluster head node itself to make a cluster entry decision. If the vehicle node classification result is consistent with the cluster head classification, it is considered that the node will join the cluster and send a cluster entry request to the cluster head. At the same time, it also includes data in the data structure such as its current location coordinates and average vehicle speed;

[0041] If the vehicle node is in three consecutive cluster head broadcast time periods T broadcast (T broadcast If a node fails to join any cluster within a time interval of 1 second, it can initiate the creation of a new cluster and become the cluster head node of the new cluster.

[0042] Step 4: After the vehicle node enters the cluster, cluster maintenance operations are performed according to the operating status;

[0043] The step 4 specifically includes:

[0044] (1) Cluster merging

[0045] When the distance between two cluster heads traveling in the same direction reaches a set threshold (the node communication distance, typically 300 meters), the two cluster heads exchange feature vectors extracted from their respective data structures and input them into the support vector machine models carried by their respective vehicle nodes for training. The results are then compared for consistency. If these two conditions are met, the two clusters can be merged. The cluster head node of the newly merged cluster selects the cluster head of the larger cluster. In other words, the cluster head of the cluster with the larger number of cluster members becomes the new cluster head node of the merged cluster, and the cluster heads of the smaller clusters become cluster members of the new cluster head node.

[0046] (2) Cluster member disconnection

[0047] If a cluster member node fails to receive a cluster head broadcast message for three consecutive 1-second cluster head broadcast cycles, it is considered to have left the cluster and become a UN node. If the cluster head node fails to receive a cluster presence report from a cluster node for three consecutive cluster head broadcast cycles, it is considered to have left the cluster. When all cluster members are disconnected, the cluster head node attempts to join the nearest cluster within its communication radius. If this joins successfully, the original cluster is considered defunct.

[0048] (3) Cluster head node disconnection

[0049] For cluster member vehicle nodes, due to some special reasons, such as the communication hardware failure of the cluster head node or leaving the road, the cluster head changes its state and actively disconnects from all nodes in the cluster. When the cluster head node leaves the cluster, if there are still nodes in the current cluster, a new cluster head is selected according to the cluster head selection algorithm.

[0050] Step 5: After a vehicle node completes its journey within the specified timeframe, the status of each vehicle node must be summarized. If a cluster contains a large number of member nodes with short survival times, the cluster head node's support vector machine parameters C and gamma must be adjusted to optimize the performance of the cluster head vehicle node. This ensures that member nodes joining the cluster on subsequent travel sections maintain a relatively long survival time.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. The present invention mainly addresses the characteristics of the high mobility of vehicle nodes in the vehicle network (VANET), which leads to unstable network topology and restricted movement trajectories. By using the support vector machine algorithm to cluster vehicle nodes, a relatively stable communication environment and reliable management methods are provided for vehicle nodes in the cluster, thereby improving the stability of the cluster. The AP clustering algorithm is improved to select cluster head nodes, making cluster head election more efficient. Compared with existing clustering algorithms, in the present invention, vehicle nodes do not need to use third-party equipment to join the cluster, and the clustering efficiency is higher.

[0053] 2. The clustering method designed in this invention uses the support vector machine algorithm to deal with the vehicle node clustering problem, rather than the deep neural network method suitable for big data scenarios, and can cope with the rapidly changing topology structure in VANET.

[0054] 3. The vehicle network clustering method based on support vector machines of the present invention is different from the clustering methods studied in the past. It is a lightweight distributed framework that does not rely on third-party RSU data centers. Vehicle nodes can complete rapid and autonomous clustering without the need for complex calculations.

[0055] 4. The vehicle network clustering method based on support vector machine of the present invention has a certain degree of adaptability and can cope with different traffic scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0057] Figure 1 This is a schematic diagram of the vehicle node data structure in the support vector machine-based VANET clustering algorithm provided in Example 1 of the present invention.

[0058] Figure 2 This is a flow chart of cluster merging in the support vector machine-based VANET clustering algorithm provided in Example 1 of the present invention.

[0059] Figure 3 This is a flow chart of support vector machine parameter adjustment in the support vector machine-based VANET clustering algorithm provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. Of course, the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0061] Example 1

[0062] See also Figures 1 to 3 This embodiment provides a VANET clustering algorithm based on a support vector machine, which specifically includes the following steps:

[0063] Step 1: Standardize the attribute parameters required for extracting vehicles from the data set and use them as feature vectors. Specifically:

[0064] Extract the attribute parameters of the vehicles in the dataset, including vehicle type vehicle_type, vehicle x coordinate vehicle_x, vehicle y coordinate vehicle_y, and vehicle speed vehicle_speed. The feature vector x_vehicle of each vehicle node is obtained as follows:

[0065] x_vehicle i =(type i ,xpos i ,ypos i , s i )

[0066] Where type i For standardized vehicle types, xposi is the standardized vehicle x coordinate, ypos i is the standardized vehicle y coordinate, s i is the normalized vehicle speed.

[0067] Step 2: Establish the vehicle node data structure based on the data set. Specifically:

[0068] The data contained in the construction of the vehicle node data structure is as follows Figure 1 As shown in the figure, the data of each vehicle in the dataset is combined with other attribute values ​​into a data structure. The data structure of each vehicle node is indexed by time.

[0069] Step 3: Initially, use the DAPC algorithm to select the cluster head node. Specifically:

[0070] S301, the initial moment feature vector x_vehicle extracted in step 1 i Input into the DAPC algorithm to obtain the category labels of 10 vehicle nodes and the index of the cluster head node;

[0071] S302: Based on the cluster head index, several cluster head nodes are determined and the corresponding data in the data structure of each cluster head vehicle node is updated. The remaining vehicles are matched to the corresponding cluster head node according to their category and added to the cluster of the cluster head node to become member nodes. At the end of the initial time, several clusters exist on the road.

[0072] Step S4: When the vehicle starts to travel, the UN nodes on the road begin to join the cluster. Specifically:

[0073] S401, traverse the current cluster head node and add each cluster head node to the neighbor table of the current UN node;

[0074] S402: Each UN node selects the optimal cluster head node in the neighbor table through the DAPC algorithm.

[0075] S403: Each UN node begins clustering: its own feature vector is input into the support vector machine model of the selected cluster head. The calculated result is evaluated. If it is consistent with the cluster head, the node is considered eligible to join the cluster. After the node joins the cluster, the relevant data in the data structure is updated. The cluster head node updates member_list, and the member node updates head_id.

[0076] Step S5: Cluster maintenance operation during vehicle node travel, specifically:

[0077] (1) Cluster merging:

[0078] S501, comparing the distances between two different cluster heads, when the distance is less than the cluster head communication radius (generally 300 meters);

[0079] S502, inputting the feature vectors of the two cluster head nodes into the support vector machine model of another node;

[0080] S503: Compare the support vector machine calculation results of the two cluster head nodes at this time. If they are consistent, it is considered that the two clusters can be merged.

[0081] S504: The cluster head of the larger cluster is used as the cluster head of the newly merged cluster, the old cluster head and its cluster members are used as cluster members of the new cluster, and the data in the data structure of each node is updated;

[0082] (2) The connection between the cluster member node and the cluster head node is disconnected, specifically:

[0083] S511. Compare the distance between the cluster member node and the cluster head node to which it belongs to determine whether it exceeds its own communication distance (generally 300 meters).

[0084] S512: If the distance is exceeded, the member node is deemed to be disconnected from the cluster head node.

[0085] S513: After the node is disconnected from the cluster head, it needs to update the data in the data structure, switch the flag to false, and set its own cluster head ID to null. The original cluster head node needs to update the member table and delete the data structure data of the node.

[0086] (2) Cluster extinction, specifically:

[0087] S521. When all cluster members in the cluster are disconnected, the cluster head node will try to join the cluster where the nearest cluster head within its communication radius is located.

[0088] S522: The cluster head node joins the cluster according to step 4 above. If the joining is successful, the original cluster is considered to be dead, and the corresponding data in the data structure is cleared.

[0089] Step S6, adjust the parameters of the support vector machine model of each cluster head node, as follows: Figure 3 shown.

[0090] S601: After the algorithm is run once, each historical cluster head is evaluated and the survival time of each cluster member is counted.

[0091] S602. If the survival time of more than half of the nodes is lower than the average survival time of the member nodes of the cluster, the parameters of the support vector machine of the cluster head are adjusted using the grid search method, where the interval of the grid search method is set to (0, m) where m is the initial random value of the gamma parameter at this time, that is, a random number within (0, 500).

[0092] S603. Repeat steps S601 and S602 until the survival time of more than half of the nodes in the cluster is higher than the average survival time, and then the algorithm stops.

[0093] Table 1 Vehicle data structure table

[0094]

[0095] Table 2 Global variable setting table

[0096]

[0097]

[0098] Table 3 Performance index table

[0099]

[0100] The vehicle network clustering method based on support vector machines in the present invention can improve the stability of clusters; in the present invention, vehicle nodes can enter clusters independently and quickly without the help of third-party equipment; the clustering method is a lightweight distributed framework with certain adaptability and can adapt to different traffic scenarios.

[0101] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A VANET clustering algorithm based on support vector machine, characterized in that: The following steps are involved: Step 1: Extract feature vectors from each vehicle node that is newly on the road, and construct a support vector machine model for each vehicle node based on the feature vectors; In the step 1: S101. A data structure is established for the vehicle structure, which includes the current time point time, which is the vehicle's travel time on the target road section; the vehicle type vehicle_type, which is a bus, car, or truck, and needs to be converted to 0, 1, and 2 in sequence, and then participates in Z normalization; the vehicle's x-coordinate or longitude vehicle_x; the vehicle's y-coordinate or latitude vehicle_y; the vehicle's speed vehicle_speed; the vehicle's ID number vehicle_id, which is an English identifier consisting of the vehicle's travel direction and the lane it is in; initially, the vehicle's entry time in_time and departure time out_time are empty. 0, when the subsequent vehicle joins or leaves the cluster while driving, the current moment is recorded; the initial value of the vehicle identification flag is false, and when the vehicle enters the cluster while driving, the flag switches to True; the initial value of the vehicle state state is UN, and when the vehicle becomes a cluster head node during driving, the state switches to the CH cluster head vehicle node, and if it becomes a cluster member node, the state switches to the CM cluster member vehicle node; the initial value of the vehicle neighbor table neighbor is empty, and the corresponding data in the node data structure is added according to the distance; the initial value of the vehicle cluster member table member is empty, and when the node becomes a cluster head node later, the corresponding data of the cluster member node is added to member; the cluster head ID head_id of the vehicle, the initial value is empty, and when the node becomes a cluster member node, it is updated to the ID number of the cluster head node to which it belongs; the vehicle support vector machine parameters C and gamma, C is the penalty parameter, set to 0.25, gamma is the kernel parameter, and the initial value is randomly set from 1 to 500; C is the penalty coefficient, which is the tolerance to error. The higher C is, the easier it is to overfit, and the smaller C is, the easier it is to underfit; S102, extracting the vehicle node's own attributes into a 4-dimensional feature vector, wherein the attributes include the vehicle type, vehicle x-coordinate, vehicle y-coordinate, and vehicle speed in the above data structure, and undergoing data normalization; S103: Build a support vector machine model for each vehicle node, which is used to extract C and gamma in the data structure for judgment when the node becomes a cluster head or enters a cluster; Step 2: In the clustering stage, the UN vehicle nodes that are newly on the road and have not yet entered the cluster are clustered using the improved AP clustering algorithm to select cluster heads; Step 3: After the optimal cluster head selection is completed, the vehicle node will enter the cluster; Step 4: After the vehicle node enters the cluster, cluster maintenance operations are performed according to the operating status; Step 5: After the vehicle node has finished traveling within the specified time, the status of each vehicle node needs to be summarized. If there are a large number of cluster member nodes with short survival time in a cluster, the parameters C and gamma of the cluster head node support vector machine need to be adjusted to optimize the performance of the cluster head vehicle node so that the member nodes added to the cluster in the subsequent driving section can maintain a relatively long survival time.

2. The VANET clustering algorithm based on support vector machine according to claim 1 is characterized in that: In S102, after data normalization, the normalization process of feature t is as follows: Where t' represents the standardized t, mean is the mean of all data of feature t, and std is the standard deviation of all data of feature t; Get the vehicle node feature vector x_vehicle i as follows: x_vehiclei=(type i ,xpos i ,ypos i ,s i ) Where type i For standardized vehicle types, xpos i is the standardized vehicle x coordinate, ypos i is the standardized vehicle y coordinate, s i is the normalized vehicle speed; In S103, the decision function is as follows: Where sign() is the sign function. When n is positive, sign(n) is 1, when n is negative, sign(n) is -1, and when n is zero, sign(n) is 0. Gamma is a parameter of the function after selecting the Gaussian kernel function as the kernel function. The larger the gamma value, the fewer support vectors, and the smaller the gamma value, the more support vectors. The number of support vectors affects the speed of the algorithm. i and α i * is the Lagrange multiplier, which is positive, y i Indicates the category of cluster head node, y j It is the category feature of the input node.

3. The VANET clustering algorithm based on support vector machine according to claim 1 is characterized in that: In the second step, in the clustering stage, the UN vehicle node that first enters the road uses the improved AP clustering algorithm to implement cluster head selection, which includes the following steps: S201. All vehicles that are newly on the road are similarly calculated using the improved AP clustering algorithm DAPC as follows: Among them, D i is the average displacement of node i at all times, that is, the mathematical expectation of the displacement value of node i, MR is the maximum communication distance of the vehicle, that is, the one-hop distance, Speed i is the mathematical expectation of the instantaneous velocity of the node, speed m is the maximum speed limit; At the initial moment, the vehicle nodes on the road are selected as cluster heads according to the following criteria: if r(i, i) + a(i, i) > 0, then node i is a cluster head node, and all nodes that meet this condition are cluster head nodes; Where r(i, k) is the attraction degree, which is the message sent from node k to node i, indicating the suitability of node i to select node j as the cluster head, and r(i, i) is the self-attraction degree; a(i, k) is the attribution degree, which is the message sent from node i to node k, indicating the suitability of node k to be the cluster head of node i, and a(i, i) is the self-attribution degree. It is calculated as follows: r(i,k)=s(i,k)-max{a(i,k′)+s(i,k′)} a(i,k)=min{0,r(k,k)+∑max(0,r(i′,k))} a(k,k)=∑max(0,r(i′,k)) Where k≠k', k' represents nodes other than k, a(i,k') represents the degree of belonging of points other than k to point i, which is initially 0; s(i,k') represents the similarity of points other than k to i, that is, other points other than i are competing for the ownership of point i; r(i',k) represents the degree to which point k is the cluster center of points other than i. The probability of k being the cluster center is obtained by taking all attraction values ​​greater than or equal to 0 and adding them. That is, the cumulative proof that data point i chooses k as its cluster center with the support of these data points with attraction values ​​greater than 0 is obtained. S202. In order to formulate the optimal cluster head selection criteria in the neighbor table, the following composite evaluation function is constructed: (1) The position scaling function is defined as follows: Node un needs to be added to the cluster where the cluster head node t is located, and its position needs to be transformed: where num t is the number of nodes in the current cluster t, x t is the x coordinate of the cluster head node, y t is the y coordinate of the cluster head node, x un is the x coordinate of the UN vehicle node, y un is the y coordinate of the vehicle node; (2) Sum of belongingness and attraction: Among them, CH i The cluster head of node i is node k, and its value is the sum of the belonging degree and the attraction degree; S203, construct the optimal cluster head selection rule in the DAPC algorithm, which is defined as follows: CH m =w1argminDF+w2argmin(-CH old ) Where w1+w2=1, the w1 value corresponding to this function should be greater than w2, and the values ​​of w1 and w2 depend on the performance of AP clustering.

4. The VANET clustering algorithm based on support vector machine according to claim 1 is characterized in that: In step 3, after the optimal cluster head selection is completed, the vehicle node will enter the cluster, which includes the following: The vehicle node uses the support vector machine model of the cluster head node to enter the cluster. At this time, the feature vector of the UN vehicle node is input into the support vector machine classifier carried by the cluster head node itself to make a cluster entry decision. If the classification result of the vehicle node is consistent with the cluster head classification, it is considered that the node will join the cluster and send a cluster entry request to the cluster head, along with its current location coordinates and the data in the average vehicle speed data structure. If the vehicle node is in three consecutive cluster head broadcast time periods T broadcast If a node fails to join any cluster successfully, it can initiate the establishment of a new cluster and serve as the cluster head node of the new cluster.

5. The VANET clustering algorithm based on support vector machine according to claim 1 is characterized in that: In step 4, after the vehicle node enters the cluster, cluster maintenance operations are performed according to the operating status; the steps include: (1) Cluster merging When the distance between two cluster heads traveling in the same direction reaches the set threshold, the two cluster heads exchange the feature vectors extracted from their respective data structures and input them into the support vector machine model carried by their respective vehicle nodes for training. The comparison results are consistent. If the above two conditions are met, the two clusters can be merged. The cluster head node of the newly merged cluster selects the cluster head of the larger cluster, that is, the cluster head of the cluster with a larger number of cluster members becomes the new cluster head node of the merged cluster, and the cluster head nodes of the smaller clusters become cluster members of the new cluster head node. (2) Cluster member disconnection If a member node in the cluster does not receive the broadcast information of the cluster head for three consecutive cluster head broadcast time periods, it is judged that the node has left the cluster and become a UN node. If the cluster head node does not receive the cluster report of a node in the cluster for three consecutive cluster head broadcast time periods, it is judged that the node has left the cluster. When all cluster members in the cluster are disconnected, the cluster head node will try to join the nearest cluster within its communication radius. If the joining is successful, the original cluster is considered to be dead. (3) Cluster head node disconnection For cluster member vehicle nodes, due to some special reasons, such as the communication hardware failure of the cluster head node or leaving the road, the cluster head changes its state and actively disconnects from all nodes in the cluster. When the cluster head node leaves the cluster, if there are still nodes in the current cluster, a new cluster head will be selected according to the cluster head selection algorithm.

Images