CN108881016B

CN108881016B - Method for constructing self-organizing network community of Internet of vehicles and balancing communication load of community nodes

Info

Publication number: CN108881016B
Application number: CN201811001998.1A
Authority: CN
Inventors: 程久军; 原桂远
Original assignee: Tongji University
Current assignee: Tongji University
Priority date: 2018-08-30
Filing date: 2018-08-30
Publication date: 2020-12-08
Anticipated expiration: 2038-08-30
Also published as: CN108881016A

Abstract

The existing community clustering method aiming at the Internet of vehicles mostly adopts the position of the vehicles at the future time such as the current speed, the driving direction and the like to estimate the time length of the connection between the nodes in the future, and the time length is used as one of clustering bases to improve the community stability. However, the position of the vehicle at the future moment is influenced by multiple factors, and the accuracy of the estimated communication connection time is low, so that the communities in the Internet of vehicles have the characteristics of short life cycle and rapid change of the attribution of the node communities. Meanwhile, the existing routing protocol of the car networking community takes key nodes in the community as main nodes for forwarding, so that the problem of overhigh load of the key nodes is caused. The invention discloses a multi-role classification community load balancing algorithm of an Internet of vehicles self-organizing network. The future time position of the vehicle is considered in the community clustering process, the role distribution is refined in the community, and the alternative nodes are selected for the key nodes to carry out load balancing, so that the community stability is improved, the key node load is balanced, and the data transmission distortion is reduced.

Description

Method for constructing self-organizing network community of Internet of vehicles and balancing communication load of community nodes

Technical Field

The invention relates to the field of car networking.

Background

(1) Aspect of Community stability

Different from the traditional community clustering, the moving speed of the vehicle nodes in the Internet of vehicles is high, and the topological relation among the nodes is changed rapidly. The traditional community clustering method is applied to the Internet of vehicles, and has the defects of short community life cycle and low stability.

In a community clustering method designed for the Internet of vehicles, researchers propose that the current position, the current speed and the end point position of each node of a node are comprehensively considered in community clustering, and the topological relation of a vehicle in a future period of time is estimated, so that the topological relation is used as a clustering basis to prolong the survival time of a community. According to the method, after the head nodes of the two communities enter a communication area, the community combination is carried out after a period of time is delayed until the head nodes of the communities keep a relatively stable communication state, so that frequent combination and splitting of the communities are reduced, and the community stability is improved. There is a document that the moving direction of a vehicle is added into a clustering basis, and the number of neighbor nodes and an untrusted factor are integrated to be used as a basis of community clustering. The method discussed above simply selects several of the most obvious vehicle location influencing factors, and discards a great deal of traffic information that may influence the future vehicle location. When the influence factors are utilized for community clustering, the stability of the community is not improved enough, and the clustered community still has the characteristics of short life cycle and fast node attribution change.

(2) Aspect of load balancing

The community clustering method for the Internet of vehicles generally performs role distribution on vehicle nodes to serve as a basis for selecting the routing protocol forwarding nodes of the Internet of vehicles community. A community clustering model has been proposed that divides vehicle nodes into head nodes (CH) and common nodes (CM). Also, one assigns different roles to nodes within the community, including a head node (CH), a gateway node (GW), and a Common Node (CN).

Disclosure of Invention

In summary, the current community clustering method for vehicle networking generally classifies vehicle nodes in a community into three categories, namely a head node, a gateway node and a common node. The head node is usually located at the central position of the community, and good connectivity is kept between the head node and each node in the community. Each community tends to have only one head node. The gateway node is a node in the community that maintains good direct connectivity with other communities. The other nodes are classified as normal nodes. According to the existing clustering method of the car networking community, in the routing protocol of the car networking community, a head node and a gateway node are used as key nodes for forwarding, and most of network loads in the community are borne. This load imbalance causes excessive data crowding at these critical nodes, which in turn causes link congestion, in heavy traffic situations.

In order to reduce the excessively high network load of the head nodes and the gateway nodes in the community, the invention is based on a vehicle position prediction model (a position prediction model construction method based on deep learning vehicle driving influence factors in a vehicle networking complex network applied by inventor of Chengdu et al in 2017, 8, 16 days (applicant: university of Tongji, patent application No. 2017107029220, publication No. 107609633A, and 2018, 01, 19 days), while considering the community stability, the role distribution is refined, and an alternative node selection strategy of the head nodes and the gateway nodes is designed, so as to balance the excessively high network load of the head nodes and the gateway nodes.

The invention improves the community stability, balances the load of key nodes and reduces the data distortion.

Therefore, the invention provides the following technical scheme for realization:

the research method is characterized in that the time length for keeping connection between future nodes is judged by estimating the position of a vehicle at the future time based on the current speed, the driving direction and the like of the existing community clustering method aiming at the Internet of vehicles, and the estimated time length is used as one of clustering bases to improve the community stability. Meanwhile, the existing routing protocol of the car networking community takes key nodes in the community as main nodes for forwarding, so that the problem of overhigh load of the key nodes is caused. Based on a vehicle position prediction model (disclosed in the prior invention application, patent application number 2017107029220), vehicle body attributes, road information, driving environment and other factors influencing vehicle driving are comprehensively considered, the relation between the vehicle driving influencing factors and the vehicle position is mined by combining a deep learning technology, the prediction model of the vehicle position is provided, the future time position of the vehicle is considered in the community clustering process, the role distribution is refined in the community, and alternative nodes are selected for key nodes to perform load balancing, so that the community stability is improved, the key node load is balanced, and the data transmission distortion is reduced.

A multi-role classification community load balancing algorithm of a self-organizing network of the Internet of vehicles is characterized by comprising three parts,

one, community attribute and community node role definition

Step 1, community attribute;

and 2, the community node has the role.

Second, community clustering and role distribution method

Step 1, forming a community;

and 2, community maintenance.

And thirdly, performing multi-role classification community load balancing algorithm.

The following detailed description

One, community attribute and community node role definition

Step (ii) of1.1, theCommunity attributes, defined as follows:

VANET is abstracted into the structure of a weighted undirected graph. The vehicle nodes serve as points in the undirected graph, and the wireless connections between vehicles serve as edges between the points. The undirected graph with weight is G ═ V_G，E_G). Wherein V_G＝{ve₁，ve₂，...ve_kDenotes the set of vertices, ve_kRepresenting vertex k. E_G＝{e₁，e₂，...e_jDenotes an edge set, e_jRepresenting edge j. The communities are represented by vertex sets and edge sets, and are abstracted into a G subgraph form and are marked as

If two vehicles are within the transmission range of each other, an edge is considered to exist between the nodes abstracted by the two vehicles.

(1) Definition 1 vehicle direct connectivity factor trf (trassmision factor): representing the reliability of the connection between the two vehicle nodes, satisfying equation (2).

Where tr (transmission range) represents the maximum transmission range of vehicle communication. dist_t(ve_i，ve_j) Representing the distance between vehicle i and vehicle j at time t. When the distance between the vehicles is greater than the maximum transmission range, the TRF is 0, indicating that there is no connection between the two vehicle nodes. Reflected on the topology map, i.e. there is no edge between two nodes. When the distance between the vehicles is less than or equal to the maximum transmission range, the TRF is inversely related to the distance between the vehicles. The closer the distance, the larger the TRF, the higher the reliability of the connection between two vehicle nodes, and the closer the connection, the greater the weight reflected to the upper side of the topological graph.

(2) Defining 2 neighbor nodes: if the vehicle A and the vehicle B satisfy TRF (ve)_A，ve_B)＞0，A and B are said to be neighbor nodes. Reflected in the topological graph, i.e. ve_AAnd ve_BThere is an edge in between.

The node neighbor connection centrality is defined to represent the number of current neighbor nodes of the vehicle and the closeness of the connection between the vehicle and the neighbor nodes.

(3) Define 3 Neighbor Connectivity Centrality (NCC): representing the sum of the direct connectivity factors of the node and the neighboring nodes. Will ve_iThe NCC value of the node at the time t is recorded as C_i，tThen, there are:

wherein the NS_iA set of neighbor nodes representing inodes. The larger the NCC, the more reliable the connection between the node and its surrounding nodes is, and the closer the connection is.

(4) Define 4 Expected Neighbor Connection Centrality (ENCC): representing a weighted average of the NCC values of the nodes for a future period of time, starting from the current time instant. Node ve at time t_iHas an ENCC value of

Then there are:

where DTS denotes a set of time intervals { dt | dt ═ 0, Δ t, 2 Δ t, 3 Δ t₁，w₂，...}， W_kRepresenting the weight corresponding to the NCC value at the moment of t + (i-1) delta t. The larger the ENCC value, the more reliable the connection of the vehicle to the surrounding nodes in the future period of time represented by the DTS, and the tighter the connection.

(5) Define 5 Community Neighbor Connection Centrality (CNCC): representing the sum of the direct connectivity factors of the node to the nodes in the specified community at the current time. Let t time ve_iNode pair community COM_kCNCC value of

Then there is

Wherein

Is denoted ve_iCOM (component object model) of node-in-community_kA set of neighboring nodes in (1). CNCC represents the quality of direct connection of the vehicle node to a community population. The bigger the CNCC value is, the more the COM value is, the more the node is currently connected with the community_kThe higher the overall quality of the direct connection, the tighter the connection.

(6) Define 6 expected community neighbor connection centrality (ENC): representing a weighted average of the CNCC values of the designated community by the node some time after the current time. Let t time ve_iNode pair community COM_kThe ENC value of (A) is recorded as

Is provided with

The larger the ENC value is, the more the COM value is, the COM value is with the community in the future_kThe higher the overall quality of the direct connection, the tighter the connection. The influence of the CNCC value at a certain moment can be expanded by adjusting the weight of the CNCC at each time point to focus on considering the CNCC value at the moment.

(7) Define 7 node remaining load capacity (AWL): indicating the buffer space remaining for the node to forward data at the current time. Will ve_iThe AWL value of the node at the time t is recorded as AWL_i，t. At time t, to ve_iNode transmission greater than AWL_i，tWill cause the packet to be lost.

Step 1.2 theCommunity node role definition

In order to distinguish the positions of different nodes in the community, the positions of the nodes in the community and the contribution degrees of the nodes in the Internet of vehicles routing are represented. The car networking community clustering model may give different role definitions to nodes in the community. Generally, a community has at least one head node (CH) and several member nodes (CMs). The head node in the community manages various information of the community, such as a community node set, the position of each node in the community, a routing table and the like. In the process of community evolution, a head node can often decide whether a node which is not the community can join the community. Because the direct connectivity of the head node is high, the communication quality is good, and various important information in the community is stored, the head node is often regarded as an important forwarding node in the routing protocol of the Internet of vehicles community.

The invention defines a head node waiting node and a gateway node waiting node for balancing loads of the head node and the gateway node, and divides node roles in a community into five classes, namely the head node, the gateway node, the head node waiting node, the gateway node waiting node and a common member node. The role of a node that is not included in the community is defined as a free node.

(1) Define 8 head node (CH): the node with the highest ENC value in the community. Is provided with

Then node ve_iIs considered to be in the future for a period of time

The connection quality is best in the intra-community. The node is selected as COM_kThe CH node of (1). And if a plurality of nodes with the highest ENC value exist, selecting the node with the lowest number as the CH node.

(2) Define 9 head node waiting node (BCH): the nodes with relatively high degree of center are adjacent to the expected community in the community. Assume COM for community_kIn other words, the CH node is ve_hThen, the BCH node of the community needs to satisfy:

h represents the number of the CH node. Here a control factor for the number of BCH nodes selected. At the beginning of community formation, the load of data transmission of the CH node and the BCH node is low, and can be set to a large value, so that the number of the BCH nodes is reduced, the connection quality of the selected BCH node in the community is closer to the CH node, and the forwarding characteristic of the BCH node is closer to the CH node. With the aggravation of the data packet forwarding task in the community, when both the CH node and the selected BCH node reach a very high data load, the number of the CH nodes can be reduced appropriately, and more BCH nodes are selected to assist in data forwarding.

(3) Defining 10 Community COM_kTo community COM_lGateway node set (GWS): defined as set GWS_k，lAnd satisfies the following conditions:

namely Community COM_kTo community COM_lAll nodes of the gateway node set belong to the community COM_kAnd is connected with the community COM_lSome or all of the nodes in (1) are directly connected.

(4) Defining 11 Community COM_kTo community COM_lGateway node GW of_k，l: defined as set GWS_k，lMiddle-to-community COM_lNode with the largest ENC value. I.e. if ve_i∈GWS_k，lThen, it needs to satisfy:

if GWS is_k，lCOM in community_lIf the number of the nodes with the maximum number of the ENCs is multiple, the node with the lowest number is selected as the community COM_kTo community COM_lA gateway node (GW).

(5) Defining 12 Community COM_kTo community COM_lGateway node keeperNode set BGWS_k，l: is defined as GWS_k，lRemoving gateway node GW_k，lAll nodes other than that, i.e.

BGWS_k，l＝{ve_i|ve_i≠GW_k，l，ve_i∈GWS_k，l} (11)

BGWSk，_lThe node in (1) is defined as a community COM_kTo community COM_lGateway node waiting node BGW_k，l。

Common member nodes (CMs) refer to all other community nodes except CH, BCH, GW, and BGW nodes in the community. The CM node conforms to the following two characteristics:

● CM node and single hop neighbor nodes that are community CH nodes. Namely for community COM_kThe CH node is CH_kThen CM node

And CH_kThere must be an edge e between_j，

● for node CM_iThe community of which is COM_kThen equation (12) is satisfied.

Nodes that do not belong to any community are defined as free nodes (SN).

II,Community clustering and role distribution method

The clustering of the communities in the Internet of vehicles is a process of dividing nodes in the Internet of vehicles into different communities. The invention divides the process of community clustering into two stages of community formation and community maintenance.

In the initialization stage of the self-organizing network of the Internet of vehicles, each vehicle node is initialized to be an SN node. After that, no matter in the community forming stage or the community maintenance stage, the node always periodically sends the neighbor nodeThe point broadcasts a heartbeat control (HB) packet. The HB data packet carries state information of the node, including the current position of the node, the future positions of 1s, 2s and 3s obtained by the node through prediction based on a vehicle position prediction model, the community to which the node belongs, the node ENCC value, the node ENC value, the node AWL value and the like. All nodes maintain a neighbor information table neighbor Table for calculating and comparing index data of neighbor nodes, such as ENCC and ENC. The individual contents of the information table are shown in table 1. The BEAT LENGTH in the table indicates the time interval between two adjacent HB packet transmissions. The neighbor node and ve are considered as if the HB data packet of longer time is missing_iThe node is disconnected and its information is removed from the neighbor information table.

Step 2.1 Community formation

Community formation begins with election of the community CH node. And when the node updates the self ENCC value, comparing the self ENCC value with the ENCC value of the neighbor node, thereby finding out the node with the maximum ENCC value. The node having the largest ENCC value is suitable to be selected as the CH node. And when a plurality of nodes with the maximum ENCC value exist, selecting the node with the minimum ID value as the CH node. And after the CH node is selected, marking the ID of the new community as the ID of the CH node, and enabling the neighbor node to know the establishment of the new community through an HB data packet broadcasted by the CH node. And if the neighbor node is the SN node, directly joining the community and informing the CH node in a mode of joining a community request (JC) data packet. If the neighbor node belongs to other nodes, whether the neighbor node is suitable for being added into a new community needs to be judged, and details are given in a community maintenance stage according to the judgment. And the CH node updates the community member table according to the received JC data packet. And if the JC data packet represents the request to join the community, adding the JC data packet sending node into a community member table. And if the JC data packet requests to join other communities, the node sending the data packet is moved out of the community member table. After several iterations, the community gradually tends to a relatively stable state. At this point, the process of community formation is completed.

Step 2.2 Community maintenance

In the community maintenance phase, the conditions of dynamic joining or leaving of the community by the nodes and merging, splitting and extinction of the community need to be processed. Meanwhile, the role distribution of the BCH, GW and BGW nodes is also carried out in the community maintenance phase.

(a) Updating node community attribution, eliminating community, merging and splitting

At this stage, the problem of node community attribution change needs to be handled. When the node receives the HB data packets broadcasted by the neighbor node, the time length of the connection between the node and the neighbor node is estimated according to the speed, the current position and the positions at the future time of 1s, 2s and 3s, and the information table of the neighbor node is updated.

After the neighbor node information table is updated, if the node finds that a new CH node appears in the neighbor node, whether the CH node is more suitable to be used as the CH node than the current community CH node is judged. Suppose node ve_iReceive ve_hTransmitted HB packets of which ve_hIs a newly appeared CH node in the neighbor nodes. Here, an improvement is made to the testclusterierheadchange algorithm (classical algorithm, commonly used for comparison experiments) in the MDMAC method (classical algorithm, commonly used for comparison experiments). Node ve_hThe following five criteria are satisfied to be considered more suitable for doing ve_iThe head node of (2):

6)ve_hthe node is a CH node of the community in which the node is located;

7) the ENCC value satisfies

8)ve_hAnd ve_iIs sufficiently long;

9)ve_iand ve_hThe included angle of the speed between the two is small enough;

10) let vei be COM_m，ve_hThe community is COM_nThen the ENC value needs to be satisfied

Meeting these criteria guarantees ve_hThe nodes have better direct connectivity, and ve is ensured_iNode and ve_hThe communities in which the nodes are located are more closely related. At the same time ensure ve_hAnd ve_iHave a longer time connection between them, can avoid ve_iFrequently joining and leaving communities.

If ve_hMore suitable as CH node, then ve_iSelectively add ve_hThe community in which the user is located. ve_iThe following operations are carried out according to the role of the user:

3) when ve_iWhen not a CH node, ve_iNotification of ve by broadcasting JC packets_hJoin the community. And notify the original community CH node ve_iLeaving the original community, and completing the attribution change of the node community;

4) when ve_iWhen it is a CH node, ve_iNotification of ve by broadcasting JC packets_hJoin the community. And other nodes in the community are informed, and the original community disappears. Other nodes within the community become SN nodes. And then according to the received HB data packets of other CH nodes, selecting to join other communities to complete the combination of the communities. Or self-electing to be a new community CH node, attracting surrounding nodes to join, and completing the division of the community.

Therefore, the node can complete the dynamic change of the node community attribution, the extinction of the community, the splitting and the combination according to the given rule.

(b) BCH, GW and BGW node role allocation

The community maintenance phase also needs to solve the role allocation problem of BCH, GW and BGW nodes. The CH and BCH nodes maintain the same community member list. And the CH and BCH nodes respectively maintain a gateway information table according to the updated information of the GW and the BGW nodes in the neighbor nodes. And the CH node maintains a community member list and knows the community status of each node in the community. And common nodes of the community do not have a community member list, so that the status of each node in the community cannot be known. Therefore, to select a node having a similar community status as the CH node as a BCH node, the BCH node is designated by the CH node. And other nodes in the community can know the adjacent community and can calculate the ENC value of the adjacent community. Therefore, the GW and BGW nodes are self-elected by each node in the community and inform the result to the CH node and the adjacent BCH node.

The BCH node is more suitable to be used as a BCH node according to the size of the ENC value and the closer to the ENC value of the CH node.

And selecting the GW and the BGW nodes according to the ENC value of the adjacent community. Suppose COM_iThere are 5 nodes and COM in the community_jNodes in the community are directly connected. The 5 nodes are represented as<ve₁，ve₂，ve₃，ve₄，ve₅>. It is to community COM_jThe ENC value of (A) is shown in Table 2.

To COM_jMaximum ve of community ENC₃The node is selected as the community COM_iTo community COM_jGW node of (1), the remaining 4 nodes being selected as community COM_iTo community COM_jThe BGW node of (1). The GW is distinguished from the BGW node only at the CH and BCH nodes because the gateway node tables of the community are maintained only at the CH and BCH nodes. In the self-election process of the gateway node, the nodes all consider themselves to be the BGW nodes.

The CH and BCH nodes need to maintain a community membership list that includes a CH node list (only one CH node in the list), a BCH list, a GW/BGW list, and a CM list. The CH node list, the BCH list and the GW/BGW list are collectively called a key node member list. In the process of choosing BCH and GW/BGW, the CH node and the corresponding BCH node update the member list of the key nodes according to the choosing result. After the CM node joins the community, the CH node and the corresponding BCH node update the CM list. And when the member node in the community leaves the community, the CH node and the BCH node remove the corresponding member, and corresponding new role selection is performed according to the role of the removed member. The member nodes leave the community can be divided into the following two cases:

1) the member node joins other communities and actively informs the CH node and the corresponding BCH node. At this time, the CH node and the BCH node may remove the member nodes leaving the community from the corresponding list.

2) The member node loses connection with the CH node. And the CH node does not receive the HB data packet broadcasted by the member node after a period of time, and the updating of the community member table is completed.

Updating NodeB except when receiving neighbor node's HB or other notification packetBesides the maintained list items, each node in the community can also periodically update the node state of the node so as to discover the condition that the neighbor nodes lose connection. Meanwhile, the CH node needs to select or update a BCH and GW/BGW list. The time interval between two active node updates is BEAT LENGTH. Community COM_kNode ve_iThe contents of the update at time t are as follows:

5) node to ENC value for current community

6) The expected remaining connection time of the neighboring node is reduced by 1. If the value drops to 0, the neighbor node is considered to be no longer connected to the current node.

7) If ve_iAnd if the node is a community CH node, selecting or updating the BCH node at the same time, and informing all nodes in the community of the selected or updated result.

8) And updating the ENC values of the nodes and other communities. If the value changes, the CH node and the BCH node are informed, and the community GW node information is updated.

Wherein, the step 2) is a processing step for ensuring the validity of the neighbor node information. Since the node may disconnect from the neighbor node at any time, the node can only acquire the neighbor node information through the HB data packet periodically broadcast by the neighbor node. Thus, upon receiving the HB packet, the node estimates the connection time ECT with the neighbor node. The ECT is decremented by 1 each time a node actively updates state. When the ECT drops to 0, it indicates that no HB packet was received from the neighbor node for a period of time. At this point, the current node is considered to be no longer connected with the neighbor node.

Considering the community stability and the load balancing problem of the community nodes,

multi-role classification community load balancing algorithm

In a routing protocol based on community clustering, a community head node and a gateway node often play an important role in forwarding a data packet, and are frequently selected as forwarding nodes of the data packet. At present, the traditional AODV routing algorithm is improved, and routing tables among communities are stored in head nodes. When a node in the community sends a data packet to a non-neighbor node, the data packet is sent to a head node firstly. And the head node judges whether the destination node is a node in the community. If yes, the data packet is directly forwarded to the destination node, and if not, the data packet is forwarded to the corresponding gateway node according to the maintained routing table information. Therefore, the head node and the gateway node become key nodes in the packet transmission path, and the load is heavy. The invention provides an algorithm for balancing loads of the CH node and the GW node according to the role definition of the community node so as to solve the problem of overhigh loads of the CH node and the GW node.

And when the CH node is required to be selected for data forwarding, the BCH node is used as an alternative node for head node load balancing. The CH nodes and all BCH nodes in the community are considered as the head node pool (CHP) of the community. And selecting nodes with relatively high ENC values and strong residual load capacity from the pool to forward data. The value of τ is designed here as the selected evaluation factor. And selecting the node with the maximum tau value as a head node for data forwarding. For time t, ve_iτ value τ of node_i，tIs defined as

To make τ_i，tAnd

and AWL_i，tAre all in positive correlation, taken here

L (AWL) is close to 1 when the node residual load capacity AWL is at a relatively high level, at which time

May become the dominant factor of choice. When AWL is below the threshold, l (AWL) approaches 0 quickly, and AWL becomes the dominant factor of choice. At this time theThe node load is already relatively heavy and the corresponding τ value is relatively small. Where AWL is in KB. According to the selection rule of the BCH node, the BCH node and the CH node have similar ENC values. Therefore, nodes with higher AWL are more likely to be selected. Therefore, the nodes in the CHP share the load mutually, and the aim of relatively balancing the load is achieved. Tau is_i，tThe curve of (a) is shown in fig. 4. In the figure, ρ is 3, μ is-1.67, and the x axis represents

The y-axis represents AWL_i，tZ-axis represents τ_i，t。

When the CH node or the BCH node needs to select the GW node for data forwarding, the BGW node may be used to share the load of the GW node. And when selecting, regarding the GW node and the BGW node as a gateway node pool (GWP). And selecting a proper evaluation factor, and finding out a node which is closely associated with the current community and the target community and has higher residual load capacity as a gateway node for data forwarding by using the evaluation factor.

Description of the attached tables

TABLE 1 simulation experiment parameters for Internet of vehicles with infrastructure

TABLE 2 node ENC value example

Table 3 experimental parameter settings

Drawings

FIG. 1 is a topology diagram of an ad hoc network of the Internet of vehicles

FIG. 2 an example of a gateway node set for an ad hoc network of the Internet of vehicles

FIG. 3 Community node role switching

FIG. 4. tau_i，tCurve of values

FIG. 5 TAPASColone dataset road topology

FIG. 6 distribution of Community population percentage over Community survival time

FIG. 7 distribution of node quantity with node community change times

FIG. 8 distribution of percentage of node number with node packet forwarding

FIG. 9 high load node packet forwarding distribution

FIG. 10 packet delivery rate comparison

FIG. 11 average end-to-end delay comparison

FIG. 12 is a flow chart of the method of the present invention

Detailed Description

The specific implementation process of the invention is shown in fig. 12, and comprises the following 7 aspects:

(ii) Community Attribute definition

② definition of community node role

Method for community clustering and role distribution

Four multi-role classification community load balancing algorithm

Experiment of

Analysis of Community stability results

Seventhly, analyzing routing load balancing performance of community nodes

①

Community attribute definition

The invention abstracts the VANET into a structure of a weighted undirected graph. The vehicle nodes serve as points in the undirected graph, and the wireless connections between vehicles serve as edges between the points. The undirected graph with weight is G ═ V_G，E_G). Wherein V_G＝{ve₁，ve₂,. } represents a set of vertices, ve_kRepresenting vertex k. E_G＝{e₁，e₂,.. } represents an edge set, e_jRepresenting edge j. The VANET topology is shown in fig. 1. The community is abstracted into the form of G subgraph, which is recorded as

If two vehicles are within the transmission range of each other, an edge is considered to exist between the nodes abstracted by the two vehicles. Weight w of edge_eIs the vehicle direct connectivity factor (see definition 1). Weight w of a node_veThe expected neighbor is connected to the centrality (see definition 4).

(2) Defining 2 neighbor nodes: if the vehicle A and the vehicle B satisfy TRF (ve)_A，ve_B) If the number is more than 0, the A and the B are called as neighbor nodes. Reflected in the topological graph, i.e. ve_AAnd ve_BThere is an edge in between.

In the internet of vehicles, there are many neighbor nodes of some vehicles and many nodes capable of directly communicating. There are few neighbor nodes of some vehicles, and there are few nodes capable of direct communication. The node neighbor connection centrality is defined to represent the number of current neighbor nodes of the vehicle and the closeness of the connection between the vehicle and the neighbor nodes.

(4) Define 4 prospective neighbor connection centrality (ENCC): representing a weighted average of the NCC values of the nodes for a future period of time, starting from the current time instant. Node ve at time t_iHas an ENCC value of

Then there are:

where DTS denotes a set of time intervals { dt | dt ═ 0, Δ t, 2 Δ t, 3 Δ t₁，w₂，...}， w_iRepresenting the weight corresponding to the NCC value at the moment of t + (i-1) delta t. The larger the ENCC value, the more reliable the connection of the vehicle to the surrounding nodes in the future period of time represented by the DTS, and the tighter the connection.

Then there is

Wherein

(6) Define 6 expected community neighbor connection centrality (ENC): representing a weighted average of the CNCC values of the designated community by the node some time after the current time. Let t time ve_iNode pair community COM_kENC value ofMaking

Is provided with

Wherein WS and DTS have the meaning as defined in definition 4. The larger the ENC value is, the more the COM value is, the more the_kThe higher the overall quality of the direct connection, the tighter the connection. The influence of the CNCC value at a certain moment can be expanded by adjusting the weight of the CNCC at each time point to focus on considering the CNCC value at the moment.

②

Community node role definition

In addition to the head node and the member nodes, in order to study the connection between the current community and other communities, the concept of a gateway node (GW) is added to the partial community clustering method. The gateway node belongs to one community and is directly connected with nodes of other communities. Generally, the number of gateway nodes in a community and the connection quality of the gateway nodes and other communities can represent the closeness of the community with other communities. In a routing protocol based on community clustering, a gateway node is often used as a relay node for inter-community communication and is responsible for forwarding data packets to other communities.

In the routing protocol based on community clustering, it is because of the special status of the head node and the gateway node, which results in high load of data forwarding. The invention defines a head node waiting node and a gateway node waiting node for balancing loads of the head node and the gateway node, and divides node roles in a community into five classes, namely the head node, the gateway node, the head node waiting node, the gateway node waiting node and a common member node. The role of a node that is not included in the community is defined as a free node.

Then node ve_iIs considered to be in the future for a period of time

here a control factor for the number of BCH nodes selected. At the beginning of community formation, the load of data transmission of the CH node and the BCH node is low, and can be set to a large value, so that the number of the BCH nodes is reduced, the connection quality of the selected BCH node in the community is closer to the CH node, and the forwarding characteristic of the BCH node is closer to the CH node. With the aggravation of the data packet forwarding task in the community, when both the CH node and the selected BCH node reach a very high data load, the number of the CH nodes can be reduced appropriately, and more BCH nodes are selected to assist in data forwarding.

namely Community COM_kTo community COM_lAll nodes of the gateway node set belong to the community COM_kAnd is connected with the community COM_lSome or all of the nodes in (a) are directly connected as shown in fig. 2.

(5) Defining 12 Community COM_kTo community COM_lGateway node waiting node set BGWS_k，l: is defined as GWS_k，lRemoving gateway node GW_k，lAll nodes other than that, i.e.

BGWS_k，l＝{ve_i|ve_i≠GW_k，l，ve_i∈GWS_k，l} (11)

BGWS_k，lThe node in (1) is defined as a community COM_kTo community COM_lGateway node waiting node BGW_k，l。

And CH_kThere must be an edge e between_j，

Nodes that do not belong to any community are defined as free nodes (SN). The role transitions of the nodes are shown in fig. 3. In one community, the CH node and the BCH node may have the identity of the GW node or the BGW node. The arrow direction in the figure indicates community node role switching, (B) CH indicates a CH or BCH node, (B) GW indicates a GW or BGW node, and (B) CH (B) GW node indicates a node having both (B) CH role and (B) GW role.

③

Community clustering and role distribution method

The clustering of the communities in the Internet of vehicles is a process of dividing nodes in the Internet of vehicles into different communities. The basis of dividing the nodes is different due to different community clustering algorithms. The invention divides the process of community clustering into two stages of community formation and community maintenance.

In the initialization stage of the self-organizing network of the Internet of vehicles, each vehicle node is initialized to be an SN node. After that, whether in the community formation phase or the community maintenance phase, the node broadcasts heartbeat control (HB) data packets to the neighbor nodes periodically. The HB data packet carries state information of the node, including the current position of the node, the future positions of 1s, 2s and 3s obtained by the node through prediction based on a vehicle position prediction model, the community to which the node belongs, the node ENCC value, the node ENC value, the node AWL value and the like. All nodes maintain a neighbor information table neighbor Table for calculating and comparing index data of neighbor nodes, such as ENCC and ENC. The individual contents of the information table are shown in table 1. The BEAT _ LENGTH in the table indicates the time interval between two adjacent HB packet transmissions. The neighbor node and ve are considered as if the HB data packet of longer time is missing_iThe node is disconnected and its information is removed from the neighbor information table.

(1) Community formation

(2) Community maintenance

After the neighbor node information table is updated, if the node finds that a new CH node appears in the neighbor node, whether the CH node is more suitable to be used as the CH node than the current community CH node is judged. Suppose node ve_iReceive ve_hTransmitted HB packets of which ve_hIs a newly appeared CH node in the neighbor nodes. Here an improvement is made to the testClusterHeadChange algorithm in the MDMAC method. Node ve_hThe following five criteria are satisfied to be considered more suitable for doing ve_iThe head node of (2):

11)ve_hthe node is a CH node of the community in which the node is located;

12) the ENCC value satisfies

13)ve_hAnd ve_iIs sufficiently long;

14)ve_iand ve_hThe included angle of the speed between the two is small enough;

15) suppose ve_iThe community is COM_m，ve_hThe community is COM_nThen the ENC value needs to be satisfied

Meeting these criteria guarantees ve_hThe node has betterDirect connectivity of, and guarantee ve_iNode and ve_hThe communities in which the nodes are located are more closely related. At the same time ensure ve_hAnd ve_iHave a longer time connection between them, can avoid ve_iFrequently joining and leaving communities.

5) when ve_iWhen not a CH node, ve_iNotification of ve by broadcasting JC packets_hJoin the community. And notify the original community CH node ve_iLeaving the original community, and completing the attribution change of the node community;

6) when ve_iWhen it is a CH node, ve_iNotification of ve by broadcasting JC packets_hJoin the community. And other nodes in the community are informed, and the original community disappears. Other nodes within the community become SN nodes. And then according to the received HB data packets of other CH nodes, selecting to join other communities to complete the combination of the communities. Or self-electing to be a new community CH node, attracting surrounding nodes to join, and completing the division of the community.

(b) BCH, GW and BGW node role allocation

Besides updating the table entries maintained by the nodes themselves when receiving the HB or other notification packets of the neighboring nodes, each node in the community also periodically updates the node state thereof to discover the condition that the neighboring nodes lose connection. Meanwhile, the CH node needs to select or update a BCH and GW/BGW list. The time interval between two active node updates is BEAT _ LENGTH. Community COM_kNode ve_iThe contents of the update at time t are as follows:

9) node to ENC value for current community

10) The expected remaining connection time of the neighboring node is reduced by 1. If the value drops to 0, the neighbor node is considered to be no longer connected to the current node.

11) If ve_iAnd if the node is a community CH node, selecting or updating the BCH node at the same time, and informing all nodes in the community of the selected or updated result.

12) And updating the ENC values of the nodes and other communities. If the value changes, the CH node and the BCH node are informed, and the community GW node information is updated.

④

Multi-role classification community load balancing algorithm

To make τ_i，tAnd

and AWL_i，tAre all in positive correlation, taken here

Can become the dominant factor of selection. When AWL is below the threshold, l (AWL) approaches 0 quickly, and AWL becomes the dominant factor of choice. At this time, the node is already relatively heavily loaded, and the corresponding τ value is relatively small. Where AWL is in KB. According to the selection rule of the BCH node, the BCH node and the CH node have similar ENC values. Therefore, nodes with higher AWL are more likely to be selected. Therefore, the nodes in the CHP share the load mutually, and the aim of relatively balancing the load is achieved. Tau is_i，tThe curve of (a) is shown in fig. 4. In the figure, ρ is 3, μ is-1.67, and the x axis represents

The y-axis represents AWL_i，tZ-axis represents τ_i，t。

⑤

Experiment of

(1) Experimental data set

The simulation experiment of the invention adopts a vehicle networking simulation framework of Venns, and the data set is TAPASColone. The TAPASColone dataset is derived from the Institute of Transportation Systems at the German Aerospace Center (ITS-DLR), and contains a road topology network of a real German city, Koron. The moving track of the microscopic vehicle nodes in the urban road can be simulated by SUMO traffic simulation software to describe the traffic flow of 24 hours in the 400 square kilometer area of the German Kolon city. The data set road topology is shown in fig. 5.

According to the invention, vehicle track data in a time period of 6: 00-8: 00(am) is adopted in an experiment to provide a vehicle node movement basis for vehicle networking simulation. In the time period, at most more than 8000 vehicle nodes can run on the road simultaneously, and the requirement of the experiment of the invention on vehicle track simulation can be met.

(2) Experimental methods

And respectively using an MDMAC method and the PMRC method to perform community clustering on vehicles in a period of 6: 00-8: 00(am) in the TAPASColone data set and dynamically updating the community form of the vehicles. And (4) counting the survival time of the VANET community formed under the two methods and the number of times of change of the node community in the community. The statistical results were compared to compare the stability of the communities formed by the PMRC and MDMAC methods. After the community is formed, a source node and a destination node for sending data packets are selected according to the same random mode, the data packets are sent periodically, and the number of data packets forwarded by each node in the community, the delivery rate of the data packets and the average end-to-end delay are recorded. And comparing the statistical results to compare the performances of the VANET community obtained by the two methods in the community routing of the Internet of vehicles. Wherein the packet delivery rate represents the ratio of the number of packets successfully arriving at the destination node to the total number of packets to be transmitted. The average end-to-end delay represents the average length of time required for a data packet to be sent by a source node to a destination node.

The selection rule of the data forwarding node is as follows:

1) if the destination node is a direct neighbor node of the current data packet holding node, directly forwarding the data packet to the destination node;

2) if the current data packet holding node is a CM node, selecting a node from the head node pool as a next hop for forwarding the data packet;

3) if the current data packet holding node is a CH node or a BCH node, whether a destination node is a node of the community is judged firstly. If so, the data packet is directly sent to the destination node. Otherwise, selecting a proper community from the neighbor communities to forward the data packet, and selecting a node from the gateway node pool as a next hop for forwarding the data packet;

4) if the current data packet holding node is a GW node or a BGW node, selecting a next forwarding community node in a neighbor node table as a next hop of data packet forwarding according to the next forwarding community information marked by the data packet;

5) if the 4 conditions are not met, the node continues to hold the data packet and waits for the four sending conditions to be met.

In the formula (15), ρ is 1 and μ is 20 KB. The remaining experimental parameter settings are shown in table 3.

⑥

Community stability result analysis

FIG. 6 shows the community proportion of different community survival time periods obtained by the PMRC method and the MDMAC method. From left to right in the figure, the community proportion of community survival time in the intervals (0, 20s ], (20s, 40s ], (480s, 500s ]) is respectively shown, as the moving speed of the vehicle nodes is high, the proportion of communities is rapidly reduced along with the increase of the community survival time, in the MDMAC community clustering, the community survival time is less than 66.76% of all communities of 20s, the community survival time obtained by the PMRC method clustering is less than 51.77% of all communities of 20s, in the MDMAC method, the average survival time of the communities is 42.91 seconds, the average survival time obtained by the PMRC method clustering is 75.95 seconds, and therefore, the PMRC method clustering has longer survival time than the communities obtained by the MDMAC method clustering.

The number of node community changes refers to the number of times the node experiences joining a new community. FIG. 7 shows the distribution of the number of nodes with the number of changes to the community of nodes. It can be seen from the figure that, as the number of times of change of the node community increases, the number of nodes corresponding to the statistical result obtained by the MDMAC and PMRC methods increases rapidly and then decreases gradually. In the MDMAC method, the number of nodes that undergo 7 community changes is the largest. The number of nodes that undergo 6 community changes in the PMRC method is the largest. The number of nodes of the PMRC method which experience community change for less than 8 times is larger than that of the MDMA method. The MDMAC method is more than the PMRC method in the number of nodes that undergo more than 8 community changes. The average community change frequency experienced by the nodes in the MDMAC method is 9.04 times, and the average community change frequency experienced by the nodes in the PMRC method is 8.45 times. Therefore, the number of community changes experienced by nodes in the PMRC method is generally less than the MDMAC method. In conclusion, the stability of the internet of vehicles community formed by the PMRC method is superior to that of the MDMA method.

⑦

Community node routing load balancing performance analysis

Fig. 8 shows a distribution diagram of the proportion of the number of nodes with the number of forwarded packets when the packet transmission rate is 0.3 packets/second (p/s). The node proportion of the node forwarding data packet quantity in the intervals of [0, 500], (500, 1000),. and., (14000, 14500] is respectively shown from left to right in the figure, it can be seen from the figure that most of the node forwarding data packets in the communities obtained by the PMRC and MDMAC methods are lower than 500.

Fig. 9 shows the distribution of the proportion of high-load nodes to all nodes with the number of forwarded packets. As can be seen from the figure, in the community formed by the MDMAC method, there is no corresponding load balancing node, and there are a few nodes that forward too many packets. The PMRC method considers load balancing, and when the node load is too high, the residual load capacity is used as a main factor for selecting the forwarding node. Therefore, in fig. 9, the PMRC method avoids the occurrence of excessively high load nodes in the community, and in turn, the proportion of nodes in the interval (3500, 5500) of the number of forwarded packets is increased.

FIG. 10 shows a comparison of packet delivery rates for community-based Internet of vehicles community routing using MDMA and PMRC methods. As can be seen from the figure, when the packet sending rate is low, the packet delivery rate of the community obtained by the PMRC and the MDMAC in the Internet of vehicles community routing is similar. The community stability formed by the PMRC method is higher than that formed by the MDMA method, and the established community routing connection is relatively more stable, so that the delivery rate corresponding to the PMRC method is higher than that of the MDMA method. With the increase of the data packet sending rate, the load of the forwarding node is increased, and the data packet delivery rates of the two methods are in a gradually-decreasing trend. As the MDMAC method does not select alternative sending nodes for the head node in the community, the data packet sending rate is over 0.4p/s, the data is excessively attached to the head node, and the data packet delivery rate is rapidly reduced. The community obtained by the PMRC method selects alternative nodes, namely BCH nodes and BGW nodes, for the head node and the gateway node. When the data sending rate is increased and the loads of the head node and the gateway node are heavy, a proper data forwarding node is selected from the CHP node pool and the GWP node pool, so that excessive data attachment set is avoided. Therefore, the data packet delivery rate is slowly reduced after the data transmission frequency is more than 0.4 p/s. When the data packet sending rate is greater than 0.7p/s, due to the limitation of the selected number of BCH nodes and BGW nodes, the nodes start to approach a load saturation state, and the data packet delivery rate also drops rapidly. At this time, the number of BCH nodes and BGW nodes in the community can be considered to be increased to deal with higher data forwarding load. In general, compared with the MDMAC method, the PMRC method forms a community with a higher packet delivery rate in the routing of the car networking community, that is, the probability of data distortion is smaller.

Fig. 11 shows the trend of average end-to-end delay versus packet transmission rate. As can be seen from the figure, as the sending rate of the data packet increases, the average end-to-end delay generated in the community routing tends to increase in the community formed by the MDMAC method and the PMRC method. When the sending rate of the data packet is lower (the sending rate is less than or equal to 0.4p/s), the node load is lower, the community forwarding characteristics formed by the two methods are similar, and the end-to-end delay difference is not large. With the increase of the sending rate of the data packets, the load of key nodes in the community formed by the MDMAC method gradually increases and tends to be saturated, so that the forwarding performance is reduced, and the average end-to-end delay is increased more rapidly than that of the PMRC. When the sending rate of the data packet is greater than 0.7p/s, the alternative nodes in the PMRC community gradually tend to be saturated in load, and when the node cannot find a proper node with lower load to forward the data packet, the node holds the data packet for a long time to wait for the occurrence of the proper node, so that the average end-to-end delay is rapidly increased. It is contemplated that the number of candidate nodes may be increased at this time.

In conclusion, the community formed by the PMRC method provided by the invention has higher stability than the community formed by the MDMAC method, can effectively balance the node load, improve the delivery rate of data packets, reduce data distortion, reduce end-to-end delay and have better community routing forwarding performance.

The innovation points are as follows: based on a vehicle position prediction model (inventor of Chengdu et al, applied in 2017, 8, 16, and applied in 'a position prediction model construction method for vehicle driving influence factors based on deep learning in a vehicle networking complex network' (applicant: university, patent application No. 2017107029220), a multi-role classification community clustering method under a vehicle networking self-organization network scene is provided, so that community stability is improved, key node load is balanced, and data transmission distortion is reduced.

The existing community clustering method aiming at the Internet of vehicles mostly adopts the position of the vehicles at the future time such as the current speed, the driving direction and the like to estimate the time length of the connection between the nodes in the future, and the time length is used as one of clustering bases to improve the community stability. However, the position of the vehicle at the future moment is influenced by multiple factors, and the accuracy of the estimated communication connection time is low, so that the communities in the Internet of vehicles have the characteristics of short life cycle and rapid change of the attribution of the node communities. Meanwhile, the existing routing protocol of the car networking community takes key nodes in the community as main nodes for forwarding, so that the problem of overhigh load of the key nodes is caused. The invention provides a multi-role classification community clustering method under a self-organizing network scene of the internet of vehicles aiming at the problems, the method comprehensively considers the factors influencing the vehicle running such as vehicle body attributes, road information, running environment and the like, and combines a deep learning technology to mine the relation between the vehicle running influencing factors and the vehicle position and provide a prediction model of the vehicle position. ) And in the community clustering process, the future time position of the vehicle is considered, the role distribution is refined in the community, and the alternative nodes are selected for the key nodes to carry out load balancing, so that the community stability is improved, the key node load is balanced, and the data transmission distortion is reduced.

Attached table of the specification

TABLE 1

TABLE 2

TABLE 3

Claims

1. A method for building a communication load balance of a self-organizing network community and community nodes of an Internet of vehicles is characterized in that,

step one, defining and constructing a self-organizing network community of the Internet of vehicles

Define one

Abstracting a VANET into a structure with a right undirected graph; the vehicle nodes are taken as points in the undirected graph, and the wireless connection between the vehicles is taken as an edge between the points; the undirected graph with weight is G ═ V_G，E_G) (ii) a Wherein V_G＝{ve₁，ve₂，...ve_kDenotes the set of vertices, ve_kRepresents vertex k; e_G＝{e₁，e₂，...e_jDenotes an edge set, e_jRepresents edge j; the communities are represented by vertex sets and edge sets, and are abstracted into a G subgraph form and are marked as

If two vehicles are in the transmission range of each other, an edge is considered to exist between the abstract nodes of the two vehicles;

(1) definition 1 vehicle direct connectivity factor, TRF: representing the reliability of the connection between two vehicle nodes, and satisfying the formula (2);

wherein TR represents a maximum transmission range of vehicle communication; dist_t(ve_i，ve_j) Represents the distance between the vehicle i and the vehicle j at the moment t; when the distance between the vehicles is greater than the maximum transmission range, the TRF is 0, and the two vehicle nodes are not connected; reflected on the topological graph, namely that no edge exists between the two nodes; when the distance between the vehicles is less than or equal to the maximum transmission range, the TRF is inversely related to the distance between the vehicles; the closer the distance is, the larger the TRF is, the higher the reliability of the connection between two vehicle nodes is, the closer the connection is, and the greater the weight reflected to the upper side of the topological graph is;

(2) defining 2 neighbor nodes: if the vehicle A and the vehicle B satisfy TRF (ve)_A，ve_B) If the number is more than 0, the A and the B are called as neighbor nodes; reflected in the topological graph, i.e. ve_AAnd ve_BAn edge is arranged between the two edges;

the node neighbor connection centrality is defined to represent the number of the current neighbor nodes of the vehicle and the contact tightness between the vehicle and the neighbor nodes;

(3) define 3 neighbor connection centrality: representing the sum of direct connectivity factors of the node and the neighbor nodes; will ve_iThe neighbor connection centrality value of the node at the time t is recorded as C_i，tThen, there are:

wherein the NS_iA set of neighbor nodes representing inodes; the larger the connection centrality of the neighbor is, the higher the reliability of the connection between the node and the nodes around the node is, and the closer the connection is;

(4) define 4 prospective neighbor connection centrality: representing the weighted average of the neighbor connection centrality values of the nodes in a future period of time from the current moment; node ve at time t_iIs recorded as the expected neighbor connection center value

Then there are:

where DTS denotes a set of time intervals { dt | dt ═ 0, Δ t, 2 Δ t, 3 Δ t₁，w₂，...}，w_kRepresenting the weight corresponding to the value of the connection center of the neighbor at the moment of t + (i-1) delta t; the larger the expected neighbor connection centrality value is, the higher the connection reliability of the vehicle and the surrounding nodes in a future period of time represented by the DTS is, and the closer the connection is;

(5) defining 5 community neighbor connection centrality: representing the sum of direct connectivity factors of the nodes in the designated community at the current moment; let t time ve_iNode pair community COM_kThe value of the connection center of the community neighbor is recorded as

Then there is

Wherein

Is denoted ve_iCOM (component object model) of node-in-community_kA set of neighboring nodes; the community neighbor connection center meter represents the direct connection quality between the vehicle node and the community; the larger the value of the connection center of the community neighbors is, the current node and the community COM are represented_kThe higher the overall direct connection quality, the tighter the connection;

(6) defining 6 prospective community neighbor connection centrality: representing the weighted average of the community neighbor connection centrality values of the designated community at a period of time after the current moment; let t time ve_iNode pair community COM_kThe expected community neighbor connection centrality value is recorded as

Is provided with

The larger the expected community neighbor connection centrality value is, the node is represented to be in the COM with the community in the future_kThe higher the overall direct connection quality, the tighter the connection;

(7) defining 7 node residual load capacity: representing the buffer space left by the node at the current moment for forwarding data; will ve_iThe node residual load capacity value of the node at the time t is recorded as AWL_i，t(ii) a At time t, to ve_iNode transmission greater than AWL_i，tWill cause the packet to be lost;

definition II:

defining a head node waiting node and a gateway node waiting node, and dividing node roles in the community into five classes, namely a head node, a gateway node, a head node waiting node, a gateway node waiting node and a common member node; defining the role of the node which is not included in the community as a free node;

(1) define 8 head nodes: connecting a node with the highest centrality value with an expected community neighbor in the community; is provided with

Node ve_iFor a period of time in the future

Is selected as COM_kThe head node of (1);

(2) defining 9 head node waiting nodes: the nodes with high adjacent centrality of the expected community in the community; for community COM_kIn other words, the head node is ve_hThen, the head node waiting node of the community needs to satisfy:

h represents the number of the head node; the head node waiting nodes select a number of control factors; at the beginning of community formation, the load of data transmission of the head node and the head node waiting node is low; with the aggravation of a data packet forwarding task in a community, a head node waiting node is added to assist data forwarding;

(3) defining 10 Community COM_kTo community COM_lThe set of gateway nodes of: defined as set GWS_k，lAnd satisfies the following conditions:

namely Community COM_kTo community COM_lAll nodes of the gateway node set belong to the community COM_kAnd is connected with the community COM_lSome or all of the nodes in the network are directly connected;

(4) defining 11 Community COM_kTo community COM_lGateway node GW of_k，l: defined as set GWS_k，lMiddle-to-community COM_lThe expected community neighbor of (1) is connected with the node with the maximum centrality value; i.e. if ve_i∈GWS_k，lThen, it needs to satisfy:

if GWS is_k，lCOM in community_lIf a plurality of nodes with the maximum connection centrality of the expected community neighbors are provided, the node with the lowest number is selected as the community COM_kTo community COM_lThe gateway node of (2);

(5) defining 12 Community COM_kTo community COM_lGateway node waiting node set BGWS_k，t: is defined as GWS_k，lRemoving gateway node GW_k，lAll nodes other than that, i.e.

BGWS_k，l＝{ve_i|ve_i≠GW_k，l，ve_i∈GWS_k，l} (11)

BGWS_k，lThe node in (1) is defined as a community COM_kTo community COM_lGateway node waiting node BGW_k，l；

The common member nodes refer to a head node, a head node waiting node, a gateway node and all other community nodes except the gateway node waiting node in the community; the common member node conforms to the following two characteristics:

the common member node is a single-hop neighbor node of the head node in the community; namely for community COM_kThe head node is CH_kThen common member node

And CH_kThere must be an edge e between_j，

For node CM_iThe community of which is COM_kThen equation (12) is satisfied;

nodes not belonging to any community are defined as free nodes;

step two, community formation and maintenance, namely performing community clustering on vehicles in the TAPASColone data set and dynamically updating the community form of the vehicles;

step 2.1 Community formation

Community formation begins with election of a community head node; when the node updates the connection center value of the expected neighbor, the connection center value of the expected neighbor is compared with the connection center value of the expected neighbor of the neighbor node, so that the node with the maximum connection center value of the expected neighbor is found; the node with the maximum expected neighbor connection centrality value is suitable to be selected as a head node; when a plurality of nodes with the maximum expected neighbor connection centrality values exist, selecting the node with the minimum ID value as a head node; after the head node is selected, marking the ID of the new community as the ID of the head node, and enabling the neighbor nodes to know the establishment of the new community through an HB data packet broadcasted by the head node; if the neighbor node is the SN node, directly joining the community, and informing the head node in the form of a request JC data packet for joining the community; if the neighbor node belongs to other nodes, judging whether the neighbor node is suitable for being added into a new community, and detailing the judgment according to the judgment in a community maintenance stage; the head node updates a community member table according to the received JC data packet; if the JC data packet represents the request to join the community, adding a JC data packet sending node into a community member table; if the JC data packet requests to join other communities, the node sending the data packet is moved out of a community member table;

step 2.2 Community maintenance

In the community maintenance stage, the conditions of dynamic addition or departure of nodes in a community and combination, splitting and extinction of the community need to be processed; meanwhile, the role distribution of the head node waiting node, the gateway node and the gateway node waiting node is also carried out in the community maintenance phase;

At this stage, the problem of node community attribution change needs to be handled; when a node receives an HB data packet broadcasted by a neighbor node, predicting the connection duration of the node and the neighbor node according to the speed, the current position and the predicted positions at the future time of 1s, 2s and 3s, and updating a neighbor node information table;

after the neighbor node information table is updated, if the node finds that a new head node appears in the neighbor node, whether the head node is more suitable for being used as a head node than the current community head node to which the head node belongs is judged; node ve_iReceive ve_hTransmitted HB packets of which ve_hA newly appeared head node in the neighbor nodes; if ve_hMore suitably as a head node, then ve_iSelectively add ve_hThe community in which the user is located; ve_iThe following operations are carried out according to the role of the user:

1) when ve_iWhen not head node, ve_iNotification of ve by broadcasting JC packets_hJoining the community; and notify the original community head node ve_iLeaving the original community, and completing the attribution change of the node community;

2) when ve_iWhen it is the head node, ve_iNotification of ve by broadcasting JC packets_hJoining the community; other nodes in the community are informed, and the original community disappears; other nodes in the community become SN nodes; then according to the received HB data packets of other head nodes, other communities are selected to be added, and the combination of the communities is completed; or self-electing to be a new community head node, attracting surrounding nodes to join, and completing community splitting;

at this moment, the node completes dynamic change of node community attribution, community extinction, split and combination according to the given rule;

(b) head node waiting node, gateway node and gateway node waiting node role allocation

In the community maintenance stage, the problem of role distribution of a head node waiting node, a gateway node and a gateway node waiting node also needs to be solved; the head node and the head node waiting node maintain the same community member list; the head node and the head node waiting node respectively maintain a gateway information table according to the gateway nodes in the neighbor nodes and the updated information of the gateway node waiting node; the head node maintains a community member list and knows the community status of each node in the community; the common nodes of the community do not store a community member list, and the status of each node in the community cannot be known; therefore, in order to select a node having a similar community status as the head node as a head node waiting node, the head node waiting node is designated by the head node; other nodes in the community can know the adjacent community and can calculate the connection central value of the expected community neighbor of the adjacent community; the gateway node and the gateway node waiting node are self-elected by each node in the community, and the result is announced to the head node and the adjacent head node waiting node;

the head node waiting node is selected to be more suitable for being used as a head node waiting node according to the size of the connection central degree value of the neighbor of the expected community, and the more close the head node is to the node of the connection central degree value of the neighbor of the expected community;

selecting the gateway node and the gateway node waiting node according to the expected community neighbor connection centrality value of the adjacent community; COM (component object model)_iThere are 5 nodes and COM in the community_jNodes in the community are directly connected; the 5 nodes are represented as<ve₁，ve₂，ve₃，ve₄，ve₅>(ii) a To COM_jCommunity anticipatory community neighbor connection centrality maximum ve₃The node is selected as the community COM_iTo community COM_jThe rest 4 nodes are selected as the community COM_iTo community COM_jThe gateway node of (2) waiting for the node; because the gateway node table of the community is maintained only at the head node and the head node waiting node, the gateway node and the gateway node waiting node are only distinguished at the head node and the head node waiting node; in the self-election process of the gateway node, the nodes all consider that the nodes are gateway node waiting nodes;

the head node and the head node waiting node need to maintain a community member list, wherein the community member list comprises a head node list, a head node waiting node list, a gateway node/gateway node waiting node list and a common member node list; the head node list, the head node waiting node list and the gateway node/gateway node waiting node list are collectively called as a key node member list; in the process of electing the head node waiting node and the gateway node/gateway node waiting node, the head node and the corresponding head node waiting node update the key node member list according to the electing result; after the common member node joins the community, the head node and the corresponding head node waiting node update the common member node list; when the member nodes in the community leave the community, the head node and the head node waiting node remove corresponding members, and corresponding new role selection is performed according to the roles of the removed members; the member nodes leave the community can be divided into the following two cases:

1) the member node joins other communities and actively informs the head node and the corresponding head node waiting node; at the moment, the head node and the head node waiting node remove the member nodes leaving the community from the corresponding list;

2) the member node loses connection with the head node; the head node does not receive the HB data packets broadcasted by the member node after a period of time, and the update of the community member table is completed;

besides updating the table entries maintained by the nodes when receiving HB (hybrid automatic repeat) or other notification data packets of the neighbor nodes, each node in the community can also periodically update the node state of the node so as to discover the condition that the neighbor nodes lose connection; meanwhile, the head node needs to select or update a head node waiting node and a gateway node/gateway node waiting node list; the time interval between two times of node active updating is BEAT _ LENGTH; community COM_kNode ve_iThe contents of the update at time t are as follows:

1) node-to-node expected community neighbor connection centrality value of current community

2) Subtracting 1 from the expected residual connection time of the neighbor node; if the value is reduced to 0, the neighbor node and the current node are considered not to be connected any more;

3) if ve_iIf the node is a community head node, selecting or updating a head node waiting node at the same time, and notifying all nodes in the community of the selected or updated result;

4) updating the connection centrality value of the node and the expected community neighbors of other communities; if the value changes, the head node and the head node waiting node are informed, and the information of the community gateway node is updated;

wherein, the step 2) is a processing step for ensuring the validity of the neighbor node information; because the node may be disconnected with the neighbor node at any time, the node can only acquire the neighbor node information through the HB data packets periodically broadcast by the neighbor node; therefore, when receiving the HB data packet, the node estimates the connection time ECT with the neighbor node; subtracting 1 from ECT every time the node actively updates the state; when the ECT is reduced to 0, the HB data packets from the neighbor nodes are not received for a period of time; at the moment, the current node is considered to be not connected with the neighbor node any more;

step three, community load balancing method

Storing a routing table of the social interval in a head node; when a node in a community sends a data packet to a non-neighbor node, the data packet is sent to a head node firstly; the head node judges whether the destination node is a node in the community; if yes, directly forwarding to the destination node, and if not, forwarding the data packet to the corresponding gateway node according to the maintained routing table information;

when a head node needs to be selected for data forwarding, a head node waiting node is used as a candidate node for head node load balancing; taking a head node and all head node waiting nodes in the community as a head node pool of the community; designing a tau value as a selected evaluation factor; selecting the node with the maximum tau value as a head node for data forwarding; for time t, ve_iτ value τ of node_i，tIs defined as

To make τ_i，tAnd

and AWL_i，tAre all in positive correlation, taken here

Let ρ be 3, μ be-1.67, and x represent

When the head node or the head node waiting node needs to select the gateway node for data forwarding, the gateway node waiting node can be used for sharing the load of the GW node.