Cooperative vehicle-road communication method and device based on vehicle-road cooperation, and storage medium

Abstract

The invention relates to the technical field of intelligent traffic, and discloses a cooperative vehicle-road communication method and device based on vehicle-road cooperation, and a storage medium, and the method comprises the steps: carrying out the dynamic alliance division of an undirected graph based on a deep reinforcement learning algorithm of a message passing neural network; cooperative communication is implemented in the divided alliances; based on the number of data downloaded from the road side units by vehicles in the alliances, the total cooperative income is calculated; deducting the cost linearly related to the alliance scale to obtain a net income, distributing the net income to each side unit through an extra income average distribution rule, and taking an income value distributed by each side unit as a weight of a corresponding node; wherein the vehicle receives the data of the road side unit in a V2R mode, and shares the data when encountering the opposite vehicle in a V2V mode, thereby realizing maximization of the data propagation efficiency in the alliance. According to the invention, through a dynamic alliance division and cooperation mechanism, the system cost is reduced, and the efficiency of vehicle-road cooperative communication is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and in particular to a cooperative vehicle-road communication method, device and storage medium based on vehicle-road collaboration. Background Art

[0002] With the continuous development of the transportation field and the continuous iteration of emerging technologies (autonomous driving, group intelligence control, and the Internet of Things), Intelligent Transportation System (ITS) technology has become increasingly mature, and vehicle-road collaboration is an important part of intelligent transportation system technology.

[0003] Existing technologies overlook the communication potential offered by V2X. Based on the potential for information sharing between RSUs, RSUs can communicate to improve data transmission efficiency. However, collaboration between RSUs increases system costs, and global collaboration often outweighs the benefits. Therefore, there is an urgent need for cooperative V2X communication technologies that fully leverage the advantages of V2X and form dynamic collaborative alliances to enhance global communication efficiency. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention provides a cooperative vehicle-road communication method, device and storage medium based on vehicle-road collaboration.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0006] A cooperative vehicle-road communication method based on vehicle-road collaboration, comprising:

[0007] Based on the backbone network topology of the consortium vehicle-road communication scenario, an undirected graph G = (V, E) is constructed, where each node in the node set E corresponds to a roadside unit (RU). Each edge in the edge set V represents the relationship between RUs, and the edge weight is the number of vehicles traveling in opposite directions and encountering each other between the two RUs at a given moment.

[0008] Dynamically partitioning the undirected graph into coalitions using a deep reinforcement learning algorithm based on a message passing neural network;

[0009] Collaborative communication is implemented within the divided alliance: the total collaborative revenue is calculated based on the amount of data downloaded from the roadside unit by vehicles within the alliance; the net revenue is obtained by deducting the cost linearly related to the alliance size, and the net revenue is distributed to each roadside unit through the average distribution rule of additional revenue. The revenue value allocated to each roadside unit is used as the weight of the corresponding node; among them, when vehicles receive data from the roadside unit through the V2R method and share data with oncoming vehicles through the V2V method, the efficiency of data dissemination within the alliance is maximized.

[0010] In one embodiment, the alliance vehicle-road communication scenario specifically includes: using roadside communication facilities as roadside units, and the roadside units communicate with each other through wired links or wireless networks to form a backbone network; vehicles can communicate with each other within a set spatial range through their own communication equipment; two roadside units can complete information transmission through interoperability based on the communication capabilities of a pair of vehicles traveling relative to each other.

[0011] In one embodiment, the deep reinforcement learning algorithm based on the message passing neural network performs dynamic coalition partitioning on the undirected graph, specifically comprising:

[0012] Randomly partition the undirected graph, define the state set as the alliance partition result, and the action set as the operation of partitioning nodes into different alliances. Calculate the reward function based on the optimization degree of the average edge cut after partitioning.

[0013] The node encoding information is iteratively updated through a message passing neural network to generate value function predictions, and the Markov decision process is used to select the optimal partitioning action with the goal of minimizing the average edge cut, outputting multiple disjoint coalitions.

[0014] In one embodiment, randomly partitioning the undirected graph specifically includes:

[0015] Perform k-ary partitioning on the undirected graph G:

[0016] P={S1,S2,…,S k};

[0017]

[0018] P is the alliance division result, S k is the kth alliance.

[0019] In one embodiment, the defined state set is the alliance division result, the action set is the operation of dividing the node into different alliances, and the reward function is calculated based on the optimization degree of the average edge cut after the division, specifically including:

[0020] Consider a Markov decision process quintuple (S, A, T, R, γ), where T: S × A × S → [0, 1] is the transition function, is the reward function, γ∈[0,1] is the discount factor, the strategy π:S→[0,1] maps the state to the action probability distribution, and the corresponding value function Q π for:

[0021]

[0022] Among them, s0, a0 are the initial state and initial action; the reward function R is as follows:

[0023]

[0024] Among them, s * ∈S represents the optimal state, Indicates s * The corresponding average edge cut.

[0025] In one embodiment, the method of iteratively updating node encoding information through a message passing neural network, generating a value function prediction, selecting the optimal partitioning action using a Markov decision process, minimizing the average edge cut, and outputting multiple mutually disjoint coalitions specifically includes:

[0026] A deep Q network is established based on the message passing neural network; each node in the undirected graph is encoded with an n-dimensional vector Instead, k is the current number of network layers; through the message passing process, the encoding vector of the node is continuously updated:

[0027]

[0028] Among them, M k ,H k They are message function and update function respectively, N(v) refers to the neighbor nodes of node v, w uv is the weight of the edge connecting node u and node v, is an intermediate variable, are the encoding vectors of node v and node u respectively; after K iterations, the prediction of the value function Q is given by the readout function Read, that is, the impact of adding or deleting an edge from the partition:

[0029]

[0030] In one embodiment, the total benefit of collaboration is calculated based on the amount of data downloaded from the roadside unit by vehicles in the alliance, specifically including:

[0031] The total collaborative revenue u(S) generated by all roadside units in a certain period is:

[0032]

[0033] Where bi represents the roadside unit RSU in a certain period of time i Select the data to be transmitted; B S represents the set of data propagated by roadside units in alliance S during a certain period of time, B S ={b1,…,b |S|}, b S For all possible B in alliance S S The set of; β represents the benefit value required for unit effective data, Wc1 represents the weight of data c1, Indicates that the average number of vehicles k that receive RSU in a certain period of time i The amount of downloaded data, M i,j Indicates that from RSU i ∈S drives towards RSU j ∈S, and from RSU j Heading towards RSU i The set of vehicles that meet each other; |M i,j | represents the vehicle set M i,j The total number of vehicles, and |M i,j |=|M j,i |=m i,j (d i,j );W bi Represents RSU i The claimed benefit weight, P ki Indicates that vehicle k is from RSU during the set period i The amount of downloaded data; K i,j Indicates that from RSU i Heading towards RSU j and from RSU j Heading towards RSU i All vehicles.

[0034] In one embodiment, the net revenue is obtained by deducting the cost linearly related to the alliance size, and the net revenue is distributed to each roadside unit according to the additional revenue equal distribution rule, specifically including:

[0035] The cost function c(S) of alliance S is:

[0036]

[0037] Among them, α represents a constant parameter; finally, the net benefit v(S) obtained by alliance S is:

[0038] v(S)=u(S)-c(S);

[0039] u(S) is the total benefit of collaboration of alliance S;

[0040] The net benefit is distributed to each roadside unit through the rule of equal distribution of additional benefits.

[0041] A computer device includes a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the method of any embodiment when executing the computer program.

[0042] A computer-readable storage medium stores a computer program, which implements the steps of the method of any embodiment when executed by a processor.

[0043] Compared with the prior art, the beneficial technical effects of the present invention are:

[0044] 1. Significantly Improved Communication Efficiency: This invention utilizes a dynamic alliance partitioning algorithm (based on deep reinforcement learning using MPNN) to minimize average edge cuts, retain connections between roadside units (RSUs) with high collaboration potential, and reduce the fragmentation of critical communication links, thereby improving data transmission efficiency in vehicle-to-vehicle (V2V) and vehicle-to-road (V2R) collaboration. RSUs within the alliance share information such as traffic flow and distance, coordinate data transmission strategies, and combine V2V and V2R communications to maximize the amount and variety of data received by vehicles.

[0045] 2. Effectively reduce system costs: Avoid global cooperation models and adopt dynamic local alliance division to reduce unnecessary communication and coordination overhead. Constrain the alliance size through cost functions to ensure a balance between cooperative benefits and resource consumption, avoiding resource waste caused by over-expansion.

[0046] 3. Resource Utilization Optimization: This invention utilizes a collaborative mechanism based on the number of vehicle encounters, prioritizing high-density vehicle encounters (such as short-distance road sections). It optimizes communication resource allocation by dynamically adjusting collaboration priorities. By fairly distributing alliance revenue, it incentivizes roadside units to participate in collaboration and enhances overall resource collaboration enthusiasm.

[0047] 4. Enhanced system flexibility and scalability: A dynamic partitioning algorithm based on a graph neural network (MPNN) can adapt to changes in road network topology and is suitable for urban traffic scenarios of various sizes. A Markov decision process (MDP) supports real-time adjustment of alliance strategies to adapt to traffic flow fluctuations, improving the system's dynamic responsiveness.

[0048] 5. Close integration of theoretical models with practical scenarios: This paper abstracts the Internet of Vehicles (IoV) as an undirected graph, transforming the complex RSU collaboration problem into a graph partitioning optimization problem, simplifying algorithm design. Formulating benefits and costs ensures the economic feasibility of the technical solution and provides a quantitative basis for actual deployment.

[0049] In summary, the present invention significantly improves the efficiency and resource utilization of vehicle-road cooperative communication while reducing system costs through dynamic alliance division and collaboration mechanism. It has both theoretical rigor and practical operability, and provides an innovative solution for the optimization of intelligent transportation systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 is a flow chart of a method in an embodiment of the present invention;

[0051] Figure 2 Schematic diagram of an interoperable alliance vehicle-road communication scenario in an embodiment of the present invention;

[0052] Figure 3 is an abstract schematic diagram of a backbone network in an embodiment of the present invention;

[0053] Figure 4 Schematic diagram of message transmission and update in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] A preferred embodiment of the present invention will be described in detail below with reference to the accompanying drawings.

[0055] like Figure 1 As shown, the present invention provides a cooperative vehicle-road communication method based on vehicle-road collaboration, comprising the following steps:

[0056] S1. Based on the topology of the backbone network in the consortium vehicle-road communication scenario, an undirected graph G = (V, E) is constructed, where each node in the node set E corresponds to a roadside unit (RU), and each edge in the edge set V represents the relationship between RUs. The edge weight is the number of vehicles traveling in opposite directions and encountering each other between the two RUs at a given moment.

[0057] S2, performing dynamic coalition partitioning on the undirected graph using a deep reinforcement learning algorithm based on a message passing neural network;

[0058] S3, implements collaborative communication within the divided alliance: calculates the total collaborative revenue based on the amount of data downloaded from the roadside unit by vehicles within the alliance; deducts the cost linearly related to the alliance size to obtain the net revenue, and distributes the net revenue to each roadside unit through the extra revenue average distribution rule, using the revenue value allocated to each roadside unit as the weight of the corresponding node; among them, vehicles receive data from the roadside unit through the V2R method and share data with oncoming vehicles through the V2V method, realizing the maximum efficiency of data dissemination within the alliance.

[0059] In one embodiment, the alliance vehicle-road communication scenario specifically includes: using roadside communication facilities as roadside units, and the roadside units communicate with each other through wired links or wireless networks to form a backbone network; vehicles can communicate with each other within a set spatial range through their own communication equipment; two roadside units can complete information transmission through interoperability based on the communication capabilities of a pair of vehicles traveling relative to each other.

[0060] The alliance vehicle-road communication scenario studied in this invention is as follows Figure 2As shown in the figure, this scenario primarily involves roadside communication infrastructure (RSU) and connected vehicles. Connected vehicles can communicate with each other within a close spatial range using their own communication equipment (V2V). Roadside communication infrastructure can communicate with each other via wired or wireless networks to form a core network. The alliance vehicle-road communication scenario involves two RSUs leveraging their potential for information sharing between vehicles traveling opposite each other, transferring information through interoperability. Interoperability refers to the interaction between the RSUs and vehicles in this scenario. For example, RSU1 transmits data C1 to vehicle V1 (V2R), while RSU2 transmits data C2 to vehicle V2. When vehicles V1 and V2 pass each other in opposite directions, they can share information via V2V, allowing both vehicles to receive data C1 and C2. RSU1, RSU2, V1, and V2 jointly complete the vehicle-road cooperative communication.

[0061] All vehicles in the road network scenario in the present invention are networked vehicles, that is, the vehicles have vehicle-to-vehicle (road) communication functions; and the vehicle-to-vehicle (road) communication environment in the present invention is ideal, that is, delays, packet loss and other phenomena are not considered.

[0062] In one embodiment, the deep reinforcement learning algorithm based on the message passing neural network performs dynamic coalition partitioning on the undirected graph, specifically comprising:

[0063] Randomly partition the undirected graph, define the state set S as the alliance partition result, and the action set A as the operation of dividing nodes into different alliances. Calculate the reward function R based on the optimization degree of the average edge cut after partitioning;

[0064] The node encoding information is iteratively updated through a message passing neural network to generate value function predictions, and the Markov decision process is used to select the optimal partitioning action with the goal of minimizing the average edge cut, outputting multiple disjoint coalitions.

[0065] In one embodiment, randomly partitioning the undirected graph specifically includes:

[0066] Perform k-ary partitioning on the undirected graph G:

[0067] P={S1,S2,…,S k};

[0068]

[0069] P is the alliance division result, S k is the kth alliance.

[0070] In one embodiment, the state set S is defined as the alliance division result, the action set A is the operation of dividing the node into different alliances, and the reward function R is calculated based on the optimization degree of the average edge cut after the division, specifically including:

[0071] Consider a Markov decision process quintuple (S, A, T, R, γ), where T: S × A × S → [0, 1] is the transition function, is the reward function, γ∈[0,1] is the discount factor, the strategy π:S→[0,1] maps the state to the action probability distribution, and the corresponding value function Q π for:

[0072]

[0073] Among them, s0, a0 are the initial state and initial action; the reward function R is as follows:

[0074]

[0075] Among them, s * ∈S represents the optimal state, Indicates s * The corresponding average edge cut.

[0076] In one embodiment, the method of iteratively updating node encoding information through a message passing neural network, generating a value function prediction, selecting the optimal partitioning action using a Markov decision process, minimizing the average edge cut, and outputting multiple mutually disjoint coalitions specifically includes:

[0077] A deep Q network is established based on the message passing neural network; each node in the undirected graph is encoded with an n-dimensional vector Instead, k is the current number of network layers; through the message passing process, the encoding vector of the node is continuously updated:

[0078]

[0079] Among them, M k ,H k They are message function and update function respectively, N(v) refers to the neighbor nodes of node v, w uv is the weight of the edge connecting node u and node v, is an intermediate variable, are the encoding vectors of node v and node u respectively; after K iterations, the prediction of the value function Q is given by the readout function Read, that is, the impact of adding or deleting an edge from the partition:

[0080]

[0081] Since alliance vehicle-road communication requires roadside units to communicate synchronously and coordinate data transmission between each other, communication cooperation consumes costs, and the cost increases with the scale of roadside units. By weighing the advantages and disadvantages of alliance vehicle-road communication, the backbone network is divided into several non-overlapping organization alliances {S1,…,S M},in This patent proposes a deep reinforcement learning algorithm based on the MPNN (Message Passing Neural Network) algorithm to divide the Internet of Vehicles into alliances.

[0082] Specifically, the abstract IOV topology can derive a corresponding undirected graph, such as Figure 3 As shown. The node set E represents the roadside unit set N in the backbone network, where each node v i Corresponding to a roadside unit RSU i ∈N, node v i The weight is RSU i The value of the income received from the distribution In the figure, the edge set V represents the set of associated edges between roadside units in the backbone network, and the edge (i, j) represents the RSU i ∈N and RSU j ∈N associations, whose weight w(i,j) represents the number of connections from RSU at a certain moment i ∈N drives towards RSU j ∈N and from RSU j ∈N drives towards RSU i The average number of vehicles encountered by vehicles ∈N, that is, w(i,j)=|M i,j |, when i=j, w(i,j)=0.

[0083] Dividing the backbone network into alliances is equivalent to finding a suitable graph partitioning algorithm. The partitioning that this invention expects to obtain is the one with the greatest alliance benefit. This requires that the waste of high M should be avoided as much as possible during the partitioning. ij For this purpose, the following definition will be added.

[0084] P gives a k-ary partitioning of the undirected graph G.

[0085]

[0086] In an undirected graph G, for the k-ary partition result P, the number of associated edges (u, v) in different node subsets is defined as the number of cut edges N. p :

[0087] N p =count(u,v),u∈S i ∧v∈S j ∧(u,v)∈E,i≠j;

[0088] Among them, count(·) represents the counting operation, u∈S i ∧v∈S j ∧(u,v)∈e,i≠j.

[0089] Furthermore, in the k-ary partition P, the sum of the weights of the edges connecting nodes in different node subsets is C p :

[0090]

[0091] On this basis, we define the average edge cut of the k-ary partition P Then, in the case of balanced communication cooperation among nodes in different alliance organizations, in order to avoid dividing two nodes with high cooperation probability into different alliances, this section minimizes the average edge cut To optimize the goal:

[0092]

[0093] To solve this problem, this section uses a reinforcement learning method based on message passing neural networks. Consider a Markov decision process with a five-tuple (S, A, T, R, γ). The parameters are explained as follows: S is the state set, which means the current agent is in a certain alliance. A is the action set, which means the current agent will be assigned to a certain alliance. T: S × A × S → [0, 1] is the transition function, is the reward function, γ∈[0,1] is the discount factor. The strategy π:S→[0,1] maps the state to the action probability distribution, and the corresponding value function Q π It is defined as follows, where s0, a0 are the initial states:

[0094]

[0095] This section builds a deep Q-network using a message passing neural network.

[0096] Each node v in the undirected graph is encoded with n dimensions Instead, k is the number of current network layers. Through the message passing process, the node's encoding information is continuously updated, such as Figure 4 As shown:

[0097]

[0098] After K iterations, the readout function Read gives a prediction of the value function, that is, the impact of adding or removing an edge from the segmentation:

[0099]

[0100] Since directly learning how to add or remove nodes from the segmentation set is too computationally expensive, we instead learn how to continuously optimize the solution strategy based on the existing solution strategy. The corresponding reward function R is as follows, where s * ∈S corresponds to the optimal state that has been given:

[0101]

[0102] In one embodiment, the total benefit of collaboration is calculated based on the amount of data downloaded from the roadside unit by vehicles in the alliance, specifically including:

[0103] The total collaborative revenue u(S) generated by all roadside units in a certain period is:

[0104]

[0105] Where bi represents the roadside unit RSU in a certain period of time i Select the data to be transmitted; B S represents the set of data propagated by roadside units in alliance S during a certain period of time, B S ={b1,…,b |S|}, b S For all possible B in alliance S S The set of; β represents the benefit value required for unit effective data, W c1 Indicates the weight of data c1, P ki Indicates that the average number of vehicles k that receive RSU in a certain period of time i The amount of downloaded data, M i,j Indicates that from RSU i ∈S drives towards RSU j ∈S, and from RSU j Heading towards RSU i The set of vehicles that meet each other; |M i,j | represents the vehicle set M i,j The total number of vehicles, and |M i,j |=|M j,i |=m i,j (d i,j ).

[0106] In one embodiment, the net revenue is obtained by deducting the cost linearly related to the alliance size, and the net revenue is distributed to each roadside unit according to the additional revenue equal distribution rule, specifically including:

[0107] The cost function c(S) of alliance S is:

[0108]

[0109] Among them, α represents a constant parameter; finally, the net benefit v(S) obtained by alliance S is:

[0110] v(S)=u(S)-c(S);

[0111] u(S) is the total benefit of collaboration of alliance S;

[0112] The net benefit is distributed to each roadside unit through the rule of equal distribution of additional benefits.

[0113] Specifically, in the alliance vehicle-road communication scenario, in order to achieve cooperative communication, it is necessary for the two roadside units on a road to have a vehicle-to-vehicle communication traffic flow. i,j Two roadside units RSU i , RSU j , and i≠j, define the number of such interoperable vehicle-to-vehicle communication pairs as m i,j (d i,j ), which has the following properties:

[0114] (1) Each pair of vehicles communicating with each other includes a slave RSU i Heading towards RSU j The vehicle and a slave RSU j Heading towards RSU i vehicles.

[0115] (2)m i,j (d i,j ) and the distance d between the two roadside units i,j Inversely proportional.

[0116] (3) The number of vehicle pairs that have relative vehicle-to-vehicle communication must be less than |K i,j |with|K j,i The minimum value in |, i.e.

[0117] m i,j (d i,j )≤min(|K i,j |,|K j,i |);

[0118] where |K i,j |From RSU in this period i Heading towards RSU j Number of vehicles in the direction, |K j,i |For RSU j Heading towards RSU i The number of vehicles.

[0119] According to the individual correlation model, we have:

[0120]

[0121] Where δ is the ratio of vehicles that meet in opposite directions between two roadside units on a road.

[0122] The present invention aims to build an alliance for roadside facilities RSU, and the alliance organization The roadside unit implements cooperative vehicle-road communication and follows the following collaboration mechanism:

[0123] (1) Two roadside units (RSUs) on the same road i and The basic equipment or network communication can be used to inform each other that the current period is from RSU i Heading towards RSU j Average traffic flow in the direction|K i,j |, from RSU j Heading towards RSU i Average traffic volume|K j,i |, each pair of RSUs on the same road i with RSU j The distance between the vehicles and other relevant traffic information.

[0124] (2) Through information exchange, each pair of RSUs on the same road in the alliance S i with RSU j The number of car alliance pairs that may encounter each other can be estimated i,j (d i,j ).

[0125] (3) With the exchange of information, the system S can integrate and coordinate each RSU i Data sent by members to passing vehicles. This data is agreed upon within the consortium S. By combining vehicle-to-vehicle (V2R) communication (vehicles downloading data from RSUs) and vehicle-to-vehicle (V2V) communication (each pair of vehicles exchanging data with each other), the amount and variety of data received by vehicles is maximized. This collaboration is limited to any two roadside units within the system.

[0126] Therefore, by adopting the above-mentioned roadside unit cooperation mechanism, The total revenue generated by all roadside units in a certain period is:

[0127]

[0128] The above formula means that the revenue generated by the entire alliance S is generated before the vehicle passes the first roadside unit and the relative vehicle-to-vehicle communication reaches the next roadside unit or leaves the alliance. This function is mainly divided into three parts. The first part represents the revenue generated by the alliance entities in these i∈S is the revenue claimed by the alliance entity on the vehicles driving to the roadside unit outside the alliance; the second part represents the revenue claimed by the alliance entity on these vehicles driving from RSU j ∈S driving from RSU i ∈S and not with RSU i The third part represents the benefits claimed by the alliance entity on these vehicles from RSU j ∈S driving from RSU i ∈S and with the i The revenue claimed when vehicles meet in the direction of ∈S (interoperability communication occurs).

[0129] Although RSUs gain additional revenue through alliances, maintaining such cooperative vehicle-road communication between RSUs inevitably incurs additional costs. This consumption limits the revenue generated by the alliance to be linearly proportional to the number of RSUs in the alliance. To describe the cost, the present invention sets a cost function. Based on some allocation rules, the relevant revenue obtained by each RSU can be obtained. In the present invention, through the average allocation rule of the additional revenue, each RSU The benefits obtained are:

[0130]

[0131] v({i}) represents the benefits of vehicles meeting and exchanging information, Represents RSU i Profits gained in Alliance S.

[0132] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown in sequence as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts of the accompanying drawings may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times. The order of execution of these steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a portion of the steps or stages in other steps.

[0133] In one embodiment, the present invention provides a computer device comprising a processor, a memory, and a network interface connected via a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is configured to store data used in the above-described method. The network interface of the computer device is configured to communicate with an external terminal via a network connection. The computer program, when executed by the processor, implements the above-described method.

[0134] In an exemplary embodiment, the present invention further provides a computer-readable storage medium, such as a memory, comprising instructions, wherein the instructions are executable by a processor to perform the above method. The storage medium may be a computer-readable storage medium, such as a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, or the like.

[0135] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. It is intended that all variations within the meaning and range of equivalents of the claims be embraced herein, and any reference signs in the claims should not be construed as limiting the claims to which they relate.

[0136] In addition, it should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. A cooperative vehicle-road communication method based on vehicle-road collaboration, characterized in that: include: Based on the backbone network topology of the consortium vehicle-road communication scenario, an undirected graph G = (V, E) is constructed, where each node in the node set E corresponds to a roadside unit (RU). Each edge in the edge set V represents the relationship between RUs, and the edge weight is the number of vehicles traveling in opposite directions and encountering each other between the two RUs at a given moment. Dynamically partitioning the undirected graph into coalitions using a deep reinforcement learning algorithm based on a message passing neural network; Collaborative communication is implemented within the divided alliance: the total collaborative revenue is calculated based on the amount of data downloaded from the roadside unit by vehicles within the alliance; the net revenue is obtained by deducting the cost linearly related to the alliance size, and the net revenue is distributed to each roadside unit through the average distribution rule of additional revenue. The revenue value allocated to each roadside unit is used as the weight of the corresponding node; among them, when vehicles receive data from the roadside unit through the V2R method and share data with oncoming vehicles through the V2V method, the efficiency of data dissemination within the alliance is maximized.

2. The cooperative vehicle-road communication method based on vehicle-road collaboration according to claim 1, characterized in that: The alliance vehicle-road communication scenario specifically includes: using roadside communication facilities as roadside units, and the roadside units communicate with each other through wired links or wireless networks to form a backbone network; vehicles can communicate with each other within a set spatial range through their own communication equipment; two roadside units can complete information transmission through interoperability based on the communication capabilities of a pair of vehicles traveling relative to each other.

3. The cooperative vehicle-road communication method based on vehicle-road collaboration according to claim 1, characterized in that: The deep reinforcement learning algorithm based on the message passing neural network performs dynamic alliance partitioning on the undirected graph, specifically including: Randomly partition the undirected graph, define the state set as the alliance partition result, and the action set as the operation of partitioning nodes into different alliances. Calculate the reward function based on the optimization degree of the average edge cut after partitioning. The node encoding information is iteratively updated through a message passing neural network to generate value function predictions, and the Markov decision process is used to select the optimal partitioning action with the goal of minimizing the average edge cut, outputting multiple disjoint coalitions.

4. The cooperative vehicle-road communication method based on vehicle-road collaboration according to claim 3, characterized in that: The random partitioning of the undirected graph specifically includes: Perform k-ary partitioning on the undirected graph G: P={S1,S2,…,S k }; S1∪S2∪…∪S k =g; P is the alliance division result, s k is the kth alliance.

5. The cooperative vehicle-road communication method based on vehicle-road collaboration according to claim 3, characterized in that: The state set is defined as the alliance division result, the action set is the operation of dividing the node into different alliances, and the reward function is calculated based on the optimization degree of the average edge cut after the division, specifically including: Consider a Markov decision process quintuple (S, A, T, R, γ), where T: S × A × S → [0, 1] is the transition function, is the reward function, γ∈[0,1] is the discount factor, the strategy π:S→[0,1] maps the state to the action probability distribution, and the corresponding value function Q π for: Among them, s0, a0 are the initial state and initial action; the reward function R is as follows: Among them, s * ∈S represents the optimal state, Indicates s * The corresponding average edge cut.

6. The cooperative vehicle-road communication method based on vehicle-road collaboration according to claim 3, characterized in that: The message passing neural network iteratively updates node encoding information, generates value function predictions, uses a Markov decision process to select the optimal partitioning action, and outputs multiple disjoint alliances with the goal of minimizing average edge cuts. Specifically, it includes: A deep Q network is established based on the message passing neural network; each node in the undirected graph is encoded with an n-dimensional vector Instead, k is the current number of network layers; through the message passing process, the encoding vector of the node is continuously updated: Among them, M k ,H k They are message function and update function respectively, N(v) refers to the neighbor nodes of node v, w uv is the weight of the edge connecting node u and node v, is an intermediate variable, are the encoding vectors of node v and node u respectively; after K iterations, the prediction of the value function Q is given by the readout function Read, that is, the impact of adding or deleting an edge from the partition:

7. The cooperative vehicle-road communication method based on vehicle-road collaboration according to claim 3, characterized in that: The total benefit of collaboration is calculated based on the amount of data downloaded from the roadside unit by vehicles in the alliance, specifically including: The total collaborative revenue u(S) generated by all roadside units in a certain period is: Where bi represents the roadside unit RSU in a certain period of time i Select the data to be transmitted; B S represents the set of data propagated by roadside units in alliance S during a certain period of time, B S ={b1,…,b |S| }, b S For all possible B in alliance S S The set of; β represents the benefit value required for unit effective data, W c1 represents the weight of data c1, Indicates that the average number of vehicles k that receive RSU in a certain period of time i The amount of downloaded data, M i,j Indicates that from RSU i ∈S drives towards RSU j ∈S, and from RSU j Heading towards RSU i The set of vehicles that meet each other; |M i,j | represents the vehicle set M i,j The total number of vehicles, and |M i,j |=|M j,i |=m i,j (d i,j );W bi Represents RSU i The claimed benefit weight, P ki Indicates that vehicle k is from RSU during the set period i The amount of downloaded data; K i,j Indicates that from RSU i Heading towards RSU j and from RSU j Heading towards RSU i All vehicles.

8. The cooperative vehicle-road communication method based on vehicle-road collaboration according to claim 3, characterized in that: The net income is obtained by deducting the costs linearly related to the alliance size, and the net income is distributed to each roadside unit through the additional income average distribution rule, specifically including: The cost function c(S) of alliance S is: Among them, α represents a constant parameter; finally, the net benefit v(S) obtained by alliance S is: v(S)=u(S)-c(S); u(S) is the total benefit of collaboration of alliance S; The net benefit is distributed to each roadside unit through the rule of equal distribution of additional benefits.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images