Cognitive Internet of Vehicles connectivity clustering algorithm based on digital twinning

Abstract

The invention belongs to the technical field of Internet of Vehicles communication, and relates to a digital twinning-based cognitive Internet of Vehicles connectivity clustering algorithm, each vehicle constructs a vehicle communication network through a Dijkstra algorithm according to a vehicle communication probability table, and a communication probability weight minimum path from the current vehicle to all vehicles in the network is calculated through path search; enabling the direct reachable neighbor vehicles and the indirect reachable neighbor vehicles on the path with the minimum communication probability weight to jointly form a reachable cluster, and enabling the direct reachable neighbor vehicles on the path with the minimum communication probability weight to form a conflict cluster; each vehicle adjusts own transmitting power according to the conflict cluster; and each vehicle perceives a global network state through a digital twin model according to the vehicles in the conflict cluster and the channel access probability of the corresponding vehicle, selects an optimal channel to complete a channel access request, and forms a connected cluster. According to the method, the communication probability between vehicles is combined in the clustering process, channel allocation is optimized, communication conflicts are reduced, and the data transmission efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of vehicle - to - everything (V2X) communication, and particularly relates to a connectivity clustering algorithm for cognitive V2X based on digital twin, which can be used in cognitive V2X networks. Background Art

[0002] Cognitive V2X refers to a new type of vehicle - mounted communication network that integrates cognitive radio technology and intelligent transportation systems, and realizes efficient wireless resource management through dynamic spectrum sensing and intelligent decision - making. Its core functions include: optimizing spectrum allocation by real - time detecting and utilizing idle spectrum resources to alleviate the spectrum congestion problem of traditional vehicle - mounted networks; automatically adjusting communication strategies according to vehicle mobility and environmental changes to ensure connection stability.

[0003] Existing clustering algorithms have limitations in many aspects. Traditional V2X clustering methods cannot process a large amount of real - time data in a short time and are difficult to meet the high - efficiency data - processing requirements of cognitive V2X; most algorithms are designed based on predictable vehicle movements and stable network results and are difficult to adapt to the high dynamicity of the spectrum requirements and topological structures of cognitive V2X; in ensuring network connectivity, existing algorithms cannot fully consider the impacts of vehicle mobility and communication - environment uncertainties.

[0004] Digital Twin (DT) technology provides a new idea for solving the above problems. By constructing a virtual mirror of the physical V2X network, the digital - twin model can real - time map information such as vehicle positions, speeds, and channel states, and optimize the clustering strategy through simulating network changes. In existing research, the combination of digital twin and V2X clustering is still in its initial stage, lacking systematic architecture design and algorithm optimization. Summary of the Invention

[0005] The present invention provides a connectivity clustering algorithm for cognitive V2X based on digital twin. During the clustering process, it combines the communication probabilities between vehicles, optimizes channel allocation, reduces communication conflicts, and improves data - transmission efficiency. At the same time, it enhances the network robustness to ensure stable communication of vehicles during movement and adapt to complex and changeable traffic environments, ultimately improving the overall performance of cognitive V2X.

[0006] To achieve the above object, the technical solution of the present invention is as follows.

[0007] A connectivity clustering algorithm for cognitive V2X based on digital twin, comprising: Constructing a cognitive V2X network and a digital - twin model; Judging whether the roadside unit and the vehicle are within the communication range, and establishing a probability table of vehicles that can communicate with the roadside unit and a communication - probability table of the roadside unit for dynamically managing the vehicle communication connections within the coverage area; Determine whether the vehicle is within the effective communication range of neighboring vehicles, and establish a vehicle reachable neighbor vehicle probability table and a vehicle communication probability table; Each vehicle constructs a vehicle communication network according to the vehicle communication probability table through the Dijkstra algorithm. In this vehicle communication network, the reciprocal of the communication probability is used as the link weight parameter, and the communication probability weight minimum path from the current vehicle to all vehicles in the network is calculated through path search; and the directly reachable neighbor vehicles and indirectly reachable neighbor vehicles on the communication probability weight minimum path jointly form a reachable cluster, and the directly reachable neighbor vehicles on the communication probability weight minimum path form a conflict cluster; Each vehicle adjusts its own transmission power according to the conflict cluster; Each vehicle selects the optimal channel to complete the channel access request according to the vehicles in the conflict cluster and the corresponding channel access probabilities of the vehicles, and perceives the global network state through the digital twin model, forming a connected cluster.

[0008] Further preferably, the method for determining whether the roadside unit and the vehicle are within the communication range is as follows: when the roadside unit receives the channel access data packet sent by the vehicle, it extracts the positioning location of the vehicle from the data packet; then, the roadside unit calculates the Euclidean distance between itself and the vehicle; subsequently, the roadside unit compares the calculated Euclidean distance with the maximum data transmission distance of the current vehicle to determine whether the roadside unit and the vehicle are within the communication range.

[0009] Further preferably, the process of establishing the roadside unit communicable vehicle probability table is as follows: when parsing the channel access data packet of the communicable vehicle, synchronously extract the communicable vehicle ID, real-time positioning coordinates and the corresponding channel access probability parameters, and write them into the roadside unit communicable vehicle probability table, and broadcast the latest communicable vehicle probability table to the whole network through the multicast mechanism.

[0010] Further preferably, the process of establishing the roadside unit communication probability table is as follows: the roadside unit first creates a blank communication probability table and completes the initialization; then, the roadside unit traverses each received channel access data packet and the roadside unit communicable vehicle probability table; during the traversal process, for each communicable vehicle, calculate the corresponding communication probability between the roadside unit and the communicable vehicle; then, write the roadside unit's own identifier, the communicable vehicle ID and the calculated communication probability into the roadside unit communication probability table of the roadside unit; the roadside unit encapsulates the roadside unit communication probability table into a communication probability data packet and sends it to the surrounding vehicles through the broadcast mechanism.

[0011] Further preferably, the method for determining whether a vehicle and a neighboring vehicle are within the effective communication range is as follows: after each vehicle receives the channel access data packet sent by the neighboring vehicle, it parses the neighboring vehicle ID and positioning coordinates in real time, calculates the Euclidean distance between the current vehicle and the neighboring vehicle, and compares it with the maximum data transmission distance of the current vehicle, and the vehicle determines whether there are effective communication conditions between the current vehicle and the neighboring vehicle.

[0012] Further preferably, the process of establishing the vehicle reachable neighboring vehicle probability table and the vehicle communication probability table is as follows: Each vehicle parses the received neighboring vehicle channel access data packet, extracts the vehicle ID, positioning coordinates and channel access probability parameters of the communicable neighboring vehicle, and stores them in the vehicle reachable neighboring vehicle probability table; Each vehicle first initializes the vehicle communication probability table as an empty structure; by semantically parsing the received channel access data packet and the vehicle reachable neighboring vehicle probability table, extracts the communicable neighboring vehicle ID, real-time positioning coordinates and corresponding channel access probability parameters, and combines the maximum effective transmission distance to dynamically calculate the communication probability between the current vehicle and the neighboring vehicle, and writes it into the vehicle communication probability.

[0013] Further preferably, after each vehicle receives the communication probability data packet transmitted by the neighboring vehicle, it constructs and updates the vehicle communication probability table in the cognitive vehicle network.

[0014] Further preferably, the process of forming the reachable cluster and the conflict cluster is as follows: Step 9a: Traverse the updated vehicle communication probability table to construct a communication probability weight table; Step 9b: In the cognitive vehicle network composed of N vehicles, the selected vehicles are defined as the selected vehicle list, and the unselected vehicles are defined as the unselected vehicle list; Step 9c: Randomly select a vehicle as the starting vehicle, fill the vehicle ID of the starting vehicle into the selected vehicle list, and at the same time delete the vehicle ID from the unselected vehicle list, thereby completing the update of the selected vehicle list and the unselected vehicle list; Step 9d: Use the Dijkstra algorithm to construct a minimum spanning tree and establish a record list; Step 9e: Loop to execute Step 9d until the unselected vehicle list is an empty set. When the iteration terminates, the cumulative triple data set in the record list constitutes the structured representation of the minimum spanning tree, that is, the set of optimal communication weight communication links between vehicles; Step 9f: Construct a vehicle communication network formed by a minimum spanning tree based on the obtained set of optimal communication weight communication links between vehicles; in the vehicle communication network, each vehicle can determine its directly reachable neighbor vehicles and indirectly reachable neighbor vehicles, and the directly reachable neighbor vehicles and indirectly reachable neighbor vehicles together form a reachable cluster; the directly reachable neighbor vehicles on the path with the minimum communication probability weight form a conflict cluster; optimize the intra-cluster path through the minimum spanning tree within the conflict cluster, and if there is a disconnection problem in the local area within the conflict cluster, introduce the Steiner algorithm to insert virtual relay vehicles to enhance the network robustness.

[0015] Further preferably, the process of each vehicle adjusting its own transmission power according to the conflict cluster is as follows: Construct a table of directly reachable neighbor vehicles of the vehicle based on the record list to clarify the directly reachable neighbor vehicles of each vehicle; then each vehicle obtains all neighbor vehicles from the table of directly reachable neighbor vehicles of the vehicle, obtains the positioning positions of neighbor vehicles from the probability table of reachable neighbor vehicles of the vehicle, calculates the distances from each neighbor vehicle using the Euclidean distance formula, filters out the directly reachable neighbor vehicle farthest from the current vehicle, and finally dynamically adjusts its own transmission power according to this result.

[0016] Further preferably, each vehicle selects the optimal channel to complete the channel access request according to the vehicles in the conflict cluster and the channel access probabilities of the corresponding vehicles, and perceives the global network state through the digital twin model to form a connected cluster. The specific steps include: Step 11a: In the cognitive vehicle-to-everything environment, there are C channel resources available for vehicles to choose to access; the current vehicle screens the currently available idle channel resource blocks within the conflict cluster range through spectrum sensing technology; the current vehicle autonomously selects and accesses an unoccupied channel resource block, and after successful access, creates and maintains a channel resource packet; broadcasts the channel resource packet to surrounding vehicles; Step 11b: Determine whether the vehicle ID in the channel resource packet is in the conflict cluster. The specific steps include: Step 11b1: When other vehicles in the conflict cluster receive the channel resource packet sent by the current vehicle, first parse the channel resource block allocation information contained in the channel resource packet; mark the neighbor vehicles that occupy the same channel, and at the same time create a channel allocation packet and broadcast it to the directly reachable neighbor vehicles of the current vehicle; if the vehicles in the conflict cluster other than the current vehicle have not been allocated channels, mark the channel allocation packet as an empty packet, and the directly reachable neighbor vehicles of the corresponding vehicle are responsible for broadcasting the channel allocation packet; if the channels have been allocated, do not mark and continue to broadcast the channel allocation packet passed by the current vehicle; Step 11b2: The vehicles in the conflict cluster sort according to the magnitude of the channel access probabilities after receiving the channel allocation packets sent by the directly reachable neighbor vehicles, and preferentially select different channels as their available channels; Step 11c: The digital twin model evaluates and optimizes the channel allocation of the actual vehicle-to-everything (V2X) network according to the channel usage in the virtual network model and the vehicle communication requirements until all vehicles are allocated channels, ensuring the reasonable utilization of channel resources and forming connected clusters. The digital twin model verifies the effectiveness of the allocation scheme and issues optimization instructions to the physical entity layer through the data interaction layer to achieve dynamic power control and spectrum resource management.

[0017] The present invention has the following advantages: The vehicle status information is uploaded to the roadside unit (RSU) in real time through vehicle-to-infrastructure (V2I) communication. The virtual network model of the cognitive V2X network is constructed by using digital twin technology, which dynamically maps the vehicle position, speed, and channel status and synchronously updates them to the digital twin layer to achieve high-precision real-time synchronization between the virtual and physical networks. On this basis, the communication probability between the vehicle and the RSU and the communication probability between vehicles are calculated based on the log-normal shadowing model, providing a quantitative basis for clustering decisions.

[0018] Combined with the Dijkstra algorithm, a three-level clustering structure (reachable cluster → conflict cluster → connected cluster) is generated: using the reciprocal of the communication probability as the communication probability weight, a reachable cluster containing directly reachable neighbor vehicles and indirectly reachable neighbor vehicles is constructed, and a conflict cluster is constructed by extracting directly reachable neighbor vehicles and indirectly reachable neighbor vehicles. Further, the Steiner algorithm is used to optimize the intra-cluster path, solve the link conflict problem of unconnected clusters, and enhance the connectivity of unconnected clusters. By modeling vehicles as a graph through the Steiner algorithm, after detecting the boundary vehicles of unconnected sub-clusters, a minimum-weight connection tree containing intermediate forwarding vehicles is constructed by using a heuristic algorithm to avoid conflict links. At the same time, multi-paths are constructed through redundant vehicles and dynamically locally reconstructed to improve the intra-cluster connectivity and anti-conflict ability.

[0019] For dynamic traffic scenarios, the vehicle transmission power is dynamically adjusted based on the communication probability weight in the virtual network model, and the farthest directly reachable neighbor vehicle is selected to establish connectivity. At the same time, the channel allocation is optimized by combining cognitive spectrum sensing technology to ensure efficient resource utilization. While ensuring communication reliability, the number of channel allocations and network energy consumption are reduced, significantly improving resource utilization.

[0020] To adapt to high-dynamic scenarios such as intersections, the algorithm pre-acts the network topology changes through digital twin technology, dynamically adjusts the power and channel resource allocation strategies, can reduce node energy consumption, reduce the required number of channels, improve the network connectivity probability, and reduce the communication interruption probability caused by high-speed vehicle movement. In addition, the digital twin layer evaluates the channel usage status in the virtual network in real time, combines the simulation results to issue optimization instructions to the physical entity layer, and maintains network connectivity and robustness through the virtual-real collaboration mechanism. Description of the Drawings

[0021] Figure 1 This is the overall flowchart of the present invention; Figure 2 This is the digital twin model diagram constructed for the present invention; Figure 3 This is the simulation verification diagram of the average channel allocation for the intersection constructed for the present invention; Figure 4 This is the simulation verification diagram of the maximum local connectivity for the intersection constructed for the present invention; Figure 5 This is the simulation verification diagram of the path efficiency factor for the intersection constructed for the present invention. Detailed implementation manners

[0022] The following will further describe the embodiments of the present invention in detail with reference to the accompanying drawings.

[0023] As Figure 1 shown, the implementation steps of the cognitive vehicular network connectivity clustering algorithm based on digital twin of the present invention are as follows: Step 1: Construct a cognitive vehicular network with C channels and N vehicles (C > 1, N ≥ 2), where each channel supports vehicles to dynamically select access, and each channel is exclusively used by an authorized vehicle. There are M authorized channels in the fixed frequency band that the authorized vehicle is permitted to use, M = {C 1 , C 2 ,..., C m}, and the same channel is not reused among authorized vehicles. A set of N vehicles based on Poisson distribution and roadside infrastructure constitute a cognitive vehicular network (C > 1, N ≥ 2). At the same time, construct a digital twin model including a physical entity layer, a data interaction layer, and a digital twin layer; Physical entity layer: It includes actual vehicles, roadside units (RSUs), and road infrastructure, and collects data such as vehicle position, speed, and channel status through sensors and V2I communication.

[0024] Data interaction layer: Responsible for uploading the data of the physical entity layer to the digital twin layer in real time, and sending the optimized clustering strategy to the physical entity layer.

[0025] Digital twin layer: Construct a virtual network model corresponding to the actual cognitive vehicular network, including initial information such as channel resources and vehicle distribution, and achieve accurate mapping of the physical network through real-time data update.

[0026] It should be noted that in this application, for simplicity of expression, unless otherwise specified, vehicles are all cognitive vehicles.

[0027] Step 2: Each vehicle writes the vehicle ID and positioning location into the channel access data packet and broadcasts it to the roadside unit (RSU) and neighboring vehicles; Specifically, in the cognitive vehicle-to-everything (V2X) network, vehicles dynamically access idle channels based on spectrum sensing technology. Each vehicle encapsulates its unique identifier (vehicle ID) and real-time positioning coordinates into a channel access request packet and synchronously transmits it to roadside units (RSUs) and neighboring vehicles within the coverage area through a multicast mechanism to achieve efficient utilization of network resources.

[0028] Step 3: The digital twin model captures vehicle information in real time through roadside units (RSUs) and updates the virtual network model to predict and analyze the vehicle's movement trajectory and communication situation. Through wireless communication technology, it establishes connections with vehicles within the communication area of the roadside unit, continuously obtains various information of vehicles on the road, including but not limited to position, speed, etc. It uploads the collected vehicle information in the target cognitive V2X network to the digital twin model and updates it in real time to update the virtual network model to ensure that the digital twin model is synchronized with the actual cognitive V2X network state. The digital twin layer predicts and analyzes the vehicle's movement trajectory, communication situation, etc., providing decision-making support for the connectivity optimization of the V2X network.

[0029] Step 4: The roadside unit determines whether both parties are within the communication range based on the received channel access data packet of the vehicle. The specific steps are as follows: Step 4a: In the cognitive V2X network, considering the combined fading effect of roadside infrastructure and neighboring vehicles on wireless signal transmission, a log-normal shadowing model is used to describe the signal fading characteristics. Under the shadow fading effect, the maximum transmission distance D of the vehicle max Can be expressed as: , Where, Represents the Euclidean distance between the current vehicle a and the neighboring vehicle b, z represents the normalized parameter under the shadow fading effect, , Represents the standard deviation of the Gaussian random variable, α represents the path loss exponent, e is the natural constant, Represents the signal strength of the neighboring vehicle b received by the current vehicle a. P(·) represents the probability that the received signal strength is greater than the receiver sensitivity, θ represents the radian angle centered on the current vehicle a, , r represents the radius of the circle centered on the current vehicle, , R represents the initial transmission radius of the vehicle, Is the complementary error function, Is the natural exponential function. Is the received power threshold, and the calculation formula is: , where, P 0 Represents the signal power at the reference position, d 0 Represents the reference distance, and the maximum data transmission distance of the current vehicle is , and the maximum data transmission distance of the roadside unit is Ri 。

[0030] Step 4b: When the roadside unit (RSU) receives the channel access data packet sent by the vehicle, it extracts the positioning location of the vehicle from the data packet. Then, the roadside unit calculates the Euclidean distance between itself and the vehicle. , and the calculation formula of this Euclidean distance is: , where (x a , y a ) represents the positioning location coordinates of the current vehicle, and (x i , y i ) are the positioning location coordinates of the roadside unit. Subsequently, the roadside unit (RSU) will compare the calculated Euclidean distance with the maximum data transmission distance of the current vehicle a. Here, satisfies the following formula: , represents the maximum transmission range of the current vehicle a, which can be obtained through the above Step 4a. If holds, it means that the transmission of this data packet has exceeded the reachable range of the vehicle, and the RSU will directly discard this data packet; if , it indicates that the vehicle currently sending the data packet is a communicable vehicle of the roadside unit.

[0031] Step 5: The roadside unit establishes a probability table of communicable vehicles of the roadside unit and a communication probability table of the roadside unit according to the received channel access data packets of the vehicles, for dynamically managing the vehicle communication connections within the coverage area. The specific steps include: Step 5a: When the roadside unit (RSU) parses the channel access data packets of the communicable vehicles, it synchronously extracts the communicable vehicle ID, real-time positioning coordinates, and the corresponding channel access probability parameters, and writes them into the probability table of communicable vehicles of the roadside unit. The probability table of communicable vehicles of the roadside unit dynamically maintains the ID identifications of all communicable neighbor vehicles within the RSU coverage area and the corresponding channel access probability values, and after completing the data update, broadcasts the latest probability table of communicable vehicles to the whole network through the multicast mechanism; Step 5b: The roadside unit (RSU) first creates a blank communication probability table and completes the initialization. After completion of the initialization, the RSU traverses the data in each received channel access data packet and the probability table of communicable vehicles of the roadside unit one by one. During the traversal process, for each communicable vehicle, taking the current vehicle a as an example, the RSU extracts its vehicle ID. At the same time, it calculates the communication probability P (a,i) between the roadside unit i and the current vehicle a. Subsequently, the identification of the RSU itself, the ID of the communicable vehicle, and the calculated communication probability P (a,i), they are written into the roadside unit communication probability table of the RSU together. After the RSU finishes traversing and writing all the received data, the roadside unit communication probability table is updated to the latest state. At this time, the RSU encapsulates the updated roadside unit communication probability table into a communication probability data packet. Finally, through the broadcast mechanism, this data packet is sent to the surrounding vehicles so that the surrounding vehicles can obtain the latest communication probability information in a timely manner. The updated roadside unit communication probability table is shown in Table 1 below: Table 1

[0032] In the table, P (b,i) is the corresponding communication probability between roadside unit i and vehicle b, and P (x,i) is the corresponding communication probability between roadside unit i and vehicle x. Here, a, b,..., x are all vehicle IDs.

[0033] Step 6: Each vehicle determines whether it is within the effective communication range with its neighboring vehicles by receiving the channel access data packets sent by the neighboring vehicles. The specific steps are as follows: Step 6a: Referring to Step 4a, under the shadow fading effect, calculate the maximum transmission distance of the vehicle.

[0034] Step 6b: After each vehicle receives the channel access data packet sent by a neighboring vehicle, it parses the neighboring vehicle ID and positioning coordinates in real time, and calculates the Euclidean distance d (a,b) between the current vehicle a and the neighboring vehicle b, and compares the Euclidean distance d (a,b) with the maximum data transmission distance R a of the current vehicle a. The vehicle determines whether there are effective communication conditions with the neighboring vehicle. Among them, the calculation formula of the Euclidean distance d (a,b) is: , where (x a , y a ) represents the positioning position coordinates of the current vehicle, and (x b , y b ) represents the positioning position coordinates of the neighboring vehicle b. And R a satisfies the following formula: . If holds, it means that the packet exceeds the reachable range, and this data packet is directly discarded; if it does not hold, that is , it means that the neighboring vehicle b is a communicable vehicle of the current vehicle a.

[0035] Step 7: Each vehicle establishes a vehicle reachable neighboring vehicle probability table and a vehicle communication probability table according to the received channel access data packets of the neighboring vehicles. The specific steps are as follows: Step 7a: Each vehicle parses the received neighbor vehicle channel access data packets, extracts the vehicle IDs, positioning coordinates, and channel access probability parameters of the communicable neighbor vehicles, and stores this information in the vehicle reachable neighbor vehicle probability table. The real-time synchronization mechanism of the vehicle reachable neighbor vehicle probability table realizes dynamic data maintenance by continuously optimizing the mapping relationship between the IDs and channel access probabilities of the current vehicle and all reachable neighbor vehicles. After completing the information iteration, the system uses a distributed multicast protocol to push the updated vehicle reachable neighbor vehicle probability table to the global network to ensure the real-time synchronization of the topological adjacency relationship in the dynamic traffic scenario.

[0036] Step 7b: Each vehicle first initializes the vehicle communication probability table as an empty structure. Subsequently, by semantically parsing the received channel access data packets and the vehicle reachable neighbor vehicle probability table, it extracts the communicable neighbor vehicle IDs, positioning coordinates, and corresponding channel access probability parameters. Based on the log-normal shadowing model, combined with the maximum effective transmission distance R a , the communication probability P com (a, b) between the current vehicle a and the neighbor vehicle b is calculated in real time. The calculation process considers the signal attenuation characteristics and the dynamic transmission environment to ensure that the evaluation results conform to the actual channel conditions. The ID of the current vehicle a, the ID of the neighbor vehicle b, and the P com (a, b) parameters are written into the vehicle communication probability table, and the vehicle communication probability table stores the communication probabilities of all communicable neighbor vehicles in the form of an adjacency matrix.

[0037] The communication probability P com (a, b) between the current vehicle a and the neighbor vehicle b is obtained by the following formula: , where, represents the traffic density, defined as the ratio of the number of vehicles per unit area to the maximum number of vehicles that can be accommodated, with a value range of 0 - 1.

[0038] Based on the adjacency matrix update mechanism, the communication probabilities between the current vehicle and its neighbor vehicles are mapped to the vehicle communication probability table to complete the dynamic maintenance of the adjacency relationship. Through this distributed computing mechanism, each vehicle can grasp the communication probability distribution of neighbor vehicles within its coverage area in real time, providing basic data support for subsequent network functions such as routing selection and resource allocation. After processing all the received data, the updated vehicle communication probability table is encapsulated into a dedicated data packet and broadcast to neighbor vehicles within the coverage area through a multicast mechanism to maintain the real-time synchronization of the adjacency relationship across the network. Through the distributed data processing and dynamic update mechanism, it is ensured that each vehicle can grasp the communication probability distribution of neighbor vehicles in real time, providing basic data support for core functions such as routing selection and resource allocation in the cognitive vehicular network. The vehicle communication probability table of the current vehicle a is shown in Table 2 below: Table 2

[0039] Among them, P com(a,x) is the communication probability between the current vehicle a and the neighboring vehicle x.

[0040] Step 8: After each vehicle receives the communication probability data packet transmitted by the neighboring vehicle, construct and update the vehicle communication probability table in the cognitive vehicle network. The specific process is as follows: Step 8a: Each vehicle checks whether its own ID exists in the communication probability data packet sent by the neighboring vehicle. If it exists, execute Step 8b; if it does not exist, mark the communication probability between the two vehicles as 0 and execute Step 8c; Step 8b: Update the neighboring vehicle ID and its corresponding communication probability in the communication probability data packet to the vehicle communication probability table of the current vehicle, and then execute Step 8c; Step 8c: Each vehicle compares the communicable vehicle IDs in its own vehicle communication probability table with the vehicle IDs in the communication probability data packet transmitted by the neighboring vehicle one by one. If there are the same vehicle IDs, execute Step 8d; if there are no same vehicle IDs, do not perform the update operation and directly broadcast and forward the received communication probability data packet; Step 8d: When there are the same vehicle IDs in the vehicle communication probability table and the neighboring vehicle communication probability data packet (assuming the ID of vehicle x is x, x is the vehicle number and x ∈ N), update the communication probability corresponding to the vehicle ID in the neighboring vehicle communication probability data packet to the vehicle communication probability table of this vehicle. Loop and execute this update operation until the comparison and update of the communication probabilities of the vehicle IDs in the vehicle communication probability table of this vehicle and the neighboring vehicle communication probability data packet are completed. The vehicle communication probability table is shown in Table 3 below: Table 3

[0041] The diagonal elements in the table are 0, indicating that there is no self-communication. P com (b,a) represents the communication probability between vehicle b and vehicle a, P com (x,a) represents the communication probability between vehicle x and vehicle a, P com (a,x) represents the communication probability between vehicle a and vehicle x, P com (b,x) represents the communication probability between vehicle b and vehicle x.

[0042] Step 9: Each vehicle constructs a vehicle communication network through the Dijkstra algorithm according to the vehicle communication probability table. In this vehicle communication network, taking the reciprocal of the communication probability as the link weight parameter, calculate the communication probability weight minimum path (i.e., the communication link with the minimum cumulative weight value) from the current vehicle to all vehicles in the network through path search; and form the reachable cluster C by combining the directly reachable neighbor vehicles and indirectly reachable neighbor vehicles on the communication probability weight minimum path R , the directly reachable neighbor vehicles on the communication probability weight minimum path form the conflict cluster C con , and the specific steps are as follows: Step 9a: By traversing the updated vehicle communication probability table, construct a communication probability weight table. An example of the communication probability weight table is shown in Table 4 below: Table 4

[0043] Among them, P com (a, b) represents the communication probability, where a and b are both vehicle IDs, and vehicle a, vehicle b, and vehicle x are all one of the N vehicles in the cognitive vehicle network. When P com (a, b) is not zero, the communication probability weight between vehicle a and vehicle b is denoted as W (a,b) = 1 / P com (a, b); if the communication probability between two vehicles in the vehicle communication probability table is 0, then set the communication probability weight between the two vehicles to ∞. W (b,a) is the communication probability weight between vehicle b and vehicle a, W (x,a) is the communication probability weight between vehicle x and vehicle a, W (x,b) is the communication probability weight between vehicle x and vehicle b, W (a,x) is the communication probability weight between vehicle a and vehicle x, W (b,x) is the communication probability weight between vehicle b and vehicle x.

[0044] Step 9b: In the cognitive vehicle network composed of N vehicles, define the selected vehicles as the selected vehicle list, and the unselected vehicles as the unselected vehicle list. Initially, the selected vehicle list is an empty set, and the unselected vehicle list contains the vehicle IDs of all vehicles.

[0045] Step 9c: Randomly select a vehicle as the starting vehicle, fill the vehicle ID of this starting vehicle into the selected vehicle list, and at the same time delete this vehicle ID from the unselected vehicle list, thereby completing the update of the selected vehicle list and the unselected vehicle list.

[0046] At this time, the selected vehicle list is shown in Table 5 below: Table 5

[0047]

[0048] The list of unselected vehicles is shown in Table 6 below: Table 6

[0049] Step 9d: Construct a minimum spanning tree using Dijkstra's algorithm. The specific operations are as follows: Step 9d1: For each vehicle, taking vehicle a as an example, each vehicle needs to find the minimum communication probability weight between itself and other vehicles according to the communication probability weight table, and mark the corresponding vehicle (assumed to be vehicle b). Subsequently, add the ID of the marked vehicle, that is, the ID of vehicle b, to the selected vehicle list, and remove this ID from the unselected vehicle list at the same time, so as to complete the update operation of the selected vehicle list and the unselected vehicle list.

[0050] At this time, the selected vehicle list is shown in Table 7 below: Table 7

[0051] The unselected vehicle list is shown in Table 8 below: Table 8

[0052] Step 9d2: Construct a temporary path table based on the selected vehicle list and the unselected vehicle list. Fill the vehicle IDs in the unselected vehicle list into the first row of data in the temporary path table; define the second row of data in the temporary path table as the minimum communication probability weight, which is used to save the communication probability weight, and define the third row of data in the temporary path table as the neighbor vehicle, which is used to save the vehicle ID corresponding to the found minimum communication probability weight.

[0053] An example of the initial temporary path table is shown in Table 9 below: Table 9

[0054] Find the vehicle ID in the first row of data in the temporary path table that corresponds to the selected vehicle list, compare the communication probability weight data of the corresponding vehicle ID recorded in the temporary path table with the original communication probability weight data. The original communication probability weight data is read from the communication probability weight table. If the corresponding communication probability weight data read from the communication probability weight table is less than the corresponding communication probability weight data recorded in the temporary path table, update the data and save the maximum communication probability weight in the minimum communication probability weight, and at the same time save the vehicle ID corresponding to the minimum communication probability weight in the neighbor vehicle; otherwise, retain the communication probability weight data recorded in the temporary path table, and the vehicle ID in the neighbor vehicle remains unchanged.

[0055] Assume that the communication probability weight between vehicle IDx and vehicle IDa is the smallest. An example of the updated temporary path is shown in Table 10 below: Table 10

[0056] Step 9d3: Establish a record list, and save the minimum communication probability weight updated each time and the vehicle ID corresponding to the minimum communication probability weight in the form of a triple (IDa, IDb, W (a,b) ) into the record list, where W (a,b) represents the minimum communication probability weight between vehicle a and b.

[0057] Step 9e: Loop through the screening and recording process of Step 9d (including Steps 9d1 - d3) until the list of unselected vehicles is an empty set. When the iteration terminates, the cumulative triple data set in the record list constitutes the structural representation of the minimum spanning tree, that is, the set of optimal communication weight communication links between vehicles. An example of the record list is shown in Table 11 below: Table 11

[0058] Step 9f: Based on the obtained record list, construct a vehicle communication network formed by the minimum spanning tree. In the vehicle communication network, each vehicle can determine its directly reachable neighbor vehicles and indirectly reachable neighbor vehicles. The directly reachable neighbor vehicles and indirectly reachable neighbor vehicles together form the reachable cluster C R ; the directly reachable neighbor vehicles on the path with the minimum communication probability weight form the conflict cluster C con . Optimize the intra - cluster path within the conflict cluster through the minimum spanning tree. If there is a disconnection problem in the local area within the conflict cluster, introduce the Steiner algorithm to insert virtual relay vehicles to enhance the network robustness.

[0059] Step 10: Each vehicle adjusts its transmission power according to the conflict cluster C con , and the specific steps are as follows: Step 10a: Based on the record list, construct a table of directly reachable neighbor vehicles of the vehicle, and clarify the directly reachable neighbor vehicles of each vehicle. The table of directly reachable neighbor vehicles of the vehicle contains vehicle ID information and the minimum communication probability weight data of the corresponding vehicle. An example of the table of directly reachable neighbor vehicles of the vehicle is shown in Table 12 below: Table 12

[0060] Step 10b: Each vehicle obtains all its neighbor vehicles from the table of directly reachable neighbor vehicles of the vehicle, acquires the positioning positions of the neighbor vehicles from the table of probabilities of reachable neighbor vehicles of the vehicle, calculates the distances to each neighbor vehicle using the Euclidean distance formula, filters out the directly reachable neighbor vehicle that is farthest from the current vehicle, and finally dynamically adjusts its own transmission power according to this result , where λ is the received signal-to-noise ratio threshold determined by the sensitivity and bit error rate requirements of the vehicle receiver, α is the path loss exponent, and R a is the maximum data transmission distance of the current vehicle.

[0061] Step 11: Each vehicle selects the optimal channel to complete the channel access request based on the vehicles in the conflict cluster C con and the channel access probabilities of the corresponding vehicles, and perceives the global network state through the digital twin model to form a connected cluster. The specific steps are as follows: Step 11a: In the cognitive vehicle-to-everything (V2X) environment, there are C channel resources available for vehicles to choose to access. The vehicle will screen the currently available idle channel resource blocks within the conflict cluster through spectrum sensing technology. The current vehicle independently selects and accesses an unoccupied channel resource block. After successful access, it creates and maintains a channel resource packet. This channel resource packet covers information such as the channel resource block identifier, vehicle ID, and the vehicle's channel access probability, and then broadcasts the channel resource packet to the surrounding vehicles; Step 11b: Determine whether the vehicle ID in the channel resource packet is in the conflict cluster C con . The specific steps are as follows: Step 11b1: When other vehicles in the conflict cluster receive the channel resource packet sent by the current vehicle, they first parse the channel resource block allocation information contained in the channel resource packet. Based on the principle of channel mutual exclusion for vehicles in the same cluster, conflict markings are made for neighbor vehicles using the same channel, and at the same time, a channel allocation packet is created and broadcast to the directly reachable neighbor vehicles of the current vehicle. If the vehicles in the conflict cluster other than the current vehicle have not been allocated channels, the channel allocation packet is marked as an empty packet, and the directly reachable neighbor vehicles of the corresponding vehicle are responsible for broadcasting the channel allocation packet; if channels have been allocated, no marking is done, and the channel allocation packet passed by the current vehicle is continued to be broadcast.

[0062] Step 11b2: The vehicles in the conflict cluster will sort according to the magnitudes of the channel access probabilities after receiving the channel allocation packets sent by their directly reachable neighbor vehicles, and preferentially select different channels as their available channels; Step 11c: The digital twin model evaluates and optimizes the channel allocation of the actual V2X network according to the channel usage situation and vehicle communication requirements in the virtual network model until all vehicles have been allocated channels, ensuring the reasonable utilization of channel resources and forming a connected cluster C N. The digital twin model verifies the effectiveness of the allocation scheme and issues optimization instructions to the physical entity layer through the data interaction layer to achieve dynamic power control and spectrum resource management.

[0063] As Figure 2 shown, the digital twin model constructed by the present invention realizes the real-time mapping and intelligent optimization of the cognitive vehicle network through the coordinated operation of the physical entity layer, the data interaction layer and the digital twin layer. Figure 2 In it, C 1 、C 2 、C 3 、C 4 、C 5 、C 6 、C 7 are the 1st, 2nd, 3rd, 4th, 5th, 6th, and 7th channels respectively. The DT-CA algorithm represents the digital-twin-based connectivity clustering algorithm for cognitive vehicle networks of the present invention.

[0064] The physical entity layer is the actual carrier of the cognitive vehicle network, which is composed of N vehicles on the road, roadside units (RSUs), sensors, and road infrastructure. As the core nodes, the vehicles are equipped with devices such as GPS and inertial sensors to collect dynamic information such as vehicle position, vehicle speed, traffic flow density, and driving direction in real time; the RSUs are deployed at key positions on the road, undertaking the functions of vehicle communication management and data relay in the area and collecting road basic information. The GPS collects basic geographic information data. Through in-vehicle sensors and V2I communication technology, the vehicle continuously senses the surrounding environment and its own communication state during driving, and uploads vehicle dynamic parameters, channel status, and spectrum resource usage in real time, providing the most original physical entity layer data for the digital twin model and being the information source of the entire digital twin system.

[0065] The data interaction layer relies on high-speed wireless communication technologies such as 5G and DSRC to build a two-way data channel between the physical entity layer and the digital twin layer. On the one hand, data such as vehicle position, channel status, and communication requests collected by the physical entity layer are uploaded to the digital twin layer with low latency; on the other hand, instructions such as clustering strategies and channel allocation schemes generated by the digital twin layer are sent to the vehicles and RSUs in the physical entity layer through this layer to achieve virtual-real interaction.

[0066] The digital twin layer constructs a virtual network model, which can generate a road network model and simulate traffic rules, realize vehicle dynamics modeling, present behaviors such as vehicle acceleration, braking, and steering, and construct a virtual network highly consistent with the physical vehicle network. Based on the data uploaded by the data interaction layer, this layer dynamically updates vehicle positions, road basic information, etc. in the virtual network to ensure real-time synchronization with the physical network. At the same time, historical data and real-time parameters are used to preview vehicle movement trajectories and network topology changes. In addition, multi-scenario simulations of the clustering algorithm are carried out through the virtual environment, and the optimal strategy is fed back to the physical entity layer, significantly improving the connectivity and resource utilization rate of the vehicle network.

[0067] The effects of the present invention are further illustrated by the following simulations: In the simulation scenario, the present invention considers the intersection scenario. Vehicles are randomly and uniformly distributed in a 1000*1000m 2 two-dimensional plane area, the transmission power of the roadside unit is 30dBm, the transmission power of the vehicle is 27dBm, the Gaussian white noise spectral density is -174dBm / Hz, and the number of vehicles is 50 - 150.

[0068] In the intersection, the clustering results generated by the present invention in the network scenario of 50 - 150 vehicles are as Figures 3 - 5 shown: Figure 3 is a comparison chart of the average number of channel allocations of the DT-CA algorithm (Digital Twin-based Cognitive Vehicle Network Connectivity Clustering Algorithm) proposed by the present invention, the GC-CA algorithm (Greedy Clustering Connectivity Algorithm), and the GCC-CA (Greedy Clustering Connectivity Algorithm Based on Conflicting Neighbors) algorithm in the case of different numbers of vehicles; Figure 4 is a comparison chart of the maximum local connectivity of the DT-CA algorithm proposed by the present invention, the GC-CA algorithm, and the GCC-CA algorithm in the case of different numbers of vehicles; Figure 5 is a comparison chart of the path efficiency factor of the DT-CA algorithm proposed by the present invention, the GC-CA algorithm, and the GCC-CA algorithm in the case of different numbers of vehicles. It can be Figures 3 - 5 seen that the present invention reduces the number of channel allocations and network energy consumption while ensuring communication reliability through the DT-CA algorithm, significantly improving the resource utilization rate.

[0069] The above are only exemplary embodiments of the present invention and are not intended to limit the scope of protection of the present invention. For those skilled in the art, any obvious deformation or replacement made through conventional technical means based on the technical solution of the present invention shall be regarded as falling within the scope of patent protection of the present invention. Any form of modification, equivalent replacement, improvement, etc., as long as it conforms to the technical concept of the present invention and the essential spirit defined by the claims, shall be included within the scope of protection of the present invention. The adaptive adjustments made by those skilled in the art without departing from the core technical solution of the present invention still belong to the scope of patent protection of the present invention.

Claims

1. A cognitive vehicle network connectivity clustering algorithm based on digital twins, characterized by: include: Build cognitive Internet of Vehicles and digital twin models; Determine whether the roadside unit and the vehicle are within the communication range, establish a roadside unit communicable vehicle probability table and a roadside unit communication probability table, which are used to dynamically manage the vehicle communication connection within the coverage area; Determine whether the vehicle and neighboring vehicles are within effective communication range, and establish a vehicle-reachable neighboring vehicle probability table and a vehicle communication probability table; Each vehicle builds a vehicle communication network through the Dijkstra algorithm according to the vehicle communication probability table; in this vehicle communication network, the inverse of the communication probability is used as the link weight parameter, and the path with the minimum communication probability weight from the current vehicle to all vehicles in the network is calculated through path search; the directly reachable neighbor vehicles and indirectly reachable neighbor vehicles on the path with the minimum communication probability weight are combined to form a reachable cluster, and the directly reachable neighbor vehicles on the path with the minimum communication probability weight form a conflict cluster; Each vehicle adjusts its own transmission power according to the conflict cluster; Each vehicle selects the optimal channel to complete the channel access request based on the channel access probability of the vehicles in the conflict cluster and the corresponding vehicles, and perceives the global network status through the digital twin model to form a connected cluster.

2. The cognitive vehicle network connectivity clustering algorithm according to claim 1 is characterized in that: The way to determine whether the roadside unit and the vehicle are within the communication range is as follows: when the roadside unit receives the channel access data packet sent by the vehicle, it extracts the location of the vehicle from the data packet; then, the roadside unit calculates the Euclidean distance between itself and the vehicle; Subsequently, the roadside unit compares the calculated Euclidean distance with the maximum data transmission distance of the current vehicle to determine whether the roadside unit and the vehicle are within the communication range.

3. The cognitive vehicle network connectivity clustering algorithm according to claim 1, characterized in that: The process of establishing the roadside unit's communicative vehicle probability table is as follows: when parsing the channel access data packet of the communicative vehicle, the communicative vehicle ID, real-time positioning coordinates and corresponding channel access probability parameters are synchronously extracted, and written into the roadside unit's communicative vehicle probability table, and the latest communicative vehicle probability table is broadcast to the entire network through the multicast mechanism.

4. The cognitive vehicle networking connectivity clustering algorithm according to claim 1, characterized in that: The process of establishing the roadside unit communication probability table is as follows: the roadside unit first creates a blank communication probability table and completes initialization; then, the roadside unit will traverse each received channel access data packet and the roadside unit's communicative vehicle probability table; during the traversal process, for each communicative vehicle, the corresponding communication probability between the roadside unit and the communicative vehicle is calculated; then, the roadside unit's own identification, the communicative vehicle ID and the calculated communication probability are written into the roadside unit's roadside unit communication probability table; the roadside unit encapsulates the roadside unit communication probability table into a communication probability data packet and sends it to the surrounding vehicles with the help of a broadcast mechanism.

5. The cognitive vehicle networking connectivity clustering algorithm according to claim 1, characterized in that: The method for determining whether a vehicle is within the effective communication range with its neighboring vehicles is as follows: after receiving the channel access data packet sent by the neighboring vehicle, each vehicle parses the neighboring vehicle ID and positioning coordinates in real time, calculates the Euclidean distance between the current vehicle and the neighboring vehicle, and compares it with the maximum data transmission distance of the current vehicle. The vehicle determines whether it has effective communication conditions with the neighboring vehicle.

6. The cognitive vehicle network connectivity clustering algorithm according to claim 1, characterized in that: The process of establishing the vehicle reachable neighbor vehicle probability table and vehicle communication probability table is as follows: Each vehicle parses the received neighbor vehicle channel access data packet, extracts the vehicle ID, positioning coordinates and channel access probability parameters of the communicable neighbor vehicle, and stores them in the vehicle reachable neighbor vehicle probability table; Each vehicle first initializes the vehicle communication probability table to an empty structure; by semantically parsing the received channel access data packets and the vehicle's reachable neighbor vehicle probability table, it extracts the communicative neighbor vehicle ID, real-time positioning coordinates and corresponding channel access probability parameters, and dynamically calculates the communication probability between the current vehicle and the neighbor vehicle in combination with the maximum effective transmission distance, and writes the vehicle communication probability.

7. The cognitive vehicle network connectivity clustering algorithm according to claim 1, characterized in that: After each vehicle receives the communication probability data packet transmitted by the neighboring vehicle, it constructs and updates the vehicle communication probability table in the cognitive Internet of Vehicles.

8. The cognitive vehicle network connectivity clustering algorithm according to claim 7, characterized in that: The process of forming reachable clusters and conflicting clusters is as follows: Step 9a: construct a communication probability weight table by traversing the updated vehicle communication probability table; Step 9b: In the cognitive vehicle network consisting of N vehicles, the selected vehicles are defined as a selected vehicle list, and the unselected vehicles are defined as an unselected vehicle list; Step 9c: randomly select a vehicle as the starting vehicle, fill the vehicle ID of the starting vehicle into the selected vehicle list, and delete the vehicle ID from the unselected vehicle list, thereby completing the update of the selected vehicle list and the unselected vehicle list; Step 9d: Use Dijkstra's algorithm to construct a minimum spanning tree and create a record list; Step 9e: loop through the screening and recording process of step 9d until the list of unselected vehicles is an empty set. When the iteration terminates, the accumulated triplet data set in the record list constitutes a structured representation of the minimum spanning tree, that is, a set of communication links with optimal communication weights between vehicles. Step 9f: constructing a vehicle communication network formed by a minimum spanning tree according to the obtained communication link set with the optimal communication weight between vehicles; In a vehicle communication network, each vehicle can determine its own directly reachable neighbor vehicles and indirectly reachable neighbor vehicles, which together constitute a reachable cluster; The directly reachable neighbor vehicles on the path with the minimum communication probability weight form a conflict cluster; The minimum spanning tree is used to optimize the intra-cluster path within the conflict cluster. If there is a disconnection problem in the local area of ​​the conflict cluster, the Steiner algorithm is introduced to insert a virtual relay vehicle.

9. The cognitive vehicle network connectivity clustering algorithm according to claim 8, characterized in that: The process by which each vehicle adjusts its own transmission power according to the conflict cluster is as follows: Build a table of directly reachable neighbor vehicles based on the record list to clarify the directly reachable neighbor vehicles of each vehicle; Each vehicle then obtains all neighbor vehicles from the vehicle's directly reachable neighbor vehicle table, obtains the neighbor vehicle's positioning position from the vehicle's reachable neighbor vehicle probability table, uses the Euclidean distance formula to calculate the distance to each neighbor vehicle, and selects the directly reachable neighbor vehicle farthest from the current vehicle. Finally, the vehicle's own transmission power is dynamically adjusted based on this result.

10. The cognitive vehicle network connectivity clustering algorithm according to claim 1, characterized in that: Each vehicle selects the optimal channel to complete the channel access request based on the channel access probability of the vehicles in the conflict cluster and the corresponding vehicles, and perceives the global network status through the digital twin model to form a connected cluster. The specific steps include: Step 11a: In the cognitive Internet of Vehicles environment, there are C channel resources available for vehicles to choose to access; the current vehicle is within the conflict cluster range, and the spectrum sensing technology is used to screen the currently available idle channel resource blocks; the current vehicle autonomously selects and accesses an unoccupied channel resource block, and after successful access, creates and maintains a channel resource package; and broadcasts the channel resource package to surrounding vehicles; Step 11b: Determine whether the vehicle ID in the channel resource package is in the conflict cluster. The specific steps include: Step 11b1: When other vehicles in the conflict cluster receive the channel resource packet sent by the current vehicle, they first parse the channel resource block allocation information contained in the channel resource packet; mark the neighbor vehicles occupying the same channel, and create a channel allocation packet and broadcast it to the directly reachable neighbor vehicles of the current vehicle; if the vehicles other than the current vehicle in the conflict cluster have not yet been allocated a channel, the channel allocation packet is marked as an empty packet, and the directly reachable neighbor vehicles of the corresponding vehicle are responsible for broadcasting the channel allocation packet; if the channel has been allocated, no marking is performed, and the channel resource packet transmitted by the current vehicle continues to be broadcast; Step 11b2: After receiving the channel allocation packet sent by the directly reachable neighbor vehicle, the vehicles in the conflict cluster sort according to the channel access probability and select different channels as their own available channels; Step 11c: The digital twin model evaluates and optimizes the channel allocation of the actual Internet of Vehicles based on the channel usage and vehicle communication requirements in the virtual network model until all vehicles have been allocated channels, ensuring the rational use of channel resources and forming a connected cluster; the digital twin model verifies the effectiveness of the allocation plan and sends optimization instructions to the physical entity layer through the data interaction layer.

Images