Cooperative caching scheme for large file transmission in Internet of Vehicles

Abstract

The invention discloses a cooperative caching scheme oriented to large file transmission in the Internet of Vehicles, and belongs to the technical field of mobile communication of the Internet of Vehicles. The method aims at solving the problems of large file downloading interruption and poor user experience caused by sparse deployment of roadside units and high-speed movement of vehicles in scenes such as highways. The core of the method is that firstly, a cooperative vehicle selection mechanism is designed, and a candidate cooperative vehicle cluster capable of keeping stable connection with a requesting vehicle in a communication blind area is dynamically screened out based on real-time speed and position information of a vehicle; and secondly, aiming at the problem that a multi-request vehicle competes for cache resources, constructing an optimization model with the purpose of maximizing the average throughput of the system, and solving an optimal cache data distribution strategy by adopting an improved particle swarm optimization algorithm. Through the strategy, the roadside unit can pre-cache part of file data to the selected cooperative vehicle; when the request vehicle leaves the coverage area of the roadside unit and enters the communication blind area, the request vehicle can continuously acquire the data which is not downloaded from the passing cooperative vehicle through vehicle-to-vehicle communication, so that a seamless cooperative transmission link is formed. According to the method, idle storage and communication resources of the vehicle can be effectively utilized, the system throughput is remarkably improved, the downloading time delay is reduced, and the resource utilization rate of the cooperative vehicle is improved.

Description

Technical Field

[0001] This invention belongs to the field of mobile communications, specifically to a collaborative caching scheme for large file transfers in the Internet of Vehicles. Background Technology

[0002] With the rapid development of Vehicle-to-Everything (IoV) technology, especially the widespread adoption of 5G / 6G networks and the extensive application of intelligent vehicles, communication between vehicles and between vehicles and their local utility units (RSUs) has become possible. Based on these communication capabilities, IoV shows great potential in providing data-intensive entertainment services such as online video and cloud gaming. However, these services typically involve downloading large-scale file content. In typical scenarios such as highways, due to the high speed of vehicles and the large intervals between RSU deployments, the file download process faces challenges such as unstable communication, limited bandwidth, and service interruptions. These issues not only severely impact the user experience but also place higher performance demands on the transfer of large files in the connected vehicle environment.

[0003] To address the aforementioned challenges, numerous domestic publications have extensively researched how to efficiently utilize communication and caching resources in the Internet of Vehicles (IoV) to improve the download performance of large files. The paper [Yu G, He Y, Wu J, et al. Mobility-aware proactive edge caching for large files in the internet of vehicles[J]. IEEE Internet of Things Journal, 2023, 10(13): 11293-11305.] proposes a mobility-aware proactive edge caching scheme. By predicting vehicle trajectories, it pre-deploys requested files to edge nodes that may be traversed in the future and designs a system recovery mechanism to address prediction errors, thereby improving the caching success rate. The paper [Fan Q, Li X, Li J, et al. PA-cache: Evolving learning-based popularity-aware content caching in edge networks[J]. IEEE Transactions on Network and Service Management, 2021, 18(2): 1746-1757.] proposes an edge node-aware content caching algorithm for large file caching. This algorithm adaptively learns the changes in content popularity over time and pre-caches popular content to edge nodes. Compared with traditional popularity caching algorithms, this strategy significantly reduces download latency during large file retrieval, thus achieving more efficient resource utilization and quality of service in the vehicular network environment. However, while this type of method effectively alleviates file caching pressure, it still inevitably interrupts the download process when a vehicle enters a communication dead zone between adjacent RSUs due to its complete reliance on RSUs. To enable vehicles to better acquire file data and fully utilize the characteristics of V2V communication, the authors in the paper [Wang S, Ma F, Yu Z. RAMC: Reverse-auction-based multilevel cooperation for large size datadownload in VANETs[J]. IEEE Internet of Things Journal, 2023, 11(6): 11087-11100.] proposed a cost-effective multilevel cooperative download mechanism. This mechanism uses a reverse auction method to incentivize vehicles to participate in cooperative caching and fully utilize the storage and bandwidth resources of candidate vehicles, thus alleviating data download pressure. Finally, it was verified that the scheme can significantly improve the throughput of requesting vehicles and promote cooperation between vehicles with less latency and overhead.To extend the communication time between vehicles, the authors in the paper [Jin Z, Song T, Jia W K. An adaptive cooperative caching strategy for vehicular networks[J].IEEE Transactions on Mobile Computing,2024,23(10):9502-9517.] proposed an adaptive cooperative caching strategy, which uses topology awareness and greedy approximation algorithms to solve the problem, thereby improving the cache hit rate and reducing the average transmission latency. Although such schemes extend the download time to some extent, they generally ignore the impact of high-speed motion on the stability of V2V links and lack a stability screening mechanism based on vehicle speed and relative position.

[0004] While existing research has made some progress in optimizing resource allocation and improving user download experience, two key shortcomings remain: First, most research schemes ignore the fact that there are communication blind spots between adjacent RSUs, or simply assume that communication can be continued within the blind spot through simple relays, failing to design a dedicated and reliable transmission guarantee mechanism for the characteristics of the blind spot. Once a vehicle leaves the coverage area of ​​an RSU, its download process will be forcibly interrupted until it enters the next RSU, which causes a serious performance bottleneck in highway scenarios with large intervals. Second, although some studies have attempted to use V2V communication for collaborative caching, they have failed to fully consider the serious impact of high-speed vehicle movement on the stability of V2V links. In high-speed moving environments, the relative positions and connection states between vehicles change rapidly. If vehicles that can communicate stably in the future are not accurately selected based on their movement states (speed, position), the established V2V connection is very likely to be quickly interrupted within the communication blind spot, leading to collaborative caching failure. Based on the above analysis, this paper proposes a collaborative caching scheme for large file transfers, which aims to systematically solve the challenges posed by communication blind spots and vehicle mobility. Summary of the Invention

[0005] This invention aims to solve the problems of the prior art and proposes a collaborative caching scheme for large file transfers in the Internet of Vehicles (IoV). The technical solution of this invention is as follows:

[0006] A collaborative caching scheme for large file transfers in the Internet of Vehicles (IoV) includes the following steps:

[0007] First, based on real-time acquired vehicle registration information (including vehicle ID, real-time speed, and location), the RSU designs a cooperative vehicle selection mechanism. This mechanism dynamically selects vehicles capable of maintaining stable V2V connections by calculating the maintainable communication time between the requesting vehicle and candidate vehicles within the communication blind zone, forming a cooperative vehicle cluster. Next, after selecting the cooperative vehicle cluster, the problem of allocating cached data is modeled as an optimization problem aiming to maximize the system's average throughput. The IPSO algorithm is used to solve this problem, obtaining the optimal cache allocation moment strategy. Finally, through this strategy, the RSU can pre-cache some data on cooperative vehicles; after the requesting vehicle leaves the RSU's coverage area, it can continue to retrieve undownloaded data from cooperative vehicles using V2V communication, effectively avoiding download interruptions caused by insufficient coverage of a single RSU.

[0008] Furthermore, the cooperative vehicle selection mechanism based on vehicle speed and location information includes the following steps: First, a cooperative matrix C = [c ij ] N×M Its element c ij This indicates whether vehicle j is a CV of vehicle i. The specific assignment rules are as follows:

[0009]

[0010] The RV must meet the following two conditions when selecting a CV: First, the CV must have remaining cache capacity to undertake collaborative caching tasks; second, when the RV enters a communication dead zone, the CV must be able to maintain communication with it to achieve data transmission.

[0011] Furthermore, a two-dimensional rectangular coordinate system is established with the RSU vertical projection point as the origin, the X-axis along the road direction, and the Y-axis perpendicular to the road direction upwards. In this coordinate system, the position and velocity of vehicle A are respectively (x... A ,y A ) and v A The position and speed of vehicle B are (x) B ,y B ) and v B The coverage area of ​​the RSU is a horizontal distance l and a height h above the ground. The ground projection distance from the position of vehicle A to the origin (RSU projection point) is... Considering the deployment height of the RSU, the actual straight-line distance between vehicle A and the RSU is calculated using formula (2):

[0012]

[0013] Furthermore, the time when vehicle A left the RSU. It can be expressed by formula (3):

[0014]

[0015] Furthermore, the initial distance d between vehicle A and vehicle B AB This can be expressed by formula (4):

[0016]

[0017] Furthermore, based on the initial distance d between the two vehicles... AB Whether it is within the communication range R between the two vehicles will be discussed in two cases:

[0018] Scenario 1: When the initial distance d between vehicle A and vehicle B is... AB When the distance is greater than R, the two vehicles cannot communicate directly. In this case, it is necessary to calculate the distance d between the two vehicles. AB Reduce the time required to R If vehicles A and B have the same speed, the distance between them will remain constant, and vehicle B cannot become the CV of vehicle A. When vehicles A and B have different speeds, their initial positions must also be considered. This paper mainly discusses scenarios where communication may be established, including two cases: (a) and (b). Case (a): Vehicle A's speed is greater than vehicle B's, and vehicle A is behind vehicle B; Case (b): Vehicle A's speed is less than vehicle B's, and vehicle A is in front of vehicle B. Under these two cases, the time required for the two vehicles to establish communication is calculated. It can be calculated using formula (5):

[0019]

[0020] Over time Afterwards, the distance between vehicle A and vehicle B decreases to R, satisfying the communication requirements. If vehicle A has entered a communication dead zone at this point, vehicle A and vehicle B can directly establish a V2V communication link for data transmission. In this case, the duration for which vehicle A and vehicle B can transmit data is... Essentially equivalent to the time the two vehicles can maintain communication. As shown in formula (6):

[0021]

[0022] In this scenario, after the two vehicles initiate communication, they do not immediately disconnect. Due to the speed difference between them, they will gradually drift apart until the distance between them again exceeds the communication range R. Therefore, the two vehicles can maintain communication for a certain period of time. The calculation depends on two core factors: the relative speeds of the two vehicles along the road direction (X-axis) |v B -v ASecondly, the effective length of the communication range R in the directions of movement of the two vehicles. It can be calculated from formula (7):

[0023]

[0024] If vehicle A has not yet entered the communication dead zone, even if the distance between the two vehicles has reached the communication range, data transmission between them will be temporarily suspended. At this time, RV only downloads data from RSU via the high-bandwidth V2I link, while CV receives buffered data distributed by RSU. In this situation, the time during which the two vehicles can transmit data is... Remaining time required for vehicle A to leave RSU coverage area The time during which communication can be maintained between the two vehicles The decision is made jointly, as expressed by formula (8):

[0025]

[0026] Scenario 2: When the initial distance d between vehicle A and vehicle B is... AB When R ≤ R, the two vehicles are initially within communication range. The duration for which the two vehicles can maintain communication at this time is... It can be calculated using formula (9):

[0027]

[0028] Since vehicle A is initially within the RSU's range and has not yet entered the communication dead zone, it is necessary to calculate the effective time during which vehicle A can transmit data with vehicle B within the communication dead zone. This time depends primarily on two parameters: the time required for vehicle A to leave the RSU coverage area. And the time during which the two vehicles can maintain communication. This can be expressed by formula (10):

[0029]

[0030] Furthermore, the time during which vehicle A and vehicle B can transmit data within a communication blind spot. It has been calculated. If... And Cap B If > 0, then determine c. AB =1. Finally, the cooperative matrix C is obtained, which serves as the input for cache allocation optimization.

[0031] Furthermore, after selecting the CV cluster, it is necessary to optimize the proportion of data cached by each CV for the RV. This paper defines the cache allocation matrix U = [u ij ] N×M , where u ijThis indicates the size of the data cached by CV for RVi. However, the cache capacity of each CV is limited, therefore the total cached data of each CV must not exceed its cache capacity Cap, and the total cached data must meet the following constraints:

[0032]

[0033] Secondly, the amount of data a vehicle can acquire from the CV (CV Controller) after entering a communication blind spot is limited by distance, transmission rate, and bandwidth within the blind spot. Therefore, the amount of data the CV provides to the RV cannot exceed the maximum transmittable data amount between the two within the communication blind spot. That is:

[0034]

[0035] in, This is the time interval during which the two can maintain an effective connection within a communication dead zone. Let be the V2V communication rate between CVj and RVi at time t. According to Shannon's formula, the transmission rate of a communication link (including V2I and V2V) can be uniformly expressed as:

[0036]

[0037] For both V2I and V2V communication modes, the values ​​of each parameter in formula (13) are defined as follows: In V2I communication, B is the V2I link bandwidth and P is the RSU transmit power; in V2V communication, B is the V2V link bandwidth and P is the vehicle transmit power. Furthermore, h is the channel gain and N0 is the Gaussian white noise power. Because the data transmission rate between vehicles changes with distance, in order to accurately calculate the maximum amount of data that can be transmitted by a vehicle in the communication blind zone, it is necessary to analyze the dynamically changing... Integrate the points.

[0038] Furthermore, based on the proposed collaborative caching scheme, the amount of data acquired by RVi within a download cycle mainly consists of two parts: the data downloaded by RVi from RSU and the data obtained by RVi through CV in the communication dead zone. Therefore, the amount of data downloaded by RVi within a download cycle is expressed by formula (14):

[0039]

[0040] Among them, D V2I,i The amount of data downloaded directly by the vehicle from the current RSU range is calculated using formula (15):

[0041]

[0042] in, This is the time it takes for RVi to download data from RSU. In this article, Equal to the time required for the vehicle to leave the current RSU This time is calculated using formula (3). V2I This refers to the data transfer rate from RSU to RV.

[0043] Furthermore, because the data transmission rate between the vehicle and the RSU varies with distance, accurate calculation requires processing the dynamically changing RSU. V2I Integrating, as shown in formula (16):

[0044]

[0045] D V2V,i The amount of data that RVi acquires via CV in the communication dead zone is calculated using formula (17):

[0046]

[0047] Where D i,j CVj represents the size of the data cached by RVi.

[0048] Furthermore, RVi's total migration time T within a download cycle i It includes two parts: the time elapsed from when RVi issues a download request to when the user leaves the current RSU coverage area. and the time the vehicle travels in the communication blind spot Right now:

[0049]

[0050] Furthermore, to fully utilize the vehicle's idle cache resources, this paper models the CV cache resource allocation problem as an optimization problem aimed at maximizing the system's average throughput. Based on formulas (17) and (18), the throughput of RVi is expressed as η. i =D i / T i Therefore, the system's average throughput can be expressed as:

[0051] The optimization variable is the cooperation matrix C = [c ij ] N×M and cache allocation matrix U = [u ij ] N×M The optimization problem can be formulated as follows:

[0052]

[0053] Specifically, c1 ensures that the amount of data cached by each CV for each RV does not exceed its own capacity limits (cache capacity, ability to retrieve data from RSU, and ability to transfer data to RV). c2 ensures that the total amount of data obtained by each CV does not exceed the CV's cache capacity and the maximum amount of data that can be downloaded within RSU coverage. c3 ensures that the total amount of data cached by all CVs for each RV does not exceed the data size requested by the RV.

[0054] Furthermore, the collaborative cache allocation problem can be modeled as a high-dimensional, multi-constrained combinatorial optimization problem, where the decision variable is a continuous value (the amount of cached data) and the constraints are complex. Traditional mathematical programming methods are difficult to solve efficiently, while Particle Swarm Optimization (PSO) is an efficient swarm intelligence search method that can handle such complex optimization problems well. However, classic PSO suffers from premature convergence and is prone to getting trapped in local optima, making it difficult to obtain satisfactory cache allocation results in the dynamic environment of vehicle networks. Therefore, this paper proposes an improved Particle Swarm Optimization (IPSO) algorithm. The core idea includes two aspects:

[0055] (1) Introduce a circular neighborhood sharing mechanism to enhance population diversity and delay premature convergence;

[0056] (2) An adaptive adjustment strategy for inertial weight based on simulated annealing is proposed to dynamically balance global search and local development capabilities.

[0057] Furthermore, in IPSO, the position X of a particle k The direct mapping is to a cooperative cache allocation matrix U, where each element represents the amount of data cached by a certain CV for a certain RV. During initialization, the particle positions must satisfy the cooperative relationship constraints as well as the cache capacity and transmission capability constraints (i.e., formulas (11) and (12)) to ensure the validity of the initial solution. The particle positions are represented by formula (21):

[0058] X k ={u 11 ,u 12 ,...,u 1M ,u 21 ,...,u NM} (twenty one)

[0059] The optimization objective of the IPSO algorithm is to maximize the fitness value; a larger fitness value results in better optimization performance, while a smaller fitness value results in worse performance. This model aims to maximize the system's average throughput. When defining the fitness function, this paper directly uses the system's average throughput as the fitness value, which intuitively achieves the goal of maximizing throughput. Therefore, the fitness function is defined as follows:

[0060]

[0061] Furthermore, to overcome the premature convergence of traditional PSO, during particle update, the traditional method of relying on the global optimum to guide particle updates is abandoned. Instead, the update direction is guided by the particle with the maximum fitness value in its neighborhood. This strategy aims to prevent the particle swarm from converging to a local optimum prematurely. The particle velocity update formula is as follows:

[0062] V k =ωV k +c1r1(pbest k -X k )+c2r2(nbest k -X k )+c3r3(gbest'-X k ) (twenty three)

[0063] Among them, nbest k It represents the optimal position in the neighborhood of each particle. 'gbest' is a variant of the global optimal solution, defined as follows:

[0064] gbest'=gbest+e -ηg / G C(0,1) (24)

[0065] Where η = 10, and C(0,1) are random numbers generated by the Cauchy distribution function.

[0066] Furthermore, drawing inspiration from Simulated Annealing (SA), this paper combines the inertia weight with the annealing temperature, making it dynamically change throughout the search process. The core idea of ​​simulated annealing is that the system is active at high temperatures, capable of accepting inferior solutions with a certain probability, thus escaping local optima; as the temperature gradually decreases, the system tends to stabilize, eventually converging to an approximate global optimum. In the improved strategy proposed in this paper, the inertia weight is defined as a function that dynamically adjusts with temperature decay:

[0067]

[0068] Where, ω min ,ω max Let represent the maximum and minimum values ​​of the inertial weight, ΔE represent the current energy difference of the particle, λ represent the Boltzmann constant, and T decreases with the number of iterations according to the cooling coefficient r. This strategy, by introducing energy change and temperature parameters, allows the inertial weight to be dynamically adjusted according to the actual situation during the search process, thereby enhancing the algorithm's global search capability and stability. Based on the above improved strategy, the particle velocity update formula is changed to:

[0069]

[0070] This strategy enables the algorithm to dynamically adjust the inertia weights based on feedback from the search status.

[0071] Furthermore, based on the iterative solution of the above algorithm, RSU can finally obtain the globally optimal cooperative caching decision matrix. This matrix precisely quantifies the proportion of data cached by each CV for the RV, forming a complete cooperative caching scheme. RSU then distributes data to designated vehicles according to this scheme.

[0072] The advantages and beneficial effects of this invention are as follows:

[0073] 1. A collaborative vehicle selection mechanism was designed. Through this mechanism, RSU can select vehicle clusters suitable for participating in cache collaboration, providing an effective collaborative vehicle foundation for subsequent collaborative caching.

[0074] 2. A cooperative caching scheme based on improved particle swarm optimization is proposed. This scheme can optimize the amount of data cached by each cooperative vehicle for the requesting vehicle, thereby maximizing the average system throughput and improving the user experience. Attached Figure Description

[0075] Figure 1 This is a system scene diagram;

[0076] Figure 2 This is a schematic diagram showing the coordinates of vehicles within the RSU area;

[0077] Figure 3 A diagram illustrating that the initial distance of the vehicle is greater than the communication distance;

[0078] Figure 4 This is a diagram illustrating that the initial distance between the vehicle and the communication distance is less than the initial distance between the two vehicles.

[0079] Figure 5 Circular neighborhood shared graph;

[0080] Figure 6 This is a graph showing the comparison of average throughput.

[0081] Figure 7 A graph showing the relationship between average throughput and communication dead zone interval;

[0082] Figure 8 A graph showing the relationship between average download latency and communication dead zone interval;

[0083] Figure 9 This is a comparative experimental diagram showing the utilization rate of collaborative vehicle resources.

[0084] Figure 10 This is a comparison chart of the algorithm's convergence performance;

[0085] Figure 11 Figures are attached to the abstract; Detailed Implementation

[0086] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0087] The technical solution of the present invention is as follows:

[0088] In the Internet of Vehicles (IoV) with integrated mobile edge caching, to address the download interruption problem in highway scenarios with sparse RSU deployment, a cooperative vehicle selection mechanism based on vehicle speed and location information is designed. Based on this, an improved particle swarm optimization cooperative caching scheme is proposed, which fully explores and utilizes the idle storage resources of the vehicle group, significantly improves the average system throughput, and reduces the overall download latency.

[0089] The collaborative caching scheme for large file transfer in the Internet of Vehicles proposed in this invention includes the following steps:

[0090] Step 1: When a requesting vehicle initiates a large file download request to the RSU, the RSU executes a cooperative vehicle selection mechanism based on real-time vehicle registration information (including vehicle ID, real-time speed, and location). This mechanism dynamically selects vehicles that can maintain a stable V2V connection with the requesting vehicle by calculating the maintainable communication time between the requesting vehicle and candidate vehicles within the communication blind zone, thus forming a cooperative vehicle cluster.

[0091] Step 2: After selecting the cooperative vehicle cluster, the RSU models the problem of allocating cached data as an optimization problem aimed at maximizing the system's average throughput. An improved particle swarm optimization algorithm is used to solve this problem, obtaining the optimal cache allocation matrix. Based on this optimal cache allocation strategy, the RSU distributes the corresponding file data to each cooperative vehicle for pre-caching. When the requesting vehicle leaves the RSU's coverage area and enters a communication dead zone, it retrieves the remaining undownloaded data from passing cooperative vehicles via V2V communication, thus achieving seamless continuous downloading.

[0092] To evaluate the performance of the proposed algorithm, a system scenario was simulated using the MATLAB simulation platform. The simulation scenario was a highway scenario with sparsely deployed RSUs, and the distance between adjacent RSUs varied between 1 and 11 km, with a fixed value of 8 km. This paper assumes that all RSUs have cached the files required by RVs, and vehicles can directly download the required data from the RSUs. Vehicles are randomly distributed on the road.

[0093] In the simulation experiment, in order to verify the effectiveness of the proposed collaborative caching scheme, three typical schemes were selected for comparison. (1) No collaborative vehicle transmission, relying on a single RSU for download, denoted as "RSU-Only". (2) CV randomly caches part of the file data for RV, denoted as "Random-CV". (3) Select the CV closest to RV as the collaborative node to cache the remaining data, which is the scheme proposed in the literature [Zhu X, Hua Q, Yan W, et al. A vehicle–road urban sensing framework for collaborative content delivery in freeway-oriented vehicular networks[J].IEEE Sensors Journal,2023,24(5):5662-5674.]. After RV leaves the coverage area of ​​RSU, data is transmitted through CV, denoted as "Nearest-CV". At the same time, the scheme proposed in this paper is denoted as "Proposed".

[0094] Figure 6 The impact of the number of vehicle cavities (CVs) and the V2V communication distance on the average throughput of the system is shown. The PSO (Program for Nearest-CV) scheme follows the same approach as the proposed scheme, using the traditional PSO algorithm in its algorithm design. Subfigure (a) shows that when the number of CVs increases from 4 to 12, the average throughput of the proposed scheme increases from approximately 2.5 Mb / s to about 3.6 Mb / s, an increase of 40%. In contrast, the traditional PSO algorithm shows a limited throughput improvement (only about 25%). This indicates that the improved PSO algorithm avoids the particle swarm from getting trapped in local optima through a cyclic neighborhood sharing mechanism. Furthermore, with a larger scale of 12 CVs, the proposed scheme shows a significant improvement over the traditional PSO algorithm, verifying the effectiveness of the dynamic learning factor and simulated annealing inertia weight adjustment. The Nearest-CV scheme, due to its reliance on neighboring CV nodes, is susceptible to vehicle mobility, resulting in large throughput fluctuations. Both the Nearest-CV and Random-CV schemes, due to the lack of global resource allocation optimization, show significantly lower throughput compared to the proposed scheme, verifying the necessity of cooperative vehicle selection mechanisms and cache allocation strategies. The average throughput of RSU-Only remains stable at around 1.8 Mb / s, significantly lower than that of the cooperative scheme. This is because vehicles in communication dead zones cannot acquire data via the V2I link, resulting in a complete download interruption. Subfigure (b) shows that the average throughput of the CV cooperative caching scheme improves as the V2V communication distance increases. This is because the greater communication distance expands the range of cooperative vehicles and allows for more available caching resources.

[0095] Figure 7The relationship between average throughput and communication dead zone interval is shown. When the communication dead zone interval is from 0 to 2 km, the average throughput of the proposed scheme increases slightly. When the dead zone interval is from 5 to 10 km, the average throughput of the proposed scheme drops from 7.8 Mb / s to 3.1 Mb / s, significantly lower than other schemes. The Nearest-CV scheme performs well with short dead zone intervals (0-4 km), but the throughput drops rapidly when the interval exceeds 4 km. The fundamental reason is that the nearest CV node is always selected, but in high-speed scenarios, nearby CVs are prone to leaving the communication range, resulting in poor stability of the cooperative link. The random allocation strategy of the Random-CV scheme leads to low resource utilization. When the RSU-Only scheme relies solely on RSU transmission, the average throughput drops sharply with the increase of the dead zone interval. This is because vehicles cannot obtain data through the V2I link within long dead zone intervals and rely entirely on the RSU to complete data download.

[0096] Figure 8 The impact of communication dead zone spacing on average download latency is shown. Average download latency is the average of the total time from request initiation to file download completion for all RVs. When the communication dead zone spacing increases from 0km to 4km, the average download latency of the proposed scheme tends to stabilize. When the communication dead zone spacing increases from 4km to 10km, the average download latency gradually increases, but the increase is significantly lower than other schemes. The RSU-Only scheme increases linearly with the dead zone spacing because it relies entirely on RSU coverage. The Nearest-CV scheme performs better with short dead zones, but the latency increases rapidly when the spacing exceeds 4km. The Random-CV scheme, with its random resource allocation, leads to resource contention and load imbalance, and the latency increases non-linearly with the increase of the dead zone. As can be seen from the figure, the proposed algorithm has the best average download latency under long-distance dead zone spacing.

[0097] Figure 9The impact of V2V communication distance and communication blind zone interval on CV resource utilization is shown. CV resource utilization is defined as the ratio of the amount of data cached by the CV to its total cache capacity. Subfigure (a) shows the relationship between CV resource utilization and V2V communication distance. As the V2V communication distance increases from 100m to 500m, the CV resource utilization of all schemes significantly improves. This is because the extended communication distance increases the coverage of inter-vehicle cooperation, thus extending the communication time window between the CV and RV, allowing for the caching of more data. The Nearest-CV scheme has lower utilization in short-distance (100-200m) scenarios because it always selects the nearest CV node, and the high-speed movement of vehicles can easily cause nearby CVs to quickly leave the communication range, limiting available resources. Random-CV, due to random allocation leading to resource contention, shows less improvement compared to the above two schemes. Subfigure (b) shows the relationship between CV resource utilization and communication blind zone interval. As the communication blind zone interval increases, the utilization of the proposed scheme and Nearest-CV tends to stabilize. This is because as the blind zone length increases, no new CVs join the collaborative cluster, and the system relies on the initially allocated CVs to complete data transmission. As can be seen from the two figures, the resource utilization of the RSU-only scheme is always 0 because no CVs participate in the transmission. The scheme proposed in this paper outperforms other schemes in terms of CV resource utilization because the improved PSO circular neighborhood sharing mechanism and inertia weight adjustment fully optimize cache allocation.

[0098] Figure 10 The convergence curves of the fitness values ​​of four algorithms during the iteration process are shown. It can be seen that the IPSO algorithm significantly outperforms the other algorithms in both convergence speed and the quality of the final solution. The traditional PSO algorithm, due to premature convergence, gets stuck in a local optimum after about 70 iterations. The GA algorithm has strong global search capabilities, but its convergence speed is slow, and the quality of its final solution is not as good as IPSO. While the SA algorithm can escape local optima in the early stages, its serial search characteristic results in the slowest convergence speed. The IPSO algorithm, benefiting from its cyclic neighborhood sharing mechanism and simulated annealing-based inertial weight adjustment strategy, effectively balances global exploration and local exploitation capabilities, resulting in not only faster convergence speed but also finding a better cache allocation strategy, ultimately achieving the highest fitness value.

[0099] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0100] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined in this invention, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0101] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0102] The above embodiments should be understood as illustrative only and not as limiting the scope of protection of the present invention. After reading the description of the present invention, those skilled in the art can make various alterations or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A collaborative caching scheme for large file transfer in vehicle networking, characterized in that, Includes the following steps:

101. The requesting vehicle initiates a file download request to the RSU. Based on the real-time motion status information of vehicles within its coverage area, the RSU executes a cooperative vehicle selection mechanism to select cooperative vehicles from the candidate vehicles that can maintain stable communication with the requesting vehicle in the communication blind spot, so as to form a cooperative vehicle cluster.

102. RSU models the cache resource allocation problem with the goal of maximizing system throughput as an optimization model, and uses an improved particle swarm optimization algorithm to solve the optimization model to obtain the optimal cache allocation strategy.

103. The RSU distributes a portion of the file requested by the requesting vehicle to the corresponding cooperating vehicles in the cooperating vehicle cluster for pre-caching according to the optimal cache allocation strategy.

104. When the requesting vehicle enters the communication blind zone, it obtains the pre-cached remaining data from the cooperating vehicles in the cooperative vehicle cluster via V2V communication to achieve seamless continuous download.

2. The collaborative caching scheme for large file transfer in a vehicle network as described in claim 1, characterized in that, The construction of the cooperative vehicle selection mechanism specifically includes:

201. Based on the real-time position and speed of the requesting vehicle and the candidate vehicle, calculate the communication time during which the two can effectively transmit data within the communication blind zone; 202. If the communication time is greater than zero, then it is determined that the candidate vehicle can become a cooperating vehicle of the requesting vehicle.

3. A collaborative caching scheme for large file transfer in a vehicle network as described in claim 2, characterized in that, Calculate the communication time during which both parties can effectively transmit data within the communication dead zone, specifically including: The proposed cooperation matrix C = [c ij ] N×M Its element c ij This indicates whether vehicle j is a CV of vehicle i. The specific assignment rules are as follows: The RV must meet the following two conditions when selecting a CV: First, the CV must have sufficient remaining buffer capacity to handle collaborative buffering tasks; second, when the RV enters a communication dead zone, the CV must be able to maintain communication to achieve data transmission. Establish a two-dimensional Cartesian coordinate system with the RSU vertical projection point as the origin, the X-axis along the road direction, and the Y-axis perpendicular to the road direction upwards. In this coordinate system, the position and velocity of vehicle A are (x...). A ,y A ) and v A The position and speed of vehicle B are respectively (x B ,y B ) and v B The coverage area of ​​the RSU is a horizontal distance l and a height h above the ground. The ground projection distance from the position of vehicle A to the origin (RSU projection point) is... Considering the deployment height of the RSU, the actual straight-line distance between vehicle A and the RSU is calculated using formula (2): The time when vehicle A leaves the RSU can be expressed by formula (3): The initial distance d between vehicle A and vehicle B AB This can be expressed by formula (4): To analyze whether vehicle A and vehicle B can establish an effective V2V communication link within a communication blind zone, this paper will consider the initial distance d between the two vehicles. AB Whether a vehicle is within the communication range R will be discussed in two cases. This article assumes that vehicle A is an RV and vehicle B is a CV, and the following sections will analyze these two cases in detail. Scenario 1: When the initial distance d between vehicle A and vehicle B is... AB When the distance is greater than R, the two vehicles cannot communicate directly. In this case, it is necessary to calculate the distance d between the two vehicles. AB Reduce the time required to R If vehicles A and B have the same speed, the distance between them will remain constant, and vehicle B cannot become the CV of vehicle A. When vehicles A and B have different speeds, their initial positions must also be considered. This paper mainly discusses scenarios where communication may be established, including two cases: (a) and (b). Case (a): Vehicle A's speed is greater than vehicle B's, and vehicle A is behind vehicle B; Case (b): Vehicle A's speed is less than vehicle B's, and vehicle A is in front of vehicle B. Under these two cases, the time required for the two vehicles to establish communication is calculated. It can be calculated using formula (5): Over time Afterwards, the distance between vehicle A and vehicle B decreases to R, satisfying the communication requirements. If vehicle A has entered a communication dead zone at this point, vehicle A and vehicle B can directly establish a V2V communication link for data transmission. In this case, the duration for which vehicle A and vehicle B can transmit data is... Essentially equivalent to the time the two vehicles can maintain communication. As shown in formula (6): In this scenario, after the two vehicles initiate communication, they do not immediately disconnect. Due to the speed difference between them, they will gradually drift apart until the distance between them again exceeds the communication range R. Therefore, the two vehicles can maintain communication for a certain period of time. The calculation depends on two core factors: the relative speeds of the two vehicles along the road direction (X-axis) |v B -v A Secondly, the effective length of the communication range R in the directions of movement of the two vehicles. It can be calculated from formula (7): If vehicle A has not yet entered the communication dead zone, even if the distance between the two vehicles has reached the communication range, data transmission between them will be temporarily suspended. At this time, RV only downloads data from RSU via the high-bandwidth V2I link, while CV receives buffered data distributed by RSU. In this situation, the time during which the two vehicles can transmit data is... Remaining time required for vehicle A to leave RSU coverage area The time during which communication can be maintained between the two vehicles The decision is made jointly, as expressed by formula (8): Scenario 2: When the initial distance d between vehicle A and vehicle B is... AB When R ≤ R, the two vehicles are initially within communication range. The duration for which the two vehicles can maintain communication at this time is... It can be calculated using formula (9): Since vehicle A is initially within the RSU's range and has not yet entered the communication dead zone, it is necessary to calculate the effective time during which vehicle A can transmit data with vehicle B within the communication dead zone. This time depends primarily on two parameters: the time required for vehicle A to leave the RSU coverage area. And the time during which the two vehicles can maintain communication. This can be expressed by formula (10):

4. A collaborative caching scheme for large file transfer in a vehicle network as described in claim 1, characterized in that, The cache resource allocation problem with the goal of maximizing system throughput is modeled as an optimization model, specifically including:

401. After selecting the CV cluster, it is necessary to optimize the proportion of data cached by each CV for RV. This paper defines the cache allocation matrix U = [u ij ] N×M , where u ij This indicates the size of the data cached by RVi (CVj). However, the cache capacity of each CV is limited, therefore the total cached data of each CV must not exceed its cache capacity (Cap), and the total cached data must meet the following constraints: Secondly, the amount of data a vehicle can acquire from the CV (CV Controller) after entering a communication blind spot is limited by distance, transmission rate, and bandwidth within the blind spot. Therefore, the amount of data the CV provides to the RV cannot exceed the maximum transmittable data amount between the two within the communication blind spot. That is: in, This is the time interval during which the two can maintain an effective connection within a communication dead zone. Let be the V2V communication rate between CVj and RVi at time t. According to Shannon's formula, the transmission rate of a communication link (including V2I and V2V) can be uniformly expressed as: For both V2I and V2V communication modes, the values ​​of each parameter in formula (13) are defined as follows: In V2I communication, B is the V2I link bandwidth and P is the RSU transmit power; in V2V communication, B is the V2V link bandwidth and P is the vehicle transmit power. Furthermore, h is the channel gain and N0 is the Gaussian white noise power. Because the data transmission rate between vehicles changes with distance, in order to accurately calculate the maximum amount of data that can be transmitted by a vehicle in the communication blind zone, it is necessary to analyze the dynamically changing... Integrate the points.

402. Based on the proposed collaborative caching scheme, the amount of data acquired by RVi in a download cycle mainly consists of two parts: the data downloaded by RVi from RSU and the data obtained by RVi through CV in the communication dead zone. Therefore, the amount of data downloaded by RVi in a download cycle is expressed by formula (14): Among them, D V2I,i The amount of data downloaded directly by the vehicle from the current RSU range is calculated using formula (15): in, This is the time it takes for RVi to download data from RSU. In this article, Equal to the time required for the vehicle to leave the current RSU This time is calculated using formula (3). V2I This refers to the data transfer rate from the RSU to the RV. Because the data transfer rate between the vehicle and the RSU varies with distance, accurate calculation requires addressing the dynamically changing RSU. V2I Integrating, as shown in formula (16): D V2V,i The amount of data that RVi acquires via CV in the communication dead zone is calculated using formula (17): Where D i,j CVj represents the size of the data cached by RVi. RVi's total migration time T within a download cycle i It includes two parts: the time elapsed from when RVi issues a download request to when the user leaves the current RSU coverage area. and the time the vehicle travels in the communication blind spot Right now:

403. Furthermore, in order to fully utilize the vehicle's idle cache resources, this paper models the CV cache resource allocation problem as an optimization problem aimed at maximizing the system's average throughput. Based on formulas (17) and (18), the throughput of RVi is expressed as η. i =D i / T i Therefore, the system's average throughput can be expressed as: The optimization variable is the cooperation matrix C = [c ij ] N×M and cache allocation matrix U = [u ij ] N×M The optimization problem can be formulated as follows:

5. A collaborative caching scheme for large file transfer in a vehicle network as described in claim 4, characterized in that, The constraints of the optimization model in step 403 include at least one of the following: The total amount of data cached by each cooperating vehicle for all requesting vehicles shall not exceed its own cache capacity; The amount of data that each cooperating vehicle caches for the requesting vehicle shall not exceed the maximum amount of data that can be transmitted between the two vehicles in the communication dead zone; The total amount of data cached by all collaborating vehicles for a single requesting vehicle does not exceed the amount of remaining data that the requesting vehicle has not yet downloaded.

6. A collaborative caching scheme for large file transfer in a vehicle network as described in claim 1, characterized in that, The improved particle swarm optimization algorithm in step S3 introduces a cyclic neighborhood sharing mechanism, which is as follows: To overcome the premature convergence of traditional PSO, this paper abandons the traditional approach of relying on the global optimum to guide particle updates. Instead, it uses the particle with the maximum fitness value in its neighborhood to guide its update direction. This strategy aims to prevent the particle swarm from converging to a local optimum prematurely. The particle velocity update formula is as follows: V k =ωV k +c1r1(pbest k -X k )+c2r2(nbest k -X k )+c3r3(gbest'-X k ) (21) Among them, nbest k It represents the optimal position in the neighborhood of each particle. gbest' is a variant of the global optimal solution, defined as follows: gbest'=gbest+e -ηg / G C(0,1) (22) Where η = 10, and C(0,1) are random numbers generated by the Cauchy distribution function.

7. A collaborative caching scheme for large file transfer in a vehicle network as described in claim 6, characterized in that, The improved particle swarm optimization algorithm also introduces an adaptive adjustment strategy for inertia weights based on simulated annealing, the strategy being: Drawing inspiration from Simulated Annealing (SA), this paper combines inertia weights with annealing temperature, making them dynamically change throughout the search process. The core idea of ​​SA is that the system is active at high temperatures, capable of accepting inferior solutions with a certain probability, thus escaping local optima; as the temperature gradually decreases, the system tends to stabilize, eventually converging to an approximate global optimum. In the proposed improvement strategy, the inertia weights are defined as a function that dynamically adjusts with temperature decay. Where, ω min ,ω max Let represent the maximum and minimum values ​​of the inertial weight, ΔE represent the current energy difference of the particle, λ represent the Boltzmann constant, and T decreases with the number of iterations according to the cooling coefficient r. This strategy, by introducing energy change and temperature parameters, allows the inertial weight to be dynamically adjusted according to the actual situation during the search process, thereby enhancing the global search capability and stability of the algorithm. Based on the above improved strategy, the particle velocity update formula is changed to: This strategy enables the algorithm to dynamically adjust inertia weights based on feedback from the search state. Compared to the traditional linear decreasing strategy, the simulated annealing inertia weight adjustment method proposed in this paper is more flexible and adaptable, and can automatically adjust the balance between exploration and development at different search stages. Especially in the complex problem of vehicular network cache optimization, this method can significantly reduce the probability of getting trapped in local optima and improve the ability to obtain the global optimum.

8. A collaborative caching scheme for large file transfer in a vehicle network as described in claim 7, characterized in that, Based on the iterative solution of the above algorithm, RSU can finally obtain the globally optimal cooperative caching decision matrix. This matrix precisely quantifies the proportion of data cached by each CV for the RV, forming a complete cooperative caching scheme. RSU then distributes data to the designated vehicles according to this scheme.

Images