Consider the edge node selection scheme for content placement

Abstract

This invention claims protection for an edge node selection scheme that considers content placement, belonging to the fields of mobile communication technology and computer network technology. It addresses the problem of effectively utilizing edge caching to shorten transmission paths in a vehicle-to-everything (V2X) edge caching environment while ensuring stable transmission. First, the network control center collects vehicle and network data at the edge node side, composed of base stations and edge servers. Second, based on the collected data, a caching strategy matrix is ​​calculated to determine the caching scheme and perform pre-caching. Third, a comprehensive edge node selection scheme is constructed by combining adaptability, received signal strength, and load rate. Next, the network control center selects the edge node with the highest score for the vehicle terminal. Finally, the vehicle terminal selects the edge node with the highest score for access.

Description

Technical Field

[0001] This invention belongs to the field of mobile communications, specifically to the edge node selection scheme in the edge cache of vehicle networking. Background Technology

[0002] With the rapid development of vehicle-to-everything (V2X) technology, passengers' demands for a superior riding experience are constantly increasing, prompting intelligent vehicles to continuously innovate and provide diversified services. To achieve efficient content delivery, edge caching technology has emerged. It stores service content on edge nodes closer to the vehicle and user, effectively shortening the content transmission path. It's worth noting that the latency for users to retrieve content depends not only on whether the edge node caches the requested content, but also directly on the edge node the vehicle chooses to connect to, such as the edge node's load rate and the signal strength the vehicle can receive. Therefore, when a vehicle selects an edge node to connect and retrieve content, in addition to considering whether the edge node caches the requested content, it should also consider whether the current network conditions of the edge node support the transmission. Combining these two factors not only reduces content retrieval latency but also ensures the stability of the transmission process.

[0003] Currently, significant progress has been made in the research of edge caching and node selection strategies for vehicle-to-everything (V2X) networks. However, existing solutions often focus on single-dimensional optimization and fail to integrate the two aspects. The authors of [Kazmi S MA, Dang TN, Yaqoob I, et al. Infotainment enabled smart cars: A joint communication, caching, and computation approach[J]. IEEE Transactions on Vehicular Technology, 2019, 68(9): 8408-8420.] jointly optimize caching and computational resources to minimize latency for smart cars when accessing entertainment content. The authors of [Mahmood A, Caseti C, Chiasserini CF, et al. Mobility-aware edge caching for connected cars[C] / / 2016 12th Annual Conference on WirelessOn-demand Network Systems and Services(WONS). IEEE, 2016: 1-8.] deeply consider vehicle mobility and edge node service time, proposing a mobility-aware edge caching strategy to ensure that content is cached on the edge nodes most needed by connected vehicles. The aforementioned edge caching research only considered the placement of cached content and failed to fully account for the impact of network transmission conditions on content retrieval. Therefore, when network conditions are poor, it may lead to a longer content retrieval delay. In contrast, in the research on node selection schemes, the literature [Rahman MA, Salih QM, Asyhari AT, et al. Traveling distance estimation to mitigate unnecessary handoff in mobilewireless networks[J]. Annals of Telecommunications, 2019, 74: 717-726.] proposes selecting appropriate edge nodes based on the received signal strength.The literature [Guo X, Omar MH, Zaini KM, et al. Access-Selection Algorithm for Heterogeneous Wireless Networks Based on Uncertain Network Attribute Values[J]. Computational Intelligence and Neuroscience, 2022, 2022(1):4646889.] proposes to select edge nodes by using received signal strength, network load, and user rate as input decision parameters to obtain a score for each edge node. However, the above node selection scheme does not consider the impact of cached content placement on content transmission, which may lead to a longer transmission path and thus a longer content retrieval delay.

[0004] To address the aforementioned shortcomings, this invention proposes an edge node selection scheme that considers content placement. This achieves effective coordination between edge caching and node selection, ensuring stable network transmission conditions while allowing user vehicles to obtain cached content from edge nodes as much as possible. This shortens the content transmission path, reduces content retrieval latency, and improves user service quality. Summary of the Invention

[0005] This invention aims to solve the problems of the prior art mentioned above. It proposes an edge node selection scheme that considers cache placement. The technical solution of this invention is as follows:

[0006] An edge node selection scheme for vehicle-to-everything (V2X) edge caching that considers cache placement includes the following steps:

[0007] First, the network control center collects vehicle and network data at the edge nodes, which consist of base stations and edge servers. Second, based on the collected data, it calculates the caching strategy matrix, determines the caching scheme, and performs pre-caching. Third, it constructs a comprehensive edge node selection scheme by combining adaptability, received signal strength, and load rate. Next, the network control center selects the edge node with the highest score for the vehicle terminal. Finally, the vehicle terminal selects the edge node with the highest score for access.

[0008] The further cache strategy matrix is ​​M = (m r,f ) R×F ,in Represents the set of edge nodes. This represents a collection of content libraries. When m r,f =1 indicates that content f has been cached on edge node r, while 0 indicates that content f has not been cached on edge node r.

[0009] Further steps to solve the cache strategy matrix include:

[0010] Content popularity definition:

[0011]

[0012] pl f Denotes the probability that content f is requested, where O f This indicates the ranking of content f in descending order of request, where 0 < θ ≤ 1 is the skewness factor.

[0013] Cache hit rate definition:

[0014]

[0015] Furthermore, solving the cache strategy matrix can be transformed into a problem of maximizing the region cache hit rate, specifically including:

[0016]

[0017] Where ω f C represents the size of the content f. r Let r represent the cache capacity of edge node r, constraint (4) represents the integer property of cache decision, and constraint (5) represents that the content cached by each edge node cannot exceed its maximum cache capacity.

[0018] Furthermore, the present invention solves problem (3) by using reinforcement learning. In order to effectively describe and simplify the modeling process of reinforcement learning, the present invention models the reinforcement learning process as a Markov decision process.

[0019] Further Markov decision-making processes specifically include:

[0020] (1) State space S:

[0021] M R×F (t): Represents the cache state matrix at time t;

[0022] C r (t): Represents the remaining cache space size of edge node r at time t.

[0023] Then the state s(t)∈S of the entire system at time t is:

[0024] s(t)=(M R×F (t),C r (t)) (6)

[0025] (2) The action space A represents the set of actions that can be taken in each state, mainly whether to cache the content f on the edge node r. Therefore, the cache action space a(t)∈A at time t is:

[0026] a(t)={m r,f (t)|m r,f (t)∈{0,1}} (7)

[0027] (3) The reward function is used to evaluate the merits of taking a certain action, and the caching strategy is adjusted by maximizing the cumulative expected reward. Since the optimization problem is to maximize the cache hit rate, the reward function is defined as a metric directly related to the hit rate:

[0028] rd(s(t),a(t))=HR(s(t),a(t)) (8)

[0029] Many mature reinforcement learning algorithms already exist. Among them, reinforcement learning methods based on proximal policy optimization are more stable and convergent because they employ clipping techniques during policy updates, limiting the magnitude of policy updates and avoiding excessive policy changes. This makes them suitable for various tasks. Based on these advantages, this patent uses a reinforcement learning algorithm based on proximal policy optimization to solve the aforementioned optimization problem.

[0030] Further reinforcement learning algorithms based on proximal optimization strategies include:

[0031] Advantage function:

[0032] AF(s(t),a(t))=δ(t)+(γλ)δ(t)+...+(γλ) T-t+1 δ(T-1),γ∈(0,1),λ∈(0,1) (9)

[0033] The measure of the quality of performing an action relative to taking the average action in a given state is δ(t) = r(t) + γV. φ (s(t+1))-V φ (s(t+1)), s(t)∈S is used to represent the difference between the actual reward at time t plus the expected future reward and the estimated reward at the current time, V φ (s(t)) is the value estimate of the value network at time t.

[0034] Policy network θ:

[0035]

[0036] A certain action will be executed, and then a corresponding caching strategy will be adopted based on that action. Here, Γ represents the degree of change in cached items by the new caching strategy compared to the old caching strategy. When AF(s(t),a(t)) > 0, it indicates that the performance of choosing action a(t) in state s(t) is better than average. In this case, the policy network should be updated as much as possible in the direction that increases the execution probability of action a(t). When AF(s(t),a(t)) < 0, it indicates that the performance of choosing action a(t) in state s(t) is worse than average. In this case, the policy network should be updated as much as possible in the direction that decreases the execution probability of action a(t). To balance the trade-off between exploring new caching strategies and utilizing already discovered caching strategies, the update range of the caching strategy is limited to (1-, 1+) ​​using the clip function, where ∈ is the clipping factor.

[0037] Value Network φ:

[0038]

[0039] It is used to evaluate the quality of the current action taken by the policy network, so that the policy network can be updated in a direction with higher value.

[0040] Furthermore, when the policy network θ and the value network φ tend to stabilize, the cached state matrix M... R×F This will also be determined.

[0041] Further definitions of the adaptability, received signal strength, and load rate include:

[0042] Fit: This measures the extent to which an edge node caches content requested by vehicles. It is related to the caching strategy matrix. The higher the fit, the more content that the edge node may cache, meaning that the vehicle is more well-suited to the edge node, and vice versa.

[0043] Received signal strength: This is a basic indicator for vehicle terminals to evaluate edge nodes. It reflects the channel quality of the network. The stronger the received signal strength, the more stable the channel is.

[0044] Load ratio: This measures the ratio of the number of resource blocks in use to the total number of resource blocks within the coverage area of ​​an edge node. A lower load ratio indicates that more resources are available for allocation.

[0045] The further calculations of fit, received signal strength, and load rate specifically include:

[0046] Compatibility:

[0047] ad k,r =pl f m r,f (12)

[0048] pl represents the fitness of vehicle k relative to edge node r, where pl f m represents the probability of vehicle request content f. r,f This indicates whether the edge node r has cached the content f.

[0049] Received signal strength:

[0050]

[0051] P represents the received signal strength between vehicle k and edge node r, where P r Let η represent the wireless signal transmit power of edge node r, η represent the path loss factor, and d represent the wireless signal transmit power of edge node r. k,r X represents the distance from vehicle k to edge node r. σ This represents Gaussian white noise with a mean of 0 and a variance of σ. This refers to a collection of vehicles.

[0052] Load rate:

[0053]

[0054] U represents the ratio of the number of used resource blocks within the network coverage area formed by edge node r to the total number of resource blocks, where U r The total number of vehicles connected to edge node r, h k,r The number of resource blocks allocated to edge node r for the k-th vehicle, H r The total number of resource blocks that can be allocated to edge node r. In this paper, if the number of vehicles connected to the edge node is below the limit, each vehicle is allocated a fixed number of resource blocks; if the number exceeds the limit, all vehicles share the total network resources equally.

[0055] Furthermore, in fuzzy decision-making, to handle fuzzy parameter information, it is usually necessary to fuzzify these parameters. The membership function is the transformation function between quantitative parameters and qualitative concepts. In this paper, the membership function maps the input to a membership degree in the interval [0,1] using the Mamdani method, representing the degree to which the parameter belongs to this fuzzy set. The larger the membership value, the higher the degree to which the parameter belongs to this set. The main steps include the following three steps:

[0056] 1) Fuzzification: The three parameters of fit, received signal strength, and load rate are selected as the three input parameters of the fuzzy logic system.

[0057] 2) Fuzzy Inference: Fuzzy rules are used to associate fuzzy inputs and outputs. Each input and output parameter is defined with three fuzzy logic levels: low, medium, and high. Therefore, the resulting fuzzy rule base contains 27 rules, meaning there are 27 possible output parameters. To ensure the vehicle selects the most suitable edge node and to differentiate the importance of the three parameters, the inference rules are designed with the following principle: edge node load rate is the most important, followed by Received Signal Strength (RSS), and then the fitness parameter's importance decreases in that order. In other words, load rate has the most significant impact on the inference result, followed by RSS, while the fitness parameter's importance is relatively low. This design more effectively reflects the different degrees of influence of edge node load, communication quality, and fitness on the decision-making process.

[0058] 3) Defuzzification: The score of the edge node is the output of the fuzzy inference engine, which is defined by three linguistic variables. These linguistic variables are defuzzified using a membership function, normalizing the edge node score to a range of 0 to 1. The defuzzified combined output value is the score Y of edge node r. r Ultimately, the edge node Y with the highest score is selected. r Remove the connection, that is:

[0059] Y r =Max{Y1,Y2,...,Y} R} (15)

[0060] The advantages and beneficial effects of this invention are as follows:

[0061] 1. In the research on edge caching, the impact of network transmission conditions on content transmission was fully considered, effectively reducing the content transmission latency caused by weak signal strength and high load rate.

[0062] 2. By introducing an adaptation parameter, effective coordination between edge caching and node selection is achieved, improving content transmission efficiency. Attached Figure Description

[0063] Figure 1 This is a network scene diagram;

[0064] Figure 2 For the fitness degree membership function graph;

[0065] Figure 3 Load factor membership function graph;

[0066] Figure 4 A membership function graph of received signal strength;

[0067] Figure 5 The graph shows the membership function for the rating.

[0068] Figure 6 A graph showing the relationship between average latency and the number of vehicles;

[0069] Figure 7 A graph showing the relationship between total latency and the number of vehicles;

[0070] Figure 8 A graph showing the relationship between congestion rate and number of vehicles;

[0071] Figure 9 A graph showing the relationship between cache hit rate and training rounds;

[0072] Figure 10 A graph showing the relationship between cache hit rate and cache capacity;

[0073] Figure 11 A graph showing the relationship between cache hit rate and the number of contents;

[0074] Figure 12 A summary diagram of the edge node selection scheme; Detailed Implementation

[0075] 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.

[0076] The technical solution of the present invention is as follows:

[0077] In the edge caching of vehicle networking, from the perspective of ensuring network transmission quality and shortening content transmission path, an edge node selection scheme that considers content placement is designed. This allows the edge node selected by the vehicle user to cache the requested content as much as possible, reducing content retrieval latency. At the same time, the impact of network conditions on content transmission is considered to avoid unstable network transmission environments caused by weak received signal strength and high edge node load.

[0078] The edge node selection scheme for content placement proposed in this invention includes the following steps:

[0079] Step 1: Based on the historical request data collected from the edge nodes, the network control center uses reinforcement learning to calculate the caching strategy matrix and pre-caches the corresponding content on the edge nodes according to the caching strategy matrix.

[0080] Step 2: The network control center collects terminal and network data, such as the location information of vehicle terminals and the number of vehicles connected to the base station. Then, based on the collected data and the solved caching strategy matrix, it calculates the vehicle's received signal strength, adaptability, and edge node load rate. Subsequently, using the received signal strength, adaptability, and load rate as inputs, a fuzzy inference system is constructed to obtain the edge node's score. Finally, the vehicle terminal selects the edge node with the highest score for access.

[0081] To evaluate the performance of the proposed solution, a simulation platform was built using Python 3.9, Pytroch 2.1, and scikit-fuzzy 0.4.2 to simulate a two-way four-lane urban core area at a 2km long intersection. It is assumed that five 5G base stations are deployed, each equipped with an edge server, which together provide content services to vehicles. In most edge caching studies, it is usually assumed that vehicles are located within the coverage area of ​​a single edge node and are connected to that node by default. In this case, the vehicle selects the edge node mainly based on the received signal strength, that is, the node selection problem is not considered. Therefore, in order to verify the effectiveness of the proposed solution, the main comparative solutions in this paper are: (1) caching is performed in advance directly through the Least Recently Used (LRU) scheme, and the vehicle selects the edge node only based on the received signal strength, which is denoted as the LR scheme. (2) The algorithm for solving the caching strategy matrix is ​​a Q-learning-based learning algorithm, and the vehicle selects the edge node only based on the received signal strength, which is denoted as the QR scheme. (3) The algorithm for solving the caching strategy matrix is ​​a learning algorithm based on PPO. The vehicle selects edge nodes only based on the received signal strength, which is denoted as the PR scheme.

[0082] Figure 7This reflects the relationship between the number of vehicles and the average latency. It can be seen that when the number of vehicles is 0-60, the system latency is mainly determined by whether the edge nodes connected to the vehicles cache the content requested by the vehicles. Since the edge node selection scheme proposed in this invention comprehensively considers the adaptability parameter, under good network conditions, vehicles can preferentially select edge nodes with higher adaptability for connection, and have a higher probability of directly retrieving the requested content from the cache without having to access a remote data center, thus significantly reducing content retrieval latency. For the other three schemes (QR, PR, LR), since the adaptability parameter is not considered, whether a vehicle can directly retrieve cached content depends on the caching strategy. Compared with the Q-learning algorithm, the PPO learning algorithm uses near-end strategy optimization during the solution process, limiting the policy update range, allowing the PPO algorithm to more flexibly adjust cached content and better meet the dynamic needs of vehicle users. Therefore, when the number of vehicles is small and network conditions are good, the latency of retrieving content by connecting to edge nodes through the PR scheme is lower than that of the QR scheme. Conversely, compared to the aforementioned learning algorithms, the LRU caching scheme cannot promptly capture dynamic changes in user requests and content access frequency, resulting in high latency even under good network conditions. As the number of vehicles gradually increases, the total latency of all four schemes shows an upward trend. However, differences in load management among the different schemes lead to significantly different latency growth characteristics. For the QR, PR, and LR schemes, because the load status of each node is not fully considered when selecting edge nodes, vehicles tend to concentrate their connections to certain edge nodes in high-density areas, causing a sharp decline in the resource allocation efficiency of these nodes, ultimately leading to network congestion and increased latency. This load imbalance reduces the service capacity of edge nodes, affecting overall network performance. In contrast, the proposed scheme comprehensively evaluates the load level of each edge node and rationally allocates the connection needs of vehicles, avoiding load imbalance between nodes and thus significantly mitigating network congestion. Specifically, by dynamically adjusting the connection strategy of edge nodes, the proposed scheme can effectively distribute vehicle connections, ensuring that the load of each edge node fluctuates within a reasonable range and avoiding latency bottlenecks caused by excessive resource utilization. Therefore, in scenarios with a large number of vehicles, the proposed solution consistently exhibits significantly lower latency than the other three algorithms, demonstrating better scalability and stability.

[0083] Figure 8 This demonstrates the relationship between total latency and the number of vehicles. Figure 7Similarly, when the number of vehicles is between 0 and 60, the proposed solution generates a lower total latency than the other three solutions. As the number of vehicles increases, the total latency of all four solutions shows an upward trend. Because the proposed solution considers the load factor, it avoids network congestion, thus resulting in a lower total latency than the other three solutions. Figure 7 and Figure 8 It can be seen that relying solely on the caching scheme while ignoring the network conditions of edge nodes may lead to significant latency when vehicles retrieve content as the number of vehicles increases. Therefore, considering network conditions is crucial for optimizing system performance.

[0084] Figure 9 Reflecting the relationship between the number of vehicles and the congestion rate, when the number of vehicles is less than 40, the congestion rate of all four schemes remains at 0% because it does not exceed the maximum number of vehicles that a single edge node can serve. However, when the number of vehicles exceeds 40, the congestion rate of the other three schemes gradually increases with the increase in vehicle density. This is mainly because vehicles rely solely on Received Signal Strength (RSS) as a criterion when selecting edge nodes, without fully considering the actual load of the edge nodes. This leads to a situation where, when a large number of vehicles are concentrated in a certain area, they tend to choose the same edge node with the strongest RSS for connection, thus exceeding the service capacity limit of that node and causing the congestion rate to rise. In contrast, the scheme proposed in this paper introduces the edge node load rate as one of the selection criteria, which can select edge nodes with lower load rates when choosing a connection target each time. This strategy effectively distributes the request pressure of vehicles on edge nodes and avoids the congestion problem caused by the overload of a single node. Therefore, even when the number of vehicles increases, the congestion rate of the proposed algorithm remains at a lower level compared to the other three algorithms.

[0085] Figure 10 The study reflects the relationship between cache hit rate and training epochs. In the first 200 training epochs, due to ample cache space, edge nodes can continuously cache frequently requested video content from vehicle users, leading to a gradual increase in cache hit rate. However, around the 200th epoch, the cache space becomes saturated, unable to cache any more content, causing the cache hit rate to stabilize. Since the edge node selection scheme proposed in this paper introduces an adaptability parameter, under good network conditions, the key factor affecting cache hit rate is whether the edge node connected to the vehicle caches the requested content. By considering the adaptability parameter, the proposed scheme increases the probability that the vehicle will connect to the edge node that caches its requested content. Therefore, compared to the other three schemes, the proposed scheme has a significant advantage in cache hit rate.

[0086] Figure 11The cache hit rate reflects the trend of cache space size. As the cache space increases, the edge nodes can cache more video content requested by users. Therefore, the cache hit rate of all four schemes increases with the increase of cache space. Figure 12 The cache hit rate reflects the trend of cache hit rate changing with the size of the content library. As the amount of content increases, the number of user-requested contents at edge nodes decreases, leading to a decrease in the cache hit rate of the four schemes as the amount of content increases. Because the scheme in this paper considers the adaptation parameter, its cache hit rate is higher than the other three algorithms.

[0087] 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.

[0088] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, 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 herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0089] 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.

[0090] 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 method for selecting edge nodes considering content placement, characterized in that, Includes the following steps:

101. At the edge node side, composed of base stations and edge servers, vehicle request data and network data are collected through the network control center. Based on the collected vehicle request data, a problem of maximizing the regional cache hit rate is constructed. A caching strategy matrix is ​​obtained by solving the problem using reinforcement learning methods to maximize the regional cache hit rate. According to the caching strategy matrix, pre-caching is performed on the edge nodes in advance; the caching strategy matrix is ​​as follows: ,in Represents the set of edge nodes. Represents a collection of content libraries; when Time indicates content Already cached on edge nodes If it is 0, it indicates that the content is... No cache on edge nodes The cache hit rate is defined as follows: ； Solving the caching strategy matrix is ​​transformed into maximizing the region cache hit rate, specifically including: ； ； ； 102. Based on the caching strategy matrix obtained in step 101 and the collected network data, the network control center calculates the vehicle-edge node fit, received signal strength, and load rate; the formula for calculating the fit is as follows: ； ； in, Indicates vehicle Relative to edge nodes Adaptability Content popularity indicates the content's popularity. The probability of being requested, where Content Ranked in descending order of requests Skewness factor; Represents edge nodes Is the content cached? The caching strategy matrix described in 101 is used to obtain the content requested by vehicles. The greater the adaptability, the more vehicle-requested content is cached by the edge nodes. With edge nodes The better the fit; 103. The fitness, load rate and received signal strength obtained in step 102 are used as inputs to the fuzzy logic system. After inference and defuzzification by the fuzzy logic system, the output results are used as the scores of each edge node. Based on the above fuzzy logic system, a comprehensive edge node selection method is constructed. The fuzzy logic system processing flow mainly includes the following three steps: 1) Fuzzification: The three parameters of fit, received signal strength, and load rate are selected as the three input parameters of the fuzzy logic system; 2) Fuzzy reasoning: Fuzzy inputs and outputs are associated through a fuzzy rule base. Each input and output parameter is defined with three fuzzy logic levels: low, medium, and high. The resulting fuzzy rule base contains 27 rules, and there are 27 possible output parameters. The edge node load rate is the most important factor, followed by the received signal strength and the adaptability. 3) Defuzzification: The score of the edge node is the output of the fuzzy inference engine, which is defined by three linguistic variables. These linguistic variables are defuzzified using a membership function, normalizing the edge node score to between 0 and 1. The defuzzified composite output value is the edge node score. Comments ; 104. Based on the edge node scores obtained in step 103, the vehicle terminal selects the edge node with the highest score for access; that is... 。 2. The edge node selection method considering content placement as described in claim 1, characterized in that, Solving the problem of maximizing the region cache hit rate employs reinforcement learning. The reinforcement learning process is modeled as a Markov process, and the Markov decision process specifically includes: (1) State space : : indicates in The cache state matrix at each time step; : indicates in Time edge node Remaining cache space size; Then in The state of the entire system at all times for: ； (2) Action space This represents the set of possible actions that can be taken in each state. Moment cache action space for: ； (3) The value of the reward function is used to evaluate the merits of taking a certain action. The strategy is adjusted by maximizing the cumulative expected reward. The reward function is defined as an indicator directly related to the hit rate: ； Reinforcement learning algorithms based on proximal optimization strategies specifically include: Advantage function: ； The measure of the quality of performing an action relative to taking the average action in a given state is denoted as follows: For discount factor, For hyperparameters, Used to indicate in The difference between the actual reward at time 1 plus the expected future reward and the estimated reward at the current time 2. It is the value network in Value estimation at this moment; Policy Network : ； It will execute a certain action, and then adopt a corresponding caching strategy based on that action. This indicates the degree of change in cached items by the new caching strategy compared to the old caching strategy. "Time" indicates the state. Select action Its performance is better than average, when "Time" indicates the state. Select action The performance is below average; the update range of the caching strategy is limited by using the clip function. Among them For pruning factors; value network : ； It is used to evaluate the quality of the policy network's current action, when the policy network... and value network When it tends to stabilize, the cache state matrix Sure.

3. The edge node selection method considering content placement as described in claim 1, characterized in that, The specific calculation formulas for the received signal strength and load rate are as follows; Received signal strength: ; Indicates vehicle Relative to edge nodes The received signal strength between, where Represents edge nodes The wireless signal transmission power, Indicates the path loss factor. Indicates vehicle To edge nodes distance, This indicates that the mean is 0 and the variance is 0. Gaussian white noise, in which Represents a set of vehicles; Load rate: ; Represents edge nodes The ratio of the number of used resource blocks to the total number of resource blocks within the network coverage area, of which To access the edge node Total number of vehicles For edge nodes For the first The number of resource blocks allocated to a vehicle For edge nodes Total number of resource blocks that can be allocated; If the number of vehicles connected to the edge node is below the rated number, each vehicle is allocated a fixed number of resource blocks. If the number exceeds the rated number, each vehicle shares the total network resources equally.

Images