Collaborative caching method for Internet of Vehicles based on social relationship perception of mobile vehicles

By predicting the vehicle movement trajectory and content popularity, selecting appropriate cache vehicles, and building a collaborative cache strategy, the problem of poor cache performance in the Internet of Vehicles is solved, and higher cache hit rate and lower content acquisition latency is achieved.

CN115314944BActive Publication Date: 2024-12-06HENAN UNIV OF SCI & TECH +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210955357.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-08-10
Publication Date
2024-12-06
Estimated Expiration
2042-08-10

AI Technical Summary

Technical Problem

In the existing Internet of Vehicles technology, the performance of mobile edge cache is affected by factors such as limited space of edge equipment, random vehicle movement, time-varying of requested content, and complexity of social relations, resulting in poor cache performance and inability to effectively improve cache hit rate and Reduce content acquisition delay.

Method used

Through a collaborative caching method based on social relationship perception of mobile vehicles, predict vehicle movement trajectory and content popularity, select appropriate cache vehicles, build collaborative caching strategies, and optimize content acquisition methods to improve cache hit rate and reduce latency.

Benefits of technology

Improves cache hit rate, reduces content acquisition latency, improves user experience quality, and shows better performance under different vehicle counts and cache capacity conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115314944B_ABST
    Figure CN115314944B_ABST
Patent Text Reader

Abstract

The present invention discloses a collaborative caching method for an Internet of Vehicles based on social relationship perception of mobile vehicles, predicts the movement trajectory of each vehicle in the Internet of Vehicles, determines the roadside unit to which the vehicle belongs at the next moment, obtains the set of vehicles within the coverage of each roadside unit, calculates the request probability of vehicles within the coverage of the roadside unit for each content, predicts the set of content requested by each vehicle at the next moment, calculates the social relationship strength of each vehicle and screens out cached vehicles, calculates the delay of each vehicle in obtaining content from cached vehicles and roadside units, and finally constructs a collaborative caching optimization problem and solves it to obtain a collaborative caching scheme under each roadside unit. The present invention predicts the roadside unit and vehicle request content at the next moment, selects a cache vehicle according to the social relationship of the vehicle, and thus formulates a correct collaborative caching strategy, thereby improving the cache hit rate and reducing the content acquisition delay.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of Internet of Vehicles, and more specifically, relates to an Internet of Vehicles collaborative caching method based on social relationship perception of mobile vehicles. Background Art

[0002] With the development of vehicle networks, in-vehicle applications such as in-vehicle intelligent terminals, autonomous driving, vehicle entertainment and safety continue to emerge, which has brought positive impacts on the development of vehicle roads, not only improving traffic efficiency, but also enhancing the user's driving experience. However, the emergence of these intelligent applications may require a lot of computing, storage and communication resources, causing serious traffic load on the core network. Mobile edge caching technology pre-caches popular content in roadside units and some vehicles. When the requesting user requests these contents, these contents are obtained through wireless transmission without going through the remote core network, which reduces the delay of content acquisition and improves the content cache hit rate, thereby improving the user's experience quality. However, considering the limited space of edge devices, the randomness of vehicle movement, the time-varying nature of requested content and the complexity of social relationships, the cache performance is poor. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a collaborative caching method for the Internet of Vehicles based on social relationship perception of mobile vehicles. The method predicts the roadside unit to which the vehicle belongs at the next moment based on the vehicle's movement trajectory, predicts the requested content based on content popularity and vehicle user interests, selects cache vehicles based on vehicle social relationships, and thus formulates a correct collaborative caching strategy, thereby improving the cache hit rate and reducing content acquisition delays.

[0004] In order to achieve the above-mentioned object of the invention, the Internet of Vehicles collaborative caching method based on social relationship perception of mobile vehicles of the present invention comprises the following steps:

[0005] S1: The set of roadside units in the Internet of Vehicles RSU = {S1,…,S n ,…,S N}, where S n represents the nth roadside unit, n=1,2,…,N, N represents the number of roadside units;

[0006] The set of content that can be requested in the Internet of Vehicles is X = {x1,…,x q ,…,x Q}, x q represents the qth content, q = 1, 2, ..., Q, Q represents the number of contents that can be requested in the Internet of Vehicles, and the content x q The size is d q ;

[0007] S2: For each vehicle in the IoV, the moving trajectory of the vehicle is predicted to obtain the predicted position at the next moment, thereby determining the roadside unit to which the vehicle belongs at the next moment; the S of each roadside unit at the next moment is obtained. n The set of vehicles within the coverage area V n = {v n,1 ,…,v n,m ,…,v n,M},v n,m Represents the roadside unit S n The mth vehicle covered, m = 1, 2, ..., M n , M n Represents the roadside unit S n The number of vehicles within the coverage area;

[0008] S3: First, count the number of requests for each content in the preset historical time period, sort the content from large to small according to the number of requests, and record the content x q The sequence number after sorting is i q ; Calculate the request probability of each content in the preset historical time period, take the content with request probability exceeding the preset threshold as popular content, and sum the request probabilities of popular content as the distribution slope ω of popular content; Use the following formula to calculate the content popularity pop of the qth content at the next moment q :

[0009]

[0010] For roadside units S n The set of vehicles within the coverage area V n , get each vehicle v n,m The content set L requested within the preset historical time period n,m Then the following calculation formula is used to calculate the interest similarity sim between vehicles m,m′ :

[0011]

[0012] Among them, φ m,m′ Indicates vehicle v n,m and vehicle v n,m′ Content Collection L n,m , L n,m′ The intersection of m′=1,2,…,M n ,|| means to find the number of contents in the set, M q Indicates that content x has been requested within the preset historical time period q The number of vehicles;

[0013] The vehicle v is calculated using the following formula n,m SIMm :

[0014]

[0015] Similarity of interest m Normalize to get the normalized interest similarity SIM′ m :

[0016]

[0017] Among them, SIM max , SIM min Respectively represent the vehicle set V n SIM of interest similarity of all vehicles in m The maximum and minimum values ​​of ;

[0018] The vehicle v is calculated using the following formula n,m Probability of interest in each content n,m :

[0019]

[0020] Content popularity pop q and vehicle v n,m Probability of interest in n,m,q Perform weighted summation to obtain the vehicle v at the next moment n,m For content x q The request probability p n,m,q :

[0021] p n,m,q =α×pop q +β×in n,m

[0022] Among them, α and β represent the preset content popularity weight value and vehicle personal interest preference weight value respectively;

[0023] Set the vehicle request content to follow the Poisson distribution, according to the vehicle v n,m For each content x q The request probability p n,m,q Predict the vehicle v n,m The content set requested at the next moment

[0024] S4: For roadside unit S n The vehicle set V n Each vehicle v in n,m , the number of vehicles that can be associated per unit time during driving is counted as the communication contact rate f n,m , integrating the interest similarity between vehicles m,m′, the social relationship strength z between vehicles is calculated pairwise using the following formula: m,m′ :

[0025] z m,m′ =f n,m ×sim m,m′

[0026] The vehicle v is calculated using the following formula n,m The strength of social relationship Z m :

[0027]

[0028] The roadside unit S n All vehicles under the cache are sorted from large to small according to the strength of social relationship, and the first K vehicles are selected as cache vehicles. The value of K is determined according to actual needs, and the cache vehicle set CV is obtained. n ={cv n,1 ,…,cv n,k ,…,cv n,K}, cv n,k Represents the roadside unit S n The kth cached vehicle under, k = 1, 2, ..., K;

[0029] S5: Calculate the latency of different content acquisition methods. The specific method is as follows:

[0030] Vehicle n,m From cache vehicle cv via V2V communication n,k Get content x q The delay T n,m,k,q for:

[0031]

[0032] Among them, rate n,m,k Indicates vehicle v n,m and cache vehicle cv n,k Communication transmission rate between

[0033] Vehicle n,m From the roadside unit S via V2R communication n Get content x q The delay T n,m,K+1,q for:

[0034]

[0035] Among them, rate n,m,K+1 Indicates vehicle v n,m and roadside units S n Communication transmission rate between

[0036] S6: Solve the following collaborative cache optimization problem to obtain each roadside unit S n The following collaborative caching scheme:

[0037]

[0038] C1:a q,k′ ∈{0,1},k′=1,2,…,K+1

[0039] C2:

[0040] C3:

[0041] C4:

[0042] Among them, A=(a q,1 ,…,a q,k ,…,a q,K+1 ) represents the cache decision vector, a q,k′ is the cache decision identifier, k′=1,2,…,K+1, a q,k =1 means that the content x q Cache to cache vehicle cv n,k , a q,k =0 means that the content x is not q Cache to cache vehicle cv n,k , a q,K+1 =1 means that the content x q Cache to roadside unit S n , a q,K+1 =0 means that the content x is not q Cache to roadside unit S n , Indicates cached vehicle cv n,k The cache capacity, Represents the roadside unit S n The cache capacity.

[0043] The present invention is based on a collaborative caching method for an Internet of Vehicles based on social relationship perception of mobile vehicles. The method predicts the movement trajectory of each vehicle in the Internet of Vehicles, determines the roadside unit to which the vehicle belongs at the next moment, obtains a set of vehicles within the coverage of each roadside unit, calculates the request probability of vehicles within the coverage of the roadside unit for each content, predicts the set of content requested by each vehicle at the next moment, calculates the social relationship strength of each vehicle and screens out cached vehicles, calculates the delay of each vehicle in obtaining content from the cached vehicles and roadside units, and finally constructs a collaborative caching optimization problem and solves it to obtain a collaborative caching solution under each roadside unit.

[0044] The present invention predicts the roadside unit to which the vehicle belongs at the next moment according to its moving trajectory, predicts the requested content based on content popularity and vehicle user interest, selects cache vehicles according to the social relationship of the vehicles, and thus formulates a correct collaborative caching strategy, thereby improving the cache hit rate and reducing content acquisition delay. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a flowchart of a specific implementation method of the Internet of Vehicles collaborative caching method based on social relationship perception of mobile vehicles of the present invention;

[0046] Figure 2 This is a diagram showing the effect of moving trajectory prediction based on the LSTM network in this embodiment;

[0047] Figure 3 It is a flowchart of reinforcement learning;

[0048] Figure 4 is a comparison chart of the hit rates of the present invention and the comparative method under different numbers of vehicles in this embodiment;

[0049] Figure 5 is a comparison diagram of the time delay of the present invention and the comparative method under different vehicle data in this embodiment;

[0050] Figure 6 is a comparison chart of cache hit rates of the present invention and the comparative method under different total cache capacities in this embodiment;

[0051] Figure 7 3 is a comparison chart of the delays of the present invention and the comparative method under different total cache capacities in this embodiment. DETAILED DESCRIPTION

[0052] The specific implementation of the present invention is described below in conjunction with the accompanying drawings so that those skilled in the art can better understand the present invention. It should be noted that in the following description, when the detailed description of known functions and designs may dilute the main content of the present invention, these descriptions will be omitted here.

[0053] Example

[0054] Figure 1 FIG. 1 is a flowchart of a specific implementation of the Internet of Vehicles collaborative caching method based on social relationship perception of mobile vehicles of the present invention. Figure 1 As shown, the specific steps of the vehicle networking collaborative caching method based on mobile vehicle social relationship perception of the present invention include:

[0055] S101: Obtaining vehicle networking parameters:

[0056] Let the set of roadside units in the vehicle network be RSU = {S1,…,S n ,…,SN}, where S n Indicates the nth roadside unit, n = 1, 2, ..., N, N represents the number of roadside units. Generally speaking, roadside units should be distributed as evenly as possible on both sides of the road, and the coverage of different roadside units should not overlap, so as to provide better services for vehicles.

[0057] The set of content that can be requested in the Internet of Vehicles is X = {x1,…,x q ,…,x Q}, x q represents the qth content, q = 1, 2, ..., Q, Q represents the number of contents that can be requested in the Internet of Vehicles, and the content x q The size is d q .

[0058] S102: Vehicle movement trajectory prediction:

[0059] For each vehicle in the IoV, the vehicle's moving trajectory is predicted to obtain the predicted position at the next moment, thereby determining the roadside unit to which the vehicle belongs at the next moment. n The set of vehicles within the coverage area V n = {v n,1 ,…,v n,m ,…,v n,M},v n,m Represents the roadside unit S n The mth vehicle covered, m = 1, 2, ..., M n , M n Represents the roadside unit S n The number of vehicles within the coverage area.

[0060] The specific method of mobile trajectory prediction can be selected according to actual needs. In the present invention, mobile trajectory prediction is performed based on LSTM (Long Short-Term Memory) network, and the specific method is:

[0061] The historical position vector of the vehicle is obtained by sampling the historical movement trajectory of the vehicle, and is input into the pre-trained LSTM network. The probability of the vehicle being located at each accessible position is output, and the accessible position corresponding to the maximum probability is taken as the predicted position of the vehicle.

[0062] The LSTM network is an improvement on the recurrent neural network. It is often used to process time series data and can effectively memorize relevant information about the vehicle's historical trajectory. Figure 2 This is a diagram showing the effect of mobile trajectory prediction based on the LSTM network in this embodiment. Figure 2 As shown in the figure, the predicted position based on the LSTM network is almost consistent with the actual position, which shows that LSTM is very accurate in predicting vehicle trajectory.

[0063] S103: Predict vehicle request content:

[0064] In order to make better decisions on content caching, it is necessary to first determine the probability of the vehicle requesting each content at the next moment, and predict the vehicle's request content based on the request probability, so as to determine which content needs to be cached. The present invention combines the popularity of the content and the personal interests of the vehicle user to determine the probability of the vehicle requesting content.

[0065] As for the content, count the number of requests for each content in the preset historical time period, sort the content from large to small according to the number of requests, and record the content x q The sequence number after sorting is i q . Calculate the request probability of each content in the preset historical time period (i.e., the ratio of the number of content requests to the total number of requests), take the content whose request probability exceeds the preset threshold as popular content, and sum the request probabilities of popular content as the distribution slope ω of popular content. Finally, the following formula is used to calculate the probability that the qth content will be requested by all vehicles at the next moment, i.e., the content popularity pop q :

[0066]

[0067] The content demand of vehicles is not only affected by the content itself, but also closely related to the personal preferences of vehicle users. Vehicle preferences can be expressed as the probability of a single vehicle requesting content at the current moment, so the probability of future content requests can be predicted based on the personal preferences of vehicle content request history information. n The set of vehicles within the coverage area V n , get each vehicle v n,m The content set L requested within the preset historical time period n,m Then the following calculation formula is used to calculate the interest similarity sim between vehicles m,m′ :

[0068]

[0069] Among them, φ m,m′ Indicates vehicle v n,m and vehicle v n,m′ Content Collection L n,m , L n,m′ The intersection of the two vehicles, that is, the set of content requested by both vehicles, m′=1,2,…,M n ,| | indicates the number of contents in the set, M q Indicates that content x has been requested within the preset historical time period q The number of vehicles, Used to represent content xq The weight of .

[0070] The vehicle v is calculated using the following formula n,m SIM m :

[0071]

[0072] Similarity of interest m Normalize to get the normalized interest similarity SIM′ m :

[0073]

[0074] Among them, SIM max , SIM min Respectively represent the vehicle set V n The maximum and minimum values ​​of interest similarity of all vehicles in .

[0075] The vehicle v is calculated using the following formula n,m Probability of interest in each content n,m :

[0076]

[0077] Content popularity pop q and vehicle v n,m Probability of interest in n,m,q Perform weighted summation to obtain the vehicle v at the next moment n,m For content x q The request probability p n,m,q :

[0078] p n,m,q =α×pop q +β×in n,m

[0079] Among them, α and β represent the preset content popularity weight value and vehicle personal interest preference weight value respectively;

[0080] Set the content of vehicle request to follow Poisson distribution, according to the vehicle v n,m For each content x q The request probability p n,m,q Predict the vehicle v n,m The content set requested at the next moment

[0081] S104: Select cached vehicles based on social relationships:

[0082] Roadside unit S n The vehicle set Vn , select the vehicles that can be cached based on the social relationships between vehicles. When the needs between vehicles are consistent and a communication link for data transmission can be established, the vehicles are said to be socially related. Therefore, social relationships have two characteristics: one is the similarity of interests between vehicles; the other is the contact rate between vehicles. For roadside units S n The vehicle set V n Each vehicle v in n,m , the number of vehicles that can be associated per unit time during driving is counted as the communication contact rate f n,m , integrating the interest similarity between vehicles m,m′ , the social relationship strength z between vehicles is calculated pairwise using the following formula: m,m′ :

[0083] z m,m′ =f n,m ×sim m,m′

[0084] The vehicle v is calculated using the following formula n,m The strength of social relationship Z m :

[0085]

[0086] The roadside unit S n All vehicles under the cache are sorted from large to small according to the strength of social relationship, and the first K vehicles are selected as cache vehicles. The value of K is determined according to actual needs, and the cache vehicle set CV is obtained. n ={cv n,1 ,…,cv n,k ,…,cv n,K}, cv n,k Represents the roadside unit S n The kth cached vehicle under, k = 1, 2, …, K.

[0087] S105: Calculate content acquisition delay:

[0088] Content acquisition latency is an important performance indicator in the cache system, which can reflect the communication quality between the requesting user and the cached content. When the latency of acquiring content is large, the content requested by the vehicle needs to be obtained from the remote content provider through the backhaul link, and cannot be obtained from the cache vehicle or roadside unit through wireless transmission, resulting in high-latency transmission. Therefore, it is necessary to calculate the latency of different content acquisition methods to provide a basis for subsequent cache decisions. The specific method is as follows:

[0089] Vehicle n,m From cache vehicle cv via V2V communication n,k Get content x q The delay Tn,m,k,q for:

[0090]

[0091] Among them, rate n,m,k Indicates vehicle v n,m and cache vehicle cv n,k The communication transmission rate between can be calculated using the following formula:

[0092]

[0093] Where B represents the transmission bandwidth, σ 2 represents additive white Gaussian noise, P n,k Indicates cached vehicle cv n,k The transmission power, h n,m,k Indicates vehicle v n,m and cache vehicle cv n,k The distance between n,m″,k Indicates vehicle v n,m″ and cache vehicle cv n,k The distance between m″≠m P n,k h n,m″,k Indicates other vehicles v n,m″ For vehicles n,m and cache vehicle cv n,k The interference between the communication, m″=1,2,…,M n And m″≠m.

[0094] Vehicle n,m From the roadside unit S via V2R communication n Get content x q The delay T n,m,K+1,q for:

[0095]

[0096] Among them, rate n,m,K+1 Indicates vehicle v n,m and roadside units S n The communication transmission rate between can be calculated using the following formula:

[0097]

[0098] Among them, P n Represents the roadside unit S n The transmission power, h n,m,K+1 Indicates vehicle v n,m and roadside units S n Since each roadside unit uses a different frequency band, the interference from other roadside units can be ignored.

[0099] S106: Collaborative caching decision:

[0100] The present invention aims to minimize content acquisition delay when making collaborative caching decisions, that is, to allow the requesting user to obtain the requested content from the network edge device as much as possible, rather than the remote content provider, so as to reduce the load of the backhaul link and reduce the content acquisition delay. Based on the above analysis, the optimization problem of collaborative caching is proposed as follows:

[0101]

[0102] C1:a q,k′ ∈{0,1},k′=1,2,…,K+1

[0103] C2:

[0104] C3:

[0105] C4:

[0106] Among them, A=(aq,1,…,aq,k,…,aq, K+1 ) represents the cache decision vector, a q,k′ is the cache decision identifier, k′=1,2,…,K+1, a q,k =1 means that the content x q Cache to cache vehicle cv n,k , a q,k =0 means that the content x is not q Cache to cache vehicle cv n,k , a q,K+1 =1 means that the content x q Cache to roadside unit S n , a q,K+1 =0 means that the content x is not q Cache to roadside unit S n , Indicates cached vehicle cv n,k The cache capacity, Represents the roadside unit S n The cache capacity.

[0107] Constraints C1 and C2 are restrictions on the values ​​of cache decision identifiers. C1 is whether the cached vehicle and roadside unit can cache or not cache content, represented by 0 or 1. In order to reasonably utilize storage capacity and reduce cache redundancy, C2 is used to ensure that each content can only be cached in one cached vehicle within the coverage area of ​​the roadside unit. C3 and C4 ensure that the content cached in each cached vehicle and roadside unit does not exceed their storage capacity.

[0108] By solving the above optimization problem, we can obtain the cache decision vector and determine the collaborative cache solution.

[0109] In this embodiment, a reinforcement learning method is used to solve the optimization problem of collaborative caching, where the state s is set to Set action a to cache decision vector A = (a q,1 ,…,a q,k ,…,a q,K+1 ), the reward function R is the delay corresponding to the cache decision vector, that is,

[0110] Figure 3 is a flowchart of reinforcement learning. Figure 3 As shown in Figure 2, the specific process of reinforcement learning is as follows:

[0111] 1) Initialize the current network Q(s,a;θ) and the target network θ and Represent the parameters of the current network and the target network respectively.

[0112] 2) Set the state s to

[0113] 3) Select a random action a with probability, or use a greedy algorithm to select the action with the largest current function value and execute.

[0114] 4) Calculate the content acquisition reward R according to the cache decision vector represented by action a and obtain the next state s′.

[0115] 5) Store this experience (s, a, R, s′) in the experience replay unit and extract small batches of samples from it for training.

[0116] 6) Update the target Q value using the target network

[0117] 7) Update the loss function using stochastic gradient descent

[0118] 8) Update the next state s′←s

[0119] 9) Stop training until the network parameters θ converge.

[0120] For the principles and specific process of reinforcement learning, please refer to the literature "Zhu H, Cao Y, Wang W, et al. Deep Reinforcement Learning for Mobile Edge Caching: Review, New Features, and Open Issues[J]. IEEE Network, 2018, 32(6): 50-57.".

[0121] In order to better illustrate the technical effect of the present invention, a specific example is used to experimentally verify the present invention. In this experimental verification, two existing caching strategies are used as comparison methods, namely the DAC algorithm and the RC algorithm. The DAC algorithm only considers the social relationship of the vehicle and does not predict the vehicle trajectory. For details, please refer to the document "Zhuo X, Li Q, Cao G, et al. Social-Based Cooperative Caching in DTNs: A Contact Duration Aware Approach [C]. 2011 IEEE Eighth International Conference on Mobile Ad-Hoc and Sensor Systems, 2011: 92-101." The RC algorithm is a random caching algorithm, that is, the roadside unit and the vehicle use the same probability to randomly cache the content.

[0122] Figure 4 : is a comparison chart of the hit rates of the present invention and the comparative method under different numbers of vehicles in this embodiment. Figure 4 As shown in the figure, the DAC algorithm does not take into account the mobility of vehicles. At this time, the cache results obtained by the static network topology cannot adapt to the actual dynamic network topology, resulting in a decrease in cache performance. The RC algorithm does not take into account social relationships, and the possibility of sharing content between vehicles is reduced, thereby reducing network performance. The present invention jointly considers vehicle mobility and sociality, so it has better performance than the two comparison methods. When the number of vehicles is 50, the cache hit rate of the method of the present invention is 14.4% and 24.9% higher than that of the RC algorithm and the DAC algorithm, respectively.

[0123] Figure 5 : is a comparison chart of the time delay of the present invention and the comparative method under different vehicle data in this embodiment. Figure 5As shown in the figure, as the number of vehicles increases, the content cache delay gradually increases, but the present invention has greater advantages in terms of delay. When the social relationship between vehicles is not considered, the possibility of sharing content between two vehicles is reduced, resulting in performance degradation. When the mobility of vehicles is not considered, the trajectory information of the vehicle cannot be predicted, and the requested content of the vehicle and the vehicle cache node cannot be determined, thereby increasing the content acquisition delay.

[0124] Figure 6 : is a comparison chart of cache hit rates of the present invention and the comparative method under different total cache capacities in this embodiment. Figure 6 As shown, the total cache capacity has an impact on the cache hit rate of the three methods, and the cache hit rate increases with the increase of the total cache capacity. This is because a larger cache capacity means that more content can be cached so that cached vehicles and roadside units can meet more requests from vehicle users. The present invention predicts the trajectory of the vehicle and the social relationship between vehicle users to more accurately select vehicles as caches. It has obvious advantages over other strategies and also obtains a better cache hit rate. When the total cache capacity is 100MB, the hit rate of the proposed strategy is 18.1% and 26% higher than that of RC and DAC, respectively.

[0125] Figure 7 : is a comparison chart of the delay of the present invention and the comparative method under different total cache capacities in this embodiment. Figure 7 As shown in Figure 2, the total content acquisition delay of the three methods decreases as the cache capacity increases. The reason is that as the total cache capacity increases, each vehicle can obtain more content from the roadside units and cache vehicles, thus reducing the total acquisition delay. Figure 7 It can be seen that the present invention has a lower total delay than the RC method and the DAC method as a whole because the present invention predicts the position of the vehicle at the next moment, and on this basis, considers the interest similarity and contact rate of the vehicle to select the cached vehicle, which increases the probability of the vehicle obtaining content from the cached vehicle. Therefore, there is a lower content acquisition delay. When the total cache capacity is 100Mb, the content acquisition delay of the present invention is reduced by 24.2% and 27.3% compared with RC and DAC, respectively.

[0126] In summary, the performance of the present invention in terms of content hit rate and delay is better than the two comparison methods.

[0127] Although the above describes the illustrative specific embodiments of the present invention to facilitate those skilled in the art to understand the present invention, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations using the concept of the present invention are protected.

Claims

1. A collaborative caching method for Internet of Vehicles based on social relationship perception of mobile vehicles, characterized in that: The following steps are involved: S1: The set of roadside units in the Internet of Vehicles RSU = {S1,…,S n ,…,S N }, where S n represents the nth roadside unit, n=1,2,…,N, N represents the number of roadside units; The set of content that can be requested in the Internet of Vehicles is X = {x1,…,x q ,…,x Q }, x q represents the qth content, q = 1, 2, ..., Q, Q represents the number of contents that can be requested in the Internet of Vehicles, and the content x q The size is d q ; S2: For each vehicle in the Internet of Vehicles, the movement trajectory of the vehicle is predicted to obtain the predicted position at the next moment, thereby determining the roadside unit to which the vehicle belongs at the next moment; Get each roadside unit S at the next moment n The set of vehicles within the coverage area V n = {v n,1 ,…,v n,m ,…,v n,M },v n,m Represents the roadside unit S n The mth vehicle covered, m = 1, 2, ..., M n , M n Represents the roadside unit S n The number of vehicles within the coverage area; S3: First, count the number of requests for each content in the preset historical time period, sort the content from large to small according to the number of requests, and record the content x q The sequence number after sorting is i q ; Calculate the request probability of each content in the preset historical time period, take the content with request probability exceeding the preset threshold as popular content, and sum the request probabilities of popular content as the distribution slope ω of popular content; Use the following formula to calculate the content popularity pop of the qth content at the next moment q : For roadside units S n The set of vehicles within the coverage area V n , get each vehicle v n,m The content set L requested within the preset historical time period n,m Then the following calculation formula is used to calculate the interest similarity sim between vehicles m,m′ : Among them, φ m,m′ Indicates vehicle v n,m and vehicle v n,m′ Content Collection L n,m , L n,m′ The intersection of m′=1,2,…,M n ,| | indicates the number of contents in the set, M q Indicates that content x has been requested within the preset historical time period q The number of vehicles; The vehicle v is calculated using the following formula n,m SIM m : Similarity of interest m Normalize to get the normalized interest similarity SIM′ m : Among them, SIM max , SIM min Respectively represent the vehicle set V n SIM of interest similarity of all vehicles in m The maximum and minimum values ​​of ; The vehicle v is calculated using the following formula n,m Probability of interest in each content n,m : Content popularity pop q and vehicle v n,m Probability of interest in n,m,q Perform weighted summation to obtain the vehicle v at the next moment n,m For content x q The request probability p n,m,q : p n,m,q =α×pop q +β×in n,m Among them, α and β represent the preset content popularity weight value and vehicle personal interest preference weight value respectively; Set the vehicle request content to follow the Poisson distribution, according to the vehicle v n,m For each content x q The request probability p n,m,q Predict the vehicle v n,m The content set requested at the next moment S4: For roadside unit S n The vehicle set V n Each vehicle v n,m , the number of vehicles that can be associated per unit time during driving is counted as the communication contact rate f n,m , integrating the interest similarity between vehicles m,m′ , the social relationship strength z between vehicles is calculated pairwise using the following formula: m,m′ : z m,m′ =f n,m ×sim m,m′ The vehicle v is calculated using the following formula n,m The strength of social relationship Z m : The roadside unit S n All vehicles under the cache are sorted from large to small according to the strength of social relationship, and the first K vehicles are selected as cache vehicles. The value of K is determined according to actual needs, and the cache vehicle set CV is obtained. n ={cv n,1 ,…,cv n,k ,…,cv n,K }, cv n,k Represents the roadside unit S n The kth cached vehicle under, k = 1, 2, ..., K; S5: Calculate the latency of different content acquisition methods. The specific method is as follows: Vehicle n,m From cache vehicle cv via V2V communication n,k Get content x q The delay T n,m,k,q for: Among them, rate n,m,k Indicates vehicle v n,m and cache vehicle cv n,k Communication transmission rate between Vehicle n,m From the roadside unit S via V2R communication n Get content x q The delay T n,m,K+1,q for: Among them, rate n,m,K+1 Indicates vehicle v n,m and roadside units S n Communication transmission rate between S6: Solve the following collaborative cache optimization problem to obtain each roadside unit S n The following collaborative caching scheme: C1:a q,k′ ∈{0,1},k′=1,2,…,K+1 Among them, A=(a q,1 ,…,a q,k ,…,a q,K+1 ) represents the cache decision vector, a q,k′ is the cache decision identifier, k′=1,2,…,K+1, a q,k =1 means that the content x q Cache to cache vehicle cv n,k , a q,k =0 means that the content x is not q Cache to cache vehicle cv n,k , a q,K+1 =1 means that the content x q Cache to roadside unit S n , a q,K+1 =0 means that the content x is not q Cache to roadside unit S n , Indicates cached vehicle cv n,k The cache capacity, Represents the roadside unit S n The cache capacity.

2. The Internet of Vehicles collaborative caching method according to claim 1, characterized in that: The vehicle movement trajectory prediction in step S2 is implemented based on the LSTM network, and the specific method is: The historical position vector of the vehicle is obtained by sampling the historical movement trajectory of the vehicle, and is input into the pre-trained LSTM network. The probability of the vehicle being located at each accessible position is output, and the accessible position corresponding to the maximum probability is taken as the predicted position of the vehicle.

3. The Internet of Vehicles collaborative caching method according to claim 1, characterized in that: In step S5, the vehicle v n,m and cache vehicle cv n,k Communication transmission rate between n,m,k The calculation formula is as follows: Where B represents the transmission bandwidth, σ 2 represents additive white Gaussian noise, P n,k Indicates cached vehicle cv n,k The transmission power, h n,m,k Indicates vehicle v n,m and cache vehicle cv n,k The distance between n,m″,k Indicates vehicle v n,m″ and cache vehicle cv n,k The distance between.

4. The Internet of Vehicles collaborative caching method according to claim 1, characterized in that: In step S5, the vehicle v n,m and roadside units S n Communication transmission rate between n,m,K+1 The calculation formula is as follows: Where B represents the transmission bandwidth, σ 2 represents additive white Gaussian noise, P n Represents the roadside unit S n The transmission power, h n,m,K+1 Indicates vehicle v n,m and roadside units S n The channel gain between .

5. The Internet of Vehicles collaborative caching method according to claim 1, characterized in that: The optimization problem of cooperative caching in step S6 is solved by using a reinforcement learning method, where the state s is set to Set action a to cache decision vector A = (a q,1 ,…,a q,k ,…,a q,K+1 ), the reward function R is the delay corresponding to the cache decision vector, that is,