A network coding-based method and system for edge node content caching in vehicle networking

Abstract

The application discloses a kind of network coding-based edge node content caching method and system for Internet of Vehicles, for the cellular Internet of Vehicles system of a certain road / region edge node by several roadside units (Road Side Unit, RSU) as: the method and system model vehicle user content request mode, predict the probability that a certain vehicle obtains the content it needs from a particular edge node, then based on network coding method, on-demand coding is carried out to the content to be cached, finally, the content block is stored in each roadside unit RSU, to improve user content request hit rate-the probability that the requested data can be directly obtained from RSU.The method proposed in the application has lower influence of uncertainty environment, higher edge node caching content hit rate and other advantages compared with existing caching methods considering vehicle movement process.

Description

Technical Field

[0001] This invention belongs to the field of network communication technology, and specifically relates to a method for caching mobile-aware encoded content based on network coding. Background Technology

[0002] With the continuous development of 5G and IoT (Internet of Things) technologies, connected transportation systems, represented by cellular vehicle-to-everything (C-V2X), will be a key infrastructure for the realization of future autonomous driving and safe transportation systems. The cellular vehicle-to-everything architecture involves key components such as cloud platforms, edge servers (ES), edge nodes (ENs), onboard units (OUs), and user smart devices. Edge nodes can be dedicated roadside units (RSUs) deployed along roadsides or at intersections, or they can be general-purpose devices such as small base stations or wireless Wi-Fi access points.

[0003] However, compared to cloud-centric storage, edge nodes have limited storage space, thus requiring efficient content caching strategies to maximize the benefits of content caching.

[0004] Since the caching strategies for vehicle-to-everything (V2X) edge nodes are typically designed for applications with latency constraints, such as streaming media, where content needs to be cached sequentially on several edge nodes and delivered according to the vehicle's movement order, traditional content caching methods (such as LRU, LFU, etc.) are not applicable.

[0005] Given this, existing research mainly falls into two categories: one is caching strategies based on vehicle trajectory prediction, which pre-caches content that may be requested by these vehicles at edge nodes by predicting their future positions. The other is determining the content to cache at edge nodes based on its popularity or the interests of the target user group. However, both of these methods require statistical analysis of each vehicle's trajectory or the interests and social tendencies of users inside the vehicle, inevitably raising concerns about user privacy violations. Furthermore, they offer relatively weak consideration for the sequential caching of vehicle movement. Summary of the Invention

[0006] Purpose of the invention: The purpose of this invention is to improve the hit rate of roadside unit cache content in cellular vehicle-to-everything (V2X) networks by introducing network coding technology.

[0007] Technical solution: A method for caching content at the edge nodes of a vehicle-to-everything (V2X) network based on network coding, characterized by comprising:

[0008] Several RSUs serve as edge nodes of the vehicle-to-everything (V2X) system. Each RSU is equipped with a memory that caches target content. The target content to be cached consists of K data blocks, and each block is identified by its sequence number k, where 1 ≤ k ≤ K.

[0009] Model the content request pattern of vehicle users to predict the probability that a vehicle will obtain the content it needs from a specific edge node. Deploy content on RSU based on network coding method and store content blocks in RSU to increase the probability that the data requested by the user will be obtained directly from RSU.

[0010] Furthermore, the specific steps for modeling vehicle user content request patterns are as follows:

[0011] Let R be the set of roadside units (RSUs) serving the target vehicle user. Suppose R is ordered according to the order in which the target vehicle will access the RSUs, that is, the set of a series of RSUs that the vehicle user will pass through: R = {R1, R2, R3…R}. i …}; Let D = {d1, d2, d3, …} be the content block of the source file to be cached, and let X be the random variable. i This indicates that the vehicle user is in R i The random variable representing the last content block number received while staying in the coverage area is defined as X0 = 0; then the vehicle user in R i The set of content blocks downloaded while staying in the coverage area is denoted as (X). i-1 ,X i For the k-th content block, it can be accessed by the vehicle user from R. i The probability of downloading within the coverage area is:

[0012] p i (k)=P(k∈(X i-1, X i ])=P(X i-1 <k≤X i )

[0013] Let random variable S i For the vehicle user in R i The total number of content blocks downloaded while staying in the coverage area, therefore S i The probability density function mainly depends on two factors: 1) the vehicle's position in R i 2) Duration of stay within the coverage area; i The actual throughput obtained by vehicle users within the coverage area depends on whether the RSU cache content is hit, as well as its own wireless data rate and channel contention.

[0014] According to S i and X i The definition is easy to obtain:

[0015]

[0016] Furthermore, the vehicle user is in R i The probability p of downloading the k-th content block is given by the coverage area. i (k) is:

[0017]

[0018] or

[0019]

[0020] And in R i Within the coverage area, the average number of file blocks downloaded by users of this vehicle is:

[0021]

[0022] Furthermore, it should be noted that the premise of the previous step does not require assuming that the number of downloaded blocks is related when the vehicle user is in different RSU coverage areas. However, if we relax the condition and assume that the time a vehicle user spends in different RSU coverage areas is independent, that is, the number of downloaded blocks for a vehicle user in different RSUs is independent, then...

[0023]

[0024] Furthermore, if the vehicle user is in a different R i The coverage area is the number of content blocks downloaded (S). i If the vehicles are independent and identically distributed, then the vehicle user will be in R. i The probability p of downloading the k-th content block is given by the coverage area. i (k) is:

[0025] p i (k)=f S *p i-1 (k)

[0026] Among them, f S is the probability density function of S, and * denotes convolution operation.

[0027] Furthermore, the network coding method assumes that a source file Y, after being segmented, contains n content blocks y1, y2, ..., y n The encoding process is denoted as E = C·Y, where C is called the encoding matrix, and E = (e1,...,e2) / (e3,...,e4) / (e5,...,e6) / (e7,...,e8) / (e9,...,e9) / (e1 ... m ) T It is the set of m coded blocks generated after the encoding operation, where m ≥ n, and any coded block e in the set iIt is generated by a linear combination of n source data blocks over a selected finite field:

[0028] e i =c i ·Y=(c i,1 ,...,c i,n )·(y1,...,y n ) T i∈[1,m]

[0029] Where, vector c i Also known as coded block e i The encoded vector has elements randomly generated from a finite field of computation.

[0030] Furthermore, the content deployment of the Roadside Unit (RSU) based on the network coding method specifically involves:

[0031] Input the file Y to be cached and the RSU set R = {R1, R2, ... R...} i …};

[0032] Divide the source file Y into blocks;

[0033] Generate the encoding vector matrix C;

[0034] Generate a set of coded blocks E;

[0035] Deploy content for each RSU in sequence.

[0036] (1) Estimate the vehicle user's location in R i Average number of blocks downloaded per hour E(S) i ), S i It is a random variable;

[0037] (2) Select E(S) from E i ) coded blocks, denoted as E i and store in R i ;

[0038] (3) E = EE i ;

[0039] The process ends when all RSUs have been cached.

[0040] A network coding-based vehicle network edge node content caching system includes a vehicle network system, edge nodes, and several roadside units (RSUs) as edge nodes of the vehicle network system. Each RSU is equipped with a memory, which caches target content. The target content to be cached consists of K data blocks, and each block is identified by its sequence number k, where 1 ≤ k ≤ K.

[0041] The vehicle-to-everything (V2X) system includes a network coding-based RSU content deployment module, which models vehicle user content request patterns and deploys content to roadside units (RSUs) based on network coding methods.

[0042] Furthermore, the network coding-based RSU content deployment module defines R as the set of roadside units (RSUs) serving the target vehicle user. Let R be ordered according to the order in which the target vehicle will access the roadside units (RSUs), that is, the set R = {R1, R2, R3…R} of the series of RSUs the vehicle user will pass through. i …}; Let D = {d1, d2, d3, …} be the content block of the source file to be cached, and let X be the random variable. i This indicates that the vehicle user is in R i The random variable representing the last content block number received while staying in the coverage area is defined as X0 = 0; then the vehicle user in R i The set of content blocks downloaded while staying in the coverage area is denoted as (X). i-1 ,X i For the k-th content block, it can be accessed by the vehicle user from R. i The probability of downloading within the coverage area is:

[0043] p i (k)=P(k∈(X i-1 ,X i ])=P(X i-1 <k≤X i )

[0044] Let random variable S i For the vehicle user in R i The total number of content blocks downloaded while staying in the coverage area is obtained as follows:

[0045]

[0046] The vehicle user is in R i The probability p of downloading the k-th content block is given by the coverage area. i (k) is:

[0047]

[0048] or

[0049]

[0050] And in R i Within the coverage area, the average number of file blocks downloaded by users of this vehicle is:

[0051]

[0052] Furthermore, the content deployment of the Roadside Unit (RSU) based on the network coding method specifically involves:

[0053] Input the file Y to be cached and the RSU set R = {R1, R2, ... R...} i …};

[0054] Divide the source file Y into blocks;

[0055] Generate the encoding vector matrix C;

[0056] Generate a set of coded blocks E;

[0057] Deploy content for each RSU in sequence.

[0058] (1) Estimate the vehicle user's location in R i Average number of blocks downloaded per hour E(S) i ), S i It is a random variable;

[0059] (2) Select E(S) from E i ) coded blocks, denoted as E i and store in R i ;

[0060] (3) E = EE i ;

[0061] The process ends when all RSUs have been cached.

[0062] Beneficial effects: This invention establishes an efficient caching strategy that stores content blocks in the RSU to ensure that vehicles can obtain content in a timely and high-throughput manner, and that the requested data has a high probability of being obtained directly from the RSU rather than from the backhaul link, thereby improving network efficiency and system fault tolerance. Attached Figure Description

[0063] Figure 1 Architecture diagram of edge caching system for connected vehicle content;

[0064] Figure 2 This is a schematic diagram of the MAP content caching algorithm;

[0065] Figure 3 A schematic diagram of the RSU content caching algorithm based on network coding;

[0066] Figure 4 This is a flowchart of the RSU content caching algorithm based on network coding. Detailed Implementation

[0067] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading the present invention, any modifications of the present invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0068] Consider as Figure 1 The illustration shows a C-V2X vehicle-to-everything (V2X) system in an urban environment, where edge nodes consist of various Remote Units (RSUs) deployed at streetlights or road intersections. Each RSU is equipped with memory to cache target content and provide services to passing vehicle users. This patent assumes that the content of the file of interest to be cached consists of K data blocks, and each block is identified by its sequence number k, therefore 1 ≤ k ≤ K.

[0069] exist Figure 1 In the scenario shown, several edge nodes near the road are controlled by a C-V2X edge server (ES). Since the ES is connected to a vehicle-to-everything (V2X) traffic flow awareness system, we assume that the ES knows or can predict the edge nodes that vehicles will pass through in the next few minutes and understands the distribution of dwell time at each edge node (e.g., generated by statistical analysis of historical information). Furthermore, the ES is aware of the available space in the RSUs cache and the content requests made by passing users. Based on this information, the ES can determine which content blocks should be cached at each edge node and download these blocks from the cloud center accordingly, deploying them to the corresponding edge nodes.

[0070] When a vehicle user enters the coverage area of ​​an RSU, the application requests a new batch of content blocks, indicating the content item it is receiving and the content block number k it expects to retrieve from it. The RSU checks whether block number k is cached or whether it needs to be downloaded from the backhaul link, which is obviously an expensive operation in terms of bandwidth and latency. As the block is transmitted, the RSU continues to send the next block, again relying on the backhaul link if it is not stored in its own cache.

[0071] Therefore, our goal is to establish an efficient caching strategy that stores content blocks in the RSU to ensure that vehicles can obtain content in a timely and high-throughput manner, and that the requested data has a high probability of being obtained directly from the RSU, rather than from the backhaul link.

[0072] The method described depends on the figure shown. Figure 2The illustrated MAP content caching strategy is a Mobility-Aware Probabilistic caching strategy (MAP) that considers vehicle movement. It involves analyzing the probability distribution of time a target vehicle user spends within the coverage area of ​​each edge node, and then deploying content blocks on these edge nodes to maximize the content hit rate. Therefore, the MAP strategy does not require statistical analysis of user or vehicle movement; instead, it analyzes the distribution of vehicle dwell time at each edge node to formulate a deployment strategy for target content on these edge nodes with the aim of maximizing the hit rate. This invention employs the MAP strategy, and the content cached at the edge nodes in the method described herein is an encoded block formed after network encoding operations, rather than the original content.

[0073] Content caching at the edge nodes of the vehicle-to-everything (V2X) network, exemplified by RSU, is performed in units of content blocks. Under the management of the edge server, the file to be cached is first divided into data blocks of fixed size. If the file cannot be divided equally, zero-padding is performed on the last block. The network coding method assumes that a source file Y, after being split, contains n content blocks y1, y2, ..., y... n The encoding process is denoted as E = C·Y, where C is called the encoding matrix, and E = (e1,...,e2) / (e3,...,e4) / (e5,...,e6) / (e7,...,e8) / (e9,...,e9) / (e1 ... m ) T It is the set of m coded blocks generated after the encoding operation, where m ≥ n, and any coded block e in the set i It is generated by a linear combination of n source data blocks over a selected finite field:

[0074] e i =c i ·Y=(c i,1 ,...,c i,n )·(y1,...,y n ) T i∈[1,m]

[0075] Where, vector c i Also known as coded block e i The encoded vectors contain elements randomly generated from a finite field of computation. Clearly, decoding the source content block from the encoded block is the process of solving a system of linear equations. As we know from linear algebra, when a user vehicle obtains encoded blocks whose encoded vectors are linearly independent, it can be decoded using algorithms for solving systems of linear equations, such as Gaussian elimination or Gauss-Jordan elimination, to recover the source file content.

[0076] by Figure 1 Taking the R1 vehicle content caching service on the road as an example, this paper compares and describes the vehicle-to-everything (V2X) RSU content caching process of the method described in this patent with the simple MAP caching strategy.

[0077] Consider the three RSUs {R1, R2, R3} that the vehicle will pass through sequentially in the next few minutes as edge nodes of the content to be cached. Assume that the file to be cached, f, contains three content blocks: f = {f1, f2, f3}, and each RSU can provide a cache space of one content block size for this vehicle user.

[0078] Simple MAP content caching strategy: The MAP caching strategy deploys the three data blocks of the source file onto the corresponding RSUs in sequence, based on the RSU order. Figure 2 As shown. Because the dwell time of a vehicle within the coverage area of ​​each RSU is affected by factors such as current traffic flow, road conditions, and driving habits, the number of content blocks that can be downloaded during each RSU period varies. If we assume that the vehicle's dwell time in coverage areas R1, R2, and R3 is 2, 1, and 2 respectively, then one content block can be downloaded per time unit.

[0079] When a vehicle is within the coverage area of ​​R1, it can first download the content block f1 from R1's storage space. However, since the vehicle resides at R1 for two time slices, the vehicle user will download f2 from the backhaul link in the second time slice. When the vehicle user arrives at R2, it needs f3, but R2 caches f2, so it still needs to be downloaded from the backhaul link. Therefore, through the MAP strategy, the vehicle user can directly obtain one content block from the RSU, with a cache hit rate of 1 / 3.

[0080] The method described in this patent (based on the MAP strategy, incorporating a network coding method): If a network coding method is used, the content blocks of the source file first need to be encoded. Let the encoded content blocks be denoted as e = {e1, e2, e3, ...}, where e = A·f, and A is the coding coefficient matrix. Since the encoding operation operates over a finite field, the size of the encoded content blocks is the same as that of the source file content blocks. After encoding, each encoded block is deployed at each RSU. Unlike the simple MAP strategy, the encoded blocks after network coding are equivalent, therefore they can be arbitrarily deployed at each RSU, as shown in the deployment result. Figure 3 As shown.

[0081] The vehicle's dwell time within the coverage areas of R1, R2, and R3 remains 2, 1, and 2 seconds, respectively. When a vehicle is within the coverage area of ​​R1, it can first download the content block e1 from the storage space of R1. However, since the vehicle will reside at R1 for two time slices, the vehicle user will download an encoded block e* from the backhaul link in the second time slice. e* can be an encoded block left over from the deployment process, such as e4, or it can be dynamically encoded by the edge server from the source content block. Without loss of generality, let's assume e* is e4. When the vehicle user arrives at R2, it downloads the encoded block e2 cached by R2. At this point, the user vehicle has obtained three encoded file content blocks and can decode them. Therefore, through the network coding algorithm, the vehicle user can directly obtain two content blocks from the RSU, with a cache hit rate of 2 / 3, which is twice as high as the simple MAP strategy.

[0082] The above-mentioned network coding-based edge content caching method for cellular vehicle networks can effectively improve the hit rate of cached content in roadside units and reduce the impact of uncertain environments. Figure 4 This is the algorithm flowchart.

[0083] This invention presents a novel probabilistic caching method for edge node content that considers vehicle movement. The method divides the content to be cached, which is of interest to vehicle users, into fixed-size blocks and performs network encoding. A vehicle movement perception model is established to predict the probability that a vehicle will obtain the content it needs from a specific edge node. Then, based on the aforementioned model, the network-encoded content is deployed to each edge node sequentially. Compared with existing caching methods that consider vehicle movement, the proposed method has advantages such as lower susceptibility to uncertainties and higher cache hit rate at edge nodes.

Claims

1. A method for caching content at the edge nodes of a vehicle-to-everything (V2X) network based on network coding, characterized in that, include: Several roadside units (RSUs) serve as edge nodes of the vehicle-to-everything (V2X) system. Each RSU is equipped with a memory that caches target content. The target content to be cached consists of K data blocks, and each block is identified by its sequence number k, where 1 ≤ k ≤ K. Model the content request pattern of vehicle users to predict the probability that a vehicle will obtain the content it needs from a specific edge node. Deploy content on RSU based on network coding method and store content blocks in RSU to increase the probability that the data requested by the user will be obtained directly from RSU. The specific steps for modeling vehicle user content request patterns are as follows: Let R be the set of Roadside Units (RSUs) serving the target vehicle user. Suppose R is ordered according to the order in which the target vehicle will access the Roadside Units (RSUs), that is, the set of a series of RSUs that the vehicle user will pass through: R = {R1, R2, R...} 3… R i… Let D be the content block D of the source file to be cached, containing {d1, d2, d3,}. … }, and denote the random variable This indicates that the vehicle user is in R i A random variable representing the last content block number received while the coverage area is occupied, defined as follows: Then the vehicle user in R i The set of content blocks downloaded while staying in the coverage area is denoted as For the k-th content block, it can be accessed by the vehicle user from R. i The probability of downloading within the coverage area is: ; Denote random variables , For the vehicle user in R i The total number of content blocks downloaded while staying in the coverage area is obtained as follows: ； The vehicle user is in R i The probability of downloading the k-th content block when the coverage area is given. for: ; or ; And in R i Within the coverage area, the average number of file blocks downloaded by users of this vehicle is: ; The network coding method assumes that a source file Y, after being segmented, contains n content blocks. The encoding process is denoted as ,in This is called the encoding matrix. It is generated after encoding operation A set of coded blocks, wherein Any coded block in the set It is generated by a linear combination of n source data blocks over a selected finite field: ； Where, vector Also known as a coded block The encoded vector, whose elements are randomly generated from a finite field of computation; The content deployment of Roadside Units (RSUs) based on the network coding method specifically involves: Input the file Y to be cached and the RSU set R = {R1, R2, ... R} i… }; Divide the source file Y into blocks; Generate the encoding vector matrix C; Generate a set of coded blocks E; Deploy content for each RSU in sequence. (1) Estimate the vehicle user's location in R i Average number of blocks downloaded per hour E(S) i ), It is a random variable; (2) Select E(S) from E. i ) coded blocks, denoted as E i and store in R i ; (3)E=EE i ; The process ends when all RSUs have been cached.

2. The method for caching content at vehicle network edge nodes based on network coding according to claim 1, characterized in that, If it is assumed that the time a vehicle user spends within the coverage area of ​​different RSUs is independent, that is, the number of download blocks a vehicle user downloads is independent across different RSUs, then... 。 3. The method for caching content at vehicle network edge nodes based on network coding according to claim 1, characterized in that, If the vehicle user is in a different R i The coverage area is the number of content blocks downloaded (S). i If the vehicles are independent and identically distributed, then the vehicle user will be in R. i The probability of downloading the k-th content block when the coverage area is given. for: , in, is the probability density function of S, and * denotes convolution operation.

4. A network coding-based vehicle-to-everything (V2X) edge node content caching system, comprising a V2X system and edge nodes, characterized in that, Several roadside units (RSUs) serve as edge nodes of the vehicle-to-everything (V2X) system. Each RSU is equipped with a memory that caches target content. The target content to be cached consists of K data blocks, and each block is identified by its sequence number k, where 1 ≤ k ≤ K. The vehicle-to-everything (V2X) system includes a network coding-based RSU content deployment module, which models vehicle user content request patterns and deploys content to roadside units (RSUs) based on network coding methods. The network coding-based RSU content deployment module defines R as the set of Roadside Units (RSUs) serving the target vehicle user. Let R be ordered according to the order in which the target vehicle will access the RSUs, i.e., the set of a series of RSUs that the vehicle user will pass through: R = {R1, R2, R...} 3… R i… Let D be the content block D of the source file to be cached, containing {d1, d2, d3,}. … }, and denote the random variable This indicates that the vehicle user is in R i A random variable representing the last content block number received while the coverage area is occupied, defined as follows: Then the vehicle user in R i The set of content blocks downloaded while staying in the coverage area is denoted as For the k-th content block, it can be accessed by the vehicle user from R. i The probability of downloading within the coverage area is: ; Denote random variables For the vehicle user in R i The total number of content blocks downloaded while staying in the coverage area is obtained as follows: ; The vehicle user is in R i The probability of downloading the k-th content block when the coverage area is given. for: ; or ; And in R i Within the coverage area, the average number of file blocks downloaded by users of this vehicle is: ; The network coding method assumes that a source file Y, after being segmented, contains n content blocks. The encoding process is denoted as ,in This is called the encoding matrix. It is generated after encoding operation A set of coded blocks, wherein Any coded block in the set It is generated by a linear combination of n source data blocks over a selected finite field: ； Where, vector Also known as a coded block The encoded vector, whose elements are randomly generated from a finite field of computation; The content deployment of Roadside Units (RSUs) based on the network coding method specifically involves: Input the file Y to be cached and the RSU set R = {R1, R2, ... R} i… }; Divide the source file Y into blocks; Generate the encoding vector matrix C; Generate a set of coded blocks E; Deploy content for each RSU in sequence. (1) Estimate the vehicle user's location in R i Average number of blocks downloaded per hour E(S) i ), It is a random variable; (2) Select E(S) from E. i ) coded blocks, denoted as E i and store in R i ; (3)E=EE i ; The process ends when all RSUs have been cached.

Images