Multi-Time Scale Content Caching and Transmission Optimization Method for NOMA-Based Vehicular Networks

Abstract

The present invention discloses an optimization method for multi-time scale content caching and transmission in a vehicle-to-everything (V2X) network based on non-orthogonal multiple access (NOMA). The method includes constructing a V2X network scenario based on NOMA, and then establishing a transmission model for vehicles in the V2X network scenario. The V2X network scenario includes a macro base station, a sub-base station, and vehicle providers equipped with caching functions. Then, a quality of service model is constructed based on the vehicle transmission model and an optimization objective is established. Finally, according to the optimization objective, the caching status, transmission status, and power of each video content provider are allocated to obtain an optimal allocation scheme. By introducing a hybrid time scale optimization model, a cooperative transmission scheme, and a pre-caching technology, the present invention overcomes the problems of poor multi-connection adaptability, unstable transmission, and excessive burden on the central network in the prior art, significantly improves the quality of service and stability of video stream transmission in the V2X network, and is applicable to high-density dynamic scenarios.

Description

Technical Field

[0001] The present invention belongs to the field of vehicle networking technology, and particularly relates to an optimization method for multi-time scale content caching and transmission in vehicle networking based on NOMA. Background Art

[0002] With the rapid development of 5G technology, vehicle networking has become an important part of intelligent transportation systems. The explosion in the number of vehicles has led to a large influx of data and an increasing demand for network connectivity. Given the dynamic and multi-connection characteristics of vehicle networking, traditional mobile network topologies may experience transmission failures due to resource limitations and data congestion. However, expanding network capacity by adding new base stations is costly. Therefore, non-orthogonal multiple access (NOMA) technology has been proposed to provide solutions for future multi-connection scenarios. NOMA differs from traditional orthogonal multiple access (OMA) technology in that it allows multiple users to share the same frequency band channel, thereby improving spectrum utilization and effectively reducing access conflicts. When applying NOMA technology for data transmission in dense vehicle networking (where multiple transmission sources cooperate to provide video transmission services to vehicle requesters), many new challenges arise. First, considering the joint decoding nature of NOMA technology, power allocation must be considered. Second, different connection strategies result in different user groups, which in turn generate inter-group and intra-group interference. All of the above factors will affect the transmission efficiency. Therefore, compared with traditional transmission methods, the NOMA framework constitutes a more complex scenario.

[0003] Although there has been some work on caching and NOMA technology, most of these works focus on frequency allocation from the perspective of resource saving. For example, they aim to minimize the average delay, increase network throughput, and improve the cache hit rate, but do not fully consider the impact of cache placement on video transmission performance. These works are not applicable to the optimization of video transmission in dense network scenarios. At the same time, the existing work on content transmission in static scenarios based on NOMA is not applicable to vehicle frameworks with frequent connection switches. In addition, previous studies have ignored the cooperation between transmission sources and the utilization of resources from different infrastructure providers (InPs) to improve transmission gain. Existing wireless caching technologies cannot be directly applied to specific high-density vehicle networking environments. The multi-connection problem in NOMA-based networks in dense transmission scenarios has attracted wide attention, but existing studies have not fully considered the problem of frequent switching between users and providers. In addition, these studies have not adopted pre-caching technology to relieve the burden on the central network. Inspired by the advantages of content caching and NOMA technology, some work has explored cache-aided transmission optimization in NOMA networks and studied two wireless caching strategies applicable to NOMA environments, especially for updating content caching during peak hours. However, these methods have limited effects on improving video transmission quality in dynamic scenarios. Summary of the Invention

[0004] In order to solve the problems existing in the background art, the purpose of the present invention is to provide an optimization method for multi-time scale content caching and transmission in a vehicle-to-everything (V2X) network based on non-orthogonal multiple access (NOMA), aiming to maximize the long-term average quality of service of in-vehicle video streams.

[0005] The technical solution adopted by the present invention includes the following steps:

[0006] Step S1: First, construct a V2X vehicle network scenario in a computer based on NOMA, and then establish a transmission model of vehicles in the V2X vehicle network scenario;

[0007] Step S2: Then, the processor in the computer constructs a quality of service model of vehicles based on the transmission model of vehicles, and then the processor establishes an optimization objective based on the quality of service model of vehicles;

[0008] Step S3: Finally, the processor processes the transmission model according to the optimization objective to obtain the optimal caching status, transmission status, and power allocation scheme of each video content provider in the V2X vehicle network scenario.

[0009] The V2X vehicle network scenario in Step S1 is constructed based on non-orthogonal multiple access (NOMA), and includes a macro base station, several sub-base stations, and vehicle providers equipped with caching functions. The macro base station is connected to a content server storing a video file library. The V2X vehicle network covers several infrastructure providers. Each infrastructure provider manages several sub-base stations and vehicle providers. The sub-base stations and vehicle providers managed by the same infrastructure provider are content providers under this infrastructure provider.

[0010] The content providers under the same infrastructure provider share the bandwidth resources of this infrastructure provider. The bandwidth resources of one infrastructure provider are orthogonal to the bandwidth resources of the other infrastructure providers. The content providers store video files and are used to provide the required video files for vehicle requesters.

[0011] The transmission model of vehicles in Step S1 mainly consists of a vehicle request caching model, a vehicle distribution model, and an in-vehicle network packet model. The vehicle request caching model is used to represent the caching status of content providers. The vehicle distribution model is used to represent the transmission status of video files between content providers and vehicle requesters. The in-vehicle network packet model is used to represent the data bit rate received by vehicle requesters on infrastructure providers.

[0012] In step S2, the average quality of service of the vehicle requester is used as the optimization goal. The specific steps of step S3 are as follows: Obtain the maximum value of the average quality of service, and then use the cache status, transmission status, and power allocation of the content provider when the average quality of service reaches the maximum value as the optimal vehicle network content caching and transmission scheme.

[0013] The expression of the vehicle request cache model in step S1 is as follows:

[0014] C p ={ c p 1 , c p 2 ,…, c p f ,…, c p F}, c p f ∈{0,1}

[0015] ∑ F f=1 c p f ≤C pmax

[0016] Among them, C p represents the cache status of content provider p; c p f represents the cache status. When content provider p caches video file f, then c p f =1. When content provider p does not cache video file f, then c p f =0; C pmax represents the cache capacity limit of content provider p; f represents the requested video file; F represents the total number of video files in the video file library;

[0017] The expression of the vehicle distribution model is as follows:

[0018] a i,p,r f (t) ≤c p f

[0019] ∑ I i=1 ∑ Pi p=1 a i,p,r f (t) ≤A max

[0020] ∑ R r=1 ai,p,r f ≤B i,p max

[0021] where a i,p,r f (t) represents the transmission status of video file f between content provider p and vehicle requester r under the i-th infrastructure provider InP in time slot t. If at time t, the i-th infrastructure provider InP i under the content provider p transmits the cached video file f to the vehicle requester r, then a i i,p,r f (t)=1; otherwise a(t)=0; I represents the total number of infrastructure providers; Pi represents the total number of content providers under the i-th infrastructure provider InP i,p,r f ; R represents the total number of vehicle requesters; A i represents the upper limit of the number of content providers that can be associated with each vehicle requester at the same time; B max represents the upper limit of the number of vehicle requesters that content provider p under infrastructure provider InP i,p max can serve; a i i,p,r f represents the transmission status of video file f between content provider p and vehicle requester r under the i-th infrastructure provider InP i in all time slots;

[0022] The expression of the in-vehicle network packet model is as follows:

[0023] R r (t)= ∑ I i=1 R i,r (t)+ r bs

[0024] R i,r (t)= W i ×log 2 (1+γ i,r (t))

[0025] γ i,r (t)= s i,r (t) / ( I i,r IGI (t)+ I i,r ICI (t)+ σ i,r (t) 2 )

[0026] ​s i,r (t) = ∑ F f=1 po f ∑ p∈Pi a i,p,r f (t) × p i,p,r ×h i,p,r

[0027] R c ≤R r (t) ≤R max

[0028] Among them, R r (t) represents the total data bit rate received by vehicle requester r within time slot t; r bs represents the data transmission bit rate dependent on the remote macro base station; R c represents the bit rate of the base layer of the SVC video stream; R max represents the bit rate corresponding to the highest video quality in the SVC video stream; R i,r (t) represents the data bit rate received by vehicle requester r from infrastructure provider InP i during time slot t; W i represents the bandwidth resource of the i-th infrastructure provider InP i ; γ i,r (t) represents the transmission signal-to-noise ratio from infrastructure provider InP i to vehicle requester r during time slot t; s i,r (t) represents the signal expectation received by vehicle requester r through bandwidth resource W i during time slot t; I i,r IGI (t) represents the in-group interference signal received by vehicle requester r through bandwidth resource Wi during time slot t; I i,r ICI (t) represents the inter-group interference signal received by vehicle requester r through bandwidth resource W i during time slot t; σ i,r (t) represents the noise power received by vehicle requester r through bandwidth resource W i during time slot t; p i,p,r represents the power allocated by content provider p under the i-th infrastructure provider InP i to vehicle requester r; h i,p,r represents the channel gain allocated by content provider p under the i-th infrastructure provider InP i to vehicle requester r; po f represents the probability that vehicle requester requests video file f.

[0029] The specific steps of step S2 are as follows:

[0030] Step S2.1: First, obtain the expression of the quality of service model according to the following formula:

[0031] QoS r (t)= a n ln(b n R r (t) / r r desir )

[0032] where QoS r (t) represents the quality of service corresponding to vehicle requester r within time slot t; a n and b n both represent preset quality of service model parameters; r r desir represents the video data bit rate expected by vehicle requester r; R r (t) represents the total data bit rate received by vehicle requester r within time slot t;

[0033] Step S2.2: Then, taking the maximization of the average quality of service of the vehicle requester as the optimization goal, obtain the optimization goal based on the quality of service model according to the following formula. The function expression of the optimization goal is as follows:

[0034] QoSa’=max{ QoSa}= max{(1 / R)∑ R r=1 QoS r (t)}

[0035] where QoSa’ represents the maximum value of the average quality of service, and QoSa represents the average quality of service;

[0036] In the above formula, the maximum value of the average quality of service is obtained by solving based on the SCA algorithm. In the SCA algorithm, the function expression of the average quality of service QoSa is as follows:

[0037] QoSa=(1 / T) (1 / R) ∑ T t=1 ∑ R r=1 a n ln(b n (r r (g z ,d z ,e z )+ h r (k)) / r r desir )

[0038] -η(M(a z) - N(a z-1 ) - ▽ a N(a z-1 ) (a z –a z-1 )) - η(O(c z ) - P(c z-1 ) - ▽ c P(c z-1 ) (c z –c z -1 ))

[0039] Among them, T represents the total number of time slots; R represents the total number of vehicle requesters; r r (g z , d z , e z ) represents the amount of data obtained by vehicle user r in the z-th iteration through transmission by surrounding vehicles; h r (k) represents a preset linear function; η represents a penalty parameter; M(a) represents the first function with respect to the transmission state variable a i,p,r f ; M(a z ) represents the value of the function M(a) in the z-th iteration; N(a) represents the second function with respect to the transmission state variable a i,p,r f ; N(a z ) represents the value of the function N(a) in the z-th iteration; O(c) represents the third function with respect to the cache state variable c p f ; O(c z ) represents the value of the function O(c) in the z-th iteration; P(c) represents the fourth function with respect to the cache state variable c p f ; P(c z ) represents the value of the function P(c) in the z-th iteration; ▽ a N(a) represents the gradient of the function N(a); ▽ a N(a z ) represents the value of the z-th iteration gradient ▽ a N(a); ▽ c P(c) represents the gradient of the function P(c); ▽ c P(c z ) represents the value of the z-th iteration gradient ▽ c P(c); The vector a represents the set of all transmission state variables a i,p,r f ; The vector c represents the set of all cache state variables c p f The set of.

[0040] In the said step S2, the expressions of the first function M(a), the second function N(a), the third function O(c), and the fourth function P(c) are as follows:

[0041] M(a)= ∑ I i=1 ∑ Pi p=1 ∑ R r=1 ∑ F f=1 a i,p,r f

[0042] N(a)= ∑ I i=1 ∑ Pi p=1 ∑ R r=1 ∑ F f=1 (a i,p,r f ) 2

[0043] O(c)= ∑ Pi p=1 ∑ F f=1 c p f

[0044] P(c)= ∑ Pi p=1 ∑ F f=1 (c p f ) 2

[0045] ▽ a N(a)= 2a i,p,r f

[0046] ▽ c P(c)= 2 c p f

[0047] Among them, I represents the total number of infrastructure providers; Pi represents the total number of content providers under the i-th infrastructure provider InP i The following content; R represents the total number of vehicle requesters; F represents the total number of video files in the video file library; a i,p,r f Represents the transmission status of the video file f between the content provider p and the vehicle requester r under the i-th infrastructure provider InP i in all time slots; cp f Indicates the cache status.

[0048] The vehicle networking content caching and transmission optimization system adopted by the present invention:

[0049] Includes a model construction module, which is used to construct a transmission model of a vehicle in a V2X vehicle networking scenario and construct a quality of service model of the vehicle based on the transmission model;

[0050] Includes an optimization objective construction module, which is used to establish an optimization objective based on the quality of service model of the vehicle;

[0051] Includes an optimal solution processing module, which is used to process the transmission model according to the optimization objective to obtain the cache status, transmission status and power allocation scheme of each video content provider.

[0052] Pre-caching is a technology that stores content on edge nodes or terminal devices in advance before actual demand occurs. In vehicle networking, pre-caching can effectively relieve the central network pressure during peak hours. By caching popular or high-demand content near users (such as roadside units or in-vehicle terminals) during off-peak hours, real-time transmission requirements can be reduced, thereby improving transmission efficiency and response speed. This method is particularly effective in the transmission of in-vehicle videos with high traffic and low-latency requirements, and can significantly reduce network latency and enhance the user experience. Cooperative distribution is a strategy for content transmission achieved through the collaboration of multiple network nodes or service providers. In NOMA vehicle networking, due to the high-speed mobility and frequent handover of vehicle users, traditional single-service provision schemes are difficult to meet the stability requirements. Cooperative distribution allows content requesters to be provided with content services by different service providers InP, thus leveraging the collaboration of multiple nodes to improve the flexibility and robustness of transmission. In practical applications, this collaboration can significantly improve bandwidth utilization, reduce transmission interruptions caused by frequent handovers, and ensure the quality of service QoS of video streams.

[0053] The sample average method is an approximation method widely used in stochastic optimization and is typically used to solve optimization problems involving future uncertainties. The sample average approximation (SAA) simulates possible future scenarios by generating a large number of random samples and uses the expectations of these samples to approximate the long-term objective, thus reducing the originally intractable long-term optimization problem to a sample-based deterministic problem. In the present invention, SAA is used for long-term caching optimization, enabling the caching policy to be reasonably approximated based on future user preferences and channel information, thereby better improving the quality of service. The successive convex approximation is an iterative algorithm for solving non-convex optimization problems. This method solves the non-convex problem by decomposing it into a series of convex sub-problems. In each iteration, the non-convex objective function is approximated as a convex function to gradually approach the optimal solution. In the present invention, due to the logarithmic form of the average quality of service (QoS) and the complexity of power allocation in the non-orthogonal multiple access (NOMA) environment, the problem is modeled as a mixed-integer non-linear programming (MINLP) problem. The successive convex approximation (SCA) algorithm solves the MINLP problem by transforming it into multiple convex sub-problems, ensuring convergence during the calculation process and improving the computational efficiency of the algorithm, making it applicable to the dynamically changing vehicular network environment.

[0054] The present invention constructs a vehicular network scenario characterized by high dynamicity, asynchronous updates, and multi-connectivity. The present invention carefully considers multiple key factors, including the actual scalable video coding (SVC) video bit rate, caching capacity, power constraints, and complex interference in NOMA. Meanwhile, to ensure video quality, the present invention designs a cooperative transmission scheme that enables vehicle requesters to be served by providers from the same or different service providers. Caching location, cooperative content distribution, and power allocation are jointly optimized on different time scales to ensure optimal transmission performance. In the long-term caching phase, the present invention uses the sample average approximation (SAA) method to approximate the long-term transmission performance through the expectation of random samples; in the short-term distribution phase, an iterative algorithm based on successive convex approximation (SCA) is adopted to transform the logarithmic form of the average quality of service (QoS) from a mixed-integer non-linear programming (MINLP) problem into a convex problem. The computational complexity of the proposed algorithm is mathematically analyzed, and a rigorous theoretical proof is provided to demonstrate its convergence.

[0055] The present invention is applicable to the vehicle networking scenario with high dynamics, asynchronous updates, and multi-connectivity to maximize the long-term average quality of service (QoS) of in-vehicle video streams. The present invention designs a cooperative transmission strategy that comprehensively considers the SVC-encoded video bit rate, cache capacity, power constraint, and NOMA interference, enabling requesters to be served by providers of the same or different Internet service providers (InPs), thereby enhancing the QoS of video streams in the NOMA vehicle networking. Joint optimization of multiple time scales: The challenge that the cache policy depends on future user preferences and transmission channel information is solved by the sample average approximation (SAA) method, optimizing the long-term caching performance. In the short term, an iterative algorithm based on successive convex approximation (SCA) is adopted to transform the QoS logarithmic-form mixed-integer non-linear programming (MINLP) problem into a convex problem, achieving efficient content distribution and power allocation in dynamic scenarios. Algorithm convergence and complexity analysis: Mathematical complexity analysis and theoretical convergence proof of the proposed algorithm are provided, verifying the computational efficiency and practical applicability of the optimization method.

[0056] The beneficial effects of the present invention are as follows:

[0057] 1. By optimizing caching and content distribution on different time scales, the present invention solves the balance problem between real-time response and long-term service in dynamic scenarios.

[0058] 2. Cooperative transmission strategy and multi-connection adaptability: The cooperative transmission strategy designed by the present invention enables requesters to be served by the same or different service providers (InPs). Through cross-service provider collaboration, the resource utilization rate in multi-connection scenarios is improved, thereby significantly enhancing transmission stability and flexibility.

[0059] 3. Using pre-caching technology to reduce network burden: In the long-term caching stage, a pre-caching strategy is adopted to reduce the dependence on the central network during peak hours, optimize resource allocation, and improve service quality.

[0060] 4. Improving the quality of service (QoS): Through the hybrid time-scale optimization strategy, the present invention can meet the requirements of real-time video stream transmission in the short term and optimize the cache distribution in the long term, thereby effectively improving the average QoS of in-vehicle video transmission.

[0061] 5. Enhancing transmission stability: Since the cooperative transmission scheme of the present invention allows requesters to obtain content from multiple service providers, it overcomes the service interruption problem caused by frequent user handovers, ensuring the continuity and stability of video streams; reducing network congestion and latency: The pre-caching strategy effectively reduces the burden on the central network during peak hours, thereby alleviating network congestion, improving the overall transmission efficiency, and reducing transmission latency.

[0062] 6. Optimization Algorithm Adapted to Dynamic Environment: Through algorithms such as Sample Average Approximation (SAA) and Successive Convex Approximation (SCA), the present invention takes into account both computational complexity and real-time performance in the optimization algorithm, ensuring fast response and adaptation to environmental changes in dynamic vehicle networking scenarios, and having high practicality.

[0063] In summary, by introducing an optimization model with a hybrid time scale, a cooperative transmission scheme, and a pre-caching technology, the present invention overcomes problems in the prior art such as poor multi-connection adaptability, unstable transmission, and excessive burden on the central network, significantly improving the quality of service and stability of video stream transmission in vehicle networking and being applicable to high-density dynamic scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a scenario diagram of a cooperative transmission system for vehicle networking based on NOMA;

[0065] Figure 2 It is a multi-time scale framework diagram;

[0066] Figure 3 It is a schematic diagram of the optimization process in the long-term content caching stage;

[0067] Figure 4 It is a schematic diagram of the optimization process in the short-term content distribution stage;

[0068] Figure 5 It is a schematic diagram of the overall process for optimizing the average quality of service by using the sample average approximation and successive convex approximation methods in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0069] The present invention will be described in detail below in conjunction with specific implementation cases. The following implementation cases will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form.

[0070] The present invention adopts a V2X (Vehicle-to-Everything) communication scheme with an auxiliary caching technology in an urban scenario with intensive vehicle requests. As Figure 1As shown in the figure, the scenario includes a macro base station MBS, several small base stations SBSs on the roadside, and vehicles equipped with caching functions. These facilities cooperate to provide services for numerous vehicle requesters. The macro base station MBS is connected to a content server storing a complete video file library. The V2X network covers multiple infrastructure providers InPs. Each infrastructure provider InP manages several small base stations SBSs and vehicle providers, and can share the bandwidth resources of this infrastructure provider InP. This bandwidth resource is orthogonal to the resources of other multiple infrastructure providers InPs, avoiding interference in the wireless frequency bands between different multiple infrastructure providers InPs. By sharing the resources of InPs, multiple InPs not only improve the system performance but also enhance the feasibility of multi-connection. Define the set of InPs as I’={1,2,…,I}, and the set of content providers belonging to the i-th infrastructure provider InPi is P i ’={1,2,…,q,…,Pi}. For the content providers of the i-th infrastructure provider InPi, the content providers from 1 to q are several small base stations SBSs, and the content providers from q + 1 to Pi are vehicle providers. In addition, there are R vehicle requesters with demands for video content, and the set of vehicle requesters is represented as R’={1,2,…,R}. To cope with the intensive connection demands, NOMA (Non-Orthogonal Multiple Access) technology is adopted to support multiple connections between vehicle requesters and content providers under the same InP. At the same time, vehicle requesters can also receive collaborative services from multiple content providers belonging to different InPs, representing the video content in the video file library.

[0071] In the NOMA-based assisted caching transmission network, caching policies and transmission policies are crucial for improving system performance. All content providers cooperate to update the content cache, aiming to improve the cache hit rate. At the same time, the transmission policy is decided based on the cache state and channel information to ensure meeting the needs of users. Given the high mobility of vehicles, the switching frequency of wireless connections is more frequent than the change in the popularity of video content. Therefore, the caching and transmission phases are updated on different time scales. The caching policy is updated in a long-term time interval and is divided into multiple short time slots T’={1,2,…,T} for content distribution update. Figure 2 shows the hybrid time scale framework of content caching and distribution, where t x，yDenote the y-th short time slot within the x-th long time scale, where each long time scale updates the content cache, and each short time slot is responsible for updating the distribution strategy, power allocation, and vehicle grouping. At the beginning of each long-term stage, the content provider fetches the video content to be cached from the remote server, and the cached content will not be updated until the next long-term stage. On this basis, based on the cache strategy optimized for the long-term stage, the transmission strategy, power allocation, and vehicle grouping are optimized in each short time slot to improve the transmission quality. However, future user requests and transmission link states affect the policy decision. To solve this anti-causal problem, the long-term caching stage is approximated by the average value method of random samples. Next, the models of content request, caching, distribution, link state, interference, and power allocation will be introduced in detail.

[0072] The embodiments of the present invention include the following steps:

[0073] Step S1: First, construct a V2X vehicle-to-everything network scenario in the computer based on NOMA, and then establish a transmission model of vehicles in the V2X vehicle-to-everything network scenario;

[0074] Step S2: Then, the processor in the computer constructs a quality-of-service model of vehicles based on the transmission model of vehicles, and then the processor establishes an optimization goal based on the quality-of-service model of vehicles;

[0075] Step S3: Finally, the processor processes the transmission model according to the optimization goal to obtain the optimal cache state, transmission state, and power allocation of each video content provider in the V2X vehicle-to-everything network scenario.

[0076] The V2X vehicle-to-everything network scenario in Step S1 is constructed based on non-orthogonal multiple access (NOMA) and includes a macro base station, several sub base stations, and vehicle providers equipped with caching functions. The sub base stations, vehicle providers, and the macro base station are all connected to a content server storing a complete video file library. The V2X vehicle-to-everything network covers several infrastructure providers. Each infrastructure provider manages several sub base stations and vehicle providers. The sub base stations and vehicle providers managed by the same infrastructure provider are content providers under this infrastructure provider. The content providers under the same infrastructure provider share the bandwidth resources of this infrastructure provider. The bandwidth resources of one infrastructure provider are orthogonal to the bandwidth resources of the other infrastructure providers. The content providers store video files and are used to provide the required video files for vehicle requesters (i.e., users).

[0077] The transmission model of vehicles in Step S1 mainly consists of a vehicle request cache model, a vehicle distribution model, and an in-vehicle network grouping model. Among them, the construction method of the vehicle request cache model is as follows:

[0078] In the V2X vehicle-to-everything scenario, the considered time slot is denoted as t. Each vehicle requester requests only one video file from the video file library in each time slot t. The probability that a video file is requested is obtained by processing according to the following formula:

[0079] po f =f –a / (∑ F n=1 n –a )

[0080] where po f represents the probability that the vehicle requester requests the video file f; f represents the requested video file; a represents the parameter of the vehicle requester's request concentration; F represents the total number of video files in the video file library; n represents the summation index variable;

[0081] The expressions and constraints of the cache state of the content provider in the V2X vehicle-to-everything scenario are as follows:

[0082] C p ={ c p 1 , c p 2 ,…, c p f ,…, c p F}, c p f ∈{0,1}

[0083] ∑ F f=1 c p f ≤C pmax

[0084] where C p represents the cache state of the content provider p; c p f represents the cache state. When the content provider p caches the video file f, then c p f =1. When the content provider p does not cache the video file f, then c p f =0; C pmax represents the cache capacity limit of the content provider p.

[0085] Specifically, the video files in the complete video file library are sorted in descending order according to the popularity of the video files. Generally, the popularity of the video files can be regarded as the requested probability po that the vehicle requester requests the video files f. Assume that each vehicle requester can only request one video file in each time slot. The popularity can be modeled by the Zipf distribution model, and po 1 >po 2 >…po F In the video content caching stage, content providers will pre-cache popular video files in their own cache space to reduce the load of the central server. However, due to the limitation of cache capacity, content providers can only cache a part of the files in the video file library.

[0086] The method for constructing the vehicle distribution model in step S1 is as follows: in a V2X vehicle networking scenario, constructing the transmission status of the video file between the content provider and the vehicle requester, and the transmission status between the content provider and the vehicle requester;

[0087] The constraints on the transmission status of the video file f between the content provider and the vehicle requester are as follows:

[0088] a i,p,r f (t) ≤c p f , c p f ∈{0,1}

[0089] ∑ I i=1 ∑ Pi p=1 a i,p,r f (t) ≤A max , f∈F,p∈Pi

[0090] ∑ R r=1 a i,p,r f ≤B i,p max , i∈I,f∈F, p∈Pi

[0091] Among them, a i,p,r f (t) represents the number of infrastructure providers InP in time slot t. i The transmission status of the video file f between the content provider p and the vehicle requester r under i The content provider p in the following situation transmits the cached video file f to the vehicle requester r, then a i,p,r f (t)=1, otherwise a i,p,r f (t) = 0; I represents the total number of infrastructure providers; Pi represents the i-th infrastructure provider InPi The total number of content providers under; R represents the total number of vehicle requesters; A max represents the upper limit of the number of content providers that can be associated with each vehicle requester simultaneously; B i,p max represents the infrastructure provider InP i the upper limit of the number of vehicle requesters that content provider p can serve under; a i,p,r f represents the transmission status of video file f between content provider p and vehicle requester r under the i-th infrastructure provider InP i during all time slots;

[0092] The constraints on the transmission status between content providers and vehicle requesters are as follows:

[0093] a i,p,r = max{ a i,p,r f}= min{∑ F f=1 a i,p,r f , 1}

[0094] max{ a i,p,r f (t)} ≤ a i,p,r (t) ≤ ∑ F f=1 a i,p,r f (t)

[0095] a i,p,r f (t) ∈ {0, 1}, a i,p,r (t) ∈ {0, 1}

[0096] where a i,p,r represents whether there is a transmission link between content provider p and vehicle requester r under the i-th infrastructure provider InP i during all time slots, regardless of whether a specific video file f is transmitted. If there is a transmission link, then a i,p,r = 1, otherwise a i,p,r = 0, a i,p,r (t) represents whether there is a transmission link between content provider p and vehicle requester r under the i-th infrastructure provider InP i during time slot t. If there is a transmission link, then a i,p,r (t) = 1, otherwise a i,p,r (t) = 0; max{} represents the maximum value function; min{} represents the minimum value function.

[0097] The NOMA distribution model of the vehicle in the present invention is based on NOMA technology. In the transmission stage, NOMA technology is adopted to meet the large-scale connection requirements. In the present invention, the NOMA grouping is only determined by the channel gain, and the NOMA grouping is dynamically adjusted based on the cache state, vehicle movement, and power allocation. And the NOMA grouping will affect the interference between different transmission links, thereby affecting the transmission efficiency. To avoid transmission waste and simplify the system complexity, the number of content providers transmitting the video file f to a single vehicle requester is limited. In addition, within each time slot t, the number of vehicle requesters that a single content provider can serve is also limited. In the proposed NOMA vehicle distribution model, each vehicle requester can be served by multiple vehicle providers and sub-base stations under different infrastructure providers. This vehicle distribution model makes full use of bandwidth resources and breaks through the boundaries between different network operators. At the same time, vehicle requesters can dynamically form NOMA groups according to their own content requirements and wireless channel states.

[0098] The construction method of the in-vehicle network grouping model in step S1 is as follows:

[0099] First, in the V2X vehicle-to-everything scenario, the data bit rate received by the vehicle requester r on the infrastructure provider InP within the time slot t is processed according to the following formula: i Continued from above:

[0100] R i,r (t) = W i × log 2 (1 + γ i,r (t))

[0101] γ i,r (t) = s i,r (t) / (I i,r IGI (t) + I i,r ICI (t) + σ i,r (t) 2 )

[0102] s i,r (t) = ∑ F f=1 po f ∑ p∈Pi a i,p,r f (t) × p i,p,r × h i,p,r

[0103] Among them, R i,r (t) represents the data bit rate received by the vehicle requester r on the infrastructure provider InP i ; W iDenote the \(i\)-th infrastructure provider InP i 's bandwidth resource; \(\gamma\) i,r (t) represents the transmission signal-to-noise ratio from the infrastructure provider InP i to the vehicle requester \(r\) within time slot \(t\); \(s\) i,r (t) represents the signal expectation received by the vehicle requester \(r\) via the bandwidth resource \(W\) i ; \(I\) i,r IGI (t) represents the in-group interference signal received by the vehicle requester \(r\) via the bandwidth resource \(W_i\); \(I\) i,r ICI (t) represents the inter-group interference signal received by the vehicle requester \(r\) via the bandwidth resource \(W\) i ; \(\sigma\) i,r (t) represents the noise power received by the vehicle requester \(r\) through the bandwidth resource \(W\) i ; \(p\) i,p,r Denote the power allocated by the content provider \(p\) under the \(i\)-th infrastructure provider InP i to the vehicle requester \(r\); \(h\) i,p,r Denote the channel gain allocated by the content provider \(p\) under the \(i\)-th infrastructure provider InP i to the vehicle requester \(r\);

[0104] Next, in the V2X vehicle networking scenario, the total data throughput received by the vehicle requester \(r\) within time slot \(t\) is processed according to the following formula:

[0105] \(R\) r (t)=\(\sum\) I i=1 \(R\) i,r (t) + \(r\) bs

[0106] \(r\) bs =\(\sum\) F f=1 \(p_o\) f (1 - min\{\(\sum\) i∈I \(\sum\) p∈Pi \(a\) i,p,r , 1\}\(v\) 0 )

[0107] \(R\) c \(\leq R\) r (t) \(\leq R\) max

[0108] where, \(R\) r (t) represents the total data bit rate (i.e., throughput) received by the vehicle requester \(r\) within time slot \(t\); \(r\) bs represents the data transmission bit rate dependent on the remote macro base station; \(v\) 0Denote the transmission rate provided by the remote macro base station to the vehicle requester; R c Denote the bit rate of the base layer of the SVC (Scalable Video Coding) video stream; R max Denote the bit rate corresponding to the highest video quality in the SVC video stream.

[0109] Specifically, in the NOMA-based cooperative transmission network, the vehicle requester uses the successive interference cancellation (SIC) technique to decode the required information and remove the interference signal. According to the principle of SIC technology, the vehicle requester within each NOMA group decodes and excludes the signals of the vehicle requesters with a lower decoding order than its own, and regards the signals of the vehicle requesters with a higher decoding order than itself as noise (i.e., the in-group interference IGI). At the same time, the signals of other vehicle requesters not in the same NOMA group as the vehicle requester are regarded as the inter-group interference ICI. The decoding order of the signal is jointly affected by the channel gain, the noise level, and the power allocation. As Figure 1 shown, the vehicles and the small base station SBS of the same color belong to the same infrastructure provider and share the bandwidth resources of this infrastructure provider. In each time slot t, the vehicle requesters sharing the same bandwidth resources are arranged in descending order according to the channel interference cancellation decoding order, that is, λ i,1 >λ i,2 >…λ i,r …>λ i,R where λ i,r represents the SIC decoding order of the vehicle requester r on the bandwidth resource W i . Assume that each content provider has complete user channel state information (CSI), and this user channel state information can be obtained through channel estimation technology:

[0110] The expression of the channel gain is as follows:

[0111] h i,p,r =G ×h i,p,r ’ ×β p,r ×d p,r –a0

[0112] where G represents the power gain factor brought by the amplifier and the antenna; h i,p,r ’~CN(0,1), h i,p,r ’ represents the complex zero-mean Gaussian variable of Rayleigh fading; β p,r follows the lognormal distribution, representing the shadow fading effect; d p,r represents the distance between the content provider p and the vehicle requester r within the time slot t, and a0 represents the path loss exponent. The position of the vehicle can be predicted based on the driving information and is regarded as unchanged during the short time slot.

[0113] According to the NOMA model, the conditions for vehicle requester r and vehicle requester r0 to belong to the same NOMA group are as follows:

[0114] ∑ Pi p=1 a i,p,r ×a i,p,r0 ≥1

[0115] Denote the set S of vehicle requesters that are in the same NOMA group as vehicle requester r and share bandwidth W i as follows: i,r defined as:

[0116] S i,r ={ r0∈R||∑ Pi p=1 a i,p,r ×a i,p,r0 ≥1}

[0117] where || represents the definition of a set, indicating that the element r0 contained in set S i,r satisfies the following condition: ∑ Pi p= 1 a i,p,r ×a i,p,r0 ≥1;

[0118] The set of in-group interference IGI of vehicle requester r with respect to set S i,r IGI is expressed as:

[0119] S i,r IGI ={ r0∈R||∑ Pi p=1 a i,p,r ×a i,p,r0 ≥1, λ i,r0 >λ i,r}

[0120] The set of inter-group interference ICI of vehicle requester r is denoted as S i,r ICI and is:

[0121] S i,r ICI ={ r0∈R||∑ Pi p=1 a i,p,r ×a i,p,r0 =0}

[0122] To calculate the received bit rate of each vehicle requester and achieve successful decoding, a series of constraints related to power allocation and signal reception must be satisfied. Here, p i,p,r (t) represents the infrastructure provider InP within time slot ti The power allocated by the content provider p to the vehicle requester r under the condition that the total power received by the connected vehicle requester complies with the following maximum power limit condition:

[0123] ∑ R r=1 p i,p,r (t) ≤P i,p ma

[0124] Where P i,p ma represents the maximum transmission power of the content provider p under the infrastructure provider InP i The expected signal received by the vehicle requester r is expressed as:

[0125] s i,r (t)= ∑ F f=1 po f ∑ p∈Pi c p f × a i,p,r f (t) × p i,p,r ×h i,p,r

[0126] = ∑ F f=1 po f ∑ p∈Pi a i,p,r f (t) × p i,p,r ×h i,p,r

[0127] The in-group interference IGI signal and the inter-group interference ICI signal received by the vehicle requester r are respectively expressed as:

[0128] I i,r IGI (t)= ∑ r0∈R , λ i,r0 >λ i,r ,min{∑ p∈Pi a i,p,r ×a i,p,r0 ,1}×∑ p∈Pi a i,p,r0 ×p i,p,r0 ×h i,p,,r

[0129] I i,r ICI (t)= ∑ r0∈R, r0≠r ,(1- min{∑ p∈Pi a i,p,r ×ai,p,r0 , 1}) × ∑ p∈Pi a i,p,r0 × p i,p,r0 × h i,p,,r

[0130] In the NOMA network environment with coordinated transmission, the decoding sequence is affected by various factors, resulting in a complex decision-making process for the NOMA network. To simplify the signal decoding process and enable the NOMA network to be determined only based on the channel gain, the following constraints must be satisfied to ensure that when λ i,j > λ i,r , the signal of vehicle requester j is decoded first:

[0131] min{∑ p∈Pi a i,p,r a i,p,r0 , 1} log 2 (1 + γ i,r ) ≤ log 2 (1 + S i,r,j VP (t) / (I i,r,j VP (t) + I i,r ICI (t) + σ i,r (t) 2 )

[0132] S i,r,j VP (t) = ∑ p∈Pi a i,p,r f × p i,p,r × h i,p,j

[0133] I i,r,j VP (t) = ∑ r0∈R , λ i,r0 > λ i,r , min{∑ p∈Pi a i,p,r × a i,p,r0 , 1} × ∑ p∈Pi a i,p,r0 × p i,p,r0 × h i,p,j

[0134] where γ i,r represents the transmission signal-to-noise ratio from the infrastructure provider InP i to vehicle requester r over all time slots; S i,r,j VP (t) represents the signal power of vehicle requester j in time slot t over the bandwidth resource W iThe decoded signal power received from the vehicle requester r; I i,r,j VP (t) represents the in-group interference IGI power of the vehicle requester j received from the vehicle requester r within the time slot t on the bandwidth resource W i The in-group interference IGI power of the vehicle requester r received from the vehicle requester r;

[0135] In addition, the signal-to-noise ratio SINR of the vehicle requester r using the infrastructure provider InP i is defined by the following formula:

[0136] γ i,r (t)= s i,r (t) / ( I i,r IGI (t)+ I i,r ICI (t)+ σ i,r (t) 2 )

[0137] Therefore, the corresponding received bit rate obtained by the vehicle requester r within the t-th time slot can be calculated. The data bit rate obtained by the vehicle requester r from the surrounding content providers is represented by R r (t). For requests for niche video content that are not cached locally, data has to be obtained from the remote base station BS. The data transmission bit rate in this case is identified by r bs .

[0138] Step S2 is specifically as follows:

[0139] Step S2.1, first, the expression of the quality of service model is obtained by processing according to the following formula:

[0140] QoS r (t)= a n ln(b n R r (t) / r r desir )

[0141] Among them, QoS r (t) represents the quality of service corresponding to the vehicle requester r within the time slot t; a n and b n both represent preset quality of service model parameters, a n >0, b n >0; r r desir represents the desired video data bit rate of the vehicle requester r;

[0142] In specific implementation, Quality of Service (QoS) is a key indicator for measuring the satisfaction of vehicle requesters (i.e., users) and the transmission effect. Considering the characteristics of video streams, QoS can be evaluated based on the received throughput. The value of the video data bit rate is based on the requested SVC-encoded video type. The present invention uses a logarithmic model to formally express the QoS perceived by the requester. When the throughput received by vehicle requester r is less than the expected video data bit rate, as the data throughput R r (t) increases, the QoS value of vehicle requester r r will increase significantly and rapidly. However, if the throughput received by vehicle requester r is greater than the expected video data bit rate, as the data throughput R r (t) increases, the growth of the QoS r value will become slow. This mechanism will promote the efficient utilization of infrastructure providers by more vehicle requesters, while avoiding over-allocation of resources to a few vehicle requesters. Therefore, this QoS model is applicable to designing wireless transmission strategies in practical scenarios.

[0143] Step S2.2: Then, taking the maximization of the average QoS of vehicle requesters as the optimization goal, based on the QoS model, the following formula is used to obtain the optimization goal, and the functional expression of the optimization goal is as follows:

[0144] QoSa’ = max{QoSa} = max{(1 / R)∑ R r=1 QoS r (t)}

[0145] where QoSa’ represents the maximum value of the average QoS, and QoSa represents the average QoS;

[0146] In the above formula, the maximum value of the average QoS is obtained by solving based on the SCA iterative algorithm. In the SCA iterative algorithm, the functional expression of the average QoS QoSa is as follows:

[0147] QoSa = (1 / T)(1 / R)∑ T t=1 ∑ R r=1 a n ln(b n (r r (g z ,d z ,e z ) + h r (k)) / r r desir )

[0148] -η(M(a z ) - N(az-1 ) - ▽ a N(a z-1 ) (a z –a z-1 )) - η(O(c z ) - P(c z-1 ) - ▽ c P(c z-1 ) (c z –c z -1 ))

[0149] Among them, T represents the total number of time slots; R represents the total number of vehicle requesters; r r (g z , d z , e z ) represents the data volume obtained by the vehicle user r in the z-th iteration through transmission by surrounding vehicles, that is, the objective function; h r (k) represents a preset linear function regarding the vector k; η represents a penalty parameter; M(a) represents the first function regarding the transmission state variable a i,p,r f of z ) represents the value of the function M(a) in the z-th iteration; N(a) represents the second function regarding the transmission state variable a i,p,r f of z ) represents the value of the function N(a) in the z-th iteration; O(c) represents the third function regarding the cache state variable c p f of z ) represents the value of the function O(c) in the z-th iteration; P(c) represents the fourth function regarding the cache state variable c p f of z ) represents the value of the function P(c) in the z-th iteration; ▽ a N(a) represents the gradient of the function N(a); ▽ a N(a z ) represents the value of the z-th iteration gradient ▽ a N(a); ▽ c P(c) represents the gradient of the function P(c); ▽ c P(c z ) represents the value of the z-th iteration gradient ▽ c P(c); The vector a represents the set of all transmission state variables a i,p,r f ; The vector c represents the set of all cache state variables c p f ; The vector k represents the set of all seventh variables k rf Set

[0150] In a specific implementation, the custom data volume r r (g z ,d z ,e z ) of the initial value r r (g 0 ,d 0 ,e 0 ) and use the initial value r r (g 0 ,d 0 ,e 0 ) as the basis for the first round of iteration. Subsequently, in each iteration, generate the function term function g r (x) according to the solution of the previous iteration. Then, by converting the objective function r r (g z ,d z ,e z ) into a convex function, obtain a new feasible solution (g z ,d z ,e z ) and the corresponding objective value.

[0151] In step S2, the expressions of the first function M(a), the second function N(a), the third function O(c), and the fourth function P(c) are as follows:

[0152] M(a)= ∑ I i=1 ∑ Pi p=1 ∑ R r=1 ∑ F f=1 a i,p,r f

[0153] N(a)= ∑ I i=1 ∑ Pi p=1 ∑ R r=1 ∑ F f=1 (a i,p,r f ) 2

[0154] O(c)= ∑ Pi p=1 ∑ F f=1 c p f

[0155] P(c)= ∑ Pi p=1 ∑ F f=1 (c p f ) 2

[0156] ▽ a N(a)= 2a i,p,r f

[0157] ▽ c P(c)= 2 c p f

[0158] where I represents the total number of infrastructure providers; Pi represents the total number of content providers under the i-th infrastructure provider InP i The following content provider; R represents the total number of vehicle requesters; F represents the total number of video files in the video file library; a i,p,r f represents the transmission status of the video file f between the content provider p and the vehicle requester r under the i-th infrastructure provider InP i over all time slots; c p f represents the cache status.

[0159] Specifically, step S3 is as follows: obtaining the maximum value of the average quality of service based on the SCA iterative algorithm, and then taking the cache status c of the content provider when the average quality of service reaches the maximum value p f , the transmission status a i,p,r f and the power allocation p i,p,r as the optimal vehicle network content caching and transmission scheme.

[0160] A vehicle network content caching and transmission optimization system for the method of the present invention:

[0161] Including a model construction module for constructing a transmission model of vehicles in a V2X vehicle network scenario and constructing a quality of service model of vehicles based on the transmission model;

[0162] Including an optimization objective construction module for establishing an optimization objective based on the quality of service model of vehicles;

[0163] Including an optimal solution processing module for processing the transmission model according to the optimization objective to obtain the cache status, transmission status, and power allocation scheme of each video content provider.

[0164] The objective of the present invention is to maximize the average quality of service QoSa of all vehicle requesters in the NOMA-based vehicle-to-everything (V2X) scenario. To achieve this objective, during T time slots, the cache placement C = [c p f

[0165] 、the content transmission policy A = { a i,p,r f (t), a i,p,r (t)} and the power allocation P = [p i,p,r will be jointly optimized. Considering that the content caching and transmission in any two consecutive long-term phases are independent of each other, the transmission optimization of the system will be performed once in each long-term time interval. For each long-term phase, the transmission optimization problem oriented to quality of service will be formulated as a problem of maximizing the average quality of service during T time slots, and the expression for maximizing the average quality of service is as follows:

[0166] max{ (1 / T) (1 / R) ∑ T t=1 ∑ R r=1 QoS r (t)}

[0167] where QoS r (t) represents the quality of service corresponding to the vehicle requester r in time slot t; without loss of generality, it is assumed that the speed of content cache update is much slower than that of content transmission and power allocation. In addition, to determine the transmission policy and power allocation, the cache information must be obtained first at the beginning of each time period. However, due to the uncertainty of future user requests and channel states in the continuous T time slots within this time period, the content caching problem cannot be directly solved. To handle this stochastic optimization problem with a hybrid time scale, the sample average approximation (SAA) method is adopted to estimate the future expected value of the average quality of service. Therefore, the entire optimization process is divided into two stages: the long-term content caching stage and the short-term content distribution stage.

[0168] I. Long-term content caching stage

[0169] In the long-term content caching stage, both the popularity of videos and the diversity of caches need to be considered. Caching the same popular video content among content providers can ensure multiple transmission sources, but it will affect the cache diversity. On the contrary, if completely different video files are cached among surrounding content providers, although it can improve the cache diversity and cache hit rate, it will lead to a reduction in the number of content providers serving users simultaneously. Therefore, under the limitation of cache capacity, formulating a cache strategy requires a trade-off between cache diversity and transmission diversity.

[0170] ​There is a close interrelationship among content caching, transmission strategy, and power allocation, and they cannot be considered separately. However, at the beginning of each large time interval, the user preferences and channel information within the consecutive T time slots of this interval cannot be obtained. To solve this problem, the sample average approximation (SAA) method is adopted, and the cached content is decided by averaging multiple random samples, as Figure 3 shown. The key of SAA is to regard the average quality of service (QoSa) as a random sample and approximate the long-term performance by taking the average of T0 random samples. Therefore, the content caching strategy is determined by the following formula:

[0171] max{ (1 / T0) (1 / R) ∑ T0 t0=1 ∑ R r=1 QoS r (t0)}

[0172] QoSa =(1 / R)∑ R r=1 QoS r (t)

[0173] (1 / T0) (1 / R) ∑ T0 t0=1 ∑ R r=1 QoS r (t0)≈E(QoSa)

[0174] where T0 represents the total number of samples; t0 represents the sample ordinal number; E( ) represents the mathematical expectation function; and the total number of samples T0 must be large enough to meet the accuracy of the mathematical expectation approximation of the average quality of service. The content transmission strategy and power allocation will also be optimized together with the random samples, but they are not the actual transmission results pursued ultimately.

[0175] II. Short-Term Content Distribution Phase

[0176] In the short-term content distribution phase, based on the content caching decision obtained in the caching phase, the distribution strategy and power allocation are jointly optimized in the actual transmission scenario. Given the high mobility of the vehicle network, the transmission connections between vehicles are updated rapidly with the change of location in each time slot. Therefore, to improve the system performance, the transmission links and vehicle groups will be adjusted according to the instant transmission information. As Figure 4 shown, the content transmission strategy and power allocation will be updated in each time slot t and optimized by the following formula:

[0177] max{(1 / R)∑ R r=1 QoS r (t)}

[0178] The above formulas are classified as mixed-integer non-linear programming MINLP problems, which are difficult to solve. These two formulas cannot be directly solved by conventional optimization methods. To find a solution within an acceptable computational time, a series of transformation processes are required. The present invention proposes an iterative algorithm based on SCA (successive convex approximation) to find the optimal solution. Specifically, first, the original problem needs to be equivalently transformed into the form of a difference-of-convex algorithm DC problem. Then, the SCA algorithm linearizes the second term in the difference-of-convex DC problem through Taylor expansion, thereby obtaining a convex objective function and constraint conditions. As this SCA-based iterative method continues, the algorithm gradually converges to a local optimal solution.

[0179] The optimization problem of the average quality of service QoSa has been equivalently transformed into a tractable form, and the specific objective function is shown as follows:

[0180] max{(1 / T) (1 / R) ∑ T t=1 ∑ R r=1 QoS r (t)}

[0181] Since the constraints are still non-convex, the problem still has complexity. A penalty parameter η>>1 can be used to incorporate the constraints into the objective function. Furthermore, in each iteration, a tractable convex problem is obtained, which can be efficiently solved in polynomial time by using the interior point method IPM of the standard solver CVX.

[0182] To standardize the expression form of the formula, the following variables are introduced:

[0183] The first variable β i,r, r0 = min{∑ p∈Pi a i,p,r ×a i,p,r0 ,1}; The second variable θ i,p,r, r0 = a i,p,r ×a i,p,r0 ;

[0184] The third variable c i,p,r = a i,p,r ×p i,p,r ; The fourth variable e i,p, r, r0 =(1-β i,r, r0 ) c i,p,r ;

[0185] The fifth variable d i,p,r, r0 =β i,r, r0 ×c i,p,r ; The sixth variable g i,p,r f = a i,p,rf ×p i,p,r

[0186] The seventh variable k r f = min{∑ i∈I ∑ p∈Pi a i,p,r f , 1};

[0187] Correspondingly, the in - group interference signal IGI and the inter - group interference signal ICI received by the vehicle requester r are reformulated as:

[0188] I i,r IGI (t)= ∑ r0∈R , λ i,r0 >λ i,r , ∑ p∈Pi d i,p,r, r0 ×h i,p,,r

[0189] I i,r,j VP (t)= ∑ r0∈R , λ i,r0 >λ i,r , ∑ p∈Pi d i,p,r, r0 ×h i,p,j

[0190] s i,r (t)= ∑ F f=1 po f ∑ p∈Pi g i,p,r f ×h i,p,r

[0191] By introducing substitution variables, the optimization problem P’(P, A, C) can be reconstructed as:

[0192] max{(1 / T)(1 / R)∑ T t=1 ∑ R r=1 a n ln(b n R r (t) / r r desir )}

[0193] -η∑ I i=1 ∑ Pi p=1 ∑ R r=1 ∑F f=1 (a i,p,r f - (a i,p,r f ) 2 )- ∑ Pi p=1 ∑ F f=1 (c p f - (c p f ) 2 )}

[0194] R r (t)= f r (x)- g r (x)+ h r (k)

[0195] h r (k)= ∑ F f=1 po f (1-k r f ) v 0

[0196] Among them, T represents the total number of time slots; R represents the total number of vehicle requesters; Pi represents the total number of content providers under the i-th infrastructure provider InP i The following content; F represents the total number of video files in the video file library; η represents the penalty parameter; R r (t) represents the total data bit rate received by vehicle requester r within time slot t; r r desir represents the video data bit rate expected by vehicle requester r; f r (x) represents the first convex function; g r (x) represents the second convex function; h r (k) represents a linear function; k r f represents the seventh variable; v 0 represents the transmission rate provided by the remote macro base station to the vehicle requester;

[0197] The functional expression of the optimization problem P’(P,A,C) is as follows:

[0198] max{(1 / T)(1 / R)∑ T t=1 ∑ R r=1 a n ln(b n H(g,d,e,k) / rr desir )}

[0199] -η∑ I i=1 ∑ Pi p=1 ∑ R r=1 ∑ F f=1 (a i,p,r f - (a i,p,r f ) 2 )- ∑ Pi p=1 ∑ F f=1 (c p f - (c p f ) 2 )}

[0200] Among them, H(g, d, e, k) represents a convex function, which is a convex function with respect to the variables d, e, and g; η represents a penalty parameter.

[0201] To convert the objective function H(g, d, e, k) into a convex function, for any feasible solution d of the variable z-1 , e z-1 , the function g r (x) can be approximated by the first-order Taylor expansion. In addition, after applying the first-order Taylor approximation, the functions and some constraints in the objective can be regarded as convex functions. For the determined points g z-1 , d z-1 , e z-1 , a z-1 , c z-1 , by solving the following convex optimization problem, the upper bound of the objective function can be obtained.

[0202] The expressions of the above functions are as follows:

[0203] r r (g z , d z , e z ) = ∑ I i=1 f i,r (g z , d z , e z ) - g i,r (g z , d z , e z )

[0204] -▽d g i,r (d z-1 ,e z-1 )(d z –d z-1 )-▽ e g i,r (d z-1 ,e z-1 )(e z -e z-1 )

[0205] f i,r (g,d,e)= W i ×log 2 (I i,r IGI (t)+ I i,r ICI (t)+ σ i,r (t) 2 + s i,r (t))

[0206] g i,r (d,e)= W i ×log 2 (I i,r IGI (t)+ I i,r ICI (t)+ σ i,r (t) 2 )

[0207] If λ i,r0 > λ i,r , p ∈ Pi, ▽ d g i,r= h i,p,r / ((ln2)(I i,r IGI (t)+ I i,r ICI (t)+ σ i,r (t) 2 ))

[0208] Otherwise, ▽ d g i,r = 0

[0209] If r0 ∈ R, r0 ≠ r, p ∈ Pi, ▽ e g i,r= h i,p,r / ((ln2)(I i,r IGI (t)+ I i,r ICI (t)+ σ i,r (t) 2))

[0210] Otherwise, ▽ e g i,r = 0

[0211] H(g, d, e, k) = f r (g, d, e) - g r (g, d, e) + h r (k)

[0212] f r (g, d, e) = ∑ I i=1 f i,r (g, d, e)

[0213] g r (g, d, e) = ∑ I i=1 g i,r (g, d, e)

[0214] Where r r (g, d, e) represents the second objective function; r r (g z , d z , e z ) represents the value of the z-th iteration function r r (g, d, e); f i,r (g, d, e) represents the first iteration function, f i,r (g z , d z , e z ) represents the value of the z-th iteration function f i,r (g, d, e); g i,r (d, e) represents the fourth convex function, g i,r (g z , d z , e z ) represents the value of the z-th iteration function g i,r (d, e); ▽ d g i,r represents the first gradient function; ▽ e g i,r represents the second gradient function;

[0215] In the above formula, the average quality of service is obtained by solving the SCA iterative algorithm. In the SCA iterative algorithm, the function expression of the average quality of service is as follows:

[0216] QoSa = (1 / T)(1 / R)∑ T t=1 ∑ R r=1 an ln(b n (r r (g z ,d z ,e z )+ h r (k)) / r r desir )

[0217] -η(M(a z )- N(a z-1 )-▽ a N(a z-1 ) (a z –a z-1 ))-η(O(c z )- P(c z-1 )-▽ c P(c z-1 ) (c z –c z -1 ))

[0218] Among them, r r (g z ,d z ,e z ) represents the amount of data obtained by vehicle user r through transmission by surrounding vehicles in the z-th iteration, that is, the objective function; h r (k) represents the amount of data transmitted by the base station and is also a linear function of k; M(a) represents a function of the transmission state variable a i,p,r f ; M(a z ) represents the value of M(a) in the z-th iteration; N(a) represents a quadratic function of the transmission state variable a i,p,r f ; N(a z ) represents the value of N(a) in the z-th iteration; O(c) represents a function of the cache state variable c p f ; O(c z ) represents the value of O(c) in the z-th iteration; P(c) represents a quadratic function of the cache state variable c p f ; P(c z ) represents the value of P(c) in the z-th iteration; ▽ a N(a) represents the gradient of the function N(a); ▽ c P(c) represents the gradient of the function P(c); The vector a represents the set of all transmission state variables a i,p,r f , and the vector c represents the set of all cache state variables c p fset, vector k represents all seventh variables k r f set.

[0219] From a feasible solution r of the optimization problem r (g 0 ,d 0 ,e 0 ) as the initial value to start iteration, and the feasible solution r generated in each round of iteration r (g z+1 ,d z+1 ,e z+1 ) is based on the feasible solution of the previous round.

[0220] To solve this complex joint optimization problem, an SCA-based iterative strategy is adopted to gradually approximate the result. In each round of iteration, the original problem is transformed into a convex problem that can be solved by traditional convex optimization algorithms such as the interior point method. It should be noted that the solution obtained through optimization is an upper bound of the solution of the initial problem. To further improve the suboptimal solution, the algorithm will continue to iterate until the preset stopping condition is met. To handle the transformed difference-of-convex (DC) approximation problem, the low-complexity algorithm applied is as follows: As Figure 5 shown, first, a set of feasible solution sets is initialized as the basis for the first round of iteration. Subsequently, in each iteration, the functional term function g r (x) and its corresponding gradient function are generated according to the solution of the previous iteration. Then, by transforming the objective function into a convex function, a new feasible solution (g z ,d z ,e z ) and the corresponding objective value are obtained, and this result is either improved or remains unchanged compared with the previous iteration. This process will continue until the convergence condition is met.

[0221] The SCA-based iterative algorithm proposed by the present invention has a polynomial time complexity. The optimization problem in each round of iteration is a convex problem and can be solved by the interior point method, with a computational complexity of O(q 3.5 ), where q represents the total number of variables in the problem. For the optimization of the short-term content distribution strategy, the time complexity of each iteration is O((3IPRR + FIPR + IRR + IPR + RF) 3.5 ), where q = 3IPRR + FIPR + IRR + IPR + RF is the number of variables. Similarly, in the process of long-term content caching optimization, the caching variable c p f is trained during T time slots, and the number of variables is (3IPRR + FIPR + IRR + IPR + RF + PF)T. Generally speaking, the SCA-based iterative algorithm provides a suboptimal solution that approximates the global optimal solution for the optimization problem.

[0222] Starting from a feasible solution \((g 0 , d 0 , e 0 ) of the optimization problem as the initial value, the feasible solution \((g z+1 , d z+1 , e z+1 ) generated in each round of iteration is based on the feasible solution of the previous round. This iterative process will continue until the set maximum number of iterations is reached or the predetermined convergence criterion is satisfied. In this process, to prove the convergence of the iterative algorithm based on SCA, the lower bound of the algorithm is derived as follows:

[0223] r i,r = f i,r (g z , d z , e z ) - g i,r (g z , d z , e z )

[0224] ≥f i,r (g z , d z , e z ) - g i,r (d z-1 , e z-1 ) - ▽ d g i,r (d z-1 , e z-1 )(d z – d z-1 ) - ▽ e g i,r (d z-1 , e z-1 )(e z - e z-1 )

[0225] = max g,d,e f i,r (g, d, e) - g i,r (d, e) - ▽ d g i,r (d z-1 , e z-1 )(d z – d z-1 ) - ▽ e g i,r (d z-1 , e z-1 )(e z - e z-1 )

[0226] ≥ f i,r (g z-1 , d z-1 , e z-1 ) - g i,r (d z-1 , e z-1 ) - ▽ d g i,r (d z-1 , e z-1 )(d z – d z-1 ) - ▽ e g i,r (d z-1 , e z-1 )(e z - e z-1 )

[0227] = f i,r (g z-1 , d z-1 , e z-1 ) - g i,r (g z-1 , d z-1 , e z-1 )

[0228] where r i,r represents the difference between the convex function f i,r (g, d, e) and the convex function g i,r (d, e); therefore, after each iteration, compared with the previous result, the average quality of service QoSa either improves or remains unchanged. The feasible solution (g z , d z , e z ) after iteration will at least converge to a local optimal solution.

[0229] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A NOMA-based multi-time-scale content caching and transmission optimization method for Internet of Vehicles, characterized in that: The following steps are involved: Step S1: First, a V2X vehicle networking scenario is constructed in a computer based on NOMA, and then a vehicle transmission model is established in the V2X vehicle networking scenario; The V2X vehicle networking scenario of step S1 is constructed based on non-orthogonal multiple access NOMA, including a macro base station, several sub-base stations and a vehicle provider equipped with a cache function. The macro base station is connected to a content server storing a video file library. The V2X vehicle networking network covers several infrastructure providers, each of which manages several sub-base stations and vehicle providers. The sub-base stations and vehicle providers managed under the same infrastructure provider serve as content providers under the infrastructure provider. The vehicle transmission model in step S1 mainly consists of a vehicle request cache model, a vehicle distribution model and a vehicle network grouping model. The vehicle request cache model is used to represent the cache status of the content provider, the vehicle distribution model is used to represent the transmission status of the video file between the content provider and the vehicle requester, and the vehicle network grouping model is used to represent the data bit rate received by the vehicle requester on the infrastructure provider; The expression of the vehicle request cache model in step S1 is as follows: C p ={ c p 1 , c p 2 ,…, c p f ,…, c p F }, c p f ∈{0,1} ∑ F f=1 c p f ≤C pmax Among them, C p Indicates the cache status of content provider p; c p f Indicates the cache status. When content provider p caches video file f, then c p f =1, when content provider p does not cache video file f, then c p f =0; C pmax represents the cache capacity limit of content provider p; f represents the requested video file; F represents the total number of video files in the video file library; The expression of the vehicle distribution model is as follows: a i,p,r f (t) ≤c p f ∑ I i=1 ∑ Pi p=1 a i,p,r f (t) ≤A max ∑ R r=1 a i,p,r f ≤B i,p max Among them, a i,p,r f (t) represents the number of infrastructure providers InP in time slot t. i The transmission status of the video file f between the content provider p and the vehicle requester r under i The content provider p in the following situation transmits the cached video file f to the vehicle requester r, then a i,p,r f (t)=1, otherwise a i,p,r f (t) = 0; I represents the total number of infrastructure providers; Pi represents the i-th infrastructure provider InP i The total number of content providers under; R represents the total number of vehicle requesters; A max Indicates the upper limit of the number of content providers that can be associated with each vehicle requester at the same time; B i,p max InP stands for Infrastructure Provider i The upper limit of the number of vehicle requesters that the content provider p can serve; a i,p,r f represents the number of infrastructure providers InP in all time slots i The transmission status of the video file f between the content provider p and the vehicle requester r under The expression of the vehicle network grouping model is as follows: R r (t)= ∑ I i=1 R i,r (t)+ r bs R i,r (t)= W i ×log2(1+γ i,r (t)) γ i,r (t)= s i,r (t) / ( I i,r IGI (t)+ I i,r ICI (t)+ σ i,r (t) 2 ) s i,r (t)= ∑ F f=1 po f ∑ p∈Pi a i,p,r f (t) × p i,p,r ×h i,p,r R c ≤R r (t) ≤R max Among them, R r (t) represents the total data bit rate received by vehicle requester r in time slot t; r bs represents the data transmission bit rate that depends on the remote macro base station; R c Indicates the bit rate of the base layer of the SVC video stream; R max Indicates the bit rate corresponding to the highest video quality in the SVC video stream; R i,r (t) represents the vehicle requester r in the infrastructure provider InP in time slot t i The data bit rate received on i represents the i-th infrastructure provider InP i Bandwidth resources; γ i,r (t) represents the number of nodes from the infrastructure provider InP in time slot t. i The transmission signal-to-noise ratio between the vehicle requester r and s i,r (t) represents the bandwidth resource W used by the vehicle requester r in time slot t i Received signal expectation; I i,r IGI (t) represents the intra-group interference signal received by the vehicle requester r via the bandwidth resource Wi in time slot t; i,r ICI (t) represents the bandwidth resource W used by the vehicle requester r in time slot t i The inter-group interference signal received; σ i,r (t) represents the bandwidth resource W used by the vehicle requester r in time slot t i Received noise power; p i,p,r represents the i-th infrastructure provider InP i The power allocated by content provider p to vehicle requester r under i,p,r represents the i-th infrastructure provider InP i The channel gain allocated by the content provider p to the vehicle requester r under f represents the probability that the vehicle requester requests the video file f; Step S2: Next, the processor in the computer constructs a service quality model of the vehicle based on the transmission model of the vehicle, and then the processor establishes an optimization target based on the service quality model of the vehicle; The step S2 is specifically as follows: Step S2.1: First, the expression of the service quality model is obtained according to the following formula: QoS r (t)= a n ln(b n R r (t) / r r desir ) Among them, QoS r (t) represents the service quality corresponding to the vehicle requester r in time slot t; a n and b n Both represent the preset service quality model parameters; r r desir R represents the video data bit rate expected by vehicle requester r; r (t) represents the total data bit rate received by vehicle requester r in time slot t; Step S2.2: Next, maximizing the average service quality of vehicle requesters is taken as the optimization goal. The optimization goal is obtained according to the following formula based on the service quality model. The function expression of the optimization goal is as follows: QoSa'=max{ QoSa}= max{(1 / R)∑ R r=1 QoS r (t)} Among them, QoSa' represents the maximum value of the average service quality, and QoSa represents the average service quality; In the above formula, the maximum value of the average service quality is obtained based on the SCA algorithm. In the SCA algorithm, the function expression of the average service quality QoSa is as follows: QoSa=(1 / T) (1 / R) ∑ T t=1 ∑ R r=1 a n ln(b n (r r (g z ,d z ,e z )+ h r (k)) / r r desir ) -η(M(a z )- No z-1 )–▽ a No z-1 ) (to z -to z-1 ))-η(O(c z )- P(c z-1 )–▽ c P(c z-1 ) (c z –c z-1 )) Where T represents the total number of time slots; R represents the total number of vehicle requesters; r r (g z ,d z ,e z ) represents the amount of data obtained by vehicle user r in the zth iteration through transmission from surrounding vehicles; h r (k) represents the preset linear function; η represents the penalty parameter; M(a) represents the transmission state variable a i,p,r f The first function of M(a z ) represents the value of the z-th iteration function M(a); N(a) represents the value of the transmission state variable a i,p,r f The second function of z ) represents the value of the z-th iteration function N(a); O(c) represents the value of the cache state variable c p f The third function of O(c z ) represents the value of the function O(c) at the zth iteration; P(c) represents the value of the cache state variable c p f The fourth function of P(c z ) represents the value of the z-th iteration function P(c); ▽ a N(a) represents the gradient of function N(a); a N(a z ) represents the zth iteration gradient ▽ a The value of N(a); c P(c) represents the gradient of the function P(c); ▽ c P(c z ) represents the zth iteration gradient ▽ c The value of P(c); vector a represents all transmission state variables a i,p,r f The vector c represents all cache state variables c p f A collection of; In step S2, the expressions of the first function M(a), the second function N(a), the third function O(c), and the fourth function P(c) are as follows: M(a)= ∑ I i=1 ∑ Pi p=1 ∑ R r=1 ∑ F f=1 a i,p,r f N(a)= ∑ I i=1 ∑ Pi p=1 ∑ R r=1 ∑ F f=1 (a i,p,r f ) 2 O(c)= ∑ Pi p=1 ∑ F f=1 c p f P(c)= ∑ Pi p=1 ∑ F f=1 (c p f ) 2 ▽ a N(a)= 2a i,p,r f ▽ c P(c)= 2 c p f Where I represents the total number of infrastructure providers; Pi represents the number of infrastructure providers of the ith infrastructure provider InP i The total number of content providers under the same file system; R represents the total number of vehicle requesters; F represents the total number of video files in the video file library; a i,p,r f represents the number of infrastructure providers InP in all time slots i The transmission status of the video file f between the content provider p and the vehicle requester r under p f Indicates the cache status; Step S3: Finally, the processor processes the transmission model according to the optimization target to obtain the optimal solution for cache status, transmission status and power allocation of each video content provider in the V2X vehicle networking scenario.

2. According to claim 1, a NOMA-based multi-time scale content caching and transmission optimization method for Internet of Vehicles, characterized in that: The content providers under the same infrastructure provider share the bandwidth resources of the infrastructure provider. The bandwidth resources of one infrastructure provider are orthogonal to the bandwidth resources of other infrastructure providers. The content provider stores video files and is used to provide the required video files to the vehicle requester.

3. According to claim 1, a NOMA-based multi-time scale content caching and transmission optimization method for Internet of Vehicles, characterized in that: In step S2, the average service quality of the vehicle requester is taken as the optimization target, and the specific steps of step S3 are: obtaining the maximum value of the average service quality, and then taking the cache state, transmission state and power allocation of the content provider when the average service quality reaches the maximum value as the optimal Internet of Vehicles content cache and transmission solution.

4. A vehicle networking content caching and transmission optimization system for implementing any of the methods described in claims 1-3, characterized in that: The system comprises a model building module for building a transmission model of a vehicle in a V2X vehicle networking scenario, and building a service quality model of the vehicle based on the transmission model; It includes an optimization target building module for establishing optimization targets based on a vehicle's service quality model; It comprises an optimal solution processing module, which is used to process the transmission model according to the optimization target to obtain the cache state, transmission state and power allocation solution of each video content provider.

5. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the content caching and transmission optimization method described in any one of claims 1 to 3 are implemented.

Images