A joint clustering, offloading and resource optimization method in ultra-dense cache networks

Abstract

The present invention discloses a method for optimizing joint clustering, offloading and resources in an ultra-dense cache network. The method comprises: obtaining basic network information of an ultra-dense Internet of Things, constructing a network system in which a clustered cache server provides auxiliary computing offloading and file downloading, and constructing an optimization problem for minimizing system delay under constraints; obtaining an initial solution according to the optimization problem and defining it as a parent population, using an improved water wave energy attenuation algorithm to adaptively explore and develop the parent population, and finally obtaining a global optimal individual; performing pre-caching, computing offloading and file downloading according to the position of the global optimal individual, satisfying the cache server clustering and cache capacity constraints, and achieving the goal of minimizing system delay.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communications, and in particular relates to a method for optimizing joint clustering, unloading and resources in an ultra-dense cache network. Background Art

[0002] With the development of mobile networks, many computationally intensive, data-intensive, and delay-sensitive applications have emerged. These applications generate a large number of computing tasks and file request tasks. In order to ensure that users have a high-quality experience, mobile devices urgently need more computing, storage, and communication resources to perform tasks.

[0003] Mobile edge computing is an effective solution to the above problems, which meets these requirements by deploying edge servers at the edge of the network. Compared with cloud computing, mobile edge computing can achieve lower latency and more flexible services. On this basis, ultra-dense mobile edge networks combine mobile edge computing and heterogeneous cellular networks to provide users with a large amount of computing resources and cache resources by deploying a large number of low-power small base stations.

[0004] In order to further utilize the resources of small base stations to alleviate communication pressure, some studies have proposed configuring cache servers in small base stations, and reducing upload time to improve computing offloading efficiency by pre-caching computing tasks, which is called an ultra-dense cache network. This solution still has certain problems. First, in reality, users not only have to transmit uplink to perform computing tasks, but also download files downlink, but related research often only focuses on one of the two situations without considering the situation where both are performed at the same time; second, a large number of small base stations often pre-cache the same files, which causes a waste of storage resources; third, computing tasks and files involve two different types of services, but both require the use of small base station cache resources, which forms a competition between the two. Reasonable allocation of cache resources is a complex optimization problem. In addition, the water wave optimization algorithm used in the prior art has structural problems, which cannot balance the exploration and development process well. There is a problem of only linear search in the solution process, which makes it difficult to solve optimization problems in complex scenarios. Summary of the invention

[0005] In view of this, the present invention provides a method for joint clustering, offloading and resource optimization in an ultra-dense cache network to solve the problem in the prior art of algorithm performance degradation caused by a lack of research on new problems arising from cache server-assisted computing offloading.

[0006] A first aspect of the present invention provides a method for joint clustering, offloading and resource optimization in an ultra-dense cache network, comprising:

[0007] Step S1: Obtain basic network information of the ultra-dense cache network, build a network system based on the basic network information, the network system includes a communication model, a cache model, and a calculation model, and build an optimization problem under the constraints of cache server clustering and cache capacity, and the optimization problem is a system delay minimization problem;

[0008] Step S2: Obtain an initial solution according to the optimization problem, and define the initial solution as the parent population. Use the water wave energy attenuation algorithm to search the parent population, and finally output the position of the global optimal individual, so as to obtain the solution with the minimum system delay. The process of searching the parent population by the water wave energy attenuation algorithm includes: initializing the population and determining the global optimal individual; under the control of water wave energy, adaptively performing improved propagation operations and refraction operations, wherein the propagation operation performs a coarse-grained search on the population, and the refraction operation performs a fine-grained search on the population; find the optimal individual of the current population, and perform the improved breaking wave operation on it to help the algorithm jump out of the local optimum;

[0009] Step S3: Perform joint clustering, offloading and resource optimization configuration of the ultra-dense cache network according to the position of the global optimal individual.

[0010] Furthermore, the step S2 includes the following sub-steps:

[0011] Step S21: Initialize the maximum number of iterations of the water wave energy attenuation algorithm , and set the current iteration number t to 1;

[0012] Step S22: Encode individual water waves and define individual water wave wavelengths for calculation of propagation operations;

[0013] Step S23: Initialize the population and construct water wave individuals Fitness function: Use the fitness function to calculate the fitness values ​​of all water wave individuals in the population, and take the individual with the highest fitness value as the global best individual;

[0014] Step S24: Determine whether the current number of iterations t is less than or equal to the maximum number of iterations , if the current number of iterations t is less than or equal to the maximum number of iterations , then based on the water wave energy, the water wave population is adaptively propagated or refracted, and then the wave breaking operation is performed on the best individual of the current population to obtain the target population; if the current iteration number t is greater than the maximum iteration number , then output the encoding of the global best individual.

[0015] Furthermore, the step S24 includes the following sub-steps:

[0016] Step S241: Calculate the energy of the water wave population under the current iteration number, generate a random number between 0 and 1 and compare it with the water wave energy. If the random number is less than or equal to the water wave energy, perform a propagation operation, i.e., step S242; if the random number is greater than the water wave energy, perform a refraction operation, i.e., step S243;

[0017] Step S242: updating the water wave individuals with the given propagation rule, and performing adaptive perturbation on the updated individuals, that is, first performing diversity-guided perturbation, and then performing adaptive perturbation according to the adaptive probability;

[0018] Step S243: updating the individual water wave using the given refraction rule, and updating the individual wavelength of the water wave according to the individual fitness value before and after refraction;

[0019] Step S244: after step S241 is executed, the water wave wavelength and breaking wave coefficient are updated, the fitness value of the water wave population is calculated using the fitness function, the best individual of the current population is found, and the breaking wave operation is performed on it;

[0020] Step S245: Compare the fitness values ​​of the best individual in the current population with the best individual in the world. If the fitness value of the former is greater than that of the latter, the best individual in the current population is used to update the best individual in the world. Otherwise, the best individual in the world is retained.

[0021] Step S246: Set the current number of iterations The value of is increased by 1, and the process returns to step S24, and the iteration is repeated until the iteration ends. Finally, the globally best individual is retained as the optimal solution to the problem.

[0022] The second aspect of the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the terminal device, wherein the processor implements the above-mentioned method for joint clustering, unloading, and resource optimization in an ultra-dense cache network when executing the computer program.

[0023] A third aspect of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned method for joint clustering, offloading and resource optimization in an ultra-dense cache network.

[0024] The present invention searches by utilizing an improved water wave energy attenuation algorithm. The water wave energy is used to control the adaptive execution of propagation and refraction operations, and the breaking wave operation is performed on the global optimal individual, which promotes the reasonable allocation of algorithm exploration and development processes, enhances the ability to solve complex optimization problems, greatly improves the performance of the method of the present invention, and can well achieve the goal of minimizing system delay. Specifically, an optimization problem is constructed based on the basic information of the ultra-dense mobile edge network, and the initial solution of the optimization problem is obtained. The solution is used as the parent population, and the improved water wave energy attenuation algorithm is used to iteratively search the parent population, and finally the position of the global optimal individual is obtained; cluster cache-assisted computational offloading and resource optimization configuration are performed according to the position of the global optimal individual. The present invention jointly optimizes user association, sub-channel selection, and task pre-caching to minimize system delay. Under the constraints of proportional allocation of computing resources, clustered deployment of cache servers, and cache capacity, this method can well achieve the goal of minimizing system delay. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0026] Figure 1 It is a flow chart of an implementation method of a combined clustering, offloading and resource optimization method in an ultra-dense cache network provided by an embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of the effect of the number of network users on the fitness value in the water wave optimization algorithm;

[0028] Figure 3 It is a schematic diagram of the effect of the number of network users on the fitness value in the water wave energy attenuation algorithm and the adaptive genetic algorithm;

[0029] Figure 4 A schematic diagram showing the effect of the number of network users on the average delay of a computing-sensitive mobile terminal disclosed by the present invention;

[0030] Figure 5 The present invention discloses a schematic diagram of the impact of the number of network users on the average delay of high-speed mobile terminals. DETAILED DESCRIPTION

[0031] The embodiments of the present invention are described in detail below. The embodiments described with reference to the accompanying drawings are exemplary. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0032] See also Figure 1This embodiment provides a method for optimizing clustering, offloading, and resources in an ultra-dense cache network, including the following steps:

[0033] Step S1: Obtain basic network information of the ultra-dense cache network, build a network system based on the basic network information, the network system includes a communication model, a cache model, and a calculation model, and build an optimization problem under the constraints of cache server clustering and cache capacity, and the optimization problem is a system delay minimization problem;

[0034] Step S2: Obtain an initial solution according to the optimization problem, and define the initial solution as the parent population. Use the water wave energy attenuation algorithm to search the parent population, and finally output the position of the global optimal individual, so as to obtain the solution with the minimum system delay. The process of searching the parent population by the water wave energy attenuation algorithm includes: initializing the population and determining the global optimal individual; under the control of water wave energy, adaptively performing improved propagation operations and refraction operations, wherein the propagation operation performs a coarse-grained search on the population, and the refraction operation performs a fine-grained search on the population; find the optimal individual of the current population, and perform the improved breaking wave operation on it to help the algorithm jump out of the local optimum;

[0035] Step S3: Perform joint clustering, offloading and resource optimization configuration of the ultra-dense cache network according to the position of the global optimal individual.

[0036] In this embodiment, step S1 includes the following sub-steps:

[0037] Step S11: In the network system, each macro cell deploys a macro base station, and small base stations, the small base station index set is Each small base station is equipped with an edge computing server and an edge cache server, and the small base station is connected to the macro base station through a wired link; there are two types of mobile terminals, one is a computing-sensitive mobile terminal that uploads computing tasks to the computing server for computing, and the other is a high-speed mobile terminal that requests file downloads from the cache server; there are There are computing-sensitive mobile terminals, and the computing-sensitive mobile terminal index set is , high-speed mobile terminals, the high-speed mobile terminal index set is .

[0038] In order to make full use of computing resources and reasonably reuse cache resources, the K-means clustering method is used to divide the small base stations into clusters. It is worth noting that all small base stations in the same cluster are interconnected through wired links to form a cache resource pool, and can share cache resources in the edge server with each other. Each high-speed mobile terminal has a file request. When the high-speed mobile terminal is associated with a small base station, if the required file has been cached by the small base station, the high-speed mobile terminal can directly download the file; otherwise, if other small base stations in the cache resource pool cache the required file, it can be obtained from other small base stations in the same cluster through wired links. In addition, each computing-sensitive mobile terminal has a computing task requirement, and its computing task adopts binary offloading, which can be chosen to be executed locally or offloaded to the small base station for execution.

[0039] In order to reduce network interference and improve spectrum efficiency, frequency division multiplexing and non-orthogonal multiple access are used to utilize resources. Uplink Band and downlink frequency band , respectively used for uplink access link and downlink access link transmission between mobile terminals and small base stations, , and The bandwidths are , and ,in is the frequency band division factor, and satisfies The uplink access link uses uplink non-orthogonal multiple access to divide the uplink frequency band According to the number of small base station clusters, it can be divided into sub-bands, and the sub-bands occupied by each cluster are divided into Uplink non-orthogonal multiple access multiplexing sub-channels, the uplink sub-channel index set is ,in Expressing rounding operations, represents the bandwidth of the uplink non-orthogonal multiple access sub-channel, Indicates the uplink subchannel; the downlink access line adopts downlink non-orthogonal multiple access, which divides the downlink frequency band Divide into sub-bands, each of which is further divided into downlink non-orthogonal multiple access multiplexing sub-channels, the downlink sub-channel index set is ,in represents the bandwidth of the downlink non-orthogonal multiple access sub-channel, Indicates the downlink subchannel;

[0040] Step S12: Build a communication model based on the basic network information. First, since the amount of downloaded calculation result data is very small, the present invention does not consider the downlink transmission of calculation-sensitive mobile terminals; in the above resource utilization method, for any mobile terminal associated with certain small base stations, there is only intra-cluster interference. According to the characteristics of uplink non-orthogonal multiple access, the mobile terminal will be interfered with by other mobile terminals with worse channel gain on the same subchannel in the uplink. After receiving the signal, it needs to be decoded in descending order of channel gain. Therefore, the signal associated with the small base station Computationally sensitive mobile devices In subchannel Uplink transmission rate for:

[0041] ,

[0042] in, and Represents computing-sensitive mobile terminals The selected subchannel index and small base station index, and Represents computing-sensitive mobile terminals The selected subchannel index and small cell index; Indicates computationally sensitive mobile terminals The uplink transmission power, Indicates computationally sensitive mobile terminals Uplink transmit power; Indicates computationally sensitive mobile terminals and small base stations The channel gain between Indicates computationally sensitive mobile terminals and small base stations The channel gain between represents the noise power; Indicates small base station The cluster to which it belongs; Indicates computing-sensitive mobile terminals and small base stations The set of users that cause interference to the uplink. In this formula, when When computing sensitive mobile terminals Affected by the uplink transmission interference of different base stations in the same cluster; When computing sensitive mobile terminals Affected by uplink transmission interference from the same cluster and base station.

[0043] The frequency band utilization method of the downlink access link adopts downlink non-orthogonal multiplexing. According to the characteristics of downlink non-orthogonal multiplexing, the mobile terminal is interfered by other mobile terminals with better channel gain on the same subchannel in the downlink. After receiving the signal, it needs to decode it in the ascending order of channel gain. Associated with small base station High-speed mobile terminal In subchannel Downlink transmission rate for:

[0044] ,

[0045] in, and Respectively represent high-speed mobile terminals The selected subchannel index and small base station index, and Respectively represent high-speed mobile terminals The selected subchannel index and small cell index; Indicates small base station The downlink transmission power is Indicates high-speed mobile terminal and small base stations The channel gain between Indicates high-speed mobile terminal and small base stations The channel gain between Indicates high-speed mobile terminals and small base stations The set of users that cause interference in the downlink. In this formula, when When high-speed mobile terminals Affected by downlink transmission interference from different base stations in the same cluster; When high-speed mobile terminals Affected by downlink transmission interference from the same cluster and base station;

[0046] Step S13: Construct cache model: There are two types of cache data in the edge cache server. One is the computing task cache, which can reduce the upload time and execution time of computing tasks by reusing the data of computing offload tasks. The other is the file cache, which can be responded to by small base stations instead of the core network to respond to the file request of the device, thereby reducing the file transmission time. It is worth noting that due to the limited capacity of the edge cache server, the two types of cache data are in a competitive relationship. In reality, file requests often obey the ZIPF distribution, which can be obtained by statistical analysis of historical data. Therefore, in the simulation process, the request probability of the file can be generated according to the ZIPF distribution.

[0047] In the above network system, any high-speed mobile terminal can obtain file data from the file set, and the file set is , and each file request can only be responded to by one small base station, any file The size is , is the number of collections. For high-speed mobile terminals To file Request index, if the high-speed mobile terminal Request File ,but ,otherwise ;set up For File In small base stations The cache index on the small base station Cached files ,but ,otherwise .

[0048] Small cells are clustered according to their locations. Small cells in each cluster collaborate with edge cache servers to share storage resources, thus forming a cache resource pool. It should be noted that in order to avoid caching the same content repeatedly, if the cache resource pool has cached files, , then there can only be one small base station cache file in the cluster Assumptions For high-speed mobile terminals With small base stations The association indication set between the high-speed mobile terminal and the small base station is , if the high-speed mobile terminal With small base stations If ,otherwise .

[0049] when , When high-speed mobile terminals Small Base Station Request File When , then the small base station Cached files , so high-speed mobile terminals From small base stations Download File Downlink access delay for:

[0050] ,

[0051] in Represents a file The size of To associate to a small base station High-speed mobile terminal In subchannel The downlink transmission rate is For high-speed mobile terminals and sub-channel When the high-speed mobile terminal When selecting a subchannel, ,otherwise .

[0052] when , When high-speed mobile terminals Small Base Station Request File When , ,in For File With small base stations Other small base stations in the same cluster The cache index on , For small base stations The cluster to which it belongs, i.e. small base station For small base stations Any other small base station in the same cluster, then the small base station File transfer via wired link Transmission to small base stations , wired transmission delay for:

[0053] ,

[0054] in Indicates the wired transmission rate between small base stations.

[0055] when , When high-speed mobile terminals Small Base Station Request File When, if , , high-speed mobile terminal Only through small base stations The wired backhaul link between the macro base station obtains files from the core network , Downlink return delay for:

[0056] ,

[0057] in Indicates the wired backhaul rate between the small base station and the macro base station.

[0058] Therefore, high-speed mobile terminals From small base stations Get File The average delay is:

[0059] ;

[0060] Step S14: Constructing a computing model: The computing tasks of any computing-sensitive mobile terminal can be executed locally or offloaded to a small base station for execution. For computing-sensitive mobile devices With small base stations The association indication set between the calculation-sensitive mobile terminal and the small base station is , if the computing-sensitive mobile terminal With small base stations If related, ,otherwise . Design for a sensitive mobile terminal The computing task data size is The calculation result data size is , calculate the number of CPU cycles required per bit for this task as , computing-sensitive mobile terminals The computing power is , small base station The computing power is .

[0061] set up Pre-cached index for computing tasks for computationally sensitive mobile terminals, when small base stations Cached calculation-sensitive mobile terminals When the computing task ,otherwise Similarly, if the cache resource pool has cached computing-sensitive mobile terminals There can only be one small base station in the cluster to cache computationally sensitive mobile terminals. In addition, suppose To calculate the index of the reused data type, when When small base stations Pre-caching computationally sensitive mobile devices The computing task itself, computing-sensitive mobile terminals No need to upload computing tasks to small base stations , but it needs to be in a small base station Perform calculation operations; when When small base stations Pre-caching computationally sensitive mobile devices The result of the calculation task, small base station Directly return the calculation results to the calculation-sensitive mobile terminal It is worth noting that since the amount of calculation result data is much smaller than the amount of task data, the time to return the calculation result can be ignored.

[0062] when When computing sensitive mobile terminals The computing tasks are executed locally, with a delay of for:

[0063] ;

[0064] when When computing sensitive mobile terminals The computing tasks in small base stations Execution, execution delay for:

[0065] ,

[0066] in For small base stations Assigned to computationally intensive mobile terminals The proportion of computing resources, under the equal proportion computing resource allocation, when the computing-sensitive mobile terminal Associated small base stations hour, It is expressed as:

[0067] ;

[0068] when , , , When For small base stations Cluster, small base station Transmitting computational multiplexing data to small base stations via wired links , small base station Then perform the calculation operation, the wired transmission delay for:

[0069] ;

[0070] when , , When computing sensitive mobile terminals The computing task needs to be uploaded to the small base station Execution, uplink access delay for:

[0071] ;

[0072] In summary, the computationally sensitive mobile terminal is completed The delay required for the computation task for:

[0073] ;

[0074] Step S15: construct an optimization problem according to the following formula: to improve the user experience of mobile terminals, minimize the total weighted delay of computationally sensitive mobile terminals and high-rate mobile terminals, that is:

[0075] ,

[0076] in and The weighted coefficients for calculating the total delay of sensitive mobile terminals and the total delay of high-speed mobile terminals are respectively; A set of association indications for computing sensitive mobile terminals and small base stations; It is a set of association indications between a high-rate mobile terminal and a small base station; To calculate the selection indication set of sensitive mobile terminals and subchannels, For computing-sensitive mobile devices With subchannel The choice between the instructions, when the calculation sensitive mobile terminal Select subchannel hour, ,otherwise ; For high-speed mobile terminals With subchannel The selection indicator set, For high-speed mobile terminals With subchannel When the high-speed mobile terminal Select subchannel hour, ,otherwise ; A pre-cached index set for files in the small base station. A pre-cached index set for small base stations to calculate data for computationally sensitive mobile terminals.

[0077] Under the constraints of computing resource ratio allocation, cache server cluster deployment, and cache capacity, the device association and subchannel selection of computing-sensitive mobile terminals and high-speed mobile terminals, file pre-caching and computing task pre-caching of small base stations are combined to minimize the total weighted delay of the system. Specifically, the optimization problem is expressed as:

[0078] ;

[0079] in, represents a collection of small base stations, Represents a file index collection, represents a set of computationally sensitive mobile terminals, Represents a collection of high-speed mobile terminals. represents the uplink subchannel set, represents a downlink subchannel set; For computing-sensitive mobile devices Association indication with small base station, For high-speed mobile terminals Association indication with small base station, For computing-sensitive mobile devices With subchannel The selection instructions, For high-speed mobile terminals With subchannel The selection instructions, For File In small base stations Pre-cached indexes, For computing-sensitive mobile devices The calculation data of small base stations Pre-cached index of ; For small base stations Assigned to computationally intensive mobile terminals The proportion of computing resources, For File The amount of data, To calculate the index of the reused data type, and are the amount of computing task data and the amount of computing result data of computing-sensitive mobile terminals, respectively. For small base stations Cache capacity; There are 14 constraints. Indicates that the total computing resources allocated by any small base station to the associated computing-sensitive mobile terminals cannot exceed the computing resources of the base station. This means that the amount of data cached by any small base station cannot exceed the storage capacity of the base station. Indicates that each small base station cluster pre-caches any file at most once , Indicates that each small base station cluster pre-caches any computationally sensitive mobile terminal at most once The computational tasks It means that each computationally sensitive mobile terminal is associated with at most one small base station. This means that each high-speed mobile terminal can only be associated with one small base station. means that each computationally sensitive mobile terminal selects at most one subchannel, Indicates that each high-speed mobile terminal can only select one sub-channel. The association indication of the computationally sensitive mobile terminal can only be 0 or 1. The association indication of a high-speed mobile terminal can only be 0 or 1. The subchannel selection indication of the computationally sensitive mobile terminal can only be 0 or 1. The subchannel selection indication for a high-rate mobile terminal can only be 0 or 1. Indicates that the pre-cache index of the small base station for the file can only be 0 or 1. Indicates that the pre-cache index of the small base station for the calculation-sensitive mobile terminal calculation data can only be 0 or 1.

[0080] From the above technical solutions, it can be seen that the construction of the optimization problem is based on the communication model, cache model and computing model in the ultra-dense cache network system. Under the constraints of small base station clustering and cache capacity, the calculation-sensitive mobile terminal association matrix, high-rate mobile terminal association matrix, calculation-sensitive mobile terminal sub-channel selection matrix, high-rate mobile terminal sub-channel selection matrix, small base station file pre-cache matrix, and small base station computing task pre-cache matrix are used as reference indicators for the optimization problem, which solves the problem of single reference indicator in the prior art, realizes the formation of a cache resource pool with clustered small base stations, thereby assisting the joint execution of computing offloading and file downloading, achieving lower system latency, and constructing a new optimization problem.

[0081] Step S2: Obtain an initial solution according to the optimization problem, and define the initial solution as the parent population. Use the water wave energy attenuation algorithm to search the parent population, and finally output the position of the global optimal individual to obtain the solution with the minimum system delay.

[0082] It should be noted that, for the optimization problem, this embodiment first uses the calculation method in the prior art to obtain the initial solution to the optimization problem, and then uses the water wave energy attenuation algorithm proposed in this embodiment to continuously iterate and optimize the initial solution to obtain the optimal solution to the optimization problem, that is, to obtain the solution with the minimum system delay. There are many ways to solve the optimization problem in the prior art, which are not described in detail in this embodiment.

[0083] Step S2 in this embodiment includes the following sub-steps:

[0084] Step S21: Initialize the maximum number of iterations of the adaptive genetic water wave algorithm , and set the current iteration number t to 1;

[0085] Step S22: Water wave individual , water wave population ,include Water waves, each water wave individual The wavelength is , used for the calculation of propagation operation, with an initial value of 0.5; each water wave is encoded and the optimized parameter , , , , , Encoded as , , , , , . Among them, the calculation-sensitive mobile terminal association indication code set , Represents water waves Computationally sensitive mobile terminals Associated base station index; high-speed mobile terminal association indication code set , Represents water waves Medium and high speed mobile terminals Associated base station index; calculation sensitive mobile terminal subchannel selection code set , Represents water waves Computationally sensitive mobile terminals Selected subchannel index; high-rate mobile terminal subchannel selection code set , Represents water waves Medium and high speed mobile terminals Selected subchannel index; Small base station end calculates data pre-cache code set , Represents water waves Virtual computing data Pre-cached indexes, ,in is the number of small base stations, To calculate the number of sensitive mobile terminals; small base station end file pre-caching code set , Represents water waves Virtual File Pre-cached indexes, ,in is the number of small base stations, is the number of files;

[0086] Step S23: Initialize the population according to the following formula:

[0087] ,

[0088] in, Represents water waves Computationally sensitive mobile terminals The initial value of the associated base station index, Represents water waves Medium and high speed mobile terminals The initial value of the associated base station index, Represents water waves Computationally sensitive mobile terminals The selected sub-channel coding initial value, Represents water waves Medium and high speed mobile terminals The selected sub-channel coding initial value, Represents water waves Virtual computing tasks The initial value of the pre-cache index, Represents water waves Virtual File The initial value of the pre-cache index, Represents from the set Output a random element in Generate a and A random number between

[0089] And construct the water wave individuals in the water wave population according to the following formula The fitness function of is:

[0090] ,

[0091] in, , Represents water waves The fitness value of represents the system delay, For File In small base stations Pre-cached indexes, Pre-cached indexes for computationally sensitive mobile terminal computing tasks, For File The amount of data, To calculate the index of the reused data type, and are the amount of computing task data and the amount of computing result data of computing-sensitive mobile terminals, respectively. For small base stations Cache capacity;

[0092] Use the fitness function to calculate the fitness values ​​of all water waves in the water wave population, and take the individual with the highest fitness value as the global best individual;

[0093] It should be noted that the parameters in the fitness function can be converted into the encoding of the optimization problem parameters according to the above encoding definition. In other words, the parameters of the optimization problem are completely equivalent to their encoding.

[0094] Step S24: Determine whether the current number of iterations t is less than or equal to the maximum number of iterations , if the current number of iterations t is less than or equal to the maximum number of iterations , then based on the water wave energy, the water wave population is adaptively propagated or refracted, and then the wave breaking operation is performed on the best individual of the current population to obtain the target population; if the current iteration number t is greater than the maximum iteration number , then output the encoding of the global best individual.

[0095] From the above technical solutions, we can know that:

[0096] Specifically, step S241: Calculate the energy of the water wave population under the current number of iterations :

[0097] ,

[0098] Generate a random number between 0 and 1 and compare it with the water wave energy. If the random number is less than or equal to the water wave energy, perform a propagation operation, i.e., step S242; if the random number is greater than the water wave energy, perform a refraction operation, i.e., step S243;

[0099] Step S242: Perform a propagation operation on the individual water waves in the water wave population, and the propagation rule is:

[0100] ,

[0101] Among them, the superscript and represents the number of iterations, express In the The value of the iteration, express In the The value of the iteration; express In the The value of the iteration, express In the The value of the iteration; express In the The value of the iteration, express In the The value of the iteration; express In the The value of the iteration, express In the The value of the iteration; express In the The value of the iteration, express In the The value of the iteration; express In the The value of the iteration, express In the The value of the iteration; Express The rounding operation of is a random number between -1 and 1. For water wave individuals The wavelength of is the number of small base stations, is the number of uplink sub-channels, is the number of downlink sub-channels.

[0102] In the water wave individuals that update the propagation rules, adaptive perturbations are performed, that is, diversity-guided perturbations are performed first, and then adaptive perturbations are performed; the diversity measure is defined as follows:

[0103] ,

[0104] in , , , , , Respectively , , , , , The diagonal length of the feasible region of is:

[0105] ,

[0106] in, Represents water waves Computationally sensitive mobile terminals The average value of the associated small base station index, Represents water waves Medium and high speed mobile terminals The average value of the associated small base station index, Represents water waves Computationally sensitive mobile terminals The average value of the selected subchannel index, Represents water waves Medium and high speed mobile terminals The average value of the selected subchannel index, Represents water waves Virtual computing tasks The average value of the pre-cached index, Indicates that from a virtual file The average value of the pre-cached index;

[0107] The diversity-guided perturbation probability is defined according to the diversity measure:

[0108] ,

[0109] in , and is a pre-set probability that satisfies , and Approaching 0, and is a threshold constant and satisfies ;

[0110] The water waves in the water wave population are obtained according to the following formula Adaptive probability:

[0111] ,

[0112] in and is satisfied The constant, Represents water waves The fitness value of and They represent the maximum and average fitness of all water wave individuals in the population respectively;

[0113] According to the diversity-guided perturbation probability and adaptive probability, the water wave individuals are updated successively, and adaptive perturbations are added to them. The perturbation rules are:

[0114] ,

[0115] in, Express The rounding operation of and is a random number between 0 and 1, used to control the degree and direction of disturbance. indicates positive disturbance, otherwise it is negative disturbance;

[0116] Step S243: performing a refraction operation on individual water waves in the water wave population, and the refraction rule is as follows:

[0117] ,

[0118] in, Express The rounding operation of The function is used to generate random numbers that follow a Gaussian distribution. The function is used to take the absolute value. The calculation-sensitive mobile terminal associated base station index representing the best individual in the current population, The high-speed mobile terminal associated base station index representing the best individual in the current population, represents the computationally sensitive mobile terminal selection subchannel index of the best individual in the current population, The high-rate mobile terminal selection subchannel index representing the best individual in the current population, represents the pre-cache index of the virtual computing task of the best individual in the current population, A virtual file pre-cached index representing the best individual of the current population.

[0119] After performing the refraction operation, the wavelength of the water wave is updated according to the updated individual fitness value of the water wave:

[0120] ,

[0121] in For water wave individuals In the The wavelength at the iteration, is the individual fitness value of the water wave after the refraction operation is updated, is the fitness value before updating;

[0122] Step S244: After step S241 is completed, the water wave wavelength and the breaking wave coefficient are updated, wherein the water wave wavelength update rule is as follows:

[0123] ,

[0124] in, For water wave individuals In the The wavelength at the iteration, is the wavelength attenuation coefficient, which is usually a constant slightly greater than 1. and Respectively represent the maximum and minimum fitness values ​​of the current population, is a sufficiently small number greater than 0; the breaking wave coefficient update rule is as follows:

[0125] ,

[0126] in, and They represent the water wave population in Second and The breaking wave coefficient at the iteration, and They represent the upper and lower bounds of the breaking wave coefficient respectively.

[0127] The fitness value function is used to calculate the fitness value of the water wave population, find the best individual in the current population, and perform the breaking wave operation on it. Randomly select half the dimension of the best individual in the current population and generate a sub-water wave according to the breaking wave rule. If the fitness value of the sub-water wave is better than the best individual in the current population, the sub-water wave is used to update the best individual in the current population, otherwise the original best individual in the current population is retained. The rule for generating sub-water waves from the position of the best individual in the current population, that is, the breaking wave rule, is as follows:

[0128] ,

[0129] in, is a random number that follows a normal distribution, , , , , , The calculation-sensitive mobile terminal associated base station index, the high-rate mobile terminal associated base station index, the calculation-sensitive mobile terminal selected sub-channel index, the high-rate mobile terminal selected sub-channel index, the virtual calculation data pre-cache index, and the virtual file pre-cache index for generating sub-water waves for the breaking wave operation are respectively.

[0130] Step S245: Compare the fitness values ​​of the best individual in the current population with the best individual in the world. If the fitness value of the former is greater than that of the latter, the best individual in the current population is used to update the best individual in the world. Otherwise, the best individual in the world is retained.

[0131] Step S246: Set the current number of iterations The value of is increased by 1, and the process returns to step S24, and the iteration is repeated until the iteration ends. Finally, the globally best individual is retained as the optimal solution to the problem.

[0132] Step S3: Perform joint clustering, offloading and resource optimization configuration of the ultra-dense cache network according to the position of the global optimal individual.

[0133] And according to the obtained solution, base station association and uplink subchannel selection of calculation-sensitive mobile terminals, base station association and downlink subchannel selection of high-rate mobile terminals, and calculation task pre-caching and file pre-caching of small base station cache servers are performed.

[0134] The effects of the embodiments of the present invention can be further illustrated by simulation.

[0135] The simulation conditions are set as follows: 1 macro base station and 20 small base stations are set in the ultra-dense cache network. The computing power of each small base station is 20GHz, the cache capacity is 100MB, and the transmission power is 24dBm; the number of small base station clusters is set to 5, the number of uplink subchannels is set to 5, and the number of downlink subchannels is set to 5; a certain number of mobile terminals are set each time and randomly divided into computing-sensitive mobile terminals and high-rate mobile terminals; the transmission power of the computing-sensitive mobile terminal is 23dBm and the computing power is 1GHz. Each computing-sensitive mobile terminal has 1 computing task to be processed, the size of which is 1-5MB, and the CPU cycle required for each bit of computing task is 50-100 rounds; the number of files available for download is 30, and the data volume of each file is 1-10MB; the wired backhaul rate between the small base station and the macro base station is bit / s, the wired transmission rate between small base stations is bit / s; the frequency band division factor is set to 0.4, the algorithm population size is 20 individuals, and the maximum number of algorithm iterations is 2000.

[0136] Figure 2 Schematic diagram of the effect of the number of network users on the fitness value in the water wave optimization algorithm. Figure 3 The figure is a schematic diagram of the effect of the number of network users on the fitness value in the water wave energy attenuation algorithm and the adaptive genetic algorithm. When other conditions remain unchanged, as the user density, that is, the number of mobile terminals in each macro cell, increases, the fitness value decreases accordingly. It is easy to find that the fitness value of the water wave energy attenuation algorithm proposed in the present invention is always higher than that of the adaptive genetic algorithm, and it changes the dilemma that the water wave optimization algorithm cannot solve the optimization problem of complex scenes.

[0137] Figure 4 The present invention discloses the effect of the number of network users on the average delay of computationally sensitive mobile terminals. Since the water wave optimization algorithm cannot satisfy the penalty term, only the water wave energy decay algorithm and the adaptive genetic algorithm are compared to see the effect on the average delay of computationally sensitive mobile terminals. It is not difficult to find that the delay solved by the water wave energy decay algorithm is lower than that of the adaptive genetic algorithm at all user densities. In addition, the average delay of computationally sensitive mobile terminals increases with the increase of user density.

[0138] Figure 5 The present invention discloses the effect of the number of network users on the average delay of high-speed mobile terminals. As above, only the effect of the water wave energy attenuation algorithm and the adaptive genetic algorithm on the average delay of high-speed mobile terminals is compared. It is not difficult to find that the delay solved by the water wave energy attenuation algorithm is lower than that of the adaptive genetic algorithm under all user densities. In addition, the average delay of high-speed mobile terminals increases with the increase of user density.

[0139] Another embodiment of the present invention provides a terminal device. The terminal device of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a program for a method for optimizing joint clustering, unloading, and resources in an ultra-dense cache network. When the processor executes the computer program, the method for optimizing joint clustering, unloading, and resources in an ultra-dense cache network is implemented.

[0140] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the method for optimizing joint clustering, offloading, and resources in the ultra-dense cache network can be implemented.

[0141] Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Mentioning "embodiment" in this article means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present embodiment application. The appearance of this phrase in various positions in the specification is not necessarily the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It can be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0142] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A method for joint clustering, offloading and resource optimization in an ultra-dense cache network, characterized in that: include: Step S1: Obtain basic network information of the ultra-dense cache network, build a network system based on the basic network information, the network system includes a communication model, a cache model, and a calculation model, and build an optimization problem under the constraints of cache server clustering and cache capacity, and the optimization problem is a system delay minimization problem; Step S2: Obtain an initial solution according to the optimization problem, and define the initial solution as the parent population. Use the water wave energy attenuation algorithm to search the parent population, and finally output the position of the global optimal individual to obtain the solution with the minimum system delay. The process of searching the parent population by the water wave energy attenuation algorithm includes: initializing the population and determining the global optimal individual; adaptively performing improved propagation and refraction operations under the control of water wave energy, wherein the propagation operation performs a coarse-grained search on the population and the refraction operation performs a fine-grained search on the population; finding the optimal individual of the current population and performing an improved breaking wave operation on it to help the algorithm jump out of the local optimum; Step S3: performing joint clustering, offloading and resource optimization configuration of the ultra-dense cache network according to the position of the global optimal individual; In the network system, each macro cell deploys a macro base station, and small base stations, the small base station index set is Each small base station is equipped with an edge computing server and an edge cache server, and the small base station is connected to the macro base station through a wired link; there are two types of mobile terminals, one is a computing-sensitive mobile terminal that uploads computing tasks to the computing server for computing, and the other is a high-speed mobile terminal that requests file downloads from the cache server; there are There are computing-sensitive mobile terminals, and the computing-sensitive mobile terminal index set is , high-speed mobile terminals, the high-speed mobile terminal index set is ; According to the location of small base stations, the K-means clustering method is used to divide the small base stations into All small base stations in the same cluster are connected to each other through wired links to form a cache resource pool, and share cache resources in the edge server. Each high-speed mobile terminal has a file request. When the high-speed mobile terminal is associated with a small base station, if the required file has been cached by the small base station, the high-speed mobile terminal directly downloads the file. Otherwise, if other small base stations in the cache resource pool cache the required file, it is obtained from other small base stations in the same cluster through wired links. Each computing-sensitive mobile terminal has a computing task requirement, and its computing task adopts binary offloading, choosing to execute locally or offload to the small base station for execution. The frequency band of the network system Uplink Band and downlink frequency band , respectively used for uplink access link and downlink access link transmission between mobile terminals and small base stations, , and The bandwidths are , and ,in is the frequency band division factor, and satisfies ; The uplink access link adopts uplink non-orthogonal multiple access, which divides the uplink frequency band According to the number of small base station clusters, it can be divided into sub-bands, and the sub-bands occupied by each cluster are divided into Uplink non-orthogonal multiple access multiplexing sub-channels, the uplink sub-channel index set is ,in Expressing rounding operations, represents the bandwidth of the uplink non-orthogonal multiple access sub-channel, Indicates the uplink subchannel; the downlink access line adopts downlink non-orthogonal multiple access, which divides the downlink frequency band Divide into sub-bands, each of which is further divided into downlink non-orthogonal multiple access multiplexing sub-channels, the downlink sub-channel index set is ,in represents the bandwidth of the downlink non-orthogonal multiple access sub-channel, Indicates the downlink subchannel; In step S1, a communication model is constructed based on the basic network information and associated with the small base station Computationally sensitive mobile devices In subchannel Uplink transmission rate for: , in, and Represents computing-sensitive mobile terminals The selected subchannel index and small base station index, and Represents computing-sensitive mobile terminals The selected subchannel index and small cell index; Indicates computationally sensitive mobile terminals The uplink transmission power, Indicates computationally sensitive mobile terminals Uplink transmit power; Indicates computationally sensitive mobile terminals and small base stations The channel gain between Indicates computationally sensitive mobile terminals and small base stations The channel gain between represents the noise power; Indicates small base station The cluster to which it belongs; Indicates computing-sensitive mobile terminals and small base stations The set of users that cause interference in the uplink; Associated to small cell High-speed mobile terminal In subchannel Downlink transmission rate for: , in, and Respectively represent high-speed mobile terminals The selected subchannel index and small base station index, and Respectively represent high-speed mobile terminals The selected subchannel index and small cell index; Indicates small base station The downlink transmission power is Indicates high-speed mobile terminal and small base stations The channel gain between Indicates high-speed mobile terminal and small base stations The channel gain between Indicates high-speed mobile terminals and small base stations The set of users causing interference in the downlink; Constructing a cache model: There are two types of cache data in the edge cache server, one is the computing task cache and the other is the file cache. In the network system, any high-speed mobile terminal obtains file data from the file collection. The file collection is , and each file request can only be responded to by one small base station, any file The size is , is the number of collections; For high-speed mobile terminals To file Request index, if the high-speed mobile terminal Request File ,but ,otherwise ;set up For File In small base stations The cache index on the small base station Cached files ,but ,otherwise ; Small base stations are clustered according to their locations. Small base stations in each cluster collaborate with edge cache servers to share storage resources, thus forming a cache resource pool. If the cache resource pool has cached files, , then there can only be one small base station cache file in the cluster ;set up For high-speed mobile terminals With small base stations The association indication set between the high-speed mobile terminal and the small base station is , if the high-speed mobile terminal With small base stations If ,otherwise ; when , When high-speed mobile terminals Small Base Station Request File When , then the small base station Cached files , so high-speed mobile terminals From small base stations Download File Downlink access delay for: , in Represents a file The size of To associate to a small base station High-speed mobile terminal In subchannel The downlink transmission rate is For high-speed mobile terminals and sub-channel When the high-speed mobile terminal When selecting a subchannel, ,otherwise ; when , When high-speed mobile terminals Small Base Station Request File When , ,in For File With small base stations Other small base stations in the same cluster The cache index on , For small base stations The cluster to which it belongs, i.e. small base station For small base stations Any other small base station in the same cluster, then the small base station File transfer via wired link Transmission to small base stations , wired transmission delay for: ， in Indicates the wired transmission rate between small base stations; when , When high-speed mobile terminals Small Base Station Request File When, if , , high-speed mobile terminal Only through small base stations The wired backhaul link between the macro base station obtains files from the core network , Downlink return delay for: ， in Indicates the wired backhaul rate between the small base station and the macro base station; Therefore, high-speed mobile terminals From small base stations Get File The average delay is: ； Build a computing model: The computing tasks of any computing-sensitive mobile terminal are executed locally or offloaded to a small base station for execution; set For computing-sensitive mobile devices With small base stations The association indication set between the calculation-sensitive mobile terminal and the small base station is , if the computing-sensitive mobile terminal With small base stations If related, ,otherwise ; Design for sensitive mobile terminals The computing task data size is The calculation result data size is , calculate the number of CPU cycles required per bit for this task as , computing-sensitive mobile terminals The computing power is , small base station The computing power is ; set up Pre-cached index for computing tasks for computationally sensitive mobile terminals, when small base stations Cached calculation-sensitive mobile terminals When the computing task ,otherwise Similarly, if the cache resource pool has cached computing-sensitive mobile terminals There can only be one small base station in the cluster to cache computationally sensitive mobile terminals. In addition, suppose To calculate the index of the reused data type, when When small base stations Pre-caching computationally sensitive mobile devices The computing task itself, computing-sensitive mobile terminals No need to upload computing tasks to small base stations , but it needs to be in a small base station Perform calculation operations; when When small base stations Pre-caching computationally sensitive mobile devices The result of the calculation task, small base station Directly return the calculation results to the calculation-sensitive mobile terminal ; when When computing sensitive mobile terminals The computing tasks are executed locally, with a delay of for: ； when When computing sensitive mobile terminals The computing tasks in small base stations Execution, execution delay for: ， in For small base stations Assigned to computationally intensive mobile terminals The proportion of computing resources, under the equal proportion computing resource allocation, when the computing-sensitive mobile terminal Associated small base stations hour, It is expressed as: ； when , , , When For small base stations Cluster, small base station Transmitting computational multiplexing data to small base stations via wired links , small base station Then perform the calculation operation, the wired transmission delay for: ； when , , When computing sensitive mobile terminals The computing task needs to be uploaded to the small base station Execution, uplink access delay for: ； Completing computationally sensitive mobile terminals The delay required for the computation task for: ； The optimization problem is formulated as follows: ， in and The weighted coefficients for calculating the total delay of sensitive mobile terminals and the total delay of high-speed mobile terminals are respectively; A set of association indications for computing sensitive mobile terminals and small base stations; It is a set of association indications between a high-rate mobile terminal and a small base station; To calculate the selection indication set of sensitive mobile terminals and subchannels, For computing-sensitive mobile devices With subchannel The choice between the instructions, when the calculation sensitive mobile terminal Select subchannel hour, ,otherwise ; For high-speed mobile terminals With subchannel The selection indicator set, For high-speed mobile terminals With subchannel When the high-speed mobile terminal Select subchannel hour, ,otherwise ; A pre-cached index set for files in the small base station. A pre-cached index set for small cell computing data for computing-sensitive mobile terminals; The optimization problem is stated as: ； in, represents a collection of small base stations, Represents a file index collection, represents a set of computationally sensitive mobile terminals, Represents a collection of high-speed mobile terminals. represents the uplink subchannel set, represents a downlink subchannel set; For computing-sensitive mobile devices Association indication with small base station, For high-speed mobile terminals Association indication with small base station, For computing-sensitive mobile devices With subchannel The selection instructions, For high-speed mobile terminals With subchannel The selection instructions, For File In small base stations Pre-cached indexes, For computing-sensitive mobile devices The calculation data of small base stations Pre-cached index of ; For small base stations Assigned to computationally intensive mobile terminals The proportion of computing resources, For File The amount of data, To calculate the index of the reused data type, and are the amount of computing task data and the amount of computing result data of computing-sensitive mobile terminals, respectively. For small base stations Cache capacity; There are 14 constraints. Indicates that the total computing resources allocated by any small base station to the associated computing-sensitive mobile terminals cannot exceed the computing resources of the base station. This means that the amount of data cached by any small base station cannot exceed the storage capacity of the base station. Indicates that each small base station cluster pre-caches any file at most once , Indicates that each small base station cluster pre-caches any computationally sensitive mobile terminal at most once The computational tasks It means that each computationally sensitive mobile terminal is associated with at most one small base station. This means that each high-speed mobile terminal can only be associated with one small base station. means that each computationally sensitive mobile terminal selects at most one subchannel, Indicates that each high-speed mobile terminal can only select one sub-channel. The association indication of the computationally sensitive mobile terminal can only be 0 or 1. The association indication of a high-speed mobile terminal can only be 0 or 1. The subchannel selection indication of the computationally sensitive mobile terminal can only be 0 or 1. The subchannel selection indication for a high-rate mobile terminal can only be 0 or 1. Indicates that the pre-cache index of the small base station for the file can only be 0 or 1. Indicates that the pre-caching index of the small base station for the calculation-sensitive mobile terminal calculation data can only be 0 or 1; The step S2 comprises the following sub-steps: Step S21: Initialize the maximum number of iterations of the water wave energy attenuation algorithm , and the current number of iterations Set to 1; Step S22: Encode individual water waves and define individual water wave wavelengths for calculation of propagation operations; Step S23: Initialize the population and construct water wave individuals Fitness function: Use the fitness function to calculate the fitness values ​​of all water wave individuals in the population, and take the individual with the highest fitness value as the global best individual; Step S24: Determine the current number of iterations Is it less than or equal to the maximum number of iterations? , if the current number of iterations Less than or equal to the maximum number of iterations , then based on the water wave energy, the water wave population is adaptively propagated or refracted, and then the wave breaking operation is performed on the best individual of the current population to obtain the target population; if the current number of iterations is Greater than the maximum number of iterations , then output the encoding of the global best individual; The step S24 includes the following sub-steps: Step S241: Calculate the energy of the water wave population under the current iteration number, generate a random number between 0 and 1 and compare it with the water wave energy. If the random number is less than or equal to the water wave energy, perform a propagation operation, i.e., step S242; if the random number is greater than the water wave energy, perform a refraction operation, i.e., step S243; Step S242: updating the water wave individuals with the given propagation rule, and performing adaptive perturbation on the updated individuals, that is, first performing diversity-guided perturbation, and then performing adaptive perturbation according to the adaptive probability; Step S243: updating the individual water wave using the given refraction rule, and updating the individual wavelength of the water wave according to the individual fitness value before and after refraction; Step S244: after step S241 is executed, the water wave wavelength and breaking wave coefficient are updated, the fitness value of the water wave population is calculated using the fitness function, the best individual of the current population is found, and the breaking wave operation is performed on it; Step S245: Compare the fitness values ​​of the best individual in the current population with the best individual in the world. If the fitness value of the former is greater than that of the latter, the best individual in the current population is used to update the best individual in the world. Otherwise, the best individual in the world is retained. Step S246: Set the current number of iterations The value of is increased by 1, and the process returns to step S24, and the iteration is repeated until the iteration ends. Finally, the globally best individual is retained as the optimal solution to the problem.

2. The method for joint clustering, offloading and resource optimization in an ultra-dense cache network according to claim 1, characterized in that: The step S22 is specifically as follows: , water wave population ,include Water waves, each water wave individual The wavelength is ; Encode each individual water wave and optimize the parameters , , , , , Encoded as , , , , , ; Among them, the calculation-sensitive mobile terminal association indication code set , Represents water waves Computationally sensitive mobile terminals Associated base station index; high-speed mobile terminal association indication code set , Represents water waves Medium and high speed mobile terminals Associated base station index; calculation sensitive mobile terminal subchannel selection code set , Represents water waves Computationally sensitive mobile terminals Selected subchannel index; high-rate mobile terminal subchannel selection code set , Represents water waves Medium and high speed mobile terminals Selected subchannel index; Small base station end calculates data pre-cache code set , Represents water waves Virtual computing data Pre-cached indexes, ,in is the number of small base stations, To calculate the number of sensitive mobile terminals; Small cell site file pre-caching code set , Represents water waves Virtual File Pre-cached indexes, ,in is the number of small base stations, The number of files.

3. The method for joint clustering, offloading and resource optimization in an ultra-dense cache network according to claim 2, characterized in that: The step S23 initializes the population according to the following formula: ， in, Represents water waves Computationally sensitive mobile terminals The initial value of the associated base station index, Represents water waves Medium and high speed mobile terminals The initial value of the associated base station index, Represents water waves Computationally sensitive mobile terminals The selected sub-channel coding initial value, Represents water waves Medium and high speed mobile terminals The selected sub-channel coding initial value, Represents water waves Virtual computing tasks The initial value of the pre-cache index, Represents water waves Virtual File The initial value of the pre-cache index, Represents from the set Output a random element in Generate a and A random number between And construct the water wave individuals in the water wave population according to the following formula The fitness function of is: ， in, , Represents water waves The fitness value of represents the system delay, For File In small base stations Pre-cached indexes, Pre-cached indexes for computationally sensitive mobile terminal computing tasks, For File The amount of data, To calculate the index of the reused data type, and are the amount of computing task data and the amount of computing result data of computing-sensitive mobile terminals, respectively. For small base stations The cache capacity.

4. The method for joint clustering, offloading and resource optimization in an ultra-dense cache network according to claim 3, characterized in that: Step S241: Calculate the energy of the water wave population at the current number of iterations : ， Generate a random number between 0 and 1 and compare it with the water wave energy. If the random number is less than or equal to the water wave energy, perform a propagation operation, i.e., step S242; if the random number is greater than the water wave energy, perform a refraction operation, i.e., step S243; Step S242: Perform a propagation operation on the individual water waves in the water wave population, and the propagation rule is: ， Among them, the superscript and represents the number of iterations, express In the The value of the iteration, express In the The value of the iteration; express In the The value of the iteration, express In the The value of the iteration; express In the The value of the iteration, express In the The value of the iteration; express In the The value of the iteration, express In the The value of the iteration; express In the The value of the iteration, express In the The value of the iteration; express In the The value of the iteration, express In the The value of the iteration; Express The rounding operation of is a random number between -1 and 1. For water wave individuals The wavelength of is the number of small base stations, is the number of uplink sub-channels, is the number of downlink sub-channels; Adaptive perturbation is performed on the water wave individuals in the propagation rule update, that is, diversity-guided perturbation is performed first, and then adaptive perturbation is performed; the diversity measure is defined as follows: ， in , , , , , Respectively , , , , , The diagonal length of the feasible region of is: ， in, Represents water waves Computationally sensitive mobile terminals The average value of the associated small base station index, Represents water waves Medium and high speed mobile terminals The average value of the associated small base station index, Represents water waves Computationally sensitive mobile terminals The average value of the selected subchannel index, Represents water waves Medium and high speed mobile terminals The average value of the selected subchannel index, Represents water waves Virtual computing tasks The average value of the pre-cached index, Indicates that from a virtual file The average value of the pre-cached index; The diversity-guided perturbation probability is defined according to the diversity measure: ， in , and is a pre-set probability that satisfies , and Approaching 0, and is a threshold constant and satisfies ; The water waves in the water wave population are obtained according to the following formula Adaptive probability: ， in and is satisfied The constant, Represents water waves The fitness value of and They represent the maximum and average fitness of all water wave individuals in the population respectively; According to the diversity-guided perturbation probability and adaptive probability, the water wave individuals are updated successively, and adaptive perturbations are added to them. The perturbation rules are: ， in, Express The rounding operation of and is a random number between 0 and 1, used to control the degree and direction of disturbance. indicates positive disturbance, otherwise it is negative disturbance; Step S243: performing a refraction operation on individual water waves in the water wave population, and the refraction rule is as follows: ， in, Express The rounding operation of The function is used to generate random numbers that follow a Gaussian distribution. The function is used to take the absolute value. The calculation-sensitive mobile terminal associated base station index representing the best individual in the current population, The high-speed mobile terminal associated base station index representing the best individual in the current population, represents the computationally sensitive mobile terminal selection subchannel index of the best individual in the current population, The high-rate mobile terminal selection subchannel index representing the best individual in the current population, represents the pre-cache index of the virtual computing task of the best individual in the current population, The virtual file pre-cached index representing the best individual in the current population; After performing the refraction operation, the wavelength of the water wave is updated according to the updated individual fitness value of the water wave: ， in For water wave individuals In the The wavelength at the iteration, is the individual fitness value of the water wave after the refraction operation is updated, is the fitness value before updating; Step S244: After step S241 is completed, the water wave wavelength and the breaking wave coefficient are updated, wherein the water wave wavelength update rule is as follows: ， in, For water wave individuals In the The wavelength at the iteration, is the wavelength attenuation coefficient, which is usually a constant slightly greater than 1. and Respectively represent the maximum and minimum fitness values ​​of the current population, is a sufficiently small number greater than 0; the breaking wave coefficient update rule is as follows: ， in, and They represent the water wave population in Second and The breaking wave coefficient at the iteration, and represent the upper and lower bounds of the breaking wave coefficient respectively; The fitness value function is used to calculate the fitness value of the water wave population, find the best individual in the current population, and perform the wave-breaking operation on it; randomly select half the dimension of the best individual in the current population and generate a sub-water wave according to the wave-breaking rule. If the fitness value of the sub-water wave is better than the best individual in the current population, the sub-water wave is used to update the best individual in the current population, otherwise the original best individual in the current population is retained; the rule for generating sub-water waves from the position of the best individual in the current population, that is, the wave-breaking rule is as follows: ， in, is a random number that follows a normal distribution, , , , , , They are the calculation-sensitive mobile terminal associated base station index, the high-rate mobile terminal associated base station index, the calculation-sensitive mobile terminal selection sub-channel index, the high-rate mobile terminal selection sub-channel index, the virtual calculation data pre-cache index, and the virtual file pre-cache index for generating sub-water waves for the breaking wave operation.

5. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and running on the terminal device, characterized in that: When the processor executes the computer program, the method for joint clustering, offloading and resource optimization in an ultra-dense cache network described in any one of claims 1-4 is implemented.

Images