A task cache allocation method and system in mobile augmented reality

Abstract

The present invention discloses a task cache allocation method and system in mobile augmented reality, including: obtaining a cache task waiting queue; allocating cache tasks to corresponding mobile terminals according to the preference values ​​of mobile terminals covered by mobile augmented reality devices for cache tasks in the cache task waiting queue, and determining the reward of cache allocation action; adjusting the size of mobile terminal cache space according to the value of the adjustment indicator and the cache task hit rate of the mobile terminal in the current decision time slice, and determining the reward of the cache space adjustment action; constructing a reward function according to the reward of the cache allocation action and the reward of the cache space adjustment action, constructing a target to maximize the cache hit rate, and training a cache allocation network in this way; using the trained cache allocation network to treat cache tasks to obtain a task cache allocation strategy. Effectively allocate the cache of mobile augmented reality tasks and improve the cache hit rate.

Description

Technical Field

[0001] The present invention relates to the field of mobile edge computing technology, and in particular to a method and system for allocating task cache in mobile augmented reality. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Mobile edge computing (MEC) is an emerging technology that provides cloud computing capabilities in the radio access network (RAN), very close to the terminal device. Mobile edge computing enables a wide variety of applications and services to run at the edge of the mobile network, not only reducing service latency but also alleviating congestion in other parts of the mobile core network.

[0004] Mobile Augmented Reality (MAR) is widely used in various scenarios, such as education, tourism, and visual assistance for special groups. Therefore, MAR as a future technology development direction has attracted more and more attention, and the market size is getting bigger and bigger. However, the most important challenge is that the computing resources on mobile devices are limited, such as the CPU, GPU, memory and battery of smartphones and smart glasses, while mobile augmented reality applications require a lot of computing resources to run computer vision calculations.

[0005] At present, the main method of calculating mobile augmented reality is to use the data processing capabilities of edge servers and cloud servers to offload computing tasks to the server. The server provides analysis of computer vision algorithms based on its powerful computing capabilities. However, server-based mobile augmented reality cannot respond quickly due to transmission delays, and fast response is an important indicator of mobile augmented reality. Therefore, while using server-based data processing, let the mobile device locally cache the recognition results processed by the server, avoiding the transmission of images to the server during repeated detection and reducing network delays.

[0006] If the content image preferences of mobile augmented reality tasks of multiple mobile terminals are similar, caching the recognition results to other mobile terminals with similar preferences can greatly improve the cache hit rate; at the same time, the cache space size of each mobile terminal also affects the cache search speed. If the cache is too small, it cannot store the required cache. If the cache is too large, it will take too long to search the cache. Summary of the invention

[0007] In order to solve the above problems, the present invention proposes a task cache allocation method and system in mobile augmented reality, which allocates cache tasks to mobile terminal devices with similar cache preferences, adjusts the cache space size, obtains corresponding rewards, constructs a reward function, and obtains the cache allocation strategy with the highest cache task hit rate by training the cache allocation network.

[0008] In order to achieve the above object, the present invention adopts the following technical solution:

[0009] In a first aspect, the present invention provides a method for allocating task cache in mobile augmented reality, comprising:

[0010] Get the cache task waiting queue;

[0011] According to the preference values ​​of the mobile terminals covered by the mobile augmented reality device for the cache tasks in the cache task waiting queue, the cache tasks are allocated to the corresponding mobile terminals, and the rewards for the cache allocation actions are determined;

[0012] According to the value of the adjustment indicator and the cache task hit rate of the mobile terminal in the current decision time slice, the size of the mobile terminal cache space is adjusted, and the reward for the cache space adjustment action is determined;

[0013] A reward function is constructed based on the reward of the cache allocation action and the reward of the cache space adjustment action, and the goal is constructed to maximize the cache hit rate to train the cache allocation network;

[0014] The cache tasks are treated with the trained cache allocation network to obtain the task cache allocation strategy.

[0015] As an optional implementation, the preference value for the cache task in the cache task waiting queue is:

[0016]

[0017] Among them, μ um is the preference value for the cache task, Typ(m, t k ) is the cache task m in the cache task topic t k The attribute function, preference function Pr(u, t k ) is the mobile terminal u for the subject t k preference; k is the kth topic, and K is the total number of topics.

[0018] As an optional implementation, the reward for the cache allocation action is: c (s, a) = Hit ul ·μ um ; Among them, r(s, a) is the reward of the cache allocation action, Hit ul represents the cache task hit rate, μum Indicates the preference value for the cache task, s is the state, and a is the action.

[0019] As an optional implementation, the reward for the cache space adjustment action is: in, Adjust the reward of the action for cache space, is the size of the cache space after adjustment, lc is the number of files in the cache space, s is the state, and a is the action.

[0020] As an optional implementation, the cache task hit rate is: Among them, h(ul) self represents the probability that mobile terminal u hits cache file l in its own cache space, u y Indicates the test result of mobile terminal u performing y caches.

[0021] As an optional implementation, the process of adjusting the size of the mobile terminal cache space includes setting an adjustment indicator and dividing the time step into multiple decision time slices. After executing the action in the decision time slice, the value of the adjustment indicator is increased by 1. If the value of the adjustment indicator is less than the total number of decision time slices and the cache task waiting queue is not empty, the cache allocation action is performed in the current decision time slice; if the value of the adjustment indicator increases to the total number of decision time slices and the cache task waiting queue is empty, the cache space size of the mobile terminal whose cache task hit rate is lower than the threshold is adjusted in the current decision time slice.

[0022] As an optional implementation, the reward function is constructed by weighting the reward for the cache allocation action and the reward for the cache space adjustment action respectively, and setting constraints, wherein the constraints are: the preference value of the cache task and the cache task hit rate are in the range of 0 to 1, the adjusted mobile terminal is optional, and the number of cache tasks does not exceed the size of the cache space.

[0023] In a second aspect, the present invention provides a task cache allocation system in mobile augmented reality, comprising:

[0024] A task acquisition module is configured to acquire a cache task waiting queue;

[0025] A first action module is configured to allocate cache tasks to corresponding mobile terminals according to preference values ​​of mobile terminals covered by the mobile augmented reality device for cache tasks in a cache task waiting queue, and determine a reward for a cache allocation action;

[0026] A second action module is configured to adjust the size of the mobile terminal cache space according to the value of the adjustment indicator and the cache task hit rate of the mobile terminal in the current decision time slice, and determine the reward of the cache space adjustment action;

[0027] A reward module, configured to construct a reward function according to a reward of a cache allocation action and a reward of a cache space adjustment action, so as to construct a target to maximize a cache hit rate, so as to train a cache allocation network;

[0028] The allocation module is configured to use the trained cache allocation network to obtain a task cache allocation strategy for the cache task.

[0029] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in the first aspect is performed.

[0030] In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, wherein when the computer instructions are executed by a processor, the method described in the first aspect is performed.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] The present invention proposes a task cache allocation method and system in mobile augmented reality, provides a more efficient cache task allocation and cache space adjustment strategy, effectively allocates the cache of mobile augmented reality tasks, unloads the cache to mobile terminals with similar cache preferences according to the preference value of the buffered tasks, and adjusts the size of the cache space to improve the cache hit rate and reduce the power consumption and time of the mobile terminal device.

[0033] The present invention proposes a method and system for task cache allocation in mobile augmented reality. A cache allocation network is constructed based on a deep reinforcement learning algorithm, cache tasks are allocated to mobile terminal devices with similar cache preferences, and the cache space size is adjusted to obtain corresponding rewards and construct a reward function. By training the cache allocation network, a cache allocation strategy with the highest cache task hit rate is obtained.

[0034] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0036] Figure 1 A schematic diagram of a task cache allocation method in mobile augmented reality provided by Embodiment 1 of the present invention;

[0037] Figure 2 A schematic diagram of the cache space provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0038] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0039] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0040] It should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "include" and "have" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0041] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0042] Example 1

[0043] At present, reinforcement learning algorithms are mostly used to solve resource allocation problems in MEC or IoT systems. Markov decision process (MDP) is a suitable theory to describe this problem. In order to solve the MDP problem, reinforcement learning provides a series of solutions, such as dynamic programming, Monte Carlo method, deep Q-network (DQN), etc. The general process of the DQN algorithm is: the agent performs actions based on the current state input, and then obtains the corresponding rewards and a new state. The original state, the executed action, the reward obtained, and the new state are stored in the memory pool, and the memory pool is repeatedly filled. During the training process, data is randomly extracted from the memory pool and the loss function is used to train the network parameters through the gradient descent method. Every fixed number of steps, the network parameters will be copied to another network. Using the characteristics of the DQN algorithm, a cache allocation strategy that maximizes the cache hit rate of MAR tasks in mobile edge computing scenarios is implemented.

[0044] Based on this, this embodiment provides a task cache allocation method in mobile augmented reality in a mobile edge computing scenario, with the goal of maximizing the cache hit rate, including the following steps: Figure 1 As shown:

[0045] Initialize the cache space of the mobile augmented reality device and the cache database of the mobile edge server.

[0046] The mobile augmented reality device sends the data processing task to the mobile edge server, and the mobile edge server sends the data processing result back to the mobile augmented reality device and stores the cache in the cache database.

[0047] Obtain a cache task waiting queue, and according to the preference values ​​of the mobile terminals covered by the mobile augmented reality device for the cache tasks in the cache task waiting queue, allocate the cache tasks to the mobile terminals with similar preference values, obtain corresponding rewards, and determine the rewards for the cache allocation action;

[0048] According to the value of the adjustment indicator and the cache task hit rate of the mobile terminal in the current decision time slice, the size of the mobile terminal cache space is adjusted, and the reward for the cache space adjustment action is determined;

[0049] A reward function is constructed based on the reward of the cache allocation action and the reward of the cache space adjustment action. A cache allocation network is constructed based on a deep reinforcement learning algorithm. The goal is to maximize the cache hit rate, so as to train the cache allocation network.

[0050] The cache tasks are treated with the trained cache allocation network to obtain the task cache allocation strategy.

[0051] In this embodiment, the cache task waiting queue is the cache tasks that are ranked high in the cache database; it can be understood that the selected ranking digit can be customized.

[0052] In this embodiment, the cache space of the MAR device is quantized into a matrix, and the cache tasks in the cache task waiting queue, the cache space matrix and the value of the adjustment indicator are used as the state input of the cache allocation network; the cache information of the mobile edge application deployed on the MEC system is assumed to be the state s∈S:

[0053]

[0054] Where C represents the cache space of the mobile terminal where the cache is placed, and the allocation of the cache space C is expressed as u×n c Matrix, n c ∈{1, 2, ..., N c} represents the amount of cache space of a single mobile terminal, u represents the number of nearby mobile terminals, represents a cache space; m∈{1, 2, ..., M} represents the cache task waiting queue of the mobile edge server; is the value of the adjustment indicator.

[0055] When a cache task with a preference close to the mobile terminal is allocated from the mobile edge server to the cache space of the mobile terminal, Indicates that this cache space is cached; among them, Indicates that MAR cache space can be allocated; Indicates that MAR cache space has been allocated; Indicates that MAR cache space cannot be allocated, such as Figure 2 shown.

[0056] In this embodiment, the tasks processed by the mobile edge server are stored in the cache database, the cache tasks with the highest ranking in the cache database are updated to the cache task waiting queue, and the preference values ​​of the cache tasks and the mobile terminals in the cache task waiting queue are calculated; specifically:

[0057] User preference reflects the user's interest in a content, and can also indirectly reflect the probability of a user requesting a content. The user's preference for content is closely related to the type of content. Assume that the cache task content has T topics, T = {t 1 , t 2 , ..., t k} represents all the topic sets of cache task content, and cache task m is in topic t k The property function is Typ(m, t k ), if cache task m contains topic t k , then Typ(m, t k ) is 1 if cache task m does not contain topic t k , then Typ(m, t k ) has a value of 0.

[0058] Users have their own preferences for each type of topic, and the preference function Pr(u, t k ) represents user u’s response to topic t k Preferences:

[0059]

[0060] Among them, X(t k ) is the subject W k The set of all items; I(X(t k );V j ) is the mutual information; p(X(t k )) is the unconditional feature probability, indicating that the entire content set contains feature w k The probability of p(X(t k )|V j ) is the conditional feature probability, that is, the historical information V of user u j Contains w k, u∈{1, 2, ..., U} represents the number of users covered by the mobile augmented reality device.

[0061] User u's preference value μ for cache task m um for:

[0062]

[0063] Among them, k is the kth topic and K is the total number of topics.

[0064] In this embodiment, the time step t is decoupled into multiple decision time slices ω. At the same time, in order to reduce the action space, two action functions are designed to reduce the action space a∈A=A α ∪A β It is divided into cache allocation action and cache space adjustment action, and selects tasks through the indicators of cache tasks in the cache task waiting queue and queue conditions.

[0065] Cache allocation action A α It refers to the action of placing cache tasks into the cache space. Each time step t corresponds to the cache to be placed. Taking this action means finding a suitable mobile terminal to place the cache for the current cache;

[0066]

[0067]

[0068]

[0069]

[0070] in, Indicates that the cache task is from Start putting it in u i up; u i ∈{1,...,U} indicates that it is cached in the user group, Indicates that the space occupied by the cache cannot exceed the user's remaining space. Indicates that the current cache space can be placed.

[0071] Cache space adjustment action Indicates that the mobile terminal u i The cache space is adjusted to

[0072]

[0073]

[0074] Where, when adjusting the indicator And when the cache task waiting queue is empty, adjust the hit rate Hit ul The number of mobile terminals u below the set value τ i The cache space size, i.e. Hit ul Mobile terminals with ≤τ; Refers to the adjustment indicator, To count the total number of decision time slices, each decision time slice Increment by 1; when When , schedule the cache task in the current decision time slice; when When the cache space of the mobile terminal is adjusted within the current decision time slice.

[0075] In this embodiment, in the time slice, through action A α or A β The state is transferred from state s to the next state s′, and the reward value is calculated as follows:

[0076]

[0077] For each decision time slice ω, the action space selection A α or A β , transfer state s to state s′, and in the transition state: 1) And the cache task waiting queue is not empty, that is, m 1 ≠0, select operation A α To assign cache actions. 2) And the cache task waiting queue is empty, that is, m 1 =0, select action A β To adjust the cache space action.

[0078] In this embodiment, the rewards of the cache allocation action and the cache space adjustment action are calculated, and a new state input is obtained at the same time; specifically:

[0079] The reward r(s, a) for the cache allocation action is a function of the state s and the action a:

[0080] r c (s, a) = Hit ul ·μ um ;

[0081] Among them, Hit ul represents the cache task hit rate, μ um Indicates preference for cache tasks.

[0082] Assume there are l files in the cache space, l∈{1, 2, ..., l c}; then the cache task hit rate is:

[0083] Among them, h(ul) self represents the probability that user u hits cache file l in its own cache space, l∈{1, 2, ..., l c};u y Indicates the test results of user u's recent y caches.

[0084] The reward for the cache space adjustment action is: Indicates the size of the cache space The reciprocal of .

[0085] Therefore, the reward function is designed as:

[0086]

[0087]

[0088] st um ∈{0, 1};

[0089] Hit ul ∈{0, 1};

[0090]

[0091] 0≤l c ≤N c ;

[0092] Among them, ω∈{0,1} is r c (s, a) and The weight coefficient between um and Hit ul Indicates that the cache task preference value and cache task hit rate range from 0 to 1; Indicates any choice in the mobile terminal device set; 0≤l c ≤N c Indicates that the number of cache files cannot exceed the size of the entire cache space.

[0093] This embodiment aims to maximize cache preference and hit rate, and maximizes cache utilization and improves mobile terminal detection efficiency by placing cache tasks in mobile terminals with similar preferences. The original state, executed actions, rewards obtained, and new states are sent to the mobile edge server and stored, and the network is trained on the mobile edge server.

[0094] Training samples are randomly selected from the mobile edge server, and the network parameters are trained through the loss function until the training is completed and the data converges; the loss function is

[0095]

[0096] Therefore, in order to maximize the cache hit rate and cache preference, the cache allocation strategy is used to allocate cache tasks to devices with similar cache file preferences. When the cache is on a device with relevant preferences and the hit rate of the last few cache accesses is high, a larger reward will be received. Otherwise, it means that the cache is offloaded to a MAR device with irrelevant preferences, and a small reward will be received. Through continuous accumulation, the mobile edge server will learn a cache allocation strategy that maximizes the cache hit rate of cache tasks.

[0097] Example 2

[0098] This embodiment provides a task cache allocation system in mobile augmented reality, including:

[0099] A task acquisition module is configured to acquire a cache task waiting queue;

[0100] A first action module is configured to allocate cache tasks to corresponding mobile terminals according to preference values ​​of mobile terminals covered by the mobile augmented reality device for cache tasks in a cache task waiting queue, and determine a reward for a cache allocation action;

[0101] A second action module is configured to adjust the size of the mobile terminal cache space according to the value of the adjustment indicator and the cache task hit rate of the mobile terminal in the current decision time slice, and determine the reward of the cache space adjustment action;

[0102] A reward module, configured to construct a reward function according to a reward of a cache allocation action and a reward of a cache space adjustment action, so as to construct a target to maximize a cache hit rate, so as to train a cache allocation network;

[0103] The allocation module is configured to use the trained cache allocation network to obtain a task cache allocation strategy for the cache task.

[0104] It should be noted that the above modules correspond to the steps described in Example 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above Example 1. It should be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer executable instructions.

[0105] In further embodiments, there is also provided:

[0106] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in Embodiment 1 is performed. For the sake of brevity, it will not be described in detail here.

[0107] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0108] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0109] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the method described in Example 1 is completed.

[0110] The method in Example 1 can be directly embodied as a hardware processor, or a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it is not described in detail here.

[0111] Those skilled in the art will appreciate that the units, i.e., algorithm steps, of the various examples described in conjunction with this embodiment can be implemented in electronic hardware or in a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0112] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A task cache allocation method in mobile augmented reality, characterized in that: include: Get the cache task waiting queue; According to the preference values ​​of the mobile terminals covered by the mobile augmented reality device for the cache tasks in the cache task waiting queue, the cache tasks are allocated to the corresponding mobile terminals, and the rewards for the cache allocation actions are determined; The cache allocation action A α It refers to the action of placing cache tasks into the cache space. Each time step t corresponds to the cache to be placed. Taking this action means finding a suitable mobile terminal for the current cache to place the cache; in, Indicates that the cache task is from Start putting it in u i up; u i ∈{1,…,U} indicates that it is cached in the user group, Indicates that the space occupied by the cache cannot exceed the user's remaining space. Indicates that the current cache space can be placed; According to the value of the adjustment indicator and the cache task hit rate of the mobile terminal in the current decision time slice, the size of the mobile terminal cache space is adjusted, and the reward for the cache space adjustment action is determined; The cache task hit rate is: Among them, h(ul) self represents the probability that mobile terminal u hits cache file l in its own cache space, u y Indicates the test result of mobile terminal u performing y caches; The cache space adjustment action Indicates that the mobile terminal u i The cache space is adjusted to Among them, when adjusting the indicator And when the cache task waiting queue is empty, adjust Hit ul Mobile terminals with ≤τ; Refers to the adjustment indicator, To count the total number of decision time slices, each decision time slice Increment by 1; when When , schedule the cache task in the current decision time slice; when When , in the current decision time slice, adjust the size of the mobile terminal cache space; The process of adjusting the size of the mobile terminal cache space includes setting an adjustment indicator, and dividing the time step into a plurality of decision time slices. After executing an action in the decision time slice, the value of the adjustment indicator is increased by 1. If the value of the adjustment indicator is less than the total number of decision time slices and the cache task waiting queue is not empty, the cache allocation action is executed in the current decision time slice; if the value of the adjustment indicator increases to the total number of decision time slices and the cache task waiting queue is empty, the cache space size of the mobile terminal whose cache task hit rate is lower than the threshold is adjusted in the current decision time slice; A reward function is constructed based on the reward of the cache allocation action and the reward of the cache space adjustment action, and the goal is constructed to maximize the cache hit rate to train the cache allocation network; The reward for the cache allocation action is: c (s,a)=Hit ul ·μ um ; Among them, r(s,a) is the reward of cache allocation action, Hit ul represents the cache task hit rate, μ um Indicates the preference value for the cache task, s is the state, and a is the action; The reward for the cache space adjustment action is: in, Adjust the reward of the action for cache space, is the size of the cache space after adjustment, l c There are l files in the cache space; s is the state, a is the action; The trained task cache allocation network is used to obtain the task cache allocation strategy for cache tasks.

2. A method for allocating task cache in mobile augmented reality according to claim 1, characterized in that: The preference values ​​for cache tasks in the cache task waiting queue are: Among them, μ um is the preference value for the cache task, Typ(m, t k ) is the cache task m in the cache task topic t k The attribute function, preference function Pr(u, t k ) is the mobile terminal u for the subject t k preference; k is the kth topic, and K is the total number of topics.

3. The method for allocating task cache in mobile augmented reality according to claim 1, characterized in that: The reward function is constructed by weighting the reward of the cache allocation action and the reward of the cache space adjustment action respectively, and setting constraints. The constraints are: the preference value of the cache task and the cache task hit rate range between 0 and 1, the adjusted mobile terminal is optional, and the number of cache tasks does not exceed the size of the cache space.

4. A task cache allocation system in mobile augmented reality, characterized in that: include: A task acquisition module is configured to acquire a cache task waiting queue; A first action module is configured to allocate cache tasks to corresponding mobile terminals according to preference values ​​of mobile terminals covered by the mobile augmented reality device for cache tasks in a cache task waiting queue, and determine a reward for a cache allocation action; The cache allocation action A α It refers to the action of placing cache tasks into the cache space. Each time step t corresponds to the cache to be placed. Taking this action means finding a suitable mobile terminal for the current cache to place the cache; in, Indicates that the cache task is from Start putting it in u i up; u i ∈{1,…,U} indicates that it is cached in the user group, Indicates that the space occupied by the cache cannot exceed the user's remaining space. Indicates that the current cache space can be placed; A second action module is configured to adjust the size of the mobile terminal cache space according to the value of the adjustment indicator and the cache task hit rate of the mobile terminal in the current decision time slice, and determine the reward of the cache space adjustment action; The cache task hit rate is: Among them, h(ul) self represents the probability that mobile terminal u hits cache file l in its own cache space, u y Indicates the test result of mobile terminal u performing y caches; The cache space adjustment action Indicates that the mobile terminal u i The cache space is adjusted to Among them, when adjusting the indicator And when the cache task waiting queue is empty, adjust Hit ul Mobile terminals with ≤τ; Refers to the adjustment indicator, To count the total number of decision time slices, each decision time slice Increment by 1; when When , schedule the cache task in the current decision time slice; when When , in the current decision time slice, adjust the size of the mobile terminal cache space; The process of adjusting the size of the mobile terminal cache space includes setting an adjustment indicator, and dividing the time step into a plurality of decision time slices. After executing an action in the decision time slice, the value of the adjustment indicator is increased by 1. If the value of the adjustment indicator is less than the total number of decision time slices and the cache task waiting queue is not empty, the cache allocation action is executed in the current decision time slice; if the value of the adjustment indicator increases to the total number of decision time slices and the cache task waiting queue is empty, the cache space size of the mobile terminal whose cache task hit rate is lower than the threshold is adjusted in the current decision time slice; A reward module, configured to construct a reward function according to a reward of a cache allocation action and a reward of a cache space adjustment action, so as to construct a target to maximize a cache hit rate, so as to train a cache allocation network; The reward for the cache allocation action is: c (s,a)=Hit ul ·μ um ; Among them, r(s,a) is the reward of cache allocation action, Hit ul represents the cache task hit rate, μ um Indicates the preference value for the cache task, s is the state, and a is the action; The reward for the cache space adjustment action is: in, Adjust the reward of the action for cache space, is the size of the cache space after adjustment, l c There are l files in the cache space; s is the state, a is the action; The allocation module is configured to use the trained cache allocation network to obtain a task cache allocation strategy for the cache task.

5. An electronic device, characterized in that: The invention comprises a memory and a processor and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method according to any one of claims 1 to 3 is completed.

6. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the method described in any one of claims 1 to 3.

Images