Mobile edge computing task unloading method and system, processing equipment and storage medium

Abstract

The invention relates to a mobile edge computing task unloading method and system, a processing device and a storage medium, and the method is characterized in that the method comprises the steps: obtaining a mobile edge computing task of a to-be-unloaded mobile device; the obtained mobile edge computing task is input into a pre-constructed DQN model, a task unloading result of the mobile device is obtained, and the task unloading result comprises the steps that the mobile edge computing task is directly processed locally, unloaded to an MEC server or forwarded to an edge server; wherein the DQN model is optimized based on the genetic algorithm, the genetic algorithm is used for guiding training of the DQN algorithm, the convergence speed can be remarkably increased, the performance and stability of the algorithm can be effectively improved, and the method can be widely applied to the fields of computers and automation.

Description

Technical Field

[0001] The present invention relates to the field of computers and automation, and in particular to a mobile edge computing task offloading method, system, processing device and storage medium. Background Art

[0002] With the rapid development of Internet of Things technology, the popularity of mobile devices and the improvement of their performance, running new and complex mobile edge computing tasks on mobile devices is not only more sensitive to latency, but also brings about a significant increase in computing and energy pressure. In this context, mobile edge computing has emerged. Compared with the traditional cloud computing model, mobile edge computing can effectively improve the response speed, reduce the network load, and provide a better user experience by using mobile edge computing to perform computing tasks at the edge of the network.

[0003] As the underlying supporting technology of mobile edge computing, task offloading is of great importance. It is responsible for offloading mobile edge computing tasks from mobile devices to edge servers or the cloud, which not only greatly improves the execution efficiency of mobile edge computing tasks and the performance of mobile applications, but also reduces energy consumption and latency. However, when specifying task offloading strategies, many factors should be considered, such as latency, energy consumption, mobility, etc. Due to the mobility of smart devices, the heterogeneity of mobile edge computing tasks (i.e., different mobile edge computing tasks require different computing resources and network resources), and the limited and heterogeneous edge node resources, it is known that the task offloading strategy solution process is an NP-hard problem. Therefore, how to comprehensively consider the optimization objectives and quickly solve the optimal offloading strategy has always been a hot topic of research.

[0004] At present, the existing mobile edge computing task offloading mostly considers latency or energy consumption issues separately, without organically combining the two, and ignoring the consideration of capacity. In addition, in the solution process, most of the heuristic algorithms are used alone, but they are prone to fall into local optimal solutions and have poor convergence speed. When the scene changes, the algorithm must be redesigned and calculated, which will cause a lot of time waste. Summary of the invention

[0005] In view of the above problems, the object of the present invention is to provide a mobile edge computing task offloading method, system, processing device and storage medium that can effectively reduce latency and overall energy consumption.

[0006] To achieve the above objectives, the present invention adopts the following technical solutions: In a first aspect, a mobile edge computing task offloading method is provided, comprising: Obtaining a mobile edge computing task of a mobile device to be offloaded; Input the acquired mobile edge computing task into the pre-built DQN model to obtain the task offloading result of the mobile device, where the task offloading result includes directly processing the mobile edge computing task locally, offloading it to the MEC server, or forwarding it to the edge server; Wherein, the DQN model is optimized based on a genetic algorithm.

[0007] Furthermore, based on the genetic algorithm, the specific process of optimizing the DQN model is as follows: Randomly generate a set of DRL algorithm parameter settings or strategies as the initial individuals of the population; Build the DQN model; Based on the obtained training data of the mobile edge computing task offloading problem, the constructed DQN model is trained; Based on the genetic algorithm, the trained DQN model is optimized and the parameters of the DQN model are updated; Check whether the training reaches the preset termination condition. If the termination condition is met, stop the training and output the optimized DQN model. Otherwise, retrain the constructed DQN model.

[0008] Furthermore, the DQN model includes a state space, an action space and a reward function; The state space includes mobile device status, MEC server status, network status, mobile edge computing task status and historical information; The action space represents offloading tasks to edge servers when there is When there is a mobile edge computing task, the action Represents a mobile edge computing task Migrate to edge servers superior:

[0009] The reward function is:

[0010] in, Are reward values, , and | ; , Indicates that the network status has improved. Indicates that the network status has deteriorated. Indicates that the network status remains unchanged.

[0011] Furthermore, the genetic algorithm is used to optimize the trained DQN model, and the parameters of the DQN model are updated, including: The population is initialized to randomly generate h chromosomes; Evaluate the DQN model based on the fitness function; Based on the evaluation results, individuals are selected as parents to reproduce the next generation; Select two parents for crossover to generate new individuals; Perform mutation operations on the selected individuals to generate new individuals; Update the parameters of the DQN model based on the generated new individuals.

[0012] Furthermore, the fitness function is:

[0013] Among them, when When , it indicates the mobile edge computing task Processed locally; otherwise, ;when When , it indicates the mobile edge computing task Processed on the main MEC server; otherwise, the mobile edge computing task will be forwarded according to resource and time constraints to the nearest edge server in the collaboration domain or remote server; Calculate the number of tasks for mobile edge computing; is the number of mobile devices; For mobile edge computing tasks Energy consumption when executed locally; is the number of MEC servers; Mobile edge computing tasks Total energy consumption; Limited by:

[0014]

[0015]

[0016]

[0017]

[0018] in, It is a set of mobile edge computing tasks; For MEC server Assign to task computing resources; is the total computing resources; for overall delay; For mobile edge computing tasks Maximum allowed delay; parameter When , it indicates the mobile edge computing task Offloaded to MEC server via subcarrier ,on the contrary .

[0019] Furthermore, the energy consumption model of the task executed locally is:

[0020] in, is the energy consumption per CPU cycle of the mobile device; is the CPU clock frequency of the mobile device; For mobile edge computing tasks workload; The energy consumption model of tasks executed on the MEC server is:

[0021] in, For mobile edge computing tasks Energy consumption of uplink transmission; is the communication energy consumption between MEC servers; It is the energy consumption of executing mobile edge computing tasks on MEC servers.

[0022] In a second aspect, a mobile edge computing task offloading system is provided, comprising: A task acquisition module, used to acquire the mobile edge computing task of the mobile device to be offloaded; A task offloading result determination module is used to input the acquired mobile edge computing task into a pre-built DQN model to obtain the task offloading result of the mobile device, where the task offloading result includes directly processing the mobile edge computing task locally, offloading it to the MEC server, or forwarding it to the edge server; Wherein, the DQN model is optimized based on a genetic algorithm.

[0023] Furthermore, based on the genetic algorithm, the specific process of optimizing the DQN model is as follows: Randomly generate a set of DRL algorithm parameter settings or strategies as the initial individuals of the population; Build the DQN model; Based on the obtained training data of the mobile edge computing task offloading problem, the constructed DQN model is trained; Based on the genetic algorithm, the trained DQN model is optimized and the parameters of the DQN model are updated; Check whether the training reaches the preset termination condition. If the termination condition is met, stop the training and output the optimized DQN model. Otherwise, retrain the constructed DQN model.

[0024] In a third aspect, a processing device is provided, comprising computer program instructions, wherein the computer program instructions, when executed by the processing device, are used to implement the steps corresponding to the above-mentioned mobile edge computing task offloading method.

[0025] In a fourth aspect, a computer-readable storage medium is provided, characterized in that computer program instructions are stored on the computer-readable storage medium, wherein the computer program instructions are used to implement the steps corresponding to the above-mentioned mobile edge computing task offloading method when executed by the processor.

[0026] The present invention adopts the above technical solution, which has the following advantages: 1. The present invention utilizes a decision model with task offloading knowledge to make efficient task offloading decisions in a timely manner according to changes in network status, which can effectively reduce latency and overall energy consumption and improve user experience.

[0027] 2. The present invention optimizes multiple objectives while minimizing energy consumption under the premise of satisfying capacity and delay constraints, and uses a genetic algorithm to guide the DQN algorithm to learn a better task offloading knowledge model. It can not only effectively improve the quality of the solution, but also solve the inherent disadvantage that the genetic algorithm will take a lot of time to iterate when the scene changes. Ultimately, it can not only reduce the latency and overall energy consumption, but also effectively improve the user experience.

[0028] 3. Compared with the reinforcement learning algorithm alone, the present invention not only combines experience and domain knowledge, but also effectively reduces the demand for training samples and the sensitivity to low hyperparameters.

[0029] 4. The present invention uses genetic algorithms to guide the training of the DQN algorithm, which can not only significantly increase the convergence speed, but also effectively improve the performance and stability of the algorithm. The learned decision model with task offloading knowledge can be used to achieve fast and efficient task offloading. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Throughout the accompanying drawings, the same reference numerals are used to represent the same components. In the accompanying drawings: Figure 1 It is a schematic diagram of a DQN model optimization process provided by an embodiment of the present invention; Figure 2 Schematic diagram of the training process of the DQN model provided by one embodiment of the present invention; Figure 3 It is a schematic diagram of the network structure of a current value network or a target value network in a DQN model provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0031] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0032] It should be understood that the terms used in the text are only for the purpose of describing specific example embodiments, and are not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "one", "an" and "said" as used in the text may also be meant to include plural forms. The terms "include", "comprise", "contain", and "have" are inclusive, and therefore specify the existence of stated features, steps, operations, elements and / or parts, but do not exclude the existence or addition of one or more other features, steps, operations, elements, parts, and / or combinations thereof. The method steps, processes, and operations described herein are not interpreted as necessarily requiring them to be performed in the specific order described or illustrated, unless the execution order is clearly indicated. It should also be understood that additional or alternative steps may be used.

[0033] Although the terms first, second, third, etc. can be used in the text to describe multiple elements, components, regions, layers and / or sections, these elements, components, regions, layers and / or sections should not be limited by these terms. These terms can only be used to distinguish an element, component, region, layer or section from another region, layer or section. Unless the context clearly indicates, terms such as "first", "second" and other numerical terms do not imply order or sequence when used in the text. Therefore, the first element, component, region, layer or section discussed below can be referred to as the second element, component, region, layer or section without departing from the teaching of the example embodiments.

[0034] At present, the existing mobile edge computing task offloading mostly considers latency or energy consumption issues separately, without organically combining the two, and ignoring the consideration of capacity. In addition, in the solution process, a single heuristic algorithm is often used, but it is easy to fall into the local optimal solution and the convergence speed is poor. When the scene changes, the algorithm must be redesigned and calculated, which will cause a lot of time waste. An embodiment of the present invention provides a mobile edge computing task offloading method, including: obtaining a mobile edge computing task of a mobile device to be offloaded; inputting the obtained mobile edge computing task into a pre-built DQN model to obtain a task offloading result of the mobile device, and the task offloading result includes directly processing the mobile edge computing task locally, offloading it to the MEC server or forwarding it to the edge server; wherein the DQN model is optimized based on a genetic algorithm. The present invention not only weighs multiple optimization objectives, but also uses a genetic algorithm to guide the training of DQN, and then optimizes the DQN algorithm model, so that the acquisition of the final task offloading decision is not only optimized by the genetic algorithm, but also does not need to go through the convergence process, effectively reducing the latency. In addition, compared with the use of reinforcement learning algorithms alone, it not only combines experience and domain knowledge, but also effectively reduces the demand for training samples and the sensitivity to low hyperparameters.

[0035] Example 1 This embodiment provides a mobile edge computing task offloading method, including the following steps: 1) If Figure 1 As shown, a DQN (an algorithm combining deep learning and reinforcement learning) model is constructed, and the constructed DQN model is optimized based on a genetic algorithm to obtain an optimized DQN model, specifically: 1.1) Initialize the population: Randomly generate a set of parameter settings or strategies of the DRL (deep reinforcement learning) algorithm as the initial individuals of the population.

[0036] 1.2) If Figure 2 As shown, build a DQN model.

[0037] Specifically, the DQN model includes state space, action space and reward function, where: State space: includes mobile device status (available computing power, memory capacity, current load, etc. of edge devices), MEC server (edge ​​server) status, network status (network latency, bandwidth, connection quality, etc.), mobile edge computing task status (computing requirements, memory requirements, execution time, etc. of mobile edge computing tasks), historical information (previous decision and reward history, used to capture sequence information), etc.

[0038] Action space: represents offloading tasks to edge servers. When there is a mobile edge computing task, the action Represents a mobile edge computing task Migrate to edge servers superior: (1) Reward function: The value of the reward function set in this embodiment is related to the state of the container cluster. When the current container cluster state improves after executing a certain action, a positive reward is given. Otherwise, a negative reward is given .

[0039] (2) in, Are reward values, , and | , the absolute value inequality of the two is to reduce invalid operations; , Indicates that the network status has improved. Indicates that the network status has deteriorated. Indicates that the network status remains unchanged.

[0040] This embodiment is designed as Figure 3 The network structure shown in the figure includes a convolutional neural network and a fully connected layer, which can effectively capture the spatial structure and features in the input data. The target value network structure is the same as the current value network structure, and their inputs are both feature matrices, and their outputs are both the Q values ​​of each action in the action space under the current state.

[0041] 1.3) Training the DQN model: Based on the obtained training data of the mobile edge computing task offloading problem, the constructed DQN model is trained.

[0042] 1.4) Based on the genetic algorithm, the trained DQN model is optimized, the parameters of the DQN model are updated, and the optimized DQN model is obtained. The genetic algorithm is an optimization algorithm based on natural selection and genetic mechanism. It simulates the biological evolution process in nature and finds the optimal solution to the problem through continuous evolution and selection. The algorithm process includes initialization, fitness evaluation, selection, crossover, mutation operation and population iteration. Specifically: 1.4.1) Initialization: The population is initialized to randomly generate h chromosomes, where the length of each chromosome is 3, corresponding to the three parameters of neural network weight, learning rate and discount factor, and their corresponding value ranges are between [-1, 1], [0.001, 0.1], and [0.9, 0.99] respectively.

[0043] 1.4.2) Determine the fitness function in advance. The specific process is as follows: ①Determine the system model MEC Server Collection , its total computing capacity is , mobile device collection , mobile edge computing task set Each mobile edge computing task is composed of four tuples, namely, mobile edge computing tasks ,in, For mobile edge computing tasks The amount of data, For mobile edge computing tasks workload, For mobile edge computing tasks The maximum delay allowed, For mobile edge computing tasks The number of CPU clocks required for each bit of data. In addition, let For MEC server Assigned to mobile edge computing tasks computing resources.

[0044] ②Determine the communication rate Communication rate For mobile devices To the MEC server The data rate is: (3) Communication rate For MEC server With other MEC servers The data rate is: (4) in, is the network bandwidth; For mobile devices The transmission power; For mobile devices and MEC server The channel gain between The interference caused by other mobile devices uploading mobile edge computing tasks to the MEC server; Interference caused by communication between different MEC servers; is the noise variance.

[0045] ③ Perform local calculations Latency model for tasks executed locally (i.e., on mobile devices): Mobile edge computing tasks Latency of local execution for: (5) in, is the CPU clock frequency of the mobile device.

[0046] Energy consumption model of tasks executed locally: mobile edge computing tasks Energy consumption when executing locally for: (6) in, is the energy consumption per CPU cycle of the mobile device.

[0047] ④Unload tasks When the mobile edge computing task is offloaded to the MEC server for execution, its total delay includes the mobile edge computing task upload delay, forwarding delay, execution delay and delivery delay, but the delivery delay is negligible compared to the upload delay.

[0048] The latency model of tasks executed on MEC servers is as follows: This embodiment sets that the main MEC server can only process the mobile edge computing task when it has all the resources to complete the mobile edge computing task. Otherwise, the mobile edge computing task will be forwarded to other nearest MEC servers or remote servers in the collaborative domain according to capacity and proximity constraints, so as to avoid the mobile edge computing task waiting for too long on a certain MEC server, resulting in resource consumption and reduced user experience.

[0049] when When , it indicates the mobile edge computing task Processed locally; otherwise, .when When , it indicates the mobile edge computing task Processed on the main MEC server; otherwise, the mobile edge computing task will be forwarded according to resource and time constraints To the nearest edge server in the collaboration domain or remote server.

[0050] Transmission delay For mobile edge computing tasks From a mobile device Transmit to MEC server Delay in the process: (7) Forwarding delay For mobile edge computing tasks From the main MEC server Forward to a nearby edge server in the collaboration domain The resulting communication delay is: (8) in, is the link capacity connecting adjacent MEC servers.

[0051] Execution delay For mobile edge computing tasks On the main MEC server Execution time on: (9) in, The CPU clock frequency of the MEC server.

[0052] The overall delay for: (10) Energy consumption model of tasks executed on MEC servers: For mobile edge computing tasks Energy consumption of uplink transmission: (11) in, For mobile devices Transferring mobile edge computing tasks to MEC servers The power required; For mobile devices Transferring mobile edge computing tasks to MEC servers The time required.

[0053] The energy consumption of communication between MEC servers is: (12) in, For MEC server and MEC server The power required for communication overhead between the mobile edge computing tasks If it is not forwarded to other MEC servers, =0.

[0054] Energy consumption for executing mobile edge computing tasks on MEC servers: (13) in, For each MEC server Energy consumption of CPU cycles; It is the CPU clock frequency of the MEC server.

[0055] Mobile edge computing tasks Total energy consumption for: (14) ⑤Determine the optimization objective function, that is, the fitness function This embodiment ultimately solves a multi-objective optimization problem, and its ultimate optimization goal is to minimize energy consumption while satisfying capacity and delay constraints. The fitness function is: (15) Limited by: (16) (17) (18) (19) (20) in, For MEC server Assign to task computing resources; is the total computing resources; For mobile edge computing tasks The maximum delay allowed; when the parameter When , it indicates the mobile edge computing task Offloaded to MEC server via subcarrier ,on the contrary .

[0056] Formula (16) limits that only one mobile device can perform mobile edge computing tasks. ; Formula (17) limits that only one MEC server can execute the task offloading strategy in the collaborative domain; Formula (18) ensures that the allocated computing resources cannot exceed the total computing resources: Formula (19) ensures that the mobile edge computing task The total delay of the mobile edge computing task must not exceed the maximum allowed delay; Formula (20) represents It can only be executed on one MEC server at most, where the mobile edge computing task allocation matrix is .

[0057] 1.4.3) Fitness evaluation: The DQN model is evaluated based on the fitness function, where the fitness function reflects the comprehensive consideration of latency, capacity, and overall energy consumption.

[0058] Specifically, the fitness value of the solution is calculated to evaluate its quality. That is, under the premise of satisfying capacity and delay constraints, the smaller the energy consumption of task offloading, the higher the fitness value: (twenty one) in, is the fitness value; for a specific candidate solution; For the task Energy consumption when executed locally; When indicating a task Process locally, otherwise ; When indicating a task Processed on the main MEC server, otherwise the task will be forwarded according to resource and time constraints to the nearest edge server in the collaboration domain or remote server; is the number of tasks; is the number of MEC servers.

[0059] 1.4.4) Based on the evaluation results, individuals are selected as parents to reproduce the next generation, where the probability of each individual being selected is proportional to its own fitness: (twenty two) in, for The probability of a parent reproducing; is the total number of candidate solutions.

[0060] 1.4.5) Select two parents for crossover to generate new individuals.

[0061] Specifically, let the two parent individuals , To cross, select the intersection point ,in, is the number of genes on a chromosome, and the new individuals are generated as: (twenty three) (twenty four) in, , For new individuals.

[0062] Specifically, in order to avoid the problem that the genes with high fitness are quickly destroyed due to the increase in the number of iterations, the present invention sets formula (7): (25) in, For gene position The probability of mutation,when the number of iterations is small, the probability of mutation is fixed, but as the number of iterations increases, the probability of mutation decreases; is a fixed value, is the preset number of iterations.

[0063] 1.4.6) Perform mutation operations on the selected individuals to generate new individuals.

[0064] Specifically, mutation operations are performed on the selected individuals to avoid falling into local optimal solutions too early and generate new individuals. : (26) in, Chromosome No. The value of each gene position; Chromosome No. The probability of this formula occurring is .

[0065] 1.4.7) Update the parameters of the DQN model according to the generated new individuals to obtain an optimized DQN model to further improve the model performance.

[0066] 1.5) Check the termination condition: Check whether the training reaches the preset termination condition. If the termination condition is met, stop the training and output the optimized DQN model, which can be used to solve the problem of mobile edge computing task offloading; otherwise, return to step 1.3) to retrain the constructed DQN model.

[0067] 2) Obtain the mobile edge computing tasks of the mobile device to be offloaded.

[0068] 3) Input the obtained mobile edge computing task into the optimized DQN model to obtain the task offloading result of the mobile device. The task offloading result includes directly processing the mobile edge computing task locally, offloading it to the MEC server, or forwarding it to the edge server.

[0069] Example 2 This embodiment provides a mobile edge computing task offloading system, including: The task acquisition module is used to obtain the mobile edge computing tasks of the mobile device to be offloaded.

[0070] The task offloading result determination module is used to input the acquired mobile edge computing task into the pre-built DQN model to obtain the task offloading result of the mobile device. The task offloading result includes directly processing the mobile edge computing task locally, offloading it to the MEC server, or forwarding it to the edge server.

[0071] Wherein, the DQN model is optimized based on a genetic algorithm.

[0072] In a preferred embodiment, the specific process of optimizing the DQN model based on the genetic algorithm is as follows: A set of parameter settings or strategies of the DRL algorithm is randomly generated as the initial individuals of the population.

[0073] Build the DQN model.

[0074] The constructed DQN model is trained based on the acquired training data of the mobile edge computing task offloading problem.

[0075] Based on the genetic algorithm, the trained DQN model is optimized and the parameters of the DQN model are updated.

[0076] Check whether the training reaches the preset termination condition. If the termination condition is met, stop the training and output the optimized DQN model. Otherwise, retrain the constructed DQN model.

[0077] The system provided in this embodiment is used to execute the above-mentioned method embodiments. Please refer to the above-mentioned embodiments for specific processes and detailed contents, which will not be repeated here.

[0078] Example 3 This embodiment provides a processing device corresponding to the mobile edge computing task offloading method provided in Embodiment 1. The processing device can be applicable to a client processing device, such as a mobile phone, a laptop computer, a tablet computer, a desktop computer, etc., to execute the method of Embodiment 1.

[0079] The processing device includes a processor, a memory, a communication interface and a bus, and the processor, the memory and the communication interface are connected through the bus to complete mutual communication. The memory stores a computer program that can be run on the processing device, and the processing device executes the mobile edge computing task offloading method provided in this embodiment 1 when running the computer program.

[0080] In some implementations, the memory may be a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk storage.

[0081] In some other implementations, the processor may be a central processing unit (CPU), a digital signal processor (DSP), or other general-purpose processors of various types, which are not limited herein.

[0082] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0083] Those skilled in the art will understand that the structure of the above-mentioned computing device is only a partial structure related to the solution of the present invention, and does not constitute a limitation on the computing device to which the solution of the present invention is applied. The specific computing device may include more or fewer components, or combine certain components, or have a different arrangement of components.

[0084] Example 4 This embodiment provides a computer program product corresponding to the mobile edge computing task offloading method provided in Embodiment 1. The computer program product may include a computer-readable storage medium on which computer-readable program instructions for executing the mobile edge computing task offloading method described in Embodiment 1 are loaded.

[0085] Computer readable storage media can be tangible devices that hold and store instructions used by instruction execution devices. Computer readable storage media can be, for example, but not limited to, electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any combination thereof.

[0086] The above embodiment provides a computer-readable storage medium, whose implementation principle and technical effect are similar to those of the above method embodiment, and will not be repeated here.

[0087] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0088] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0089] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0090] The above embodiments are only used to illustrate the present invention, wherein the structure, connection mode and manufacturing process of each component may be changed. Any equivalent transformations and improvements based on the technical solution of the present invention should not be excluded from the protection scope of the present invention.

Claims

1. A mobile edge computing task offloading method, characterized in that: include: Obtaining a mobile edge computing task of a mobile device to be offloaded; Input the acquired mobile edge computing task into the pre-built DQN model to obtain the task offloading result of the mobile device, where the task offloading result includes directly processing the mobile edge computing task locally, offloading it to the MEC server, or forwarding it to the edge server; Wherein, the DQN model is optimized based on a genetic algorithm.

2. A mobile edge computing task offloading method according to claim 1, characterized in that: Based on the genetic algorithm, the specific process of optimizing the DQN model is as follows: Randomly generate a set of DRL algorithm parameter settings or strategies as the initial individuals of the population; Build the DQN model; Based on the obtained training data of the mobile edge computing task offloading problem, the constructed DQN model is trained; Based on the genetic algorithm, the trained DQN model is optimized and the parameters of the DQN model are updated; Check whether the training reaches the preset termination condition. If the termination condition is met, stop the training and output the optimized DQN model. Otherwise, retrain the constructed DQN model.

3. A mobile edge computing task offloading method as claimed in claim 2, characterized in that: The DQN model includes a state space, an action space, and a reward function; The state space includes mobile device status, MEC server status, network status, mobile edge computing task status and historical information; The action space represents offloading tasks to edge servers when there is When there is a mobile edge computing task, the action Represents a mobile edge computing task Migrate to edge servers superior: The reward function is: in, Are reward values, , and | ; , Indicates that the network status has improved. Indicates that the network status has deteriorated. Indicates that the network status remains unchanged.

4. A mobile edge computing task offloading method as claimed in claim 2, characterized in that: The method of optimizing the trained DQN model based on the genetic algorithm and updating the parameters of the DQN model includes: The population is initialized to randomly generate h chromosomes; Evaluate the DQN model based on the fitness function; Based on the evaluation results, individuals are selected as parents to reproduce the next generation; Select two parents for crossover to generate new individuals; Perform mutation operations on the selected individuals to generate new individuals; Update the parameters of the DQN model based on the generated new individuals.

5. A mobile edge computing task offloading method as claimed in claim 4, characterized in that: The fitness function is: Among them, when When , it indicates the mobile edge computing task Processed locally; otherwise, ;when When , it indicates the mobile edge computing task Processed on the main MEC server; otherwise, the mobile edge computing task will be forwarded according to resource and time constraints to the nearest edge server in the collaboration domain or remote server; Calculate the number of tasks for mobile edge computing; is the number of mobile devices; For mobile edge computing tasks Energy consumption when executed locally; is the number of MEC servers; Mobile edge computing tasks Total energy consumption; Limited by: in, It is a set of mobile edge computing tasks; For MEC server Assign to task computing resources; is the total computing resources; for overall delay; For mobile edge computing tasks Maximum allowed delay; parameter When , it indicates the mobile edge computing task Offloaded to MEC server via subcarrier ,on the contrary .

6. A mobile edge computing task offloading method as claimed in claim 5, characterized in that: The energy consumption model of the task executed locally is: in, is the energy consumption per CPU cycle of the mobile device; is the CPU clock frequency of the mobile device; For mobile edge computing tasks workload; The energy consumption model of tasks executed on the MEC server is: in, For mobile edge computing tasks Energy consumption of uplink transmission; is the communication energy consumption between MEC servers; It is the energy consumption of executing mobile edge computing tasks on MEC servers.

7. A mobile edge computing task offloading system, characterized in that: include: A task acquisition module, used to acquire the mobile edge computing task of the mobile device to be offloaded; A task offloading result determination module is used to input the acquired mobile edge computing task into a pre-built DQN model to obtain the task offloading result of the mobile device, where the task offloading result includes directly processing the mobile edge computing task locally, offloading it to the MEC server, or forwarding it to the edge server; Wherein, the DQN model is optimized based on a genetic algorithm.

8. A mobile edge computing task offloading system as claimed in claim 7, characterized in that: Based on the genetic algorithm, the specific process of optimizing the DQN model is as follows: Randomly generate a set of DRL algorithm parameter settings or strategies as the initial individuals of the population; Build the DQN model; Based on the obtained training data of the mobile edge computing task offloading problem, the constructed DQN model is trained; Based on the genetic algorithm, the trained DQN model is optimized and the parameters of the DQN model are updated; Check whether the training reaches the preset termination condition. If the termination condition is met, stop the training and output the optimized DQN model. Otherwise, retrain the constructed DQN model.

9. A processing device, characterized in that: The method comprises computer program instructions, wherein the computer program instructions are used to implement the steps corresponding to the mobile edge computing task offloading method according to any one of claims 1 to 6 when the computer program instructions are executed by a processing device.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, wherein the computer program instructions, when executed by a processor, are used to implement steps corresponding to the mobile edge computing task offloading method described in any one of claims 1 to 6.

Images