Low-latency and high-reliability task offloading and scheduling method in self-powered mobile edge computing scenarios

Abstract

The present invention discloses a low-latency and highly reliable task unloading and scheduling method in a self-powered mobile edge computing scenario. The method realizes efficient task unloading and scheduling by establishing a mobile edge computing system architecture of multiple devices and multiple servers, combined with Lyapunov optimization and reinforcement learning algorithms. The system architecture of the present invention includes a base station, multiple heterogeneous edge servers with different computing capabilities, and multiple mobile devices equipped with energy collection modules, which can effectively cope with energy-constrained mobile computing scenarios. In the present invention, the mobile device obtains renewable energy from the environment through the energy collection module to power the device. The system is managed by time slot division, and the scheduling cycle is evenly divided into several equal discrete time slots. In each time slot, the system makes task unloading decisions and optimizes resource scheduling based on the current state to minimize task processing delays while ensuring the stability of device energy and the reliability of edge servers.

Description

Technical Field

[0001] The present invention belongs to the field of mobile edge computing technology, and in particular to a low-latency and high-reliability task offloading and scheduling method in a self-powered mobile edge computing scenario. Background Art

[0002] With the rapid development of mobile edge computing technology, offloading computationally intensive tasks from mobile devices to nearby edge servers has become an important means to improve the computing power of mobile devices. However, as battery-powered embedded systems, traditional mobile devices often face the problem of insufficient power during task execution and offloading. Although energy harvesting technology provides a possible solution to such problems, how to balance task latency, energy stability, and service reliability in a dynamic environment remains an important challenge.

[0003] At present, the academic community has conducted extensive research on task offloading and scheduling issues in mobile edge computing. For example, the literature (Z. Kuang, L. Li, J. Gao, L. Zhao and A. Liu, "Partial offloadingscheduling and power allocation for mobile edge computing systems," IEEE IoT, vol. 6, no. 4, pp. 6774-6785, 2019.) proposed a task offloading and transmission power allocation scheme for single-user mobile edge computing systems, mainly focusing on reducing the energy consumption of mobile devices; the literature (L. Yu, J. Zheng, Y. Wu, F. Zhou and F. Yan, " A DQN-based joint spectrum and computing resource allocation algorithm for MEC networks," GLOBECOM, pp. 5135-5140, 2022.) introduced a collaborative communication and computing resource allocation method based on reinforcement learning to maximize system throughput. However, these existing technical solutions have the following shortcomings: First, most research works only focus on static task offloading strategies and do not consider the dynamic characteristics of self-powered systems; second, existing solutions often ignore the reliability constraints of the system, which is crucial to ensure the correct execution of tasks; finally, in a multi-device and multi-server architecture, the problem of how to simultaneously optimize task delays, ensure energy stability and meet reliability requirements has not been effectively solved. Summary of the invention

[0004] The purpose of the present invention is to provide a low-latency and high-reliability task offloading and scheduling method in a self-powered mobile edge computing scenario in order to address the defect in the prior art that dynamic task scheduling in a self-powered mobile edge computing system is insufficiently considered.

[0005] The technical solution to achieve the purpose of the present invention is: on the one hand, a low-latency and high-reliability task offloading and scheduling method in a self-powered mobile edge computing scenario is provided, and the method comprises:

[0006] Step 1, establishing a self-powered mobile edge computing system architecture including a base station, multiple heterogeneous edge servers with different computing capabilities, and multiple mobile devices, defining the execution rules of dynamic task offloading in the self-powered mobile edge computing system architecture, and defining a communication model, a computing model, a reliability model, an energy consumption model, and a utility function of mobile application delay under the constraints of ensuring the long-term energy stability of the battery and the reliability of the server; wherein the mobile device includes an energy collection module for obtaining renewable energy from the environment to power the device;

[0007] Step 2: Using the Lyapunov drift plus penalty-based optimization method, the long-term battery energy constraint problem is decoupled into a series of deterministic optimization problems within a single time slot, controlling the difference between energy collection and consumption in each time slot;

[0008] Step 3: Design a task scheduling scheme: Based on the given task offloading strategy, provide the optimal reliability allocation of tasks to edge servers, device frequency adjustment, and optimal allocation of computing resources to achieve the best trade-off between latency and energy stability.

[0009] Step 4, design a task offloading algorithm based on reinforcement learning, and use the task scheduling scheme to find the optimal offloading decision that achieves minimum delay and meets energy stability and reliability constraints.

[0010] On the other hand, a low-latency and high-reliability task offloading and scheduling system in a self-powered mobile edge computing scenario is provided, the system comprising:

[0011] The first module is used to establish a self-powered mobile edge computing system architecture including a base station, multiple heterogeneous edge servers with different computing capabilities and multiple mobile devices, define the execution rules of dynamic task offloading in the self-powered mobile edge computing system architecture, and define a communication model, a computing model, a reliability model, an energy consumption model and a utility function of mobile application delay under the constraints of ensuring the long-term energy stability of the battery and the reliability of the server; wherein the mobile device includes an energy collection module for obtaining renewable energy from the environment to power the device;

[0012] The second module is used to decouple the long-term battery energy constraint problem into a series of deterministic optimization problems within a single time slot using an optimization method based on Lyapunov drift plus penalty, and to control the difference between energy collection and consumption within each time slot;

[0013] The third module is used to design a task scheduling scheme: based on a given task offloading strategy, it provides the optimal reliability allocation of tasks to edge servers, device frequency adjustment, and optimal allocation of computing resources to achieve the best trade-off between latency and energy stability;

[0014] The fourth module is used to design a task offloading algorithm based on reinforcement learning, and use the task scheduling scheme to find the optimal offloading decision that achieves minimum delay and meets energy stability and reliability constraints.

[0015] Compared with the prior art, the present invention has the following significant advantages: it proposes a dynamic task offloading decision method that takes into account the uncertainty of energy collection, designs an energy management strategy based on Lyapunov optimization, decouples the long-term battery energy constraint problem into a series of deterministic optimization problems within a single time slot, develops a reliability-aware task allocation algorithm, and implements a reinforcement learning optimization method under multi-objective constraints. This effectively reduces the processing delay of mobile applications, improves the energy utilization efficiency of the system, enhances the reliability of task execution, and realizes dynamic optimization of system performance.

[0016] The present invention is further described in detail below in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of the low-latency and high-reliability task offloading and scheduling method in the self-powered mobile edge computing scenario of the present invention.

[0018] Figure 2 It is a schematic diagram of the system architecture model of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. 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.

[0020] It should be noted that if there are descriptions involving "first", "second", etc. in the embodiments of the present invention, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0021] The present invention proposes a low-latency and highly reliable task offloading and scheduling method in a self-powered mobile edge computing scenario. The method achieves efficient task offloading and scheduling by establishing a multi-device and multi-server mobile edge computing system architecture, combined with Lyapunov optimization and reinforcement learning algorithms. The system architecture of the present invention includes a base station, multiple heterogeneous edge servers with different computing capabilities, and multiple mobile devices equipped with energy collection modules, which can effectively cope with energy-constrained mobile computing scenarios.

[0022] In the present invention, the mobile device obtains renewable energy from the environment through the energy collection module to power the device. The system adopts time slot division for management, and evenly divides the scheduling cycle into several equal discrete time slots. In each time slot, the system makes task offloading decisions and optimizes resource scheduling based on the current state to minimize task processing delays while ensuring the stability of device energy and the reliability of edge servers.

[0023] In one embodiment, in combination Figure 1 , provides a low-latency and high-reliability task offloading and scheduling method in a self-powered mobile edge computing scenario, the method comprising:

[0024] Step 1, Combine Figure 2 , establish a self-powered mobile edge computing system architecture including a base station, multiple heterogeneous edge servers with different computing capabilities and multiple mobile devices, define the execution rules of dynamic task offloading in the self-powered mobile edge computing system architecture, and define a communication model, a computing model, a reliability model, an energy consumption model and a utility function of mobile application delay under the constraints of ensuring the long-term energy stability of the battery and the reliability of the server; wherein the mobile device includes an energy harvesting module for obtaining renewable energy from the environment to power the device;

[0025] Step 2: Using the Lyapunov drift plus penalty-based optimization method, the long-term battery energy constraint problem is decoupled into a series of deterministic optimization problems within a single time slot, controlling the difference between energy collection and consumption in each time slot;

[0026] Step 3: Design a task scheduling scheme: Based on the given task offloading strategy, provide the optimal allocation of tasks to edge servers in terms of reliability and computing resources to achieve the best trade-off between latency and energy stability.

[0027] Step 4, design a task offloading algorithm based on reinforcement learning, and use the task scheduling scheme to find the optimal offloading decision that achieves minimum delay and meets energy stability and reliability constraints.

[0028] Furthermore, in one embodiment, step 1 specifically includes:

[0029] Step 1.1, define the description information of mobile devices in the self-powered mobile edge computing system: H represents the set of mobile devices, each mobile device Includes an energy harvesting module;

[0030] Define the description information of the server in the self-powered mobile edge computing system: S represents the set of edge servers;

[0031] Define the description information of the base station in the self-powered mobile edge computing system: the base station is represented by BS;

[0032] Step 1.2, define the task model in the self-powered mobile edge computing system: Represents a collection of tasks, using a four-tuple Represents a mobile device The tasks to be processed in time slot t ,in yes The vulnerability factor when subjected to soft errors, yes The amount of data, yes The amount of data in the result, is the number of CPU cycles required to process a unit of data;

[0033] Step 1.3, define the dynamic task offloading rules in the self-powered mobile edge computing system: define the triple For the task The execution mode of and Respectively represent tasks Execute on mobile devices and offload to edge servers for execution; Indicates the task Abandoned due to insufficient energy; when tasks are offloaded to the base station, they will be assigned to different edge servers for parallel execution, and each task can only be assigned to one edge server, and tasks on the same edge server share the computing resources of the edge server; Assign a strategy to the task, where is a binary variable. Assigned to the mth edge server When , its value is 1; is the resource allocation strategy, where is assigned to computing resources; for a given task offloading decision , the set of tasks to be processed in time slot t is divided into the local task set , Uninstall Task Set and discard task set ;

[0034] Step 1.4, define the communication model of tasks in the self-powered mobile edge computing system: According to the Shannon-Hartley theorem, mobile devices using orthogonal frequency division multiple access technology Upload rate at time t and download speed for:

[0035]

[0036]

[0037] In the formula, and They are assigned to mobile devices at time t. Uplink bandwidth and downlink bandwidth, Is a mobile device The upload transmission power, is the transmission power of the base station; Is a mobile device The channel gain between the BS and the BS is usually considered as a constant within a time slot; is the noise power;

[0038] Step 1.5, define the computational model of tasks in the self-powered mobile edge computing system: For each task ,There are three modes of operation: local execution, offload remote execution, or discard;

[0039] Defining processing tasks The total delay is The total energy consumption is ;

[0040] (1) When the task is executed locally on the mobile device, the frequency of the mobile device remains fixed within a time slot; Execute tasks locally Delay for:

[0041]

[0042] In the formula, For mobile devices The frequency in time slot t, yes The amount of data, is the number of CPU cycles required to process a unit of data, then the corresponding energy consumption for:

[0043]

[0044] in, is a constant related to CMOS circuits;

[0045] (2) Uninstall remote execution or discard

[0046] Remotely executed uninstall delays Upload time , Execution time and download time sum:

[0047]

[0048] mobile device Uninstall Tasks Energy consumed Energy consumption for uploading and download energy consumption sum:

[0049]

[0050] Processing tasks Total delay and energy consumption They are:

[0051]

[0052]

[0053] Step 1.6, define the reliability model of tasks in the self-powered mobile edge computing system: including execution reliability and transmission reliability;

[0054] (1) The execution reliability is the probability of successfully executing a task in the presence of potential soft errors.

[0055] The average soft error rate is expressed as :

[0056]

[0057] in, and are the maximum and minimum frequencies of a mobile device or edge server, respectively, t is a hardware-related parameter, For mobile devices or edge servers at the highest frequency The error rate when working is a frequency variable;

[0058] The task is executed on the mobile device or edge server The execution reliability is expressed as :

[0059]

[0060] In the formula, when the task is executed on the mobile device , when the task is executed on the edge server ; yes Vulnerability factor when subjected to soft errors;

[0061] (2) The transmission reliability refers to the probability of successful transmission of a data request between a mobile device and an edge server;

[0062] Task The transmission reliability is expressed as :

[0063]

[0064] In the formula, bit error rate is a constant, indicating the bit error rate per unit time; Indicates the task Total transmission time;

[0065] In time slot t, the task Reliability for:

[0066]

[0067] (3) System reliability of time slots Depends on the successful transmission and execution of all tasks, expressed as:

[0068]

[0069] The unloading decision at time slot 𝑡 When confirmed, uninstall the task set and local task set The system reliability is determined by the local task set Reliability and offloaded task set Reliability composition:

[0070]

[0071]

[0072]

[0073] Step 1.7, define the energy consumption model of tasks in the self-powered mobile edge computing system:

[0074] Each mobile device is equipped with an energy harvesting module that collects renewable energy such as solar energy from the environment and converts it into usable electrical energy, which is then stored in a battery. In the system, the stored energy is the only source of energy supply for the mobile device.

[0075] The change of battery energy between adjacent time slots is expressed as:

[0076]

[0077] In the formula, and The mobile device at the beginning of time slot t and time slot t+1 is of battery energy, During time slot t, the mobile device The number of energy units collected by the energy collection module, which will be stored in the battery and can be used in the next time period. Represents a mobile device The energy consumed in time slot t;

[0078] Step 1.8, define the task execution cost utility function:

[0079] The optimization problem of minimizing the long-term average delay is:

[0080]

[0081] When the battery power is insufficient to meet the local execution of the task or the load transfer, the task will be discarded;

[0082] Considering the negative impact of task abandonment on the optimization objective, a weight is introduced for each mobile device , the long-term average delay and the penalty together constitute the execution cost, then the execution cost minimization problem is expressed as:

[0083]

[0084]

[0085]

[0086]

[0087]

[0088]

[0089]

[0090] In the formula, the C1 constraint indicates that a task can only have one operation mode; the C2 constraint indicates that the energy consumption of the mobile device in a time slot cannot exceed the remaining energy of its battery; the C3 constraint indicates that the battery will not violate its discharge constraint, where, and For mobile devices The minimum and maximum discharge energy of the battery; the C4 constraint indicates that a task can be assigned to at most one edge server for processing, the C5 constraint indicates that the computing resources allocated to a task on an edge server should be within the capacity of the server, and the C6 constraint indicates the server reliability Cannot be below the threshold , N is the total number of mobile devices, and M is the total number of edge servers.

[0091] Furthermore, in one embodiment, the upload time is the time required to upload task data from the mobile device to the edge server, calculated as:

[0092]

[0093] In the formula, Is a mobile device The upload rate, yes The amount of data;

[0094] The execution time is the time required for the edge server to process the task, calculated as:

[0095]

[0096] In the formula, The edge server is assigned to the task of computing resources, yes The amount of data, is the number of CPU cycles required to process a unit of data;

[0097] The download time is the time required to download the processed results from the edge server back to the mobile device, calculated as:

[0098]

[0099] In the formula Is a mobile device Download rate, yes The amount of data in the result.

[0100] Furthermore, in one embodiment, the uploading energy consumption It is the energy consumed by the mobile device in the process of uploading task data to the edge server. The calculation formula is:

[0101]

[0102] In the formula, For mobile devices The transmission power during data transmission, yes The amount of data, Is a mobile device Upload rate;

[0103] The download energy consumption is the energy consumed by the mobile device in the process of downloading the processing results from the edge server, and the calculation formula is:

[0104]

[0105] In the formula, For mobile devices The received power during data transmission, Is a mobile device Download rate, yes The amount of data in the result.

[0106] Further, in one embodiment, step 2 specifically includes:

[0107] Step 2.1: Define the energy queue model of the mobile device, including the actual battery energy queue and the virtual energy queue ;

[0108] Preferably, in some embodiments, the virtual energy queue is constructed by a weighted perturbation method, and the calculation formula is:

[0109]

[0110] In the formula, For mobile devices The disturbance parameter.

[0111] The disturbance parameter For a satisfaction The constant, Represents a mobile device The upper bound of energy consumption, , is the control parameter.

[0112] Step 2.2, construct the Lyapunov function , which is used to measure the total backlog of all energy queues in time slot t, is defined as:

[0113]

[0114] Define the Lyapunov drift function , which indicates the queue backlog change between adjacent time slots; if for all time slots , corresponding to Minimize, the backlog of the virtual energy queue reaches a stable state, and the calculation formula is:

[0115]

[0116] In the formula, For time slot +1 corresponds to the Lyapunov function;

[0117] Step 2.3, construct the upper bound of the Lyapunov drift function and prove that there is a constant satisfy:

[0118]

[0119] can be calculated individually based on the information available in the current time slot. , minimize and ensure the stability of the virtual energy queue;

[0120] In the formula, ;

[0121] Step 2.4: Construct Lyapunov drift plus penalty function As the new optimization target P2, the execution cost is optimized while maintaining the stability of the queue. The function is:

[0122]

[0123]

[0124] Further, in one embodiment, step 3 specifically includes:

[0125] Step 3.1, optimal allocation of tasks to edge servers:

[0126] Considering that ideally, each offloaded task occupies all computing resources on the assigned server ; Assign to server The task set meets ;

[0127] Based on this, Written in the following form:

[0128]

[0129]

[0130] definition For edge servers The vulnerability index, For task set Vulnerability index;

[0131] Rewrite the above equation as:

[0132]

[0133]

[0134] in, and , For the Mth edge server The vulnerability index, For task set Vulnerability index; The sum of the vulnerability indexes of all tasks can be calculated based on server parameters to determine the offload task set. The latter is a constant;

[0135] Given the unloading decision for each time slot t , Server Vulnerability Index Satisfy the conditions , Satisfy the conditions , To reach the minimum, Reach the maximum value;

[0136] A heuristic algorithm is proposed for task allocation, which assigns highly vulnerable tasks to reliable servers with low vulnerability to improve the overall execution reliability of the offloaded task set.

[0137] Step 3.2, optimal resource allocation for edge servers:

[0138] For each edge server, the resource allocation problem is formulated as a subproblem :

[0139]

[0140]

[0141] It can be observed that The objective function is a convex function.

[0142] Convert the above formula into an unconstrained augmented Lagrangian function , as shown below:

[0143]

[0144] In the formula, is the Lagrange multiplier vector;

[0145] Based on the Lagrange multiplier method and Karush-Kuhn-Tucker condition, the edge server To the task The optimal allocation of computing resources is expressed as :

[0146]

[0147] Further, in one embodiment, step 4 specifically includes:

[0148] Step 4.1, the sub-problem of task offloading is described as follows:

[0149]

[0150] C1, C3, C4

[0151] It is a combinatorial optimization problem, and the dual-depth Q network algorithm is used to solve the sub-problem of task offloading. ;

[0152] Step 4.2, define the state space, the state vector in time slot t Described as:

[0153]

[0154] In the formula, is the battery virtual energy level, To collect energy, is the data size of the task, To calculate the result, is the channel gain, is the uplink bandwidth of the device on the network, is the downlink bandwidth;

[0155] Step 4.3, define the action space, represented as the offloading decisions of N mobile devices , The uninstallation decision for the Nth mobile device;

[0156] Step 4.4, define the reward function, set it to the opposite of the objective function, that is ;

[0157] Step 4.5, through iterative training of the DDQN algorithm, we finally obtain the optimal task offloading decision that can minimize the execution cost and energy level drift.

[0158] In one embodiment, a low-latency and high-reliability task offloading and scheduling system in a self-powered mobile edge computing scenario is provided, the system comprising:

[0159] The first module is used to establish a self-powered mobile edge computing system architecture including a base station, multiple heterogeneous edge servers with different computing capabilities and multiple mobile devices, define the execution rules of dynamic task offloading in the self-powered mobile edge computing system architecture, and define a communication model, a computing model, a reliability model, an energy consumption model and a utility function of mobile application delay under the constraints of ensuring the long-term energy stability of the battery and the reliability of the server; wherein the mobile device includes an energy collection module for obtaining renewable energy from the environment to power the device;

[0160] The second module is used to decouple the long-term battery energy constraint problem into a series of deterministic optimization problems within a single time slot using an optimization method based on Lyapunov drift plus penalty, and to control the difference between energy collection and consumption within each time slot;

[0161] The third module is used to design a task scheduling scheme: based on a given task offloading strategy, it provides the optimal allocation of tasks to edge servers in terms of reliability and the optimal allocation of computing resources to achieve the best trade-off between latency and energy stability;

[0162] The fourth module is used to design a task offloading algorithm based on reinforcement learning, and use the task scheduling scheme to find the optimal offloading decision that achieves minimum delay and meets energy stability and reliability constraints.

[0163] For the specific limitations of the low-latency and high-reliability task offloading and scheduling system in the self-powered mobile edge computing scenario, please refer to the limitations of the low-latency and high-reliability task offloading and scheduling method in the self-powered mobile edge computing scenario mentioned above, which will not be repeated here. Each module in the low-latency and high-reliability task offloading and scheduling system in the above-mentioned self-powered mobile edge computing scenario can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0164] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following is achieved:

[0165] Step 1, establishing a self-powered mobile edge computing system architecture including a base station, multiple heterogeneous edge servers with different computing capabilities, and multiple mobile devices, defining the execution rules of dynamic task offloading in the self-powered mobile edge computing system architecture, and defining a communication model, a computing model, a reliability model, an energy consumption model, and a utility function of mobile application delay under the constraints of ensuring the long-term energy stability of the battery and the reliability of the server; wherein the mobile device includes an energy collection module for obtaining renewable energy from the environment to power the device;

[0166] Step 2: Using the Lyapunov drift plus penalty-based optimization method, the long-term battery energy constraint problem is decoupled into a series of deterministic optimization problems within a single time slot, controlling the difference between energy collection and consumption in each time slot;

[0167] Step 3: Design a task scheduling scheme: Based on the given task offloading strategy, provide the optimal reliability allocation of tasks to edge servers, device frequency adjustment, and optimal allocation of computing resources to achieve the best trade-off between latency and energy stability.

[0168] Step 4, design a task offloading algorithm based on reinforcement learning, and use the task scheduling scheme to find the optimal offloading decision that achieves minimum delay and meets energy stability and reliability constraints.

[0169] For the specific limitations of each step, please refer to the above limitations on low-latency and high-reliability task offloading and scheduling methods in self-powered mobile edge computing scenarios, which will not be repeated here.

[0170] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the computer program implements:

[0171] Step 1, establishing a self-powered mobile edge computing system architecture including a base station, multiple heterogeneous edge servers with different computing capabilities, and multiple mobile devices, defining the execution rules of dynamic task offloading in the self-powered mobile edge computing system architecture, and defining a communication model, a computing model, a reliability model, an energy consumption model, and a utility function of mobile application delay under the constraints of ensuring the long-term energy stability of the battery and the reliability of the server; wherein the mobile device includes an energy collection module for obtaining renewable energy from the environment to power the device;

[0172] Step 2: Using the Lyapunov drift plus penalty-based optimization method, the long-term battery energy constraint problem is decoupled into a series of deterministic optimization problems within a single time slot, controlling the difference between energy collection and consumption in each time slot;

[0173] Step 3: Design a task scheduling scheme: Based on the given task offloading strategy, provide the optimal reliability allocation of tasks to edge servers, device frequency adjustment, and optimal allocation of computing resources to achieve the best trade-off between latency and energy stability.

[0174] Step 4, design a task offloading algorithm based on reinforcement learning, and use the task scheduling scheme to find the optimal offloading decision that achieves minimum delay and meets energy stability and reliability constraints.

[0175] For the specific limitations of each step, please refer to the above limitations on low-latency and high-reliability task offloading and scheduling methods in self-powered mobile edge computing scenarios, which will not be repeated here.

[0176] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A low-latency and high-reliability task offloading and scheduling method in a self-powered mobile edge computing scenario, characterized in that: The method comprises: Step 1, establish a self-powered mobile edge computing system architecture including a base station, multiple heterogeneous edge servers with different computing capabilities and multiple mobile devices, define the execution rules of dynamic task offloading in the self-powered mobile edge computing system architecture, and define a communication model, a computing model, a reliability model, an energy consumption model and a utility function of mobile application delay under the constraints of ensuring the long-term energy stability of the battery and the reliability of the server; wherein the mobile device includes an energy collection module for obtaining renewable energy from the environment to power the device; specifically including: Step 1.1, define the description information of the mobile device in the self-powered mobile edge computing system; Step 1.2, define the task model in the self-powered mobile edge computing system; Step 1.3, defining dynamic task offloading rules in the self-powered mobile edge computing system; Step 1.4, define the communication model of tasks in the self-powered mobile edge computing system; Step 1.5, defining the computing model of the tasks in the self-powered mobile edge computing system; Step 1.6, define the reliability model of the task in the self-powered mobile edge computing system; specifically including: execution reliability and transmission reliability; (1) The execution reliability is the probability of successfully executing a task in the presence of potential soft errors. The average soft error rate is expressed as : ； in, and are the maximum and minimum frequencies of a mobile device or edge server, respectively, t is a hardware-related parameter, For mobile devices or edge servers at the highest frequency The error rate when working is a frequency variable; The task is executed on the mobile device or edge server The execution reliability is expressed as : ； In the formula, when the task is executed on the mobile device , when the task is executed on the edge server ; yes Vulnerability factor when subjected to soft errors; For mobile devices The frequency in time slot t; yes The amount of data; is the number of CPU cycles required to process a unit of data; (2) The transmission reliability refers to the probability of successful transmission of a data request between a mobile device and an edge server; Task The transmission reliability is expressed as : ； In the formula, the bit error rate is a constant, indicating the bit error rate per unit time; Indicates the task The total transmission time, , The task upload time and download time are ; In time slot t, the task Reliability for: ； In the formula, and Respectively represent tasks Execute on mobile devices and offload to edge servers for execution; (3) System reliability of time slots Depends on the successful transmission and execution of all tasks, expressed as: ； System reliability is determined by the local task set Reliability and offloaded task set Reliability composition: ； ； ； In the formula, and Mobile devices using OFDMA technology The upload rate and download rate at time t; yes The amount of data in the results; Step 1.7, define the energy consumption model of the tasks in the self-powered mobile edge computing system; Step 1.8, define the task execution cost utility function; Step 2: Using the Lyapunov drift plus penalty-based optimization method, the long-term battery energy constraint problem is decoupled into a series of deterministic optimization problems within a single time slot, controlling the difference between energy collection and consumption in each time slot; Step 3: Design a task scheduling scheme: Based on the given task offloading strategy, provide the optimal allocation of tasks to edge servers in terms of reliability and computing resources to achieve the best trade-off between latency and energy stability. Step 4, design a task offloading algorithm based on reinforcement learning, and use the task scheduling scheme to find the optimal offloading decision that achieves minimum delay and meets energy stability and reliability constraints.

2. The low-latency and high-reliability task offloading and scheduling method in the self-powered mobile edge computing scenario according to claim 1 is characterized in that: Step 1 specifically includes: In step 1.1, the description information of the mobile devices in the self-powered mobile edge computing system is defined, including: H represents the set of mobile devices, each mobile device Includes an energy harvesting module; Define the description information of the server in the self-powered mobile edge computing system: S represents the set of edge servers; Define the description information of the base station in the self-powered mobile edge computing system: the base station is represented by BS; The task model in the self-powered mobile edge computing system is defined in step 1.2, including: Represents a collection of tasks, using a four-tuple Represents a mobile device The tasks to be processed in time slot t ,in yes The vulnerability factor when subjected to soft errors, yes The amount of data, yes The amount of data in the result, is the number of CPU cycles required to process a unit of data; Step 1.3 defines the dynamic task offloading rules in the self-powered mobile edge computing system, specifically including: defining the triple For the task The execution mode of and Respectively represent tasks Execute on mobile devices and offload to edge servers for execution; Indicates the task Abandoned due to insufficient energy; when tasks are offloaded to the base station, they will be assigned to different edge servers for parallel execution, and each task can only be assigned to one edge server, and tasks on the same edge server share the computing resources of the edge server; Assign a strategy to the task, where is a binary variable. Assigned to the mth edge server When , its value is 1; is the resource allocation strategy, where is assigned to computing resources; for a given task offloading decision , the set of tasks to be processed in time slot t is divided into the local task set , Uninstall Task Set and discard task set ; The communication model of the tasks in the self-powered mobile edge computing system is defined in step 1.4, including: According to the Shannon-Hartley theorem, mobile devices using orthogonal frequency division multiple access technology Upload rate at time t and download speed for: ； ； In the formula, and They are assigned to mobile devices at time t. Uplink bandwidth and downlink bandwidth, Is a mobile device The upload transmission power, is the transmission power of the base station; Is a mobile device Channel gain between the RF and the base station; is the noise power; Step 1.5 defines the computational model of tasks in the self-powered mobile edge computing system, including: ,There are three modes of operation: local execution, offload remote execution, or discard; Defining processing tasks The total delay is The total energy consumption is ; (1) When the task is executed locally on the mobile device, the frequency of the mobile device remains fixed within a time slot; Execute tasks locally Delay for: ； In the formula, For mobile devices The frequency in time slot t, yes The amount of data, is the number of CPU cycles required to process a unit of data, then the corresponding energy consumption for: ； in, is a constant related to CMOS circuits; (2) Uninstall remote execution or discard Remotely executed uninstall delays Upload time , Execution time and download time sum: ； mobile device Uninstall Tasks Energy consumed Energy consumption for uploading and download energy consumption sum: ； Processing tasks Total delay and energy consumption They are: ； ； The energy consumption model of the tasks in the self-powered mobile edge computing system is defined in step 1.7, including: Each mobile device is equipped with an energy harvesting module to collect solar renewable energy from the environment and convert it into usable electrical energy, which is then stored in a battery. The change in battery energy between adjacent time slots is expressed as: ； In the formula, and The mobile device at the beginning of time slot t and time slot t+1 is of battery energy, During time slot t, the mobile device The number of unit energies collected by the energy harvesting module, Represents a mobile device The energy consumed in time slot t; In step 1.8, the task execution cost utility function is defined, including: The optimization problem of minimizing the long-term average delay is: ； When the battery power is insufficient to meet the local execution of the task or the load transfer, the task will be discarded; A weight is introduced for each mobile device , the long-term average delay and the penalty together constitute the execution cost, then the execution cost minimization problem is expressed as: ； ； ； ； ； ； ； In the formula, the C1 constraint indicates that a task can only have one operation mode; the C2 constraint indicates that the energy consumption of the mobile device in a time slot cannot exceed the remaining energy of its battery; the C3 constraint indicates that the battery will not violate its discharge constraint, where, and For mobile devices The minimum and maximum discharge energy of the battery; the C4 constraint indicates that a task can be assigned to at most one edge server for processing, the C5 constraint indicates that the computing resources allocated to a task on an edge server should be within the capacity of the server, and the C6 constraint indicates the server reliability Cannot be below the threshold , N is the total number of mobile devices, and M is the total number of edge servers.

3. The low-latency and high-reliability task offloading and scheduling method in the self-powered mobile edge computing scenario according to claim 2 is characterized in that: The upload time is the time required to upload task data from the mobile device to the edge server, calculated as: ； In the formula, Is a mobile device The upload rate, yes The amount of data; The execution time is the time required for the edge server to process the task, calculated as: ； In the formula, The edge server is assigned to the task of computing resources, yes The amount of data, is the number of CPU cycles required to process a unit of data; The download time is the time required to download the processed results from the edge server back to the mobile device, calculated as: ； In the formula Is a mobile device Download rate, yes The amount of data in the result.

4. The low-latency and high-reliability task offloading and scheduling method in the self-powered mobile edge computing scenario according to claim 2 is characterized in that: The upload energy consumption It is the energy consumed by the mobile device in uploading task data to the edge server. The calculation formula is: ； In the formula, For mobile devices The transmission power during data transmission, yes The amount of data, Is a mobile device Upload rate; The download energy consumption is the energy consumed by the mobile device in the process of downloading the processing results from the edge server, and the calculation formula is: ； In the formula, For mobile devices The received power during data transmission, Is a mobile device Download rate, yes The amount of data in the result.

5. The low-latency and high-reliability task offloading and scheduling method in the self-powered mobile edge computing scenario according to claim 2 is characterized in that: Step 2 specifically includes: Step 2.1: Define the energy queue model of the mobile device, including the actual battery energy queue and the virtual energy queue ; Step 2.2, construct the Lyapunov function , which is used to measure the total backlog of all energy queues in time slot t, is defined as: ； Define the Lyapunov drift function , which indicates the queue backlog change between adjacent time slots; if for all time slots , corresponding to Minimize, the backlog of the virtual energy queue reaches a stable state, and the calculation formula is: ； In the formula, For time slot +1 corresponds to the Lyapunov function; Step 2.3, construct the upper bound of the Lyapunov drift function and prove that there is a constant satisfy: ； By minimizing , minimize and ensure the stability of the virtual energy queue; In the formula, ; Step 2.4: Construct Lyapunov drift plus penalty function As the new optimization target P2, the execution cost is optimized while maintaining the stability of the queue. The function is: ； 。 6. The low-latency and high-reliability task offloading and scheduling method in the self-powered mobile edge computing scenario according to claim 5 is characterized in that: The virtual energy queue in step 2.1 is constructed by weighted perturbation method, and the calculation formula is: ； In the formula, For mobile devices The disturbance parameter.

7. The low-latency and high-reliability task offloading and scheduling method in the self-powered mobile edge computing scenario according to claim 6 is characterized in that: The disturbance parameter For a satisfaction The constant, Represents a mobile device The upper bound of energy consumption, , is the control parameter.

8. The low-latency and high-reliability task offloading and scheduling method in the self-powered mobile edge computing scenario according to claim 2 is characterized in that: Step 3 specifically includes: Step 3.1, optimal allocation of tasks to edge servers: Considering that ideally, each offloaded task occupies all computing resources on the assigned server ; Assign to server The task set meets ; Based on this, Written in the following form: ； ； definition For edge servers The vulnerability index, For task set Vulnerability index; Rewrite the above equation as: ； ； in, and , For the Mth edge server The vulnerability index, For task set Vulnerability index; Given the unloading decision for each time slot t , Server Vulnerability Index Satisfy the conditions , Satisfy the conditions , To reach the minimum, Reach the maximum value; Step 3.2, optimal resource allocation for edge servers: For each edge server, the resource allocation problem is formulated as a subproblem : ； ； Convert the above formula into an unconstrained augmented Lagrangian function , as shown below: ； In the formula, is the Lagrange multiplier vector; Based on the Lagrange multiplier method and Karush-Kuhn-Tucker condition, the edge server To the task The optimal allocation of computing resources is expressed as : 。 9. The low-latency and high-reliability task offloading and scheduling method in the self-powered mobile edge computing scenario according to claim 8 is characterized in that: Step 4 specifically includes: Step 4.1, the sub-problem of task offloading is described as follows: ； C1,C3,C4; It is a combinatorial optimization problem, and the dual-depth Q network algorithm is used to solve the sub-problem of task offloading. ; Step 4.2, define the state space, the state vector in time slot t Described as: ； In the formula, is the battery virtual energy level, To collect energy, is the data size of the task, To calculate the result, is the channel gain, is the uplink bandwidth of the device on the network, is the downlink bandwidth; Step 4.3, define the action space, represented as the offloading decisions of N mobile devices , The uninstallation decision for the Nth mobile device; Step 4.4, define the reward function, set it to the opposite of the objective function, that is ; Step 4.5, through iterative training of the DDQN algorithm, we finally obtain the optimal task offloading decision that can minimize the execution cost and energy level drift.

10. A low-latency and high-reliability task offloading and scheduling system in a self-powered mobile edge computing scenario based on the method according to any one of claims 1 to 9, characterized in that: The system comprises: The first module is used to establish a self-powered mobile edge computing system architecture including a base station, multiple heterogeneous edge servers with different computing capabilities and multiple mobile devices, define the execution rules of dynamic task offloading in the self-powered mobile edge computing system architecture, and define a communication model, a computing model, a reliability model, an energy consumption model and a utility function of mobile application delay under the constraints of ensuring the long-term energy stability of the battery and the reliability of the server; wherein the mobile device includes an energy collection module for obtaining renewable energy from the environment to power the device; The second module is used to decouple the long-term battery energy constraint problem into a series of deterministic optimization problems within a single time slot using an optimization method based on Lyapunov drift plus penalty, and to control the difference between energy collection and consumption within each time slot; The third module is used to design a task scheduling scheme: based on a given task offloading strategy, it provides the optimal allocation of tasks to edge servers in terms of reliability and computing resources to achieve the best trade-off between latency and energy stability; The fourth module is used to design a task offloading algorithm based on reinforcement learning, and use the task scheduling scheme to find the optimal offloading decision that achieves minimum delay and meets energy stability and reliability constraints.

Images