An edge computing offloading method based on an internet of things scene

Abstract

The application discloses an edge computing offloading method based on an Internet of Things scene, belongs to the technical field of Internet of Things task offloading, and constructs a cloud-edge cooperative Internet of Things edge computing system model; initializes relevant parameters of the edge computing system model; establishes a time delay model and an energy consumption model for the cloud and the edge server; constructs a system benefit function according to different mobile device offloading modes; and uses an improved sparrow search algorithm to continuously iteratively update sparrow positions and obtain the best offloading strategy and system benefit value. The method introduces a sine-cosine algorithm thought to improve the ability of the algorithm to jump out of a local optimal solution. The introduction of Levy flight strategy and random walk strategy can comprehensively explore in the search space, ensure that each user device in the system model can offload tasks to the cloud or the edge server, improve the system benefit, and reduce the task discard ratio.

Description

Technical Field

[0001] This invention belongs to the field of IoT computing offloading technology, specifically a computing offloading method for mobile edge computing in IoT scenarios. Background Technology

[0002] In recent years, with the continuous promotion of fifth-generation mobile communication technology (5G), the number of Internet of Things (IoT) devices has also been increasing, which places stringent demands on cloud infrastructure and the computing performance of the devices themselves. A large number of intensive computing requests will inevitably accelerate the energy consumption of mobile devices and shorten their lifespan. Offloading tasks to cloud servers also presents significant challenges due to the stringent requirements for high precision and low latency. Furthermore, factors such as limited spectrum resources, long backhaul link distances, and poor channel quality can lead to excessive latency and high network load, reducing user experience and network performance.

[0003] Therefore, Mobile Edge Computing (MEC) has emerged as an emerging technology that can effectively overcome the problems of high power consumption in mobile devices and the high latency that cloud servers may cause. Compared to cloud servers, edge servers use general-purpose computing platforms implemented directly at cellular base stations (BS) or local wireless access points (APs). With this characteristic, MEC allows applications to be executed near the terminal device, thereby significantly reducing end-to-end latency and the burden on the backhaul network. In view of this, offloading computing tasks to edge servers can bring ultra-low latency and flexible computing to intensive computing requests from mobile users. With the advent of MEC, the ability of resource-constrained terminal devices to offload computing tasks to edge servers is expected to be widely used in fields such as augmented reality, virtual reality, and autonomous driving. Currently, most research on edge computing offloading aims to optimize either latency or energy consumption. However, in real-world scenarios, latency directly affects the reliability of user devices, and energy consumption is also an important factor for enterprises; lower energy consumption means lower costs. Therefore, this invention will simultaneously consider both of these indicators for optimization to obtain the optimal offloading strategy that maximizes network benefits.

[0004] Solving unloading strategies typically requires transforming the problem into an integer nonlinear optimization problem, making direct solutions to the optimal decision difficult. Heuristic algorithms offer a better solution for this problem. Their advantage lies in their higher efficiency compared to general search algorithms, achieving the optimal unloading strategy with fewer iterations. However, the biggest drawback of heuristic search algorithms is that their convergence speed and the network benefit value of the final result cannot be guaranteed due to changes in the step size during each iteration. Sparrow Search Algorithm (SSA) is a novel swarm intelligence optimization algorithm inspired by the foraging and anti-predation behavior of sparrows. It simulates the process of sparrows searching for food and compares fitness values ​​to find the optimal solution to the objective function. SSA boasts advantages such as fewer control parameters, high solution accuracy, and ease of integration with other algorithms. Compared to most swarm intelligence algorithms and heuristic algorithms, SSA has certain advantages in optimizing most problems, but it still suffers from low convergence accuracy and difficulty escaping local optima. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides an edge computing offloading method based on IoT scenarios. First, a cloud-edge collaborative IoT edge computing network model is constructed. Based on the established network model, an edge server computing model and a cloud computing model are obtained, leading to a network efficiency function that simultaneously considers latency and energy consumption. Then, an improved sparrow search algorithm is used to offload computation for each IoT device, identifying the optimal computation offloading strategy that maximizes the network efficiency function. This ensures that each IoT device in the system can reasonably offload tasks to the cloud or edge server for processing, improving system efficiency and reducing the task drop rate.

[0006] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0007] An edge computing offloading method in an IoT scenario includes the following steps:

[0008] S1. Construct a cloud-edge collaborative IoT edge computing system model: This model includes a set of mobile devices U, a set of edge servers S, and a cloud server M with computing capabilities.

[0009] S2. Initialize the relevant parameters of the edge computing system model;

[0010] S3. Establish latency and energy consumption models for cloud and edge servers;

[0011] S4. When the edge server in a certain region is idle, the mobile devices in that region can directly offload the task to the edge server; if the edge server in a certain region is not idle, the mobile devices can choose to offload the task to the cloud server M. Based on this, the system benefit function is constructed.

[0012] S5. The sine-cosine SCA algorithm, Levy flight strategy, and random walk strategy are introduced to improve the sparrow position update strategy. The improved sparrow search algorithm is used to continuously update the sparrow position and obtain the optimal unloading strategy and system benefit value.

[0013] Furthermore, S2 specifically includes:

[0014] S21, the sets of mobile devices and edge servers are represented by U = {1, 2, ..., u} and S = {1, 2, ..., s}, respectively. Each mobile device u ∈ U generates one computational task L at a time. u = < c L ,d L ,T a ,T b >, where c L d represents the workload required to complete the computational task. L T represents the amount of input data required to execute the program from the mobile device to the server. a Let T represent the ideal time delay, respectively. b Indicates the maximum tolerable delay;

[0015] S22. Define the unloading strategy based on the network model. Where Q represents the set of computing tasks generated by the mobile device, It is a binary variable. This indicates that task L is unloaded to s∈S. This indicates that task L has been unloaded to the cloud M;

[0016] S23, The location of the mobile device is determined by This indicates that the location of the edge server is determined by... It means that x u y u These represent the x and y coordinates of the mobile device, respectively, with the height of the mobile device assumed to be 0; x s y s The x and y coordinates of the edge server are represented by , and H represents the height of the edge server.

[0017] Furthermore, S3 specifically includes:

[0018] S31. The channel gain from mobile device u∈U to MEC server s∈S in the uplink can be expressed as:

[0019]

[0020] In the formula, h0 represents the channel power gain when the transmission power is 1W and the transmission distance is 1m;

[0021] S32. Assuming that the locations of the mobile device and the edge server do not change within a time interval, the transmission rate from the mobile device to the edge server can be expressed as:

[0022]

[0023] In the formula, u′≠u, B and σ 2 Let p represent the uplink system bandwidth and noise power, respectively, where p = {p us |u∈U,s∈S} represents the transmission power from the mobile device u∈U to the edge server s∈S;

[0024] S33. The latency and energy consumption of the edge server and cloud server M are obtained through the network model established by S1.

[0025] Furthermore, the specific details regarding latency and energy consumption of the edge server and cloud server M in step S33 include:

[0026] S331. The latency of unloading a task to the edge server can be expressed as:

[0027]

[0028] Where R us (t) represents the uplink transmission rate; C us This indicates the computing power of the edge server;

[0029] S332. The latency of unloading the task to the cloud server can be expressed as:

[0030]

[0031] In the formula d L This represents the amount of input data required to execute a program from a mobile device to the server; the relative distance between the cloud server and the mobile device is approximately constant, and the rate at which the mobile user unloads the data to the cloud is R. um For a constant value, C um Indicates the computing power in the cloud;

[0032] S333, The energy consumption of offloading tasks to cloud servers or edge servers can be expressed as:

[0033] The energy consumption of the offloading task includes the energy consumption of the transmission task and the energy consumption of the processing task. The energy consumption of the mobile device u∈U for task transmission is defined as:

[0034]

[0035] The energy consumption of edge servers or cloud servers for processing tasks is expressed as follows:

[0036]

[0037] Where p s p represents the average power consumption of the edge server. m This indicates the average power consumption of the cloud server.

[0038] Furthermore, step S4 specifically includes:

[0039] S41. When mobile devices in the same area compete for the edge server in that area to complete the task offloading within the ideal latency, define the edge system efficiency η. L for:

[0040]

[0041] in: T a and T b T represents the ideal delay and the maximum tolerable delay, respectively. avg Indicates the average delay;

[0042] S42. When a mobile device in a certain region is processing a task, if the mobile device in that region chooses to offload the task to cloud server M, its offloading cost is defined as:

[0043] o L =θ·η L +(1-θ)E pro

[0044] Where θ represents the weighting coefficient, which is 0.5;

[0045] S43. Based on S3 and S4, the network benefits of constructing a computation offloading system model for an IoT scenario are as follows:

[0046]

[0047] Where e L This represents the normalized energy consumption.

[0048] Furthermore, step S5 specifically includes:

[0049] S51. Set the sparrow population in the algorithm and initialize the relevant parameters;

[0050] S52. After multiple iterations, the position of the sparrow in each optimization is continuously updated, and finally the position of the sparrow with the best fitness value is found. This position is then input into the computational unloading system model built in step S1, and the network benefits are obtained according to S43.

[0051] Furthermore, the step S52 of updating the sparrow's position in each optimization specifically includes:

[0052] S521, the position update strategy for the sparrow discoverer adopts the Sine-Cosine (SCA) algorithm and introduces a nonlinear sinusoidal learning factor; the formulas for the learning factor and the improved discoverer position formula are as follows:

[0053] ω=ω min +(ω max -ω min )·sin(tπ / iter max )

[0054]

[0055] In the formula, r1 is a random number in [0, 2π], r2 is a random number in [0, 2], and ω is the learning factor. Let R2∈[0,1] and ST∈[0.5,1] represent the position of the i-th sparrow in the j-th dimension when the number of iterations is t, and let R2∈[0,1] and ST∈[0.5,1] represent the warning value and the safety value, respectively.

[0056] S522: Combine the Levy flight strategy and the random walk strategy to explore the entire search space while maintaining the search for local optima.

[0057] Introducing Levy flight into the sparrow joiner update formula, the improved formula is as follows:

[0058]

[0059] In the formula, d is the vector dimension.

[0060] The calculation formula for the Levy flight strategy is as follows:

[0061]

[0062]

[0063] In the formula, Γ(x) = (x-1)!, Q is a random number that follows a normal distribution, r3, r4 are random numbers in the range [0,1], and ξ takes the value 1.5;

[0064] The formula for updating the position of sparrow joiners after introducing a random walk strategy is as follows:

[0065]

[0066] in, and Let represent two random solutions at iteration number t, where τ is a scaling factor and follows a uniform distribution in (0,1);

[0067] S523. Obtain the maximum network benefit value based on the sparrow positions in this round, and record the corresponding sparrow positions;

[0068] S524. Compare the maximum network benefit value obtained in this round with the global best network benefit value. If the network benefit value in this round is better than the global best benefit value, update the sparrow's position and update the global best network benefit value to the current network benefit value.

[0069] S525. Determine whether the number of iterations in this round has reached the maximum number of iterations. If not, repeat S52. Otherwise, output the optimal network efficiency value and the corresponding sparrow position.

[0070] The technical solution of the present invention can produce the following technical effects:

[0071] (1) The edge computing offloading method in the Internet of Things (IoT) scenario provided by this invention first constructs a computing offloading model in the IoT scenario, then converts the solution of the computing offloading strategy into an integer nonlinear optimization problem, and combines the sparrow search algorithm to iteratively solve the problem to obtain the optimal computing offloading strategy. The sine and cosine algorithm is introduced during the algorithm's execution to improve the algorithm's ability to escape local optima. (2) Introducing the Levy flight strategy and random walk strategy allows for comprehensive exploration in the search space while maintaining the search for local optima, thereby improving the algorithm's optimization capability and ensuring that each user device in the system model can reasonably offload tasks to the cloud or edge server, improving system efficiency and reducing the task drop rate. Attached Figure Description

[0072] Figure 1 This is a flowchart illustrating an edge computing offloading method in an Internet of Things (IoT) scenario according to the present invention.

[0073] Figure 2 This is a schematic diagram of an edge computing system model in the Internet of Things (IoT) scenario of this invention;

[0074] Figure 3 A comparison chart of the network efficiency of task unloading under four algorithms. Detailed Implementation

[0075] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0076] S1. Construct a cloud-edge collaborative IoT edge computing system model: This model includes a set of mobile devices U, a set of edge servers S, and a cloud server M with ultra-large computing power.

[0077] S2. Initialize the relevant parameters of the edge computing system model.

[0078] S21, the sets of mobile users and edge servers are represented by U = {1, 2, ..., u} and S = {1, 2, ..., s}, respectively. Each mobile device u ∈ U generates one computational task L at a time. u = < c L ,d L ,T a ,T b >, where c L d represents the workload required to complete the computational task. L T represents the amount of input data required to execute the program from the mobile device to the server. a and T b These represent the ideal latency and the maximum tolerable latency, respectively.

[0079] S22. Define the uninstallation strategy based on the system model. Where Q represents the set of computing tasks generated by the mobile device, It is a binary variable. This indicates that task L is unloaded to s∈S. This indicates that task L has been unloaded to the cloud M.

[0080] S23, The location of the mobile device is determined by This indicates that the location of the edge server is determined by... It means that x u y u These represent the x and y coordinates of the mobile device, respectively, with the height of the mobile device assumed to be 0; x s y s The x and y coordinates of the edge server are represented by , and H represents the height of the edge server.

[0081] S3. Establish latency and energy consumption models for cloud servers and edge servers.

[0082] S31. The channel gain from mobile device u∈U to MEC server s∈S in the uplink can be expressed as:

[0083]

[0084] In the formula, h0 represents the channel power gain when the transmission power is 1W and the transmission distance is 1m.

[0085] S32. Assuming that the locations of the mobile device and the edge server do not change within a time interval, the transmission rate from the mobile device to the edge server can be expressed as:

[0086]

[0087] In the formula, u′≠u, B and σ 2 Let p represent the uplink system bandwidth and noise power, respectively, where p = {p us |u∈U,s∈S} represents the transmission power from mobile user u∈U to edge server s∈S.

[0088] S33. The latency and energy consumption of the edge server and the cloud are obtained through the system model established by S1.

[0089] S331. The latency of unloading a task to the edge server can be expressed as:

[0090]

[0091] Where R us (t) represents the uplink transmission rate; C us This indicates the computing power of the edge server.

[0092] S332. The latency of unloading the task to the cloud server can be expressed as:

[0093]

[0094] In the formula d L This represents the amount of input data required to execute the program from the mobile device to the server. Considering the cloud server's location remains constant and is far from the mobile device, the relative distance between them can be approximated as a constant, since we can assume the rate R of the mobile device offloading to the cloud is... um It is a fixed value. C um This indicates the computing power in the cloud.

[0095] S333, The energy consumption of offloading tasks to cloud servers or edge servers can be expressed as:

[0096] The energy consumption of the offloading task includes the energy consumption of the transmission task and the energy consumption of the processing task. We define the transmission energy consumption of the mobile device u∈U as...

[0097]

[0098] The energy consumption of tasks processed by edge servers or cloud servers can be expressed as

[0099]

[0100] Where p s p represents the average power consumption of the edge server. m This indicates the average power consumption of the cloud server.

[0101] S4. When the edge server in a certain region is idle, the mobile devices in that region can directly offload the task to the edge server; if the edge server in a certain region is not idle, the mobile devices can choose to offload the task to the cloud server M. Based on this, the system benefit function is constructed.

[0102] S41. When mobile devices in the same area compete for the edge server in that area to complete the task offloading within the ideal latency, define the edge system efficiency η. L for:

[0103]

[0104] in: T a and T b T represents the ideal delay and the maximum tolerable delay, respectively. avg Indicates the average delay;

[0105] S42. When a large number of tasks are waiting to be unloaded by mobile users in a certain area, the waiting time for these tasks increases. In some extreme cases, computing tasks may be discarded. In this case, mobile users can choose to offload the computing tasks to a cloud server, but they will have to pay a cost, which is defined as:

[0106] o L =θ·η L +(1-θ)E pro

[0107] Where θ represents the weighting coefficient, which is 0.5 in this paper.

[0108] S43. Based on S3 and S4, construct a computation offloading system model for an IoT scenario. The network benefit is...

[0109]

[0110] Where e L This represents the normalized energy consumption.

[0111] S5. Use an improved sparrow search algorithm to continuously update sparrow positions and obtain the optimal unloading strategy and system efficiency value.

[0112] S51. Set the sparrow population in the algorithm and initialize the relevant parameters;

[0113] S52. After multiple iterations, the position of the sparrow in each optimization is continuously updated, and finally the position of the sparrow with the best fitness value is found. This position is then input into the spectrum allocation model built in step S1, and the system benefits are obtained according to S43.

[0114] S521. In the original sparrow search algorithm, as the number of iterations increases, the dimensionality of individual sparrows gradually decreases, leading to a reduction in the search space and making it prone to getting trapped in local optima. To improve this problem, the sparrow discoverer's position update strategy adopts the Sine-Cosine (SCA) algorithm and introduces a nonlinear sine learning factor. In the early stages of the search, the learning factor has a larger value, which is beneficial for global exploration, while in the later stages of the search, the learning factor has a smaller value, which is beneficial for improving local exploration ability and accuracy. Specifically, the learning factor formula and the improved discoverer position formula are as follows:

[0115] ω=ω min +(ω max -ω min )·sin(tπ / iter max )

[0116]

[0117] S522. In optimization problems, different search strategies can be adopted to improve search efficiency. The Levy flight strategy allows individuals to be widely distributed throughout the search space, which is beneficial for finding the global optimum; while the random walk strategy allows individuals to search within relatively dense regions, which is beneficial for finding local optima. Combining these two strategies allows for a comprehensive exploration of the search space while maintaining the search for local optima, thereby improving the algorithm's optimization capability.

[0118] (1) Levy flight strategy

[0119] In the process of population optimization, a larger search step size is needed in the early stages to improve the search ability for the global optimum, while the step size is reduced in the later stages to improve the search ability for the local optimum. The Levy flight strategy is a non-Gaussian stochastic gait that allows for alternating exploration using high-frequency short-distance and low-frequency long-distance methods. Therefore, this invention introduces Levy flight into the sparrow joiner update formula to improve the optimization effect. The improved formula is as follows:

[0120]

[0121] In the formula, d is the vector dimension, and the calculation formula for the Levy strategy is as follows:

[0122]

[0123]

[0124] In the formula, Γ(x) = (x-1)!, Q is a random number that follows a normal distribution, r3, r4 are random numbers in the range [0,1], and ξ takes the value 1.5.

[0125] (2) Random walk strategy

[0126] The random walk strategy enhances population diversity by introducing hybrid mutation and crossover, significantly improving the algorithm's ability to find local optimizations. Introducing it into intelligent optimization algorithms can accelerate the search for the optimal solution. The formula for the movement of sparrow entrants after incorporating the random walk strategy is as follows:

[0127]

[0128] in, and Let represent two random solutions at iteration number t, where τ is a scaling factor and follows a uniform distribution in (0,1).

[0129] S523. Obtain the maximum network benefit value based on the position of the sparrow in this round, and record the corresponding sparrow position.

[0130] S524. Compare the maximum network benefit value obtained in this round with the global best network benefit value. If the network benefit value in this round is better than the global best benefit value, update the sparrow's position and update the global best network benefit value to the current network benefit value.

[0131] S525. Determine whether the number of iterations in this round has reached the maximum number of iterations. If not, repeat S52. Otherwise, output the optimal network efficiency value and the corresponding sparrow position.

[0132] Experimental simulation was conducted on an edge computing offloading method in an Internet of Things (IoT) scenario according to the present invention:

[0133] The algorithm was simulated using MATLAB 2019b. The simulation scenario mainly included: one cloud server, multiple edge servers, and multiple user devices. Performance differences among edge devices were ignored, and the locations of user devices and edge devices were randomly generated. The parameters were set as follows: number of user devices U = 32, number of edge servers S = 8, and uplink channel bandwidth B = 20 × 10⁻⁶. 6 Hz, maximum transmission power of user equipment P us =5, the computing power of edge servers and cloud servers are C respectively. us =60GHz, C um =120GHz.

[0134] Simulation results are as follows Figure 3The improved Sparrow Search Algorithm (ISSA) is compared with Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Basic Sparrow Search Algorithm (SSA) through simulation experiments, resulting in a higher overall network efficiency. The simulation results show that the overall network efficiency increases with the number of iterations and gradually stabilizes at a constant. Because the improved Sparrow Search Algorithm is more likely to escape local optima, it achieves the highest overall network efficiency and converges faster than other algorithms.

[0135] The preferred embodiments of the present invention have been described in detail above, but they do not constitute any limitation on the present invention. Any equivalent substitutions or modifications made by those skilled in the art to the technical solutions and content disclosed in the present invention without departing from the scope of the present invention shall be deemed to have remained within the protection scope of the present invention.

Claims

1. A method for offloading edge computing in an Internet of Things (IoT) scenario, characterized in that, It includes the following steps: S1. Construct a cloud-edge collaborative IoT edge computing offloading system model: This model includes a set of mobile devices. A group of edge servers A cloud server with computing power ; S2. Initialize the relevant parameters of the edge computing unloading system model; S3. Establish latency and energy consumption models for cloud and edge servers; S4. When the edge servers in a certain region are idle, mobile devices in that region can directly offload tasks to the edge servers; if the edge servers in a certain region are not idle, mobile devices can choose to offload tasks to cloud servers. Based on this, a system benefit function is constructed; S5. The sine-cosine SCA algorithm, Levy flight strategy, and random walk strategy are introduced to improve the sparrow position update strategy. The improved sparrow search algorithm is used to iteratively update the sparrow position and obtain the optimal unloading strategy and system benefit value. Specifically: S51. Set the sparrow population in the algorithm and initialize the relevant parameters; S52. After multiple iterations, the position of the sparrow in each optimization is continuously updated, and finally the position of the sparrow with the best fitness value is found. This position is then fed into the computational unloading system model built in step S1 to obtain the network benefits. S2 specifically includes: S21, the collection of mobile devices and edge servers are respectively composed of and This indicates that each mobile device One computation task is generated at a time. ,in This indicates the workload required to complete the calculation task. This indicates the amount of input data required to execute the program from the mobile device to the server. Represents the ideal time delay. Indicates the maximum tolerable delay; S22. Define the uninstallation strategy based on the system model. ,in Represents the set of computing tasks generated by mobile devices, where It is a binary variable. Indicates task Uninstalled to , Indicates task Uninstalled to the cloud ; S23, The location of the mobile device is determined by This indicates that the location of the edge server is determined by... It means that among them , These represent the horizontal and vertical coordinates of the mobile device, respectively, with the height of the mobile device assumed to be 0. , These represent the x and y coordinates of the edge server, respectively. Indicates the height of the edge server; S4 specifically includes: S41. When mobile devices in the same area compete for edge servers in that area to complete task offloading within an ideal latency, define edge system efficiency. for: ， in: ; and Let represent the ideal latency and the maximum tolerable latency, respectively. Indicates the average delay; S42. When a mobile device in a certain region is processing a task, if the mobile device in that region chooses to offload the task to a cloud server... When the unloading cost is defined as: ， in This represents the weighting coefficient, which is 0.

5. S43. Based on S3 and S4, the network benefits of constructing a computation offloading system model for an IoT scenario are as follows: ， in This represents the normalized energy consumption. .

2. The edge computing offloading method in an IoT scenario according to claim 1, characterized in that, S3 specifically includes: S31, Mobile devices in the uplink To MEC server The channel gain is expressed as: ， In the formula This represents the channel power gain when the transmission power is 1W and the transmission distance is 1m. S32. Assuming that the locations of the mobile device and the edge server do not change within a time interval, the transmission rate from the mobile device to the edge server is expressed as: ， In the formula , and These represent the uplink system bandwidth and noise power, respectively. Indicates from mobile device To the edge server The magnitude of the transmission power; S33. The system model established through S1 yields the edge server and cloud server respectively. The latency and energy consumption.

3. The edge computing offloading method in an IoT scenario according to claim 2, characterized in that, The edge server and cloud server in S33 The specific details regarding latency and energy consumption include: S331, The latency of unloading the task to the edge server is expressed as: ， in Indicates the uplink transmission rate; This indicates the computing power of the edge server; S332. The latency of unloading the task to the cloud server is expressed as follows: ， In the formula This represents the amount of input data required to execute the program from the mobile device to the server; the relative distance between the cloud server and the mobile device is approximately constant, and the rate at which the mobile device is offloaded to the cloud... For a constant value, This indicates the computing power of the cloud server; S333, The energy consumption of offloading tasks to cloud servers or edge servers is expressed as follows: The energy consumption of the offloading task includes the energy consumption of the transmission task and the energy consumption of the processing task, which involves the mobile device. Task transmission energy consumption is defined as: ， The energy consumption of edge servers or cloud servers for processing tasks is expressed as follows: ， in This represents the average power consumption of the edge server. This indicates the average power consumption of the cloud server.

4. The edge computing offloading method in an IoT scenario according to claim 1, characterized in that, The specific steps in S52 to update the sparrow's position in each optimization process include: S521, the position update strategy for the sparrow discoverer adopts the Sine-Cosine (SCA) algorithm and introduces a nonlinear sinusoidal learning factor; the formulas for the learning factor and the improved discoverer position formula are as follows: ， In the formula, A random number within the range [0, 2π]. It is a random number within the range [0, 2]. As a learning factor, Indicates the number of iterations. At that time, the first The sparrow in the first Position in each dimension and These represent the warning value and the safe value, respectively. S522: Combine the Levy flight strategy and the random walk strategy to explore the entire search space while maintaining the search for local optima. Introducing the Levy flight strategy into the sparrow joiner update formula, the improved formula is as follows: ， In the formula For vector dimensions, The calculation formula for the Levy flight strategy is as follows: ， ， In the formula Q is a random number that follows a normal distribution. Random numbers within a range The value is 1.5; The formula for updating the position of sparrow joiners after introducing a random walk strategy is as follows: ， in, and These represent the number of iterations respectively. Two random solutions at time, Scaling factor, obeying Uniform distribution; S523. Obtain the maximum network benefit value based on the sparrow positions in this round, and record the corresponding sparrow positions; S524. Compare the maximum network benefit value obtained in this round with the global best network benefit value. If the network benefit value in this round is better than the global best benefit value, update the sparrow's position and update the global best network benefit value to the current network benefit value. S525. Determine whether the number of iterations in this round has reached the maximum number of iterations. If not, repeat S52. Otherwise, output the optimal network efficiency value and the corresponding sparrow position.

Images