End-edge-cloud cooperative computing task unloading method for thermodynamic Internet of Things

Abstract

The invention relates to an end-edge-cloud cooperative computing task unloading method for a thermodynamic Internet of Things, and the method comprises the steps: building a system structure of the thermodynamic Internet of Things, initializing system parameters, and respectively building a time delay model and an energy consumption model; constructing a cost function according to the time delay model and the energy consumption model, and taking the cost function as an optimization objective function; and performing iterative operation based on the improved mayfly naiad optimization algorithm to generate a task unloading scheme. The method has the beneficial effects that a mathematical model for task unloading of different data volumes under multiple edge users and multiple cloud servers is established, different task unloading strategies are searched, a better task unloading scheme is obtained based on the improved mayfly naiad optimization algorithm, and the overall time delay of the system is reduced under the constraint of energy consumption.

Description

Technical Field

[0001] The present invention relates to the technical field of thermal Internet of Things, and more precisely, it relates to an edge-cloud collaborative computing task offloading method for thermal Internet of Things. Background Art

[0002] With the acceleration of urbanization and the improvement of people's requirements for the quality of life, more and more thermal terminals are connected to the thermal Internet of Things, which inevitably leads to a rapid increase in data traffic and continuous increase in network pressure. To meet such challenges and demands, cloud computing technology has emerged, offloading massive data and computing tasks to the cloud for unified processing. However, while cloud computing solves the problem of insufficient computing resources at the edge, it also brings many problems. First, the process of transmitting massive data generated by terminal devices to the cloud computing center will inevitably bring high network latency and energy consumption. Second, more and more devices at the edge are connected to the cloud, which will cause congestion in the transmission link from the devices at the edge to the cloud center. To solve these problems, by artificially sinking storage and computing, edge computing technology has emerged, enabling devices at the edge to have a certain computing ability, and different tasks can be selected to be executed locally or offloaded to the cloud for execution.

[0003] At the same time, the introduction of edge computing technology also brings a series of challenges. For example, in the case of a large single task volume, the devices at the edge may require a relatively long time to perform calculations, while in the case of a small single task volume, transmitting the task to the cloud will bring relatively high energy consumption. Coupled with the concurrency of multi-user tasks, how to select the execution location of different tasks affects the latency and energy consumption of the entire system. Therefore, it is necessary to design a reasonable task offloading optimization method to minimize the task execution latency as much as possible while meeting the energy consumption requirements.

[0004] Regarding the task offloading optimization problem in the edge-cloud collaborative mode for the thermal Internet of Things scenario, many scholars have used methods such as game theory and hierarchical optimization algorithms to achieve optimization. However, the traditional algorithms have poor robustness, complex and time-consuming calculation processes, and do not have global search capabilities; some scholars have used intelligent algorithms to achieve the optimization of task offloading under single-user scenarios, but have not modeled the task offloading problem in the scenario of multi-user and multi-server, and are not applicable to actual scenarios. Summary of the Invention

[0005] The object of the present invention is to propose an edge-cloud collaborative computing task offloading method for thermal Internet of Things in view of the deficiencies of the prior art.

[0006] In a first aspect, there is provided an edge-cloud collaborative computing task offloading method for thermal Internet of Things, including:

[0007] S1. Establish the structure of the thermal Internet of Things system, initialize the system parameters, and construct a delay model and an energy consumption model respectively;

[0008] S2. Construct a cost function according to the delay model and the energy consumption model, and use the cost function as the optimization objective function;

[0009] S3. Perform iterative operations based on an improved mayfly optimization algorithm to generate a task offloading scheme.

[0010] Preferably, in S1, the structure of the thermal Internet of Things system includes: a cloud server, an edge server, and end-side devices; among them, the end-side devices are connected to the edge server, and the edge server is connected to the cloud server; the computing tasks generated by the end-side devices are offloaded to the corresponding edge server or cloud server for execution; the end-side devices are intelligent thermal sensors.

[0011] Preferably, in S1, the delay model includes local computing delay, edge computing delay, and cloud computing delay; the energy consumption model includes local device energy consumption and edge server energy consumption.

[0012] Preferably, S3 includes:

[0013] S301. Initialize the population parameters and generate an initial population, which includes male mayflies and female mayflies;

[0014] S302. Calculate the fitness of all individuals, and record the optimal individuals and positions of female and male mayflies respectively;

[0015] S303. When the current iteration number is less than the maximum iteration number, update the velocities of female and male individuals;

[0016] S304. Execute a hybrid crossover strategy to generate offspring; and perform an adaptive Gaussian mutation operation on the offspring;

[0017] S305. Perform a fine local search on the male population by integrating simulated annealing;

[0018] S306. Merge the new population and the original population to form an extended population; sort according to the cost value and perform truncation retention;

[0019] S307. When the number of iterations at which the global optimal solution stagnates exceeds a preset threshold, reset the non-elite individuals and retain the optimal solution; then update the algorithm parameters;

[0020] S308. Repeat S303 - S307 until the iteration number reaches the maximum iteration number, and output the global optimal solution and the task offloading scheme.

[0021] Preferably, in S303, when updating the female and male populations, randomly select some dimensions to update the velocities.

[0022] Preferably, in S304, the hybrid crossover strategy includes differential crossover and shuffle crossover.

[0023] Preferably, in S304, the perturbation dimension and perturbation range of the adaptive Gaussian mutation operation vary randomly.

[0024] In a second aspect, a device for end-edge-cloud collaborative computing task offloading for the thermal Internet of Things is provided, which is used to execute any of the methods in the first aspect, including:

[0025] A building module, configured to build the structure of the thermal Internet of Things system, initialize system parameters, and respectively construct a delay model and an energy consumption model;

[0026] A construction module, configured to construct a cost function according to the delay model and the energy consumption model, and use the cost function as an optimization objective function;

[0027] An iterative operation module, configured to perform iterative operations based on an improved mayfly optimization algorithm to generate a task offloading scheme.

[0028] In a third aspect, a computer storage medium is provided, in which a computer program is stored; when the computer program runs on a computer, the computer is enabled to execute any of the methods in the first aspect.

[0029] In a fourth aspect, an electronic device is provided, including:

[0030] A memory, configured to store a computer program;

[0031] A processor, configured to execute the computer program to implement any of the methods in the first aspect.

[0032] The beneficial effects of the present invention are as follows: The present invention provides a task offloading optimization method for a cloud-edge collaborative mode in a multi-user scenario. The method establishes a mathematical model for task offloading with different data volumes under multiple edge users and multiple cloud servers, searches for different task offloading strategies, and obtains a relatively optimal task offloading scheme based on an improved mayfly optimization algorithm, reducing the overall system delay under energy consumption constraints. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a schematic structural diagram of the end-edge-cloud thermal Internet of Things system provided by the present application;

[0034] Figure 2 It is a general flowchart of a method for end-edge-cloud collaborative computing task offloading for the thermal Internet of Things provided by the present application;

[0035] Figure 3 It is a flowchart of female update provided by the present application;

[0036] Figure 4 The male update flowchart provided for this application;

[0037] Figure 5 The mating operation flowchart provided for this application;

[0038] Figure 6 The mutation operation flowchart provided for this application;

[0039] Figure 7 The simulated annealing local search flowchart provided for this application;

[0040] Figure 8 The schematic diagram of the offloading result of the computing task under 100 task concurrency provided for this application;

[0041] Figure 9 The schematic diagram of the offloading result of the computing task under 500 task concurrency provided for this application. Detailed implementation manners

[0042] The following further describes the present invention in conjunction with embodiments. The description of the following embodiments is only used to help understand the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several modifications can still be made to the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

[0043] Embodiment 1:

[0044] To solve the problems of the prior art, Embodiment 1 of this application provides an end-edge-cloud collaborative computing task offloading method for the thermal Internet of Things, which can reduce the decision delay of the thermal Internet of Things system and reduce the energy consumption of the terminal, including:

[0045] S1. Establish the structure of the thermal Internet of Things system, initialize the system parameters, and respectively construct a delay model and an energy consumption model.

[0046] In S1, as Figure 1 shown, the structure of the thermal Internet of Things system includes: cloud - several cloud servers, edge - several edge servers, end - several intelligent thermal sensors, and each end-side device is connected to a corresponding edge server, and all edge servers are connected to the cloud server, that is, each task generated by each device on the end side can be offloaded to the corresponding edge server, and each task generated by each device on the edge side can be offloaded to any cloud server for execution, and each task generated on the end side can be offloaded to any cloud server through the edge server for execution.

[0047] In addition, the system parameters include: offloading decision variables , (1 represents local, 2 represents edge, 3 represents cloud); Transmission resource allocation , (indicating the bandwidth allocation ratio of task i at the edge); Computing resource allocation , (indicating the CPU allocation ratio of task i at the edge).

[0048] In S1, the delay model includes local computing delay, edge computing delay, and cloud computing delay; the energy consumption model includes local device energy consumption and edge server energy consumption.

[0049] Specifically, the total delay ( ) consists of local computing delay , edge computing delay , and cloud computing delay , expressed as:

[0050]

[0051] where N is the number of concurrent tasks, , is the computational amount (CPU cycles) of task , is the computing resource of the local device (cycle / s), ; is the transmission amount (bit) of task i, is the computational amount (CPU cycles) of task , is the computing resource of the edge server (cycle / s), k is the number of tasks of the current edge server, is the maximum number of tasks that the edge server can process simultaneously, is the single waiting time (related to MaxS), is the transmission resource allocation ratio ( ), is the channel gain, is the noise power spectral density. , is the fixed delay of cloud computing (seconds).

[0052] The total energy consumption ( ) consists of local energy consumption and edge energy consumption , expressed as:

[0053]

[0054] where, , is the computing resource of the local device (cycle / s), is the inherent energy consumption coefficient of the local device, is the task computing volume (CPU cycles); , is the transmission power, is the transmission data volume of the i-th task, W is the channel bandwidth, is the transmission resource allocation ratio ( ), is the channel gain, is the noise power spectral density.

[0055] In addition, this application also considers constraint conditions, including the maximum communication resources of the edge server, the maximum computing resources of the edge server, the mayfly flight speed limit, and when no communication resources are allocated to a certain local device, the computing task must be executed on the local device. Specifically, the constraint conditions are as follows:

[0056] Constraint condition 1: , indicating that there is a maximum upper limit for all the communication resources of the edge server, and the communication resources allocated to a certain local device cannot exceed this upper limit;

[0057] Constraint condition 2: , indicating that there is a maximum upper limit for all the computing resources of the edge server, and the computing resources allocated to a certain local device cannot exceed this upper limit;

[0058] Constraint condition 3: , requiring that when no communication resources are allocated to a certain local device, the computing task must be executed on the local device;

[0059] Constraint condition 4: Discrete decision uniqueness: ;

[0060] Constraint condition 5: Parallel processing limit: ;

[0061] Constraint condition 6: Resource allocation normalization: .

[0062] S2. Construct a cost function according to the delay model and the energy consumption model, and use the cost function as the optimization objective function.

[0063] The cost function can be expressed as:

[0064]

[0065] represents the system benefit, represents the delay part, represents the energy consumption part, and the system benefit is the weighted sum of the delay benefit and the energy consumption benefit, and are their respective weights, and the sum of the weights is equal to 1; represents the minimum value; the purpose of the present invention is to minimize the total system benefit price.

[0066] S3. Perform iterative operations based on the improved mayfly optimization algorithm to generate a task offloading scheme.

[0067] The mayfly optimization algorithm (MA) is a newly emerging heuristic optimization algorithm based on the behaviors of organisms in nature. It mimics the life cycle and behavioral characteristics of mayflies. The core idea of the mayfly optimization algorithm stems from the observation of the life habits of mayflies, especially their flight patterns, foraging behaviors, and reproductive strategies. In the algorithm, each mayfly individual represents a potential solution in the solution space, and the entire mayfly population searches for the optimal solution by simulating the survival competition in nature. At the beginning of the algorithm, a certain number of mayfly individuals are randomly generated as the initial population, and then the positions of these individuals are continuously updated through the iterative process to find the global optimal solution. As a swarm intelligence algorithm, it has shown great potential in solving complex optimization problems. Currently, the mayfly optimization algorithm has been applied to solve problems in multiple fields. For example, in the engineering field, the mayfly optimization algorithm can be used to solve problems such as mechanical design and electrical system optimization. For example, in the reactive power optimal allocation of a power system, this algorithm can help find the best output scheme of generating units, thereby reducing network losses and improving voltage stability; in the field of machine learning, the mayfly optimization algorithm can be used to optimize the architecture and weights of neural networks. By searching a large number of possible network configurations, the algorithm can identify the best-performing model configuration, thereby improving the prediction accuracy or classification effect; in addition to the above applications, the mayfly optimization algorithm has also been applied to multiple fields such as resource allocation, production scheduling, and logistics distribution. In these scenarios, the algorithm usually needs to handle a large number of variables and constraints, and its powerful global search ability makes it an ideal choice for solving these problems.

[0068] This application improves the mayfly optimization algorithm. Specifically, as Figure 2 shown, S3 includes:

[0069] S301. Initialize the population parameters and generate an initial population, where the initial population includes male mayflies and female mayflies.

[0070] Exemplarily, let the positions of male and female mayflies in the d-dimensional solution space be , and evaluate the performance according to the cost function defined above. Let the flight speed of mayfly individuals be , and divide them into male and female groups.

[0071] S302. Calculate the initial cost function value, and record the optimal individuals and positions of female and male mayflies respectively.

[0072] S303. As Figure 3 and Figure 4 shown, when the current iteration number t is less than the maximum iteration number T, update the velocities of female and male individuals.

[0073] In S303, when updating the female and male populations, the velocities are updated by randomly selecting some dimensions.

[0074] The flight direction of each mayfly is jointly affected by individual and social flight experiences. Each mayfly adjusts its trajectory towards the individual best position so far and the group best position reached by all mayflies in the group.

[0075] Male mayflies tend to gather in groups, and the position of each male mayfly is adjusted according to its own and neighboring experiences. Let be the current position of mayfly i searching the current space at the t-th iteration. When updating the position, the velocity at the (t + 1)-th iteration plus the sum of the positions at the t-th iteration, and its position expression is: ; Considering the continuous movement of mayflies, they perform dances within a certain distance on the water surface, and its velocity is updated as:

[0076] pbestij - gbestj -

[0077] where is the position of mayfly i at the j-th dimension at the t-th iteration; and are the attraction coefficients of the swimming behavior of mayflies; pbest is the individual best position, gbest is the global optimal position, is the visibility coefficient, which is used to control the visibility range of mayflies, represents the distance between the current position and pbest; represents the distance between the current position and gbest. Its distance calculation formula is ; In order to obtain the optimal position, mayflies must continuously update their velocities, and its velocity is updated as: , where d represents the dance hypothesis coefficient, constantly attracting the opposite sex, is a random number.

[0078] In addition, through probability control update, generate a random array with the same dimension as the variable (range [0, 1]). Generate a random threshold in the range [0, 0.5]. is a boolean array that marks whether each variable dimension needs to be updated: if the random number is less than the threshold, update that dimension; otherwise, keep it unchanged. This avoids premature convergence caused by synchronous updates of all dimensions and enhances population diversity.

[0079] The update of female mayflies is as follows:

[0080] The difference between female mayflies and male mayflies is that male mayflies tend to gather while female mayflies do not gather in groups, but female mayflies will fly to mate with male mayflies. Assume is the position of the female mayfly at the t-th iteration, and its position update is expressed as: ; The speed update of female mayflies is as follows:

[0081]

[0082] where, is the speed of the mayfly, is the position of the female mayfly at the t-th iteration, represents the attraction coefficient, is the visibility coefficient, represents the distance between male and female mayflies, is the random walk coefficient, which only works when the female mayfly is not under attack, is a random number.

[0083] The speed of each step of iteration of male and female mayflies completely inherits the speed of the previous step of iteration, which will lead to a rapid increase in the speed of mayfly individuals. Sometimes, it is necessary to reduce the acceleration so that the speed evolution of mayfly individuals slows down in order to better control the balance between the exploration and exploitation capabilities of mayflies. Therefore, referring to the speed update rule in the particle swarm algorithm, an inertia factor is added to the speed update of the above standard mayfly algorithm. Therefore, the speed update formula of male mayflies can be modified to:

[0084] pbestij - gbestj -

[0085] At the same time, the speed update formula of female mayflies can be modified to:

[0086] , represents the ordinate of the current position of the female mayfly, represents the abscissa of the current position of the female mayfly.

[0087] Similar to the male update, the update dimension is controlled by .

[0088] The present invention adopts a method of updating the speed by probability control, so that when the male and female populations are updated, not all dimensions are updated synchronously, but only some dimensions are updated. Therefore, it overcomes the destruction of excellent genes in the process of updating the speeds of male and female mayflies in the standard mayfly algorithm, avoids premature convergence caused by synchronous updating of all dimensions, enhances the population diversity, thus retains more excellent genes for the subsequent mating process, and further enhances the global search ability and the stability of the optimization result.

[0089] S304. Execute the hybrid crossover strategy to generate offspring; and perform adaptive Gaussian mutation operation on the offspring.

[0090] S305. Perform fine local search on the male population by fusing simulated annealing.

[0091] S306. Merge the newborn population and the original population to form an extended population; sort according to the cost value and perform truncation retention.

[0092] S307. When the number of iterations at which the global optimal solution stagnates exceeds the preset threshold, reset the non-elite individuals and retain the optimal solution; then update the algorithm parameters.

[0093] S308. Repeat S303 - S307 until the number of iterations reaches the maximum number of iterations, and output the global optimal solution. The global optimal solution includes the offloading method for each task, that is, determining the optimal task offloading scheme, and the corresponding minimum total system cost. The offloading decision is determined by the offloading decision variable to determine, the offloading decision variable , (1 represents local, 2 represents edge, 3 represents cloud).

[0094] Embodiment 2:

[0095] Based on Embodiment 1, Embodiment 2 of the present application provides a more specific end-edge-cloud collaborative computing task offloading method for the thermal Internet of Things, including:

[0096] S1. Establish the system structure of the thermal Internet of Things, initialize the system parameters, and respectively construct the delay model and the energy consumption model.

[0097] S2. Construct a cost function according to the delay model and the energy consumption model, and use the cost function as the optimization objective function.

[0098] S3. Perform iterative operations based on the improved mayfly optimization algorithm to generate a task offloading scheme.

[0099] S3 includes:

[0100] S301. Initialize the population parameters and generate an initial population, where the initial population includes male mayflies and female mayflies.

[0101] S302. Calculate the fitness of all individuals, and record the optimal individuals and positions of female and male mayflies respectively.

[0102] S303. When the current iteration number is less than the maximum iteration number, update the velocities of female and male individuals.

[0103] S304. As shown in Figure 5 , execute the hybrid crossover strategy to generate offspring; as shown in Figure 6 , and perform adaptive Gaussian mutation operation on the offspring.

[0104] Female and male mayflies mate to produce the next generation. The mating process is as follows: Select one parent from female and male respectively, and the method of selecting female and male samples is the same as that of male attracting female. In the standard mayfly algorithm, the optimal individual of male mayfly mates with the optimal individual of female mayfly, and the sub-optimal individual of male mayfly mates with the sub-optimal individual of female mayfly. After mating, two offspring, namely the optimal and sub-optimal ones, are obtained.

[0105] This application adds shuffle crossover to form a hybrid crossover. Through probability control, 50% of the crossover method remains unchanged, and 50% adopts shuffle crossover. The crossover method of the original mayfly algorithm is suitable for continuous optimization, while shuffle crossover is suitable for optimizing discrete decisions, such as the offloading target , and the hybrid crossover can optimize both continuous and discrete variables and adapt to complex problems.

[0106] The offspring of the original standard mayfly algorithm are:

[0107]

[0108] where is a random number, male is the parent, and female is the mother.

[0109] The offspring of shuffle crossover are:

[0110]

[0111] In addition, in S304, an adaptive Gaussian mutation operation is also performed on the positions of the offspring mayflies. Different from the Gaussian mutation with a fixed range used in the standard mayfly algorithm, in the present invention, the perturbation dimension and mutation range are controlled. Some dimensions are randomly selected, and the perturbation range changes randomly.

[0112] The specific operation is as follows: , , is the mutated individual.

[0113] In S304, the present invention introduces a hybrid crossover strategy in the mating operation, including two methods: differential crossover and shuffle crossover. Differential crossover is applicable to continuous variable optimization, while shuffle crossover is more suitable for dealing with discrete decision variables. This method overcomes the problem that a single crossover method cannot effectively handle the mixed variables in the thermal Internet of Things, and improves the ability of the algorithm to solve the actual thermal Internet of Things end-edge-cloud collaborative task offloading problem.

[0114] In addition, in the mutation operation stage, the standard mayfly algorithm usually uses mutation with a fixed range, lacking flexibility. The present invention adopts an adaptive Gaussian mutation operation, in which the perturbation dimension and range vary randomly, rather than using mutation with a fixed range. This improvement overcomes the problem of lack of flexibility in the mutation operation of the standard mayfly algorithm, enhances the local search ability of the algorithm, and helps to find a better solution.

[0115] S305. As Figure 7 shown, perform a fine local search on the male population by fusing simulated annealing. The purpose of fusion is to perturb the solution within a local range and jump out of the local optimum by probabilistically accepting a worse solution.

[0116] Specifically, randomly select 10% of the male individuals for annealing operation.

[0117] Generate a new solution by perturbation: ; The perturbation intensity is related to the current temperature: ;

[0118] Acceptance criterion: .

[0119] In S305, the present invention fuses the simulated annealing mechanism into the male population to refine the local search ability. This method allows the algorithm to accept a worse solution with a certain probability, which can avoid falling into the local optimum and enhance the global search ability. And through dynamic balance, it biases towards exploration at high temperatures and towards careful search at low temperatures. The standard mayfly algorithm lacks the annealing mechanism and only relies on genetic operations, making it vulnerable to local extrema.

[0120] S306. Merge the newborn population and the original population to form an extended population; sort according to the cost value and perform truncation retention.

[0121] S307. When the number of iterations at which the global optimal solution stagnates exceeds a preset threshold, reset the non-elite individuals and retain the optimal solution; then update the algorithm parameters.

[0122] Exemplarily, determine whether the number of iterations at which the global optimal solution stagnates exceeds 10% of the maximum number of iterations (for example, a total of 200 iterations with no change for 20 consecutive times). If so, retain the optimal solution, randomly generate all other solutions, that is, reset the non-elite individuals, update the algorithm parameters, and then increment the current iteration number by 1, denoted as: t = t + 1.

[0123] Among them, the update algorithm parameters are as follows:

[0124] Inertia weight:

[0125] Dance factor:

[0126] Temperature attenuation:

[0127] In S307, when the global optimal solution of this application stagnates for more than 10% of the iteration times, the non-elite individuals are reset and the historical optimum is retained, which overcomes the problem that the male and female mayflies in the standard mayfly algorithm follow each other and will inevitably fall into a local optimum and cannot jump out after a certain number of iterations, avoids premature convergence, and ensures that the algorithm can continuously explore new possible solution spaces.

[0128] S308. Repeat S303 - S307 until the iteration times reach the maximum iteration times, and output the global optimal solution and the task offloading scheme.

[0129] Finally, by adopting the above-mentioned measures, the total cost of the thermal Internet of Things edge-cloud collaborative system of the present invention is reduced by 9.69% compared with the standard mayfly algorithm in the case of 100 task concurrency, and is reduced by 2.25% - 16.40% compared with mainstream algorithms such as PSO and BWO; the total cost of the system is reduced by 12.75% compared with the standard mayfly algorithm in the case of 500 task concurrency, and is reduced by 10.37% - 22.44% compared with mainstream algorithms such as PSO and BWO, and the global stability is significantly enhanced, which is applicable to the actual thermal Internet of Things scenarios with high concurrency and strong constraints.

[0130] Table 1 records the comparison data of seven algorithms in the thermal Internet of Things edge-cloud collaborative computing task offloading experiment, where the seven algorithms are the improved mayfly algorithm IMA, the beluga whale optimization algorithm BWO, the particle swarm optimization algorithm PSO, the mayfly algorithm MA, the chicken swarm optimization algorithm CSO, the grasshopper optimization algorithm GOA, and the dung beetle optimization algorithm DBO.

[0131] Table 1 Comparison experiment of seven algorithms for edge-cloud collaborative task offloading (total system price)

[0132]

[0133] Figure 8 and Figure 9 are the best offloading system cost curves of task offloading instances using the improved algorithm under different task concurrency; the abscissa represents the population iteration times of the algorithm, and the ordinate represents the system cost function Values; The legend respectively represents: the improved mayfly algorithm IMA, the beluga whale optimization algorithm BWO, the particle swarm optimization algorithm PSO, the mayfly algorithm MA, the chicken swarm optimization algorithm CSO, the grasshopper optimization algorithm GOA, and the dung beetle optimization algorithm DBO of the present invention.

[0134] Through the result analysis and comparison among the algorithms, it can be found that the total price benefit of the improved mayfly algorithm system is the lowest and continuously decreases during the iteration process, indicating strong global optimization ability, fine local search, and not easy to fall into local optimum; it shows that the improved mayfly algorithm described in the present invention is applicable to the end-edge-cloud collaborative computing task offloading of the thermal Internet of Things, reduces the energy consumption of local devices, and reduces the system delay.

[0135] It should be noted that the parts that are the same or similar to those in Embodiment 1 in this embodiment can be referred to each other and will not be elaborated in this application.

[0136] Embodiment 3:

[0137] Based on Embodiment 2, Embodiment 3 of the present application provides an end-edge-cloud collaborative computing task offloading device for the thermal Internet of Things, including:

[0138] A building module, configured to build a thermal Internet of Things system structure, initialize system parameters, and respectively construct a delay model and an energy consumption model;

[0139] A construction module, configured to construct a cost function according to the delay model and the energy consumption model, and use the cost function as an optimization objective function;

[0140] An iterative operation module, configured to perform iterative operations based on the improved mayfly optimization algorithm to generate a task offloading scheme.

[0141] It should be noted that the device provided in this embodiment is the device corresponding to the method provided in Embodiment 2. Therefore, the parts that are the same or similar to those in Embodiment 2 in this embodiment can be referred to each other and will not be elaborated in this application.

Claims

1. A method for end-edge-cloud collaborative computing task offloading in the thermal Internet of Things, characterized in that Including: S1. Establish the structure of the thermal Internet of Things system, initialize the system parameters, and respectively construct a delay model and an energy consumption model; S2. Construct a cost function based on the delay model and the energy consumption model, and use the cost function as the optimization objective function; S3. Perform iterative operations based on an improved mayfly optimization algorithm to generate a task offloading scheme.

2. The end-edge-cloud collaborative computing task offloading method for the thermal Internet of Things according to claim 1, wherein, In S1, the thermal Internet of Things system structure includes: a cloud server, an edge server, and end-side devices; among them, the end-side devices are connected to the edge server, and the edge server is connected to the cloud server; the computing tasks generated by the end-side devices are offloaded to the corresponding edge server or cloud server for execution; the end-side devices are intelligent thermal sensors.

3. The edge-cloud collaborative computing task offloading method for the thermal Internet of Things according to claim 2, wherein In S1, the delay model includes local computing delay, edge computing delay, and cloud computing delay; the energy consumption model includes local device energy consumption and edge server energy consumption.

4. The end-edge-cloud collaborative computing task offloading method for the thermal Internet of Things according to claim 3, wherein S3 Including: S301. Initialize the population parameters and generate an initial population, which includes male mayflies and female mayflies; S302. Calculate the fitness of all individuals, and respectively record the optimal individuals and positions of male and female mayflies; S303. When the current iteration number is less than the maximum iteration number, update the speeds of male and female individuals; S304. Execute a hybrid crossover strategy to generate offspring; and perform an adaptive Gaussian mutation operation on the offspring; S305. Perform fine local search on the male population by integrating simulated annealing; S306. Merge the new population and the original population to form an extended population; sort according to the cost value and perform truncation retention; S307. When the iteration number of the stagnant global optimal solution exceeds a preset threshold, reset the non-elite individuals and retain the optimal solution; then update the algorithm parameters; S308. Repeat S303 - S307 until the iteration number reaches the maximum iteration number, and output the global optimal solution and the task offloading scheme.

5. The edge-cloud collaborative computing task offloading method for the thermal Internet of Things according to claim 4, characterized in that In S303, when updating the female and male populations, randomly select some dimensions to update the speed.

6. The end-edge-cloud collaborative computing task offloading method for the thermal Internet of Things according to claim 4, wherein In S304, the hybrid crossover strategy includes differential crossover and shuffle crossover.

7. The edge-cloud collaborative computing task offloading method for the thermal Internet of Things according to claim 4, wherein In S304, the perturbation dimensions and perturbation ranges of the adaptive Gaussian mutation operation change randomly.

8. An edge-cloud collaborative computing task offloading device for the thermal Internet of Things, characterized in that, For executing the method according to any one of claims 1 to 7, including: An establishment module, configured to establish the structure of the thermal Internet of Things system, initialize the system parameters, and respectively construct a delay model and an energy consumption model; A construction module, configured to construct a cost function based on the delay model and the energy consumption model, and use the cost function as the optimization objective function; An iterative operation module, configured to perform iterative operations based on an improved mayfly optimization algorithm to generate a task offloading scheme.

9. A computer storage medium, characterized in that, The computer storage medium stores a computer program; when the computer program runs on a computer, the computer is made to execute the method according to any one of claims 1 to 7.

10. An electronic device, characterized in that, Including: A memory, configured to store the computer program; A processor, configured to execute the computer program to implement the method according to any one of claims 1 to 7.

Images