Cloud-edge collaborative offloading scheduling and resource scheduling method and device based on synergistic computing

Abstract

The present invention discloses a cloud-edge-terminal collaborative unloading scheduling and resource scheduling method and device based on synaesthesia, which divides the coverage of a cloud into multiple sub-areas according to different base station coverages, and selects one of three unloading methods for each terminal to unload the sensing data; the sum of the energy consumption of the terminal and the delay of task completion of the three TDMA-based unloading methods is formulated as the sum of the cost of task completion, and based on this, the task cost optimization problem is modeled according to different unloading methods and the time slot allocation of sensing and unloading; the unloading method and TDMA unloading time slot of each terminal are determined by solving the optimization problem using linear programming and particle swarm methods; and finally, the purpose of minimizing the cost of multiple terminals completing the inspection task is achieved through cloud-edge-terminal collaborative unloading and scheduling optimization of sensing unloading time slots. It can effectively save the cost of multiple terminals in the collaborative execution of tasks under different unloading conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent industrial Internet, and specifically to a cloud-edge-terminal collaborative unloading scheduling and resource scheduling method and device based on telepathic computing. Background Art

[0002] As a new type of intelligent infrastructure in the context of a new round of industrial revolution, the traditional manual operation and maintenance mode cannot support the rapid and accurate positioning of basic equipment network faults in modern factories, and there are problems of low operation and maintenance efficiency and high operation and maintenance costs. Through the operation and maintenance intelligent terminal, intelligent operation and maintenance services such as real-time inspection and remote maintenance can be realized. Therefore, the intelligent terminal needs to transform into an intelligent and automated active mode, which can perceive the status of surrounding equipment and sensor information, and process and convert it into knowledge to perform operation and maintenance work autonomously. Since the intelligent terminal needs to handle tasks such as equipment abnormal status detection, path planning and obstacle avoidance, and data recording and forwarding in the intelligent factory, the intelligent terminal needs to have functions such as real-time perception, navigation positioning, and intelligent decision-making. These tasks and functions have high requirements for computing power and real-time performance. The computing power limitations of onboard functions and traditional cloud computing cannot support the knowledge understanding and decision-making capabilities required for network intelligence, and it is difficult to meet the accuracy and timeliness requirements of diversified network tasks. The service capabilities of cloud computing platforms alone are ultimately limited. Therefore, in order to meet the latency and computing power requirements of smart terminals to complete tasks, edge computing technology is applied. Due to its positioning advantage of being close to smart terminals, it can provide terminal users with local computing services, collaborate with cloud computing, and make reasonable offloading plans for various computing tasks to save task costs.

[0003] Usually when studying edge computing issues, it is assumed that data is generated by the terminal itself. However, when the smart terminal is performing business, the equipment in the actual factory cannot be directly connected to the local area network. The smart terminal needs to perceive data from the environment to generate tasks. In the process of perception-then-offloading, the device first perceives data from the environment, and then offloads the perceived data to the edge server for calculation. This creates a trade-off between perception and offloading. Therefore, the impact of the perception and collection stage of data on the overall offloading decision, latency, and energy consumption cannot be ignored. Therefore, how to reasonably arrange the resource allocation of perception, communication, and offloading according to the actual application scenario, and take the appropriate position for task offloading and calculation to achieve overall high operating efficiency is a basic problem for technicians in this field. Summary of the invention

[0004] The present invention proposes a cloud-edge collaborative unloading scheduling and resource scheduling method, device and storage medium based on synergistic computing, which can solve at least one of the technical problems in the background technology.

[0005] To achieve the above object, the present invention adopts the following technical solution: comprising the following steps: S1, scene description; S2, model establishment; S3, model solution;

[0006] In step S1, the scenario description includes dividing a cloud coverage area into multiple sub-areas according to different base station coverage areas, and selecting one of three offloading modes for each terminal to offload the sensing data;

[0007] In step S2, the model building includes formulating the sum of the energy consumption of the terminal and the delay of task completion of the three TDMA-based offloading methods, which is defined as the sum of the cost of task completion, and accordingly modeling it into a task cost optimization problem according to different offloading methods and the time slot allocation of sensing and offloading;

[0008] In step S3, the model solution includes using low-complexity linear programming and particle swarm methods to solve the optimization problem, determine the unloading method of each terminal, and determine the TDMA unloading time slot of each terminal; ultimately, the purpose of minimizing the cost of multiple terminals completing inspection tasks is achieved through cloud-edge collaborative unloading and scheduling optimization of perceived unloading time slots.

[0009] Furthermore, in step S1, there are three types of offloading locations, namely, offloading calculation locally, offloading calculation at a base station, and offloading calculation at a cloud;

[0010] Among them, local offload computing refers to the perception data collected by the terminal being directly calculated locally at the terminal, base station offload computing refers to the perception data collected by the terminal being offloaded to the base station for calculation without occupying the computing resources of the terminal, and cloud offload computing refers to the perception data collected by the terminal being first transmitted to the base station, and then transmitted to the cloud by the base station, using the computing resources of the cloud to process and calculate the task;

[0011] The selection of the three offloading methods requires comprehensive consideration of the weighted sum of the latency and energy consumption of the terminal's task execution.

[0012] Furthermore, the decision to offload the local computation includes when the cost of offloading the local computation of the sensory data collected by the terminal is less than the cost of offloading the computation in the base station or the cloud, that is, when ,and , choose not to uninstall and perform calculations locally, , indicating the The weighted sum of the energy consumption of task delay is The terminal selection is in The data is offloaded to the local computer for calculation in each time slot. Indicates that when the terminal Choose to offload the calculation locally. The total energy consumed by the terminal in the process of completing the perception offloading and calculation. It means that the terminal chooses to offload the calculation locally. It means that the weighted sum of the task time cost and energy cost is kind, It means that when the terminal When choosing to offload calculations at the base station or cloud, the total energy consumed by the terminal in completing the perception offloading and calculation process, It means that the terminal chooses to offload the calculation at the base station. Indicates that the terminal selected for offloading calculations locally The total time to complete the perception offloading calculation process, It means that the terminal that selects to offload the calculation at the base station The total time to complete the perception offloading calculation process, It means that the terminal that chooses to offload computing in the cloud The total time to complete the perception offloading calculation process, It means that the terminal chooses to offload computing in the cloud. The weight represents the weighted sum of the terminal time cost and energy consumption cost, which changes according to the different requirements of different tasks for the importance of delay and energy consumption.

[0013] Furthermore, the base station offloading calculation includes when the cost of offloading the local calculation of the perception data collected by the terminal is greater than the cost of offloading the calculation in the base station or the cloud, that is, when ,and , and when the cost of offloading calculations at the base station is less than the cost of offloading calculations in the cloud, that is, When , choose to offload the calculation at the base station, let , indicating the The weighted sum of the energy consumption of task delay is The terminal selection is in time slots to offload data to the base station for calculation

[0014] Furthermore, the decision to offload the computation in the cloud includes the following: if the cost of offloading the local computation of the sensory data collected by the terminal is greater than the cost of offloading the computation in the base station or the cloud, that is, when ,and , and when the cost of offloading calculations at the base station is greater than the cost of offloading calculations in the cloud, that is, When , choose to offload the calculation at the base station, let , indicating the The weighted energy consumption of task delay and weight type is The terminal selection is in The data is offloaded to the cloud for computation in each time slot.

[0015] Furthermore, step S2 includes an optimal cost selection scheme represented as a weighted sum of optimal terminal energy consumption and task completion delay, specifically as follows:

[0016] The total delay energy consumption weighted sum of the terminals to complete the sensing task collaboratively :

[0017]

[0018] in, Indicates the execution delay energy consumption weighted and weight type is The terminal group with the task of cooperating to complete the perception offloading and computing energy consumption of the inspection business, Indicates the delay of the terminal with the longest task completion delay in the terminal group. The weight value representing the weighted sum of the terminal time cost and energy consumption cost; , It means that when the terminal chooses to offload the calculation locally, the total energy consumed by the terminal in completing the perception offloading and calculation process is It represents the total energy consumed by the terminal in the process of completing the perception offloading and calculation when the terminal chooses to offload the calculation at the base station or the cloud. , , , respectively represent the terminal's offloading decision variables at the local, base station, and cloud, which are 0-1 variables. When , it means that the terminal is performing calculation offloading locally. When , it means that the terminal is performing calculation offloading at the base station. When , it means that the terminal is offloading computing in the cloud; ,in It represents the total time it takes for the terminal that chooses to perform offload calculation locally to complete the perception offload calculation process, where It represents the total time for the terminal that chooses to perform offload calculation at the base station to complete the perception offload calculation process, where It represents the total time it takes for the terminal that chooses to offload computing in the cloud to complete the perception offload computing process. It is the minimum value among the three, indicating the minimum delay for the terminal group to complete this sensing task;

[0019] The best The modeling of the solution is the following optimization problem:

[0020] ;

[0021] ;

[0022] ;

[0023] ;

[0024] ;

[0025] ;

[0026] ;

[0027] ;

[0028] ;

[0029] ;

[0030] ;

[0031] Among them, the constraints Indicates: The terminal is performing a task The total amount of sensing data collected by a terminal group within the coverage area of ​​a base station must be greater than or equal to the total amount of sensing data collected by a task. The amount of data required, Indicates The total amount of perception data required by the task, Indicates execution of The first terminal group of the class task The amount of sensory data collected by each terminal, constraints Indicates that the terminal If and only if a place is selected locally, at the edge base station, or in the cloud to offload the task;

[0032] Constraints This shows that in each time slot, there are terminals that choose to offload data at the edge base station or the cloud. Indicates The terminal is in time slots for data unloading, the constraint and constraints This shows that no matter whether the terminal is offloading tasks locally, at the base station, or in the cloud, the energy consumption of the entire process of completing the task is less than that of the terminal. The maximum total energy of the ; Constraints It indicates that the data collected during the sensing process of each terminal must be completely unloaded within the allocated unloading time slot, where Indicated in unloading time slots, Indicates the terminal's perception rate of surrounding information. Indicates terminal In the The amount of data unloaded in each time slot is expressed as ,in, Indicates the uplink transmission rate from the terminal to the edge base station.

[0033] Further, step S3 includes:

[0034] The specific steps for transforming the objective function are as follows:

[0035] Since there is a nonlinear term in the objective function This makes the objective function very complex and difficult to handle;

[0036] Then introduce the slack variable , converting nonlinear constraints into linear terms, where , in order to make the solution of the original problem the same, The following constraints should be met:

[0037]

[0038]

[0039]

[0040] Therefore, the question Converts to:

[0041]

[0042]

[0043]

[0044]

[0045]

[0046] An effective linear programming and particle swarm optimization algorithm are set up to obtain the suboptimal solution. The method is carried out in two stages: first, 1) unloading allocation problem: solving the optimization problem of unloading locations after given task unloading slots and unloading order, 2) resource allocation problem: solving the unloading slot allocation problem after given unloading locations under the optimal solution of unloading allocation solution;

[0047] According to the above two stages, the problem It is divided into two sub-problems:

[0048] Sub-questions ,Given the perception offloading time slot allocation, the particle swarm optimization algorithm is ,adopted to solve the offloading allocation problem;

[0049]

[0050]

[0051]

[0052]

[0053]

[0054]

[0055]

[0056]

[0057]

[0058]

[0059]

[0060]

[0061] Sub-questions , given the allocation of sensing offloading locations, an optimization method is used to solve the sensing offloading time slot allocation problem;

[0062]

[0063]

[0064]

[0065]

[0066]

[0067] ;

[0068] The above two sub-problems are iteratively solved by linear programming and particle swarm method, where the specific steps of linear programming and particle swarm method are as follows:

[0069] S31, randomly initialize the unloading time slot of the terminal, and use the particle swarm algorithm to solve the optimal unloading position problem of the terminal given the sensing time slot and unloading time slot of the terminal , when the cost of offloading the local computing of the perception data collected by the initialization terminal is less than the cost of offloading the computing in the base station or the cloud, choose to offload the computing locally, and let , when the cost of offloading the local computing of the sensory data collected by the terminal is greater than the cost of offloading the computing at the base station and the cloud, and when the cost of offloading the computing at the base station is less than the cost of offloading the computing at the cloud, choose to offload the computing at the base station, and let , when the cost of offloading the local computing of the sensory data collected by the terminal is greater than the cost of offloading the computing in the base station and the cloud, and when the cost of offloading the computing in the base station is greater than the cost of offloading the computing in the cloud, choose to offload the computing in the cloud, let ; Among them, the specific process of the particle swarm algorithm is as follows:

[0070] Initialization: Let , the maximum number of iterations is K, set create particles, each particle has a dimension of { Total 3 dimension, is the number of terminals; the position and velocity information of the particle are randomly determined by random numbers, and the optimal position of the particle is defined as p , the optimal position of the particle swarm is ,in ;

[0071] Check and adjust the position of the particles. When the conditions are met:

[0072] (a) Yes

[0073] Calculate the current cost function ,if Update the optimal position of the particle and the optimal position of the particle swarm

[0074]

[0075] (b) Yes

[0076] Update particle velocity: ;

[0077] Update particle positions: ;

[0078] in, is the inertia weight, and is the learning weight, vector and Assigned by uniform random distribution between -1 and 1;

[0079] (c) If the global optimal solution satisfies all constraints in P2.1, update ;

[0080] (d)

[0081] (e) Go to step (a);

[0082] Returns the best position of the particle swarm and minimum cost ;

[0083] S32. Select the unloading location with the lowest cost for each terminal to perform the task unloading and obtain the unloading arrangement;

[0084] S33. Bring the lowest cost unloading location into the unloading time slot to solve the sub-problem In this paper, the optimal perception and unloading time slot arrangement is solved by optimization. While ensuring that the total amount of perception data is sufficient to complete the task, all the perceived perception data can be unloaded under TDMA and the cost of the terminal group can be minimized. is a linear programming problem, then there must be an optimal solution at an extreme point. According to the definition of extreme point, Everywhere linearly independent activity constraints; then, by making the perception data constraint and the car energy consumption constraint active, we can get the closed-form optimal solution ; Its expression is as follows:

[0085] ,

[0086]

[0087] ,

[0088] in,

[0089] The optimal unloading time slot arrangement is obtained. After the intelligent terminals collect the perception data of the task collaboratively, they will use TDMA to unload the perception data. When one terminal is unloading, the other terminals will continue to collect perception data in parallel. Therefore, the perception time slot of the current terminal is the sum of the unloading time slots of all terminals that have unloaded tasks before. Therefore, the terminal The perception time for the perception task is: , which leads to each terminal The sensing time slot arrangement;

[0090] S34. After solving the optimal perception and unloading time slot arrangement, the current optimal perception and unloading time slot arrangement is brought into the particle swarm algorithm to continue solving the optimal unloading position, and two sub-problems are iterated until the cost function converges to obtain the optimal perception unloading time slot allocation and unloading position.

[0091] On the other hand, the present invention further discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the above method.

[0092] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0093] It can be seen from the above technical solutions that the cloud-edge-terminal collaborative unloading scheduling and resource scheduling method based on telepathy of the present invention realizes an optimization method for task unloading and resource allocation based on telepathy integration and TDMA mode. Through the collaboration between terminals, the cloud-edge-terminal unloading and computing are combined to minimize the cost and energy consumption of the terminal group to complete the total task while meeting the energy consumption constraint. A two-step method is proposed to decompose the optimization goal into resource allocation and task unloading. It can effectively save the cost of multi-terminal collaborative execution of tasks under different unloading conditions.

[0094] Specifically, in the present invention, unlike the usual edge computing problems, the default data is generated at the terminal, and the perception stage of the intelligent terminal sensing data from the environment to generate tasks when performing business is jointly considered, thereby better considering the impact of the perception and acquisition stage of data on the overall unloading decision, delay and energy consumption, and more reasonably arranging the resource allocation of perception, communication, and unloading to achieve overall high operating efficiency. At the same time, the edge computing technology is combined with traditional cloud computing, and the cloud-edge joint unloading and computing are adopted to realize local computing services for terminal users, which can better meet the accuracy and timeliness requirements of diversified network tasks. In addition, the TDMA unloading mode is adopted to coordinate the scheduling of joint perception and unloading between multiple terminals, so as to achieve efficient use of limited spectrum resources. It can achieve the maximization of multi-terminal system performance under different unloading conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] Figure 1 It is a flowchart of the present invention;

[0096] Figure 2 4 is a cost comparison chart of different unloading methods according to an embodiment of the present invention. DETAILED DESCRIPTION

[0097] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0098] like Figure 1 As shown, the cloud-edge collaborative unloading scheduling and resource scheduling method based on synergistic computing described in this embodiment includes the following steps: S1. Scenario description; S2. Model establishment; S3. Model solving.

[0099] In step S1, the scenario description refers to dividing the coverage area of ​​a cloud into multiple sub-areas according to different base station coverage areas, and selecting one of the three offloading modes for each terminal to offload the perception data.

[0100] In step S2, model building refers to formulating the sum of the energy consumption of the terminal and the delay of task completion of the three TDMA-based offloading methods, which is defined as the sum of the costs of task completion, and based on this, modeling it into a task cost optimization problem according to different offloading methods and the time slot allocation of perception and offloading.

[0101] In step S3, model solving refers to solving the optimization problem using low-complexity linear programming and particle swarm methods to determine the unloading method of each terminal and the TDMA unloading time slot of each terminal. Ultimately, the purpose of minimizing the cost of multiple terminals completing inspection tasks is achieved through cloud-edge-end collaborative unloading and scheduling optimization of perceived unloading time slots.

[0102] The following are specific instructions:

[0103] The present invention discloses a cloud-edge-end collaborative unloading scheduling and resource scheduling method based on synergistic computing, comprising a scenario description module, a model building module and a model solving module.

[0104] The scenario description module is mainly used to divide the area within the cloud coverage into sub-areas. Within the coverage of a cloud, the cloud range is divided into sub-areas covered by multiple base stations according to the coverage of multiple base stations. In each sub-area covered by a base station, multiple smart terminals collaborate to perform a task. After the smart terminals collaboratively collect the perceived data volume of the task, the perceived data will be unloaded using the TDMA method. When one terminal is unloading, the other terminals will continue to collect perceived data in parallel until the unloading scheduler arranges the terminal to unload data. There are three types of unloading locations: local unloading calculation, base station unloading calculation, and cloud unloading calculation.

[0105] Among them, local offload computing refers to the perception data collected by the terminal being directly calculated locally at the terminal; base station offload computing refers to the perception data collected by the terminal being offloaded to the base station for calculation without occupying the terminal's computing resources; cloud offload computing refers to the perception data collected by the terminal being first transmitted to the base station, and then transmitted to the cloud by the base station, using the computing resources of the cloud to process and calculate the task.

[0106] Among them, the selection of the three offloading methods needs to comprehensively consider the weighted sum of the latency and energy consumption of the terminal's task execution.

[0107] The decision to offload local computing is when the cost of offloading the local computing of the sensory data collected by the terminal is less than the cost of offloading the computing in the base station or the cloud. ,and , choose not to uninstall and perform calculations locally, , indicating the The weighted sum of the energy consumption of task delay is The terminal selection is in time slots to offload data to local for computation, where Indicates that when the terminal Choose to offload the calculation locally. The total energy consumed by the terminal in the process of completing the perception offloading and calculation. It means that the terminal chooses to offload the calculation locally. It means that the weighted sum of the task time cost and energy cost is Class, where It means that when the terminal When choosing to offload calculations at the base station or cloud, the total energy consumed by the terminal in completing the perception offloading and calculation process, where: It means that the terminal chooses to offload the calculation at the base station. Indicates that the terminal selected for offloading calculations locally The total time to complete the perception offloading calculation process, It means that the terminal that selects to offload the calculation at the base station The total time to complete the perception offloading calculation process, It means that the terminal that chooses to offload computing in the cloud The total time to complete the perception offloading calculation process, where: It means that the terminal chooses to offload computing in the cloud. The weight represents the weighted sum of the terminal time cost and energy consumption cost, which changes according to the different requirements of different tasks for the importance of delay and energy consumption.

[0108] The decision of base station offloading calculation refers to: if the cost of offloading the local calculation of the perception data collected by the terminal is greater than the cost of offloading the calculation at the base station or in the cloud, that is, when ,and , and when the cost of offloading calculations at the base station is less than the cost of offloading calculations in the cloud, that is, When , choose to offload the calculation at the base station, let , indicating the The weighted sum of the energy consumption of task delay is The terminal selection is in The data is offloaded to the base station for calculation in each time slot.

[0109] The decision to offload computing in the cloud means that if the cost of offloading computing the sensory data collected by the terminal locally is greater than the cost of offloading computing in the base station or the cloud, that is, when ,and , and when the cost of offloading calculations at the base station is greater than the cost of offloading calculations in the cloud, that is, When , choose to offload the calculation at the base station, let ; indicates the The weighted energy consumption of task delay and weight type is The terminal selection is in The data is offloaded to the cloud for computation in each time slot.

[0110] The different unloading locations of the terminal affect the minimum cost at which the terminal ultimately completes the task.

[0111] At the same time, since the present invention jointly considers the perception, offloading and calculation processes, and adopts the perception-then-offloading process, after the intelligent terminal collaboratively collects the perception data volume of the task, the perception data will be offloaded in a TDMA manner. In the case of TDMA, the time length of each time slot is recorded as , the unloading time slot is recorded as ,exist In the time slot, Each terminal senses and unloads data respectively. When one terminal is unloading, the other terminals will continue to collect sensed data in parallel. That is, the sensing time slot of the current terminal is the sum of the unloading time slots of all terminals that have unloaded tasks before. Therefore, the sensing time slot of the terminal is the sum of the unloading time slots of all terminals that have unloaded tasks before. The perception time for the perception task is: Therefore, multiple terminals need to coordinate and consider the arrangement and optimization of their own perception time slots and unloading time slots to achieve the minimum task cost of perception and communication integration.

[0112] The model building module mainly formulates the task cost of each terminal based on TDMA task offloading according to different perception data offloading locations and the time slot allocation of perception and communication integration, and then models it into a task cost optimization problem according to different offloading methods and the time slot allocation of perception and offloading.

[0113] Among them, the optimal cost selection scheme is expressed as the weighted sum of the optimal terminal energy consumption and task completion delay, as follows:

[0114] The total delay energy consumption weighted sum of the terminals to complete the sensing task collaboratively :

[0115]

[0116] in, Indicates the execution delay energy consumption weighted and weight type is The terminal group with the task of cooperating to complete the perception offloading and computing energy consumption of the inspection business, Indicates the delay of the terminal with the longest task completion delay in the terminal group. Represents the weight of the weighted sum of the terminal time cost and energy consumption cost. ,in It represents the total energy consumed by the terminal in the process of completing the perception offloading and calculation when the terminal chooses to offload the calculation locally. It represents the total energy consumed by the terminal in the process of completing the perception offloading and calculation when the terminal chooses to offload the calculation at the base station or the cloud. , , , respectively represent the terminal's offloading decision variables at the local, base station, and cloud, which are 0-1 variables. When , it means that the terminal is performing calculation offloading locally. When , it means that the terminal is performing calculation offloading at the base station. , it means that the terminal is offloading computing in the cloud. ,in It represents the total time it takes for the terminal that chooses to perform offload calculation locally to complete the perception offload calculation process, where It represents the total time for the terminal that chooses to perform offload calculation at the base station to complete the perception offload calculation process, where It represents the total time it takes for the terminal that chooses to offload computing in the cloud to complete the perception offload computing process. It is the maximum value among the three, indicating the minimum delay for the terminal group to complete this sensing task.

[0117] Optimal The modeling of the solution is the following optimization problem:

[0118]

[0119]

[0120]

[0121]

[0122]

[0123]

[0124]

[0125]

[0126]

[0127]

[0128]

[0129] Among them, the constraints Indicates: The terminal is performing a task The total amount of sensing data collected by a terminal group within the coverage area of ​​a base station must be greater than or equal to the total amount of sensing data collected by a task. The amount of data required, Indicates The total amount of perception data required by the task, Indicates execution of The first terminal group of the class task The amount of sensory data collected by each terminal, constraints Indicates that the terminal If and only if a place is selected for task offloading locally, at the edge base station, or in the cloud, the constraint It shows that in each time slot, there are terminals that choose to offload data at the edge base station or the cloud. Indicates The terminal is in time slots for data unloading, the constraint and constraints This shows that no matter whether the terminal is offloading tasks locally, at the base station, or in the cloud, the energy consumption of the entire process of completing the task is less than that of the terminal. The maximum total energy of the ; Constraints It indicates that the data collected during the sensing process of each terminal must be completely unloaded within the allocated unloading time slot, where Indicated in unloading time slots, Indicates the terminal's perception rate of surrounding information. Indicates terminal In the The amount of data unloaded in each time slot is expressed as ,in, Indicates the uplink transmission rate from the terminal to the edge base station.

[0130] According to the above model, an embodiment of the present invention describes a cost-minimizing resource-optimizing offloading method for multi-terminal collaborative tasks.

[0131] The model solving module mainly designs methods to solve the model established by the model building module. According to the nonlinear mixed integer characteristics of the model, slack variables are introduced to transform the objective function, and then low-complexity linear programming and particle swarm methods are designed to solve the optimization problem, determine the unloading method of each terminal, and determine the TDMA unloading time slot of each terminal. Finally, the purpose of minimizing the cost of multiple terminals to complete inspection tasks is achieved through cloud-edge-end collaborative unloading and scheduling optimization of perception unloading time slots.

[0132] Among them, the specific steps of transforming the objective function are as follows:

[0133] Since there is a nonlinear term in the objective function This makes the objective function very complex and difficult to handle. Therefore, the present invention introduces a slack variable , transforming this nonlinear constraint into a linear term, where , in order to make the solution of the original problem the same, The following constraints should be met:

[0134]

[0135]

[0136]

[0137] Therefore, the question can be converted into:

[0138]

[0139]

[0140]

[0141]

[0142]

[0143] Due to the problem It is a difficult mixed integer linear programming problem. To this end, this paper develops an effective low-complexity linear programming and particle swarm algorithm to obtain a suboptimal solution. The method is divided into two stages: first, 1) unloading allocation problem: solving the optimization problem of the unloading position after the given task unloading time slot and unloading order, 2) resource allocation problem: under the optimal solution of the unloading allocation solution, solving the unloading time slot allocation problem after the given unloading position. According to the above two stages, the problem It is divided into two sub-problems:

[0144] Sub-questions ,Given the perception offloading time slot allocation, the particle swarm optimization algorithm is ,adopted to solve the offloading allocation problem;

[0145]

[0146]

[0147]

[0148]

[0149]

[0150]

[0151]

[0152]

[0153]

[0154]

[0155]

[0156] Sub-questions , given the allocation of sensing offloading locations, an optimization method is used to solve the sensing offloading time slot allocation problem;

[0157]

[0158]

[0159]

[0160]

[0161]

[0162]

[0163] The two sub-problems are solved iteratively using low-complexity linear programming and particle swarm methods. Figure 1 As shown, the specific steps are as follows:

[0164] S1. Randomly initialize the unloading time slot of the terminal. Given the sensing time slot and unloading time slot of the terminal, use the particle swarm algorithm to solve the optimal unloading position problem of the terminal. , when the cost of offloading the local computing of the perception data collected by the initialization terminal is less than the cost of offloading the computing in the base station or the cloud, choose to offload the computing locally, and let , when the cost of offloading the local computing of the sensory data collected by the terminal is greater than the cost of offloading the computing at the base station and the cloud, and when the cost of offloading the computing at the base station is less than the cost of offloading the computing at the cloud, choose to offload the computing at the base station, and let , when the cost of offloading the local computing of the sensory data collected by the terminal is greater than the cost of offloading the computing in the base station and the cloud, and when the cost of offloading the computing in the base station is greater than the cost of offloading the computing in the cloud, choose to offload the computing in the cloud, let ; Among them, the specific process of the particle swarm algorithm is as follows:

[0165] 1. Initialization: Let , the maximum number of iterations is K, set create particles, each particle has a dimension of { Total 3 dimension, is the number of terminals; the position and velocity information of the particle are randomly determined by random numbers, and the optimal position of the particle is defined as p , the optimal position of the particle swarm is ,in ;

[0166] 2. Check and adjust the position of the particles. When the conditions are met:

[0167] (a) Yes

[0168] Calculate the current cost function ,if Update the optimal position of the particle and the optimal position of the particle swarm

[0169]

[0170] (b) Yes

[0171] Update particle velocity:

[0172] Update particle positions:

[0173] in, is the inertia weight, and is the learning weight, vector and Assigned by a uniform random distribution between -1 and 1

[0174] (c) If the global optimal solution satisfies all constraints in P2.1, update

[0175] (d)

[0176] (e) Go to step (a)

[0177] 3. Return to the optimal position of the particle swarm and minimum cost

[0178] S2. Select the unloading location with the lowest cost for each terminal to unload the task and obtain the unloading schedule;

[0179] S3. Bring the lowest cost unloading location into the unloading time slot to solve the sub-problem In this paper, the optimal perception and unloading time slot arrangement is solved by optimization. While ensuring that the total amount of perception data is sufficient to complete the task, all the perceived perception data can be unloaded under TDMA and the cost of the terminal group can be minimized. is a linear programming problem, then there must be an optimal solution at an extreme point. According to the definition of extreme point, Everywhere Then, by making the perception data constraint and the car energy consumption constraint active, we can get the closed-form optimal solution . Its expression is as follows:

[0180] ,

[0181]

[0182] ,

[0183] in,

[0184] The optimal unloading time slot arrangement is obtained. After the intelligent terminals collect the perception data of the task collaboratively, they will use TDMA to unload the perception data. When one terminal is unloading, the other terminals will continue to collect perception data in parallel. Therefore, the perception time slot of the current terminal is the sum of the unloading time slots of all terminals that have unloaded tasks before. Therefore, the terminal The perception time for the perception task is: , which leads to each terminal The perception time slot arrangement.

[0185] S4. After solving the optimal sensing and unloading time slot arrangement, the current optimal sensing and unloading time slot arrangement is brought into the particle swarm algorithm to continue solving the optimal unloading position, and two sub-problems are iterated until the cost function converges to obtain the optimal sensing unloading time slot allocation and unloading position.

[0186] like Figure 2 As shown, the optimization method implemented by the present invention, when compared with other different offloading methods, namely, all local offloading calculations, all base station offloading calculations, and all cloud offloading calculations, the collaborative offloading and resource allocation method of the present invention can achieve the minimum task cost when collaborative perception tasks are performed by different numbers of terminals, proving the effectiveness of the method of the present invention in minimizing costs.

[0187] On the other hand, the present invention further discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the above method.

[0188] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0189] In another embodiment provided in the present application, a computer program product comprising instructions is also provided. When the computer is run on a computer, the computer executes any of the cloud-edge collaborative unloading scheduling and resource scheduling methods based on synaesthesia in the above embodiments.

[0190] It is understandable that the system, device and storage medium provided in the embodiments of the present invention correspond to the method provided in the embodiments of the present invention, and the explanation, examples and beneficial effects of the relevant contents can refer to the corresponding parts in the above methods.

[0191] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.

[0192] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0193] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0194] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A cloud-edge-device collaborative offloading scheduling and resource scheduling method based on synergistic computing, characterized in that: The following steps are involved: S1, scenario description; S2, model building; S3, model solution; In step S1, the scenario description includes dividing a cloud coverage area into multiple sub-areas according to different base station coverage areas, and selecting one of three offloading modes for each terminal to offload the sensing data; In step S2, the model building includes formulating the sum of the energy consumption of the terminal and the delay of task completion of the three TDMA-based offloading methods, which is defined as the sum of the cost of task completion, and accordingly modeling it into a task cost optimization problem according to different offloading methods and the time slot allocation of sensing and offloading; In step S3, the model solution includes solving the optimization problem using low-complexity linear programming and particle swarm methods to determine the unloading method of each terminal and the TDMA unloading time slot of each terminal; ultimately, the purpose of minimizing the cost of multiple terminals completing inspection tasks is achieved through cloud-edge-terminal collaborative unloading and scheduling optimization of perceived unloading time slots; There are three types of offloading methods in step S1, namely, offloading calculation locally, offloading calculation at the base station, and offloading calculation at the cloud; Among them, local offload computing refers to the perception data collected by the terminal being directly calculated locally at the terminal, base station offload computing refers to the perception data collected by the terminal being offloaded to the base station for calculation without occupying the computing resources of the terminal, and cloud offload computing refers to the perception data collected by the terminal being first transmitted to the base station, and then transmitted to the cloud by the base station, using the computing resources of the cloud to process and calculate the task; The selection of the three offloading methods needs to comprehensively consider the weighted sum of the latency and energy consumption of the terminal’s task execution; Step S2 includes the optimal cost selection scheme represented as the weighted sum of the optimal terminal energy consumption and the task completion delay, which is as follows: The total delay energy consumption weighted sum of the terminals to complete the sensing task collaboratively : ,in, Indicates the execution delay energy consumption weighted and weight type is The terminal group with the task of cooperating to complete the perception offloading and computing energy consumption of the inspection business, Indicates the delay of the terminal with the longest task completion delay in the terminal group. The weight value representing the weighted sum of the terminal time cost and energy consumption cost; , It means that when the terminal chooses to offload the calculation locally, the total energy consumed by the terminal in completing the perception offloading and calculation process is It represents the total energy consumed by the terminal in the process of completing the perception offloading and calculation when the terminal chooses to offload the calculation at the base station or the cloud. , , , respectively represent the terminal's offloading decision variables at the local, base station, and cloud, which are 0-1 variables. When , it means that the terminal is performing calculation offloading locally. When , it means that the terminal is performing calculation offloading at the base station. When , it means that the terminal is offloading computing in the cloud; ,in It represents the total time it takes for the terminal that chooses to perform offload calculation locally to complete the perception offload calculation process, where It represents the total time for the terminal that chooses to perform offload calculation at the base station to complete the perception offload calculation process, where It represents the total time it takes for the terminal that chooses to offload computing in the cloud to complete the perception offload computing process. It is the minimum value among the three, indicating the minimum delay for the terminal group to complete this sensing task; The best The modeling of the solution is the following optimization problem : ； ； ； ； ； ； ； ； ； ； ； Among them, the constraints Indicates: The terminal is performing a task The total amount of sensing data collected by a terminal group within the coverage area of ​​a base station must be greater than or equal to the total amount of sensing data collected by a task. The amount of data required, Indicates The total amount of perception data required by the task, Indicates execution of The first terminal group of the class task The amount of sensory data collected by each terminal, constraints Indicates that the terminal If and only if a place is selected locally, at the edge base station, or in the cloud to offload the task; Constraints This shows that in each time slot, there are terminals that choose to offload data at the edge base station or the cloud. Indicates The terminal is in time slots for data unloading, the constraint and constraints This shows that no matter whether the terminal is offloading tasks locally, at the base station, or in the cloud, the energy consumption of the entire process of completing the task is less than that of the terminal. The maximum total energy of the ; Constraints It indicates that the data collected during the sensing process of each terminal must be completely unloaded within the allocated unloading time slot, where Indicated in unloading time slots, Indicates the terminal's perception rate of surrounding information. Indicates terminal In the The amount of data unloaded in each time slot is expressed as ,in, Indicates the uplink transmission rate from the terminal to the edge base station; Step S3 includes: right The specific steps of the conversion are as follows: because There is a nonlinear term in Make Very complex and difficult to handle; then introduce slack variables , converting nonlinear constraints into linear terms, where , in order to make the solution of the original problem the same, The following constraints should be met: , , , so the optimization problem Converts to: , , , An effective linear programming and particle swarm algorithm are set up to obtain a suboptimal solution. The effective linear programming and particle swarm algorithm are performed in two stages: first, 1) unloading allocation problem: solving the optimization problem of the unloading position after a given task unloading time slot and unloading order, 2) resource allocation problem: solving the unloading time slot allocation problem after a given unloading position under the optimal solution of the unloading allocation solution; According to the above two stages, the problem It is divided into two sub-problems: Sub-questions ,Given the perception offloading time slot allocation, the particle swarm optimization algorithm is ,adopted to solve the offloading allocation problem; , , , , , , , , , , , sub-problem , given the allocation of sensing offloading locations, an optimization method is used to solve the sensing offloading time slot allocation problem; ， ， ， ， ， ； The above two sub-problems are iteratively solved by linear programming and particle swarm method, where the specific steps of linear programming and particle swarm method are as follows: S31, randomly initialize the unloading time slot of the terminal, and use the particle swarm algorithm to solve the optimal unloading position problem of the terminal given the sensing time slot and unloading time slot of the terminal , when the cost of offloading the local computing of the perception data collected by the initialization terminal is less than the cost of offloading the computing in the base station or the cloud, choose to offload the computing locally, and let , when the cost of offloading the local computing of the sensory data collected by the terminal is greater than the cost of offloading the computing at the base station and the cloud, and when the cost of offloading the computing at the base station is less than the cost of offloading the computing at the cloud, choose to offload the computing at the base station, and let , when the cost of offloading the local computing of the sensory data collected by the terminal is greater than the cost of offloading the computing in the base station and the cloud, and when the cost of offloading the computing in the base station is greater than the cost of offloading the computing in the cloud, choose to offload the computing in the cloud, let ; Among them, the specific process of the particle swarm algorithm is as follows: Initialization: Let , the maximum number of iterations is K, set create particles, each particle has a dimension of { Total 3 dimension, is the number of terminals; the position and velocity information of the particle are randomly determined by random numbers, and the optimal position of the particle is defined as p , the optimal position of the particle swarm is ,in ; Check and adjust the particle position. When the conditions are met: (a) Yes Calculate the current cost function ,if Update the optimal position of the particle and the optimal position of the particle swarm , , (b) yes Update particle velocity: ； Update particle positions: ; in, is the inertia weight, and is the learning weight, vector and Assigned by uniform random distribution between -1 and 1; (c) If the global optimal solution satisfies all constraints in P2.1, update ; (d) , (e) go to step (a); Returns the best position of the particle swarm and minimum cost ; S32. Select the unloading location with the lowest cost for each terminal to perform the task unloading and obtain the unloading arrangement; S33. Bring the lowest cost unloading location into the unloading time slot to solve the sub-problem In the method, the optimal perception and unloading time slot arrangement is solved by optimization. While ensuring that the total amount of perception data that completes the task meets the preset requirements, all the perceived perception data can be unloaded under TDMA and the cost of the terminal group can be minimized. The closed-form optimal solution can be obtained by making the perception data constraint and the terminal energy consumption constraint active. Its expression is as follows: , , , ,in, , from which the optimal unloading time slot arrangement is obtained. After the intelligent terminals collect the perception data of the task collaboratively, they will use TDMA to unload the perception data. When one terminal is unloading, the other terminals will continue to collect perception data in parallel. Therefore, the perception time slot of the current terminal is the sum of the unloading time slots of all terminals that have unloaded tasks before. Therefore, the terminal The perception time for the perception task is: , which leads to each terminal The sensing time slot arrangement; S34. After solving the optimal perception and unloading time slot arrangement, the current optimal perception and unloading time slot arrangement is brought into the particle swarm algorithm to continue solving the optimal unloading position, and two sub-problems are iterated until the cost function converges to obtain the optimal perception unloading time slot allocation and unloading position.

2. The cloud-edge-device collaborative offloading scheduling and resource scheduling method based on synergistic computing according to claim 1 is characterized by: The decision of local offloading calculation includes when the cost of offloading calculation of the sensory data collected by the terminal locally is less than the cost of offloading calculation in the base station or cloud. , and , choose not to uninstall and perform calculations locally, , indicating the The weighted sum of the energy consumption of task delay is The terminal selection is in The data is offloaded to the local computer for calculation in each time slot. Indicates that when the terminal Choose to offload the calculation locally. The total energy consumed by the terminal in the process of completing the perception offloading and calculation. It means that the terminal chooses to offload the calculation locally. It means that the weighted sum of the task time cost and energy cost is kind, It means that when the terminal When choosing to offload calculations at the base station or cloud, the total energy consumed by the terminal in completing the perception offloading and calculation process, It means that the terminal chooses to offload the calculation at the base station. Indicates that the terminal selected for offloading calculations locally The total time to complete the perception offloading calculation process, It means that the terminal that selects to offload the calculation at the base station The total time to complete the perception offloading calculation process, It means that the terminal that chooses to offload computing in the cloud The total time to complete the perception offloading calculation process, It means that the terminal chooses to offload computing in the cloud. The weight represents the weighted sum of the terminal time cost and energy consumption cost, which changes according to the different requirements of different tasks for the importance of delay and energy consumption.

3. The cloud-edge-device collaborative offloading scheduling and resource scheduling method based on synergistic computing according to claim 1 is characterized by: The base station offloading calculation includes when the cost of offloading the local calculation of the perception data collected by the terminal is greater than the cost of offloading the calculation at the base station or in the cloud. , and , and when the cost of offloading calculations at the base station is less than the cost of offloading calculations in the cloud, that is, When , choose to offload the calculation at the base station, let , indicating the The weighted sum of the energy consumption of task delay is The terminal selection is in The data is offloaded to the base station for calculation in each time slot.

4. The cloud-edge-device collaborative offloading scheduling and resource scheduling method based on synergistic computing according to claim 1 is characterized by: The decision to offload computing in the cloud includes the following: when the cost of offloading computing locally for the sensory data collected by the terminal is greater than the cost of offloading computing in the base station or the cloud, that is, when ,and , and when the cost of offloading calculations at the base station is greater than the cost of offloading calculations in the cloud, that is, When choosing to offload computing in the cloud, , indicating the The weighted energy consumption of task delay and weight type is The terminal selection is in The data is offloaded to the cloud for computation in each time slot.

5. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method according to any one of claims 1 to 4.

Images