Edge network real-time imaging task allocation method, task allocation system and imaging system

Abstract

The invention discloses an edge network real-time imaging task allocation method, a task allocation system and an imaging system, and belongs to the technical field of edge computing task scheduling. Comprising the steps of obtaining a task and subtasks in the task; sub-tasks are divided according to a dependency relationship, the dependency relationship comprises time dependency and score dependency, and an optimal task allocation scheme is obtained by taking minimization of total completion time of real-time imaging as a target under the condition that the dependency relationship condition is met; and executing the optimal task allocation scheme, predicting the completion time of the sensor nodes on the subtasks, and dynamically scheduling the sensor nodes to which the subtasks are allocated according to the completion time. The problems of high data transmission delay and low resource utilization rate when imaging processing is carried out on a central server in a traditional seismic exploration imaging method are solved, sub-tasks are allocated, efficient utilization of resources is realized, and the real-time performance of imaging is improved.

Description

Technical Field

[0001] The present application relates to the field of edge computing task scheduling technology, specifically to an edge network real-time imaging task allocation method, a task allocation system, and an imaging system. Background Art

[0002] In the field of seismic exploration, real-time imaging technology is a hot research topic. Traditional seismic data processing methods are typically performed on central servers, resulting in high data transmission latency, slow processing speeds, and low resource utilization. With the development of edge computing technology, this technology is gradually being used in seismic exploration.

[0003] Edge computing reduces data transmission latency and improves system response speed by performing processing close to the data source. However, in seismic exploration, real-time imaging tasks involve multiple subtasks, such as data preprocessing, autopower spectrum calculation, cross-correlation coefficient calculation, and Bessel function fitting. The complex dependencies between these subtasks complicate task scheduling and allocation.

[0004] Existing task allocation methods are primarily based on traditional algorithms, such as genetic algorithms and particle swarm optimization. While these methods address the task allocation problem to some extent, their relatively simple iterative processes are prone to falling into local optimal solutions, making them unsuitable for complex real-time imaging tasks. While improving traditional algorithms or employing methods such as deep reinforcement learning for task allocation have improved task allocation to some extent, these algorithms fail to consider the dependencies between the numerous subtasks of real-time imaging, limiting their practical application. Summary of the Invention

[0005] The embodiments of the present application provide a method for allocating real-time imaging tasks in an edge network, a task allocation system, and an imaging system, which solve the problems of high data transmission delay and low resource utilization in traditional seismic exploration imaging methods that perform imaging processing on a central server.

[0006] The technical solutions adopted in this application are as follows: According to a first aspect of an embodiment of the present application, a method for allocating real-time imaging tasks in an edge network is provided, comprising: Get tasks and subtasks in tasks; Divide subtasks according to dependency relationships, including time dependency and score dependency. The time dependency indicates that the output of the most recently completed subtask serves as the input of the next subtask, and the score dependency distinguishes whether subtasks are from the same task. Under the condition of satisfying the dependency relationship, an optimal task allocation scheme is obtained with the goal of minimizing the total completion time of real-time imaging, and the optimal task allocation scheme is used to determine the sensor nodes to which the subtasks are allocated; Execute the best task allocation scheme, predict the completion time of sensor nodes for subtasks, and dynamically schedule the sensor nodes to which subtasks are assigned based on the completion time.

[0007] Furthermore, under the dependency conditions, the optimal task allocation scheme is obtained with the goal of minimizing the total completion time of real-time imaging, including: An objective function is established with the goal of minimizing the total completion time of real-time imaging; The objective function is optimized with computing resource limitations as constraints. The computing resource limitations include the computing resources of the edge server, the number of sensor nodes managed by the edge server and the limited computing resources, as well as the fact that a subtask can only be assigned to one sensor node and a sensor node can only process one subtask at a time.

[0008] Furthermore, optimizing the objective function includes: solving the objective function using an improved mutation particle swarm optimization algorithm, and dynamically adjusting the inertia weight during the solution process to balance the transmission delay between edge servers and the computing resources of the edge servers.

[0009] Furthermore, dependency conditions include: Assign subtasks with score dependencies among sensor nodes managed by the same edge server; For subtasks with time dependency, when the remaining computing resources do not meet the computing resources of the subtask, the sensor nodes managed by the edge server with the smallest transmission delay are found for allocation based on the transmission delay between the edge server and other edge servers with remaining computing resources.

[0010] Furthermore, dynamically scheduling the sensor nodes to which the subtasks are assigned according to the completion time includes: arranging the incoming subtasks to the sensor nodes with the shortest completion time or arranging the incoming subtasks to idle sensor nodes.

[0011] According to a second aspect of an embodiment of the present application, a real-time imaging task allocation system for an edge network is provided, which is used in an edge server and includes: Input module, used to obtain tasks and subtasks within tasks; A task division module is used to divide subtasks according to dependency relationships, wherein the dependency relationships include time dependency and score dependency. The time dependency is used to indicate a subtask in which the output of the most recently completed subtask serves as the input of the next task. The score dependency is used to distinguish whether subtasks are from the same task. A task allocation scheme calculation module is used to obtain an optimal task allocation scheme with the goal of minimizing the total completion time of real-time imaging under the condition of satisfying dependency relationships. The optimal task allocation scheme is used to determine the sensor nodes to which the subtasks are allocated; Execution module, used to execute the best task allocation plan; The time prediction module is used to predict the completion time of the subtasks of the sensor nodes in executing the best task allocation scheme; The execution module dynamically schedules the sensor nodes to which the subtasks are assigned according to the completion time.

[0012] Furthermore, the task allocation scheme calculation module is also used to: Establishing an objective function with the goal of minimizing the total completion time of real-time imaging, wherein the objective function includes the start time of the subtask and the transmission delay between the edge servers; The objective function is optimized with computational resource constraints, including the computational resources of the edge server, the number and computational resources of sensor nodes managed by the edge server, and the fact that a subtask can only be assigned to one sensor node and a sensor node can only process one subtask at a time. Among them, optimizing the objective function includes: using an improved mutation particle swarm optimization algorithm to solve the objective function, and dynamically adjusting the inertia weight during the solution process to balance the transmission delay between edge servers and the computing resources of the edge servers.

[0013] Furthermore, the dependency conditions include: Assign subtasks with score dependencies among sensor nodes managed by the same edge server; For subtasks with time dependency, when the remaining computing resources do not meet the computing resources of the subtask, the sensor nodes managed by the edge server with the smallest transmission delay are found for allocation based on the transmission delay between the edge server and other edge servers with remaining computing resources.

[0014] Furthermore, the execution module dynamically schedules the sensor nodes to which the subtasks are assigned according to the completion time, including: arranging the incoming subtasks to the sensor nodes with the shortest completion time or arranging the incoming subtasks to idle sensor nodes.

[0015] According to a third aspect of an embodiment of the present application, an edge network real-time imaging system is provided, comprising: multiple edge servers, and multiple sensor nodes managed by each edge server. The multiple edge servers each include an edge network real-time imaging task allocation system, wherein each edge server manages multiple sensor nodes to perform subtask calculations and interactively allocates subtasks to sensor nodes managed by other edge servers through other edge servers.

[0016] The technical solution provided by the embodiments of the present application brings at least the following beneficial effects: taking into account the dependency relationship between the subtasks of real-time imaging, the subtasks are allocated, and dynamically adjusted by predicting the specific processing time, thereby improving the utilization rate of computing resources, reducing imaging delays, realizing imaging at the edge of the seismic exploration network, and improving the real-time performance of imaging. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic diagram of the structure of an implementation environment involved in each embodiment of the present application; Figure 2 This is a flow chart of a method for allocating real-time imaging tasks in an edge network according to an embodiment of the present application; Figure 3 A flowchart of an optimal task allocation solution for minimizing the total completion time of real-time imaging according to an embodiment of the present application; Figure 4 This is an edge network real-time imaging task allocation system involved in the embodiments of the present application; Figure 5 This is a schematic diagram of the edge network real-time imaging system involved in an embodiment of the present application used in an actual scenario. DETAILED DESCRIPTION

[0018] In order to more clearly understand the above-mentioned objectives, features and advantages of the present application, the present application is described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application.

[0019] See also Figure 1 FIG. 1 is a schematic diagram of a structure of an implementation environment involved in various embodiments of the present application, and the implementation environment includes: Edge server and multiple sensor nodes managed by the edge server.

[0020] An edge server can be a single server or a cluster of multiple servers. Multiple edge servers communicate with each other through a wired network or a wireless network. An edge server can be an electronic device with data processing capabilities, such as a desktop computer, tablet computer, etc.

[0021] An edge server manages multiple sensor nodes, communicating via either wired or wireless networks. Sensor nodes utilize electronic devices with data computing capabilities. Edge servers utilize quad-core 64-bit ARM chips, boasting greater computing power than sensor nodes. These include, but are not limited to, SD cards, Ethernet cards, display units, GPS modules, and WiFi modules. Sensor nodes utilize core processors equipped with Ethernet cards and SD memory cards, enabling data storage and retrieval. These core processors possess sufficient computing power to handle subtasks assigned by the edge server. Sensor nodes include, but are not limited to, high-precision data acquisition units, GPS modules, and WiFi modules.

[0022] The above is only an example of a usage scenario. It is understandable that any scenario related to task allocation can utilize the solution of this application. This embodiment is not limited in any way.

[0023] The embodiments of the present application are particularly outstanding in seismic exploration applications. The real-time imaging task of seismic exploration involves multiple subtasks, such as data preprocessing, autopower spectrum calculation, cross-correlation coefficient calculation, and Bessel function fitting. There are complex dependencies between these subtasks, which makes task scheduling and allocation complicated. The embodiments of the present application divide subtasks into time dependency and score dependency, allocate subtasks with dependencies to sensor nodes, and arrange specific processing time for each subtask by predicting the processing time of the subtasks. The purpose of this task allocation method is to reduce the completion time of real-time imaging tasks, reduce task delays, and improve imaging effects. It solves the problem of real-time imaging task allocation, realizes the effective division and allocation of subtasks, and improves real-time performance and imaging quality.

[0024] See also Figure 2 As shown, in an embodiment of the present application, a flow chart of a method for allocating real-time imaging tasks in an edge network is disclosed. The method for allocating real-time imaging tasks in an edge network includes: S201 obtains the task and subtasks in the task; In the seismic exploration edge network, edge servers are usually deployed near field seismic monitoring sites, connected to seismographs or other sensor nodes, and multiple sensor nodes are connected in the embodiment of the present application. These edge servers are close to the data source and can process and analyze seismic data in real time. By distributing edge servers in different monitoring points and areas, a distributed network is formed. Edge servers can be interconnected in an interactive manner. Any edge server aggregates data collected by multiple sensor nodes and can manage multiple sensor nodes to perform complex task calculations.

[0025] A task can contain at least one subtask. This can be set on an edge server or from another edge server. A subtask can contain only one subtask or contain subtasks.

[0026] S202 divides the subtasks according to dependency relationships, including time dependency and score dependency. The time dependency indicates that the output of the most recently completed subtask serves as the input of the next subtask, and the score dependency indicates whether the subtasks are from the same task. The real-time imaging process is complex, typically consisting of multiple tasks, including data preprocessing, autopower spectrum calculation, cross-correlation coefficient calculation, Bessel function fitting, and shear wave velocity imaging. Each task has multiple subtasks. For example, shear wave velocity imaging requires calculating dispersion curves and inverting them using a global optimization algorithm to obtain the subsurface shear wave velocity structure. Furthermore, in seismic exploration edge networks, each edge server has limited computing resources and can only manage a limited number of sensor nodes simultaneously. Therefore, it is necessary to partition dependencies for real-time imaging.

[0027] Temporal dependencies are defined as connections between subtasks in time and space. For example, consider subtasks A and B. Subtask B depends on the output of subtask A, and subtask A's output serves as an essential parameter in the computation of subtask B. In the processing flow, subtask A must complete before subtask B can be started. This relationship is called a temporal dependency.

[0028] Some subtasks come from the same task, and fractional dependency is used to represent this spatial dependency.

[0029] It should be noted that subtasks can be further divided into multiple subtasks, or they can be inseparable. Temporal and fractional dependencies between subtasks can exist between inseparable subtasks, or between inseparable and divisible subtasks. For divisible subtasks, subtasks must be from the same task and therefore have fractional dependencies.

[0030] In one embodiment, the subtasks are divided according to the dependency relationship based on the Dependency-Aware Task Allocation (DA) algorithm, and the subtasks are divided into time dependency and score dependency: the time dependency is determined by Find the most recently completed subtask and use its output as the input of the next subtask. The time dependency of the subtask whose output is the input of the next subtask is given as 1, where For the The first task subtasks, Indicates the In the second task The output of each subtask; score dependency is used to distinguish whether the subtasks come from the same task. If the subtasks come from the same task, they have score dependency.

[0031] S203, under the dependency condition, obtaining an optimal task allocation scheme with the goal of minimizing the total completion time of real-time imaging, wherein the optimal task allocation scheme is used to determine the sensor nodes to which the subtasks are allocated; Determining the optimal task allocation scheme requires following the principle of dependencies. Based on different dependencies, the sensor nodes to which subtasks should be assigned are determined. Besides following dependencies, determining the optimal task allocation scheme also requires considering the edge server's computing resources, the computing resources required to complete the subtask, the number of subtasks, the number of edge servers, the transmission delay between edge servers, and the computing resources of other edge servers. Sensor nodes on other edge servers may have tasks already processing but excess computing resources, or they may be idle. When considering subtasks assigned to sensor nodes on other edge servers, in addition to computing resources, the transmission delay between edge servers must also be considered.

[0032] The optimal task allocation scheme is to minimize the total completion time of real-time imaging. The task allocation scheme that minimizes the total completion time of real-time imaging is regarded as the optimal task allocation scheme.

[0033] In one embodiment, various factors that influence minimizing the total completion time of real-time imaging are comprehensively considered. An objective function is constructed, and an optimization algorithm is used to solve the objective function with the goal of minimizing the total completion time of real-time imaging. The solution to the objective function that minimizes the total completion time of real-time imaging is the optimal task allocation solution. The total completion time of real-time imaging refers to the time when the output of the last subtask is completed.

[0034] The optimal task allocation scheme is used to determine the sensor node to which the subtask is assigned. This sensor node can be a sensor node in the edge server that sets the task, or a sensor node managed by another edge server with excess computing resources.

[0035] S204 executes the best task allocation plan, predicts the completion time of the sensor nodes for the subtasks, and dynamically schedules the sensor nodes to which the subtasks are allocated according to the completion time.

[0036] The entities implementing the optimal task allocation scheme include one or more sensor nodes in one or more edge servers. This requires determining the processing time for each subtask. Because time-dependent subtasks can only be executed after the output of the previous subtask has been delivered, and subtasks with score dependencies are assigned to sensor nodes managed by the same edge server, even if the output of the previous subtask has arrived, if the sensor node is already reserved for other subtasks, the subtask can only be processed after the reserved subtask has completed, causing delays.

[0037] Therefore, it is necessary to predict the completion time of the subtasks in each sensor node and schedule the upcoming subtasks to the sensor node with the shortest completion time to reduce the delay.

[0038] According to the predicted completion time of the subtask, the sensor node to which the subtask is assigned is dynamically scheduled, that is, the sensor node to which the subtask is originally assigned is changed, and the subtask is reallocated to the sensor node with the shortest completion time.

[0039] For example, the edge server assigns data preprocessing and autopower spectrum calculation to sensor nodes managed by the same edge server. Because these two tasks have a fractional dependency, assigning them to the same edge server negligible latency. The edge server also predicts the completion times of these two subtasks and schedules the next subtask for processing after their completion, thereby reducing latency. Because sensor nodes managed by the same edge server prioritize tasks with fractional dependencies, provided computing resources are available, the cross-correlation coefficient calculation is assigned to the sensor node managed by the edge server with the lowest transmission latency. The completion time of this subtask is also calculated, and the next subtask is scheduled after its completion. Furthermore, since some edge servers still have excess computing resources to execute time-dependent subtasks after scheduling subtasks with fractional dependencies, the impact of transmission latency between different edge servers is mitigated. During the MPSO algorithm solution, inertia weights are dynamically adjusted to balance transmission latency between edge servers and their computing resources. Therefore, inertia weights are set to optimally balance task allocation when multiple imaging systems collaborate, ensuring that real-time imaging always maintains the lowest latency. The edge server has stronger computing power, so it is responsible for task allocation and aggregating the calculation results of sensor nodes, and performing imaging, thereby achieving the purpose of real-time imaging in the edge network.

[0040] See also Figure 3The flowchart of obtaining an optimal task allocation solution with the goal of minimizing the total completion time of real-time imaging is shown. In one embodiment, satisfying dependency conditions and obtaining an optimal task allocation solution with the goal of minimizing the total completion time of real-time imaging includes: S301 establishes an objective function with the goal of minimizing the total completion time of real-time imaging; S302 optimizes the objective function with computing resource limitations as constraints, including: computing resources of the edge server, the number of sensor nodes managed by the edge server and the limited computing resources, as well as the fact that a subtask can only be assigned to one sensor node and a sensor node can only process one subtask at a time.

[0041] The objective function is expressed as: , in, For the last subtask Allotted time, Indicates calculation of The time required for each subtask, represents the transmission delay between edge servers; according to the limited resources of edge servers and sensor nodes, constraints are added to the objective function, including: edge servers Due to its limited computing power, it can only manage a fixed number of sensor nodes. ,in Binary decision variables, representing subtasks Is it Edge Server Processing; a subtask should be assigned to a sensor node, with ; A sensor node can only process one subtask at a time. .

[0042] In one embodiment, the dependency conditions need to satisfy the basic principles that need to be met in determining the optimal task allocation solution, namely: Assign subtasks with score dependencies among sensor nodes managed by the same edge server; For subtasks with time dependency, when the computing resources do not meet the computing resources of the subtask, the sensor nodes managed by the edge server with the smallest transmission delay are found for allocation based on the transmission delay between the edge server and other edge servers with remaining computing resources.

[0043] Specifically, for subtasks with score dependencies, in order to ensure the integrity of the tasks during execution, such subtasks are assigned to the sensor nodes managed by the same edge server for execution. The computing resources of the edge server are the sum of the computing resources of the sensor nodes it manages. The same edge server here can be the edge server that sets the task, or it can be another edge server that interacts with the edge server that sets the task; the remaining computing resources of the edge server that receives the subtask must meet the computing resources required by the subtask. For subtasks with time dependencies, since the output of the previous subtask will be used as the input of the next subtask, Calculate edge server and edge servers The maximum transmission delay between , from which the maximum transmission delay is selected Minimum two edge servers and edge servers To allocate, For edge servers and edge servers The transmission delay between them is usually due to the fact that the edge server that receives the subtasks of other edge servers also has its own tasks to process. The edge server that can receive the subtask is selected by calculating the remaining computing resources. Indicates a task Neutron Mission transmission delay.

[0044] In one embodiment, the optimization algorithm uses the MPSO algorithm (improved mutant particle swarm optimization algorithm) to find the best task allocation solution under the dependency relationship. The specific optimization process includes: (1) Use the MPSO algorithm to find the best allocation solution under the dependency relationship, and build the MPSO algorithm population around the set of maximum completion time Makespan values. In the MPSO algorithm, the maximum completion time Makespan value refers to the total time span from the start of the first subtask to the completion of all subtasks. Each Makespan value is a different particle in the population, and each particle is determined by its speed. and location Characterization. Initialize the particle's velocity and position to ,in Representation interval The uniform distribution within and are the lower and upper bounds of the solution space respectively; (2) Inertia weight in MPSO algorithm It is an important parameter that affects the performance of the algorithm in finding the optimal solution. The definition is made so that the MPSO algorithm can more accurately solve the problem of allocating subtasks with time dependency and score dependency during actual operation. In one embodiment, the definition ,in, It is the computing resource of the edge server Normalized value of transmission delay between the edge server and the edge server The weighted sum of normalized values ​​defines a normalization function To balance the transmission delay of edge servers and computing resources , and Represents variables The maximum and minimum values ​​of ; Inertia weights are set to balance task allocation when multiple real-time imaging systems collaborate to ensure that real-time imaging always maintains the lowest latency. Edge servers, with their greater computing power, are responsible for task allocation, aggregating the computational results of sensor nodes, and performing imaging, thereby achieving real-time imaging at the edge network. (3) Use a mutation strategy that randomly changes the particle position to improve the diversity within the population, prevent it from falling into the local optimum, and encourage particles to continue searching for the best global solution. During the iteration process, a mutation rate mutationRate is introduced for each particle in the population. If the generated random number is less than the mutation rate mutationRate, the position attribute of the particle will be randomly modified. The particle position attribute modified by the mutation strategy is , usually the mutation rate mutationRate is configured to 0.2, where Indicates the The particle in The value of the dimension, Represents a random number; (4) Update all particles in the population, including calculating new fitness values ​​and updating individual optimal positions and the global optimal position , and update the velocity of each particle and location At this time, the fitness value is equal to the current maximum completion time Makespan value. Therefore, the update speed and location The operations are: , in, represents the value of the best position of the individual after update, represents the value of the current individual's best position, Represents the particle position value, represents the global optimal position value after update, represents the current global optimal position value, and Respectively represent the updated particle speed and position values, and Represent the velocity and position values ​​of the particles respectively. represents the inertia weight, and is the learning factor, and is a random number in the range [0,1], providing randomness to explore the solution space; (5) Record the best fitness value and the global best position , and obtain the best task allocation solution.

[0045] In one embodiment, according to the optimal task allocation scheme, an LSTM network (long short-term memory network) is used to determine the processing time of each subtask in the sensor node. After the MPSO algorithm determines the optimal task allocation scheme, the processing time of each subtask needs to be determined. Since subtasks with time dependencies can only be executed after the output of the previous subtask is delivered, and subtasks with score dependencies are assigned to sensor nodes managed by the same edge server. Therefore, although the output of the previous subtask has arrived, if the sensor node has been reserved for other subtasks, the subtask can only be processed after the reserved subtask is completed, causing delays. Furthermore, the LSTM network is used to predict the completion time of the subtasks in each sensor node, and the upcoming subtasks are arranged to the sensor node with the shortest completion time, thereby reducing delays. By Calculate the most recent output arrival time ,in Indicates the The first task Subtasks are assigned to Sensor nodes The processing time required in the output will be the arrival time As input variables, the output arrival times of the previous five times To predict the next output arrival time , and so on, recombining the data and predicting the recombined data again until the iteration ends, and the predicted value of the completion time of each subtask is obtained.

[0046] In one embodiment, Figure 4 As shown, the embodiment of the present application discloses a real-time imaging task allocation system for an edge network. The system is arranged in an edge server and includes: Input module, used to obtain tasks and subtasks within tasks; A task division module is used to divide subtasks according to dependency relationships, wherein the dependency relationships include time dependency and score dependency. The time dependency is used to indicate a subtask in which the output of the most recently completed subtask serves as the input of the next task. The score dependency is used to distinguish whether subtasks are from the same task. A task allocation scheme calculation module is used to obtain an optimal task allocation scheme with the goal of minimizing the total completion time of real-time imaging under the condition of satisfying dependency relationships. The optimal task allocation scheme is used to determine the sensor nodes to which the subtasks are allocated; Execution module, used to execute the best task allocation plan; The time prediction module is used to predict the completion time of the subtasks of the sensor nodes in executing the best task allocation scheme; The execution module dynamically schedules the sensor nodes to which the subtasks are assigned according to the completion time.

[0047] It is understandable that the edge network real-time imaging task allocation system disclosed in this embodiment can be deployed in multiple edge servers that can work together.

[0048] In one embodiment, the task allocation solution calculation module is further configured to: Establishing an objective function with the goal of minimizing the total completion time of real-time imaging, wherein the objective function includes the start time of the subtask and the transmission delay between the edge servers; The objective function is optimized with computational resource constraints, including the computational resources of the edge server, the number of sensor nodes managed by the edge server, the limited computational resources, and the fact that a subtask can only be assigned to one sensor node and a sensor node can only process one subtask at a time. Among them, optimizing the objective function includes: using an improved mutation particle swarm optimization algorithm to solve the objective function, and dynamically adjusting the inertia weight during the solution process to balance the transmission delay between edge servers and the computing resources of the edge servers.

[0049] In one embodiment, the dependency condition in the dependency condition is satisfied, including: Assign subtasks with score dependencies among sensor nodes managed by the same edge server; For subtasks with time dependency, when the remaining computing resources do not meet the computing resources of the subtask, the sensor nodes managed by the edge server with the smallest transmission delay are found for allocation based on the transmission delay between the edge server and other edge servers with remaining computing resources.

[0050] In one embodiment, the execution module dynamically schedules the sensor nodes to which the subtasks are assigned according to the completion time, including: arranging the incoming subtasks to the sensor nodes with the shortest completion time or arranging the incoming subtasks to idle sensor nodes.

[0051] In one embodiment, the present application discloses an edge network real-time imaging system, including multiple edge servers and multiple sensor nodes managed by each edge server. The multiple edge servers each include an edge network real-time imaging task allocation system. Each edge server manages multiple sensor nodes to perform subtask calculations, and interactively allocates subtasks to sensor nodes managed by other edge servers through other edge servers, and receives subtasks from other edge servers.

[0052] It is understood that, based on the optimal task allocation solution calculated by the edge network real-time imaging task allocation system, multiple edge servers can simultaneously execute tasks in their own servers and collaboratively execute tasks on other edge servers. The edge network real-time imaging task allocation system also dynamically schedules subtasks to multiple sensor nodes in multiple edge servers based on the predicted completion time of the sensor nodes for the subtasks.

[0053] See also Figure 5 , which is a schematic diagram of the edge network real-time imaging system disclosed in this application being used in an actual scenario, and is used for real-time imaging in a seismic exploration edge network; Multiple edge servers are deployed along each measurement line, and each edge server has multiple sensor nodes. After the sensor node completes data collection, real-time imaging calculation is performed. Taking one of the edge servers as an example, after setting the task, the subtask division is performed through the edge network real-time imaging task allocation system, and the subtasks are divided into time dependence and score dependence. The dependent subtasks are assigned to the sensor nodes. Among them, data preprocessing, autopower spectrum calculation, autocorrelation coefficient calculation, and dispersion curve calculation are score-dependent, and are assigned to the sensor nodes in the same edge server for calculation. The mutual correlation coefficient depends on the input of the autopower spectrum calculation, which is time-dependent, and is score-dependent with the global optimization. Therefore, it is assigned to the sensor nodes in another edge server for calculation at the same time.

[0054] During the calculation process, it is necessary to predict the completion time, calculate the next task after the current task is completed, and dynamically schedule the sensor nodes to which the subtasks are assigned according to the completion time.

[0055] Finally, the calculation results are summarized to the edge server that performs task allocation for real-time imaging.

[0056] In one embodiment, the present application discloses a computer-readable storage medium storing computer instructions. When the instructions are executed by a processor, the edge network real-time imaging task allocation method of any of the above embodiments is implemented.

[0057] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A method for allocating real-time imaging tasks in an edge network, characterized in that: include: Get tasks and subtasks in tasks; Divide the subtasks according to dependency relationships, wherein the dependency relationships include time dependency and score dependency. The time dependency is used to indicate that the output of the most recently completed subtask serves as the input of the next subtask, and the score dependency is used to distinguish whether the subtasks are from the same task. Under the condition of satisfying the dependency relationship, an optimal task allocation scheme is obtained with the goal of minimizing the total completion time of real-time imaging, and the optimal task allocation scheme is used to determine the sensor nodes to which the subtasks are allocated; Execute the best task allocation scheme, predict the completion time of sensor nodes for subtasks, and dynamically schedule the sensor nodes to which subtasks are assigned according to the completion time.

2. The edge network real-time imaging task allocation method according to claim 1, characterized in that: Under the condition of satisfying the dependency relationship and minimizing the total completion time of real-time imaging, the best task allocation solution is obtained, including: An objective function is established with the goal of minimizing the total completion time of real-time imaging; The objective function is optimized with the limited computing resources as constraints, wherein the limited computing resources include: the limited computing resources of the edge server, the number of sensor nodes managed by the edge server and the limited computing resources, and the limited nature that a subtask can only be assigned to one sensor node and a sensor node can only process one subtask at the same time.

3. The edge network real-time imaging task allocation method according to claim 1, characterized in that: The optimization objective function includes: using an improved mutation particle swarm optimization algorithm to solve the objective function, and dynamically adjusting the inertia weight during the solution process to balance the transmission delay between edge servers and the computing resources of the edge servers.

4. The edge network real-time imaging task allocation method according to claim 1, characterized in that: Dependency conditions, including: Assign subtasks with score dependencies among sensor nodes managed by the same edge server; For subtasks with time dependency, when the remaining computing resources do not meet the computing resources of the subtask, according to the transmission delay between the edge server and other edge servers with remaining computing resources, the sensor nodes managed by the edge server with the smallest transmission delay are found for allocation.

5. The edge network real-time imaging task allocation method according to claim 1, characterized in that: The sensor nodes to which the subtasks are assigned are dynamically scheduled according to the completion time, including: arranging the incoming subtasks to the sensor nodes with the shortest completion time or arranging the incoming subtasks to the idle sensor nodes.

6. An edge network real-time imaging task allocation system, characterized in that: Used in edge servers, including: Input module, used to obtain tasks and subtasks in tasks; A task division module, used to divide subtasks according to dependency relationships, wherein the dependency relationships include time dependency and score dependency, wherein the time dependency is used to indicate a subtask in which the output of the most recently completed subtask is used as the input of a subsequent task, and the score dependency is used to distinguish whether subtasks are from the same task; A task allocation scheme calculation module is used to obtain an optimal task allocation scheme with the goal of minimizing the total completion time of real-time imaging under the condition of satisfying the dependency relationship, and the optimal task allocation scheme is used to determine the sensor nodes to which the subtasks are allocated; An execution module, used to execute the best task allocation plan; The time prediction module is used to predict the completion time of the subtasks of the sensor nodes in executing the best task allocation scheme; The execution module dynamically schedules the sensor nodes to which the subtasks are assigned according to the completion time.

7. The edge network real-time imaging task allocation system according to claim 6, characterized in that: The task allocation scheme calculation module is also used to: Establishing an objective function with the goal of minimizing the total completion time of real-time imaging, wherein the objective function includes the start time of the subtask and the transmission delay between the edge servers; The objective function is optimized with limited computing resources as constraints, wherein the limited computing resources include: computing resources of the edge server, the number and computing resources of sensor nodes managed by the edge server, and that a subtask can only be assigned to one sensor node and a sensor node can only process one subtask at a time; The optimization objective function includes: solving the objective function by using an improved mutation particle swarm optimization algorithm, and dynamically adjusting the inertia weight during the solution process to balance the transmission delay between edge servers and the computing resources of the edge servers.

8. The edge network real-time imaging task allocation system according to claim 6, characterized in that: The dependency conditions include: Assign subtasks with score dependencies among sensor nodes managed by the same edge server; For subtasks with time dependency, when the remaining computing resources do not meet the computing resources of the subtask, according to the transmission delay between the edge server and other edge servers with remaining computing resources, the sensor nodes managed by the edge server with the smallest transmission delay are found for allocation.

9. The edge network real-time imaging task allocation system according to claim 6, characterized in that: The execution module dynamically schedules the sensor nodes to which the subtasks are assigned according to the completion time, including: arranging the incoming subtasks to the sensor nodes with the shortest completion time or arranging the incoming subtasks to the idle sensor nodes.

10. An edge network real-time imaging system, characterized in that: include: Multiple edge servers, multiple sensor nodes managed by each edge server, each of the multiple edge servers includes the edge network real-time imaging task allocation system as described in any one of claims 7 to 9, each edge server manages multiple sensor nodes to calculate subtasks, and interactively allocates subtasks to sensor nodes managed by other edge servers through other edge servers.

Images