Task deployment method, model training method, equipment and storage medium

Abstract

The invention provides a task deployment method, a model training method, equipment and a storage medium, and relates to the technical field of industrial internet, and the method can fully utilize communication resources and computing resources to deploy industrial tasks on the basis of meeting the service quality. The method comprises the following steps: acquiring task information of a to-be-executed task of terminal equipment and environment information of an edge-end collaborative scene; the to-be-executed task is any one of all to-be-executed tasks in one calculation period in the edge-end collaborative scene; determining a deployment scheme of the to-be-executed task based on the task information, the environment information and a trained task deployment model; the deployment scheme of the to-be-executed task is used for indicating to deploy the to-be-executed task on the local computing device and / or the edge computing device, so that resources are utilized to the maximum extent under the condition that the processing delay of the to-be-executed task meets a delay threshold value required by the to-be-executed task; and deploying the to-be-executed task based on the deployment scheme.

Description

Technical Field

[0001] This application relates to the technical field of industrial Internet, and in particular, to a task deployment method, a model training method, a device, and a storage medium. Background Art

[0002] The traditional pyramid architecture of industrial Internet is no longer applicable to emerging intelligent applications that have emerged in factories one after another, such as image recognition, intelligent sorting, robot inspection, etc. Emerging industrial intelligent applications require efficient transfer of various data in factories and more powerful computing capabilities to support the operation of intelligent algorithms.

[0003] The cloud-edge-end industrial Internet architecture combines cloud computing and mobile edge computing (MEC) technologies, providing a new computing paradigm for industrial control systems. However, the computing power of traditional industrial control systems mainly depends on traditional programmable logic controllers (PLCs), without considering other resources in the industrial Internet scenario, resulting in insufficient utilization of resources. Summary of the Invention

[0004] This application provides a task deployment method, a model training method, a device, and a storage medium, which realizes the deployment of industrial tasks by making full use of communication resources and computing resources on the basis of meeting service quality.

[0005] In a first aspect, this application provides a task deployment method, which is applied to an edge-end collaboration scenario. The edge-end collaboration scenario includes: a terminal device, a local computing device connected to the terminal device, and an edge computing device; the local computing device communicates with the edge computing device through a base station; the method includes: obtaining task information of a to-be-executed task of the terminal device and environment information of the edge-end collaboration scenario; the to-be-executed task is any one of all to-be-executed tasks in a computing cycle in the edge-end collaboration scenario; determining a deployment plan for the to-be-executed task based on the task information, the environment information, and a trained task deployment model; the deployment plan for the to-be-executed task is used to indicate the deployment of the to-be-executed task on the local computing device and / or the edge computing device, so as to maximize the utilization of resources when the processing delay of the to-be-executed task meets the delay threshold required by the to-be-executed task; deploying the to-be-executed task based on the deployment plan.

[0006] It can be understood that the traditional industrial task deployment method mainly relies on traditional PLCs for task deployment, without making full use of the communication resources of the base stations and the computing resources of edge computing devices in the industrial Internet scenario. A task deployment method provided by this application determines the deployment plan for the task to be executed based on task information, environmental information, and a trained task deployment model, and allocates corresponding communication resources and computing resources according to the deployment plan. On the basis of meeting the delay threshold required by the task, the communication resources of the base stations, the computing resources of local computing devices, and the computing resources of edge computing devices in the industrial task scenario are maximally utilized, enabling the deployment of industrial tasks by making full use of communication resources and computing resources on the basis of meeting service quality.

[0007] A possible implementation manner is that the task deployment model is trained based on the task information of the test task, the environmental information of the edge-side collaboration scenario before the execution of the test task, and the delay threshold required by the test task; among them, the delay threshold required by the test task is used as supervision information.

[0008] Another possible implementation manner is that the types of tasks to be executed include at least one of the following: motion control tasks, machine-to-machine control tasks, and video monitoring tasks.

[0009] Another possible implementation manner is that the task deployment model includes: a first task deployment model for determining the deployment plan for control tasks; the deployment plan for control tasks is used to indicate the deployment of control tasks on local computing devices or edge computing devices; a second task deployment model for determining the deployment plan for video monitoring tasks; the deployment plan for video monitoring tasks is used to indicate the deployment of video monitoring tasks on local computing devices and / or edge computing devices.

[0010] Another possible implementation manner is that the task information of the task to be executed includes at least one of the following: the task type of the task to be executed, the number of task data packets of the task to be executed, the data packet length of the task to be executed, the delay threshold required by the task to be executed, and the minimum data transmission rate required by the task to be executed.

[0011] Another possible implementation manner is that the environmental information of the edge-side collaboration scenario includes at least one of the following: the maximum transmission power of each terminal device in the edge-side collaboration scenario, the data volume of the current task in the edge-side collaboration scenario, the delay threshold required by the current task in the edge-side collaboration scenario, the computing power of the local computing device in the edge-side collaboration scenario, the network channel state of the edge-side collaboration scenario, the number of resource blocks that can be allocated by the base station in the edge-side collaboration scenario, and the total computing power of the edge computing device in the edge-side collaboration scenario.

[0012] Another possible implementation. When the deployment plan for the task to be executed is used to indicate the deployment of the task to be executed on the local computing device and the edge computing device, the deployment plan further includes: the amount of task data deployed on the local computing device and the amount of task data deployed on the edge computing device for the task to be executed.

[0013] Another possible implementation. The amount of task data deployed on the local computing device for the task to be executed is determined based on the deployment ratio of the task to be executed on the local computing device, the number of data packets of the task to be executed, and the length of the data packets of the task to be executed; the amount of task data deployed on the edge computing device for the task to be executed is determined based on the deployment ratio of the task to be executed on the edge computing device, the number of data packets of the task to be executed, and the length of the data packets of the task to be executed.

[0014] Another possible implementation. The amount of task data deployed on the local computing device for the task to be executed satisfies the following formula:

[0015]

[0016] where, is the amount of task data deployed on the local computing device for the task to be executed, is the deployment ratio of the task to be executed on the edge computing device, is the deployment ratio of the task to be executed on the local computing device, is the number of data packets of the task to be executed, is the length of the data packets of the task to be executed.

[0017] Another possible implementation. The amount of task data deployed on the edge computing device for the task to be executed satisfies the following formula:

[0018]

[0019] where, is the amount of task data deployed on the edge computing device for the task to be executed, is the deployment ratio of the task to be executed on the edge computing device, is the number of data packets of the task to be executed, is the length of the data packets of the task to be executed.

[0020] Another possible implementation. The processing delay of the task to be executed is determined based on at least one of the following: local computing delay, edge computing delay, communication delay.

[0021] Another possible implementation. The local computing delay is determined based on the amount of task data deployed on the local computing device for the task to be executed, the computing complexity of the task to be executed, and the computing resources of the local computing device.

[0022] Another possible implementation is that the local computing delay satisfies the following formula:

[0023]

[0024] where is the local computing delay, is the amount of task data for the task to be executed deployed on the local computing device, and C n is the computational complexity of the task to be executed, and f is the computing resource of the local computing device.

[0025] Another possible implementation is that the edge computing delay is the difference between the delay threshold required by the task to be executed and the communication delay.

[0026] Another possible implementation is that the communication delay is determined based on the following parameters: the amount of task data that the task to be executed needs to transmit, the number of resource blocks allocated for the task to be executed, the transmission power, the channel gain, and the distance between the terminal device and the base station.

[0027] Another possible implementation is that the communication delay satisfies the following formula:

[0028]

[0029] where represents the amount of task data that the task to be executed needs to transmit, represents the number of resource blocks allocated for the task to be executed, and p max represents the transmission power, represents the channel gain, and d n represents the distance between the terminal device and the base station, α represents the path loss coefficient, and σ 2 represents the Gaussian white noise power.

[0030] Another possible implementation is that the number of resource blocks allocated for the task to be executed includes: the basic allocation number and the supplementary allocation number; the basic allocation number is determined based on the minimum data transmission rate required by the task to be executed, the number of bandwidths carried by each resource block, the transmission power, the channel gain, the distance between the terminal device and the base station, and the allocation frequency of the resource blocks; the supplementary allocation number is determined based on the total number of resource blocks included in the base station, the number of tasks to be executed within a computing cycle, and the sum of the basic allocation numbers of all tasks to be executed within a computing cycle.

[0031] Another possible implementation is that the basic allocation number satisfies the following formula:

[0032]

[0033] where represents the minimum data transfer rate required for the task to be executed, w represents the number of bandwidths carried by each resource block, p max represents the transmission power represents the channel gain, d n represents the distance between the terminal device and the base station, α represents the path loss coefficient, σ 2 represents the Gaussian white noise power, and x is the allocation frequency of the resource block.

[0034] Another possible implementation manner is that the supplementary allocation number satisfies the following formula:

[0035]

[0036] w rb represents the total number of resource blocks included in the base station, N represents the number of tasks to be executed within a computing cycle, represents the sum of the basic allocation numbers of all tasks to be executed within a computing cycle.

[0037] In a second aspect, the present application provides a model training method, which is applied to an edge-cloud collaboration scenario and includes: a terminal device, a local computing device connected to the terminal device, and an edge computing device; the local computing device communicates with the edge computing device through a base station; the method includes: obtaining task information of a test task, environmental information of the edge-cloud collaboration scenario before the execution of the test task, and a delay threshold required by the test task; the test task is any one of all test tasks within a computing cycle in the edge-cloud collaboration scenario; training a task deployment model to be trained based on the task information of the test task, the environmental information of the edge-cloud collaboration scenario before the execution of the test task, and the delay threshold required by the test task, to obtain a trained task deployment model; the trained task deployment model is used to determine a task deployment plan for the task to be executed by the terminal device; wherein, the task deployment plan is used to indicate deploying the task to be executed on the local computing device and / or the edge computing device, so as to maximize the utilization of resources while the processing delay of the task to be executed meets the delay threshold required by the task to be executed.

[0038] The model training method provided by the present application trains a task deployment model that can determine an optimal task deployment plan by comprehensively considering the task information of the test task, the environmental information of the edge-cloud collaboration scenario, and the delay threshold required by the task. This method enables the task to be executed to maximize the utilization of the resources of the local computing device and / or the edge computing device while meeting the delay requirements, thereby improving the overall computing efficiency and resource utilization rate in the edge-cloud collaboration scenario.

[0039] A possible implementation method is to obtain the task information of the test task, the environmental information of the front-end collaboration scenario before the execution of the test task, and the latency threshold required by the test task, and train the task deployment model to be trained to obtain a trained task deployment model, including: inputting the task information of the test task and the environmental information of the front-end collaboration scenario before the execution of the test task into the task deployment model to be trained to obtain the deployment plan of the test task; determining the processing latency of the test task based on the deployment plan of the test task; training the task deployment model based on the processing latency of the test task and the latency threshold required by the test task to obtain a trained task deployment model.

[0040] Another possible implementation method is to train the task deployment model based on the difference between the processing latency of the test task and the latency threshold required by the test task to obtain a trained task deployment model, including: setting a reward function based on the processing latency of the test task and the latency threshold required by the test task; the reward function is used to give a positive reward when the processing latency of the test task is less than or equal to the latency threshold required by the test task, and give a negative reward when the processing latency of the test task is greater than the latency threshold required by the test task; iteratively training the task deployment model to be trained based on the reward function until the reward function converges to obtain a trained task deployment model.

[0041] Another possible implementation method is that the reward function satisfies the following formula:

[0042]

[0043] where R represents the total reward value of the reward function, T represents the number of calculation cycles, R t represents the reward value of all test tasks in the t-th calculation cycle, N represents the number of test tasks, represents the reward value of the n-th test task in the t-th calculation cycle, represents the processing latency of the test task, represents the latency threshold required by the test task.

[0044] Another possible implementation method is that when the deployment plan of the test task is used to indicate the deployment of the test task on the local computing device and the edge computing device, the deployment plan further includes: the amount of task data deployed on the local computing device and the amount of task data deployed on the edge computing device for the test task.

[0045] Another possible implementation method is that the processing latency of the test task is determined based on at least one of the following: local computing latency, edge computing latency, communication latency.

[0046] Another possible implementation manner is that the local computing delay is determined based on the amount of task data of the test task deployed on the local computing device, the computing complexity of the test task, and the computing resources of the local computing device.

[0047] Another possible implementation manner is that the edge computing delay is the difference between the delay threshold required by the test task and the communication delay.

[0048] Another possible implementation manner is that the communication delay is determined based on the following parameters: the amount of task data that the test task needs to transmit, the number of resource blocks allocated for the test task, the transmission power, the channel gain, and the distance between the terminal device and the base station.

[0049] Another possible implementation manner is that the number of resource blocks allocated for the test task includes: the basic allocation number and the supplementary allocation number; the basic allocation number is determined based on the minimum data transmission rate required by the test task, the bandwidth number carried by each resource block, the transmission power, the channel gain, the distance between the terminal device and the base station, and the allocation frequency of the resource blocks; the supplementary allocation number is determined based on the total number of resource blocks included in the base station, the number of test tasks within a computing cycle, and the sum of the basic allocation numbers of all tasks to be executed within a computing cycle.

[0050] Another possible implementation manner is that the type of the test task includes at least one of the following: motion control tasks, machine - to - machine control tasks, video surveillance tasks.

[0051] Another possible implementation manner is that the task deployment model includes: a first task deployment model for determining the deployment plan of control tasks; the deployment plan of control tasks is used to indicate the deployment of control tasks on the local computing device or the edge computing device; a second task deployment model for determining the deployment plan of video surveillance tasks; the deployment plan of video surveillance tasks is used to indicate the deployment of video surveillance tasks on the local computing device and / or the edge computing device.

[0052] Another possible implementation manner is that the task information of the test task includes at least one of the following: the task type of the test task, the task data packet of the test task, the data packet length of the test task, the delay threshold required by the test task, the minimum data transmission rate required by the test task.

[0053] Another possible implementation manner is that the environmental information of the edge - terminal cooperation scenario includes at least one of the following: the maximum transmission power of each terminal device in the edge - terminal cooperation scenario, the data volume of the current test task in the edge - terminal cooperation scenario, the delay threshold required by the current test task in the edge - terminal cooperation scenario, the computing power of the local computing device in the edge - terminal cooperation scenario, the network channel state of the edge - terminal cooperation scenario, the number of resource blocks that can be allocated by the base station in the edge - terminal cooperation scenario, the total computing power of the edge computing device in the edge - terminal cooperation scenario.

[0054] In a third aspect, the present application provides a task deployment device, which is applied to an edge-cloud collaboration scenario. The device includes: an acquisition module, a decision-making module, and a deployment module; the acquisition module is configured to acquire task information of a to-be-executed task of a terminal device and environmental information of the edge-cloud collaboration scenario; the to-be-executed task is any one of all to-be-executed tasks within a computing cycle in the edge-cloud collaboration scenario; the decision-making module is configured to determine a deployment plan for the to-be-executed task based on the task information, the environmental information, and a trained task deployment model; the deployment plan for the to-be-executed task is used to indicate deploying the to-be-executed task on a local computing device and / or an edge computing device, so as to maximize resource utilization while the processing delay of the to-be-executed task meets the delay threshold required by the to-be-executed task; the deployment module is configured to deploy the to-be-executed task based on the deployment plan.

[0055] In a possible implementation, the task deployment model is trained based on task information of a test task, environmental information of the edge-cloud collaboration scenario before the test task is executed, and the delay threshold required by the test task; wherein, the delay threshold required by the test task is used as supervision information.

[0056] In another possible implementation, the types of the to-be-executed tasks include at least one of the following: motion control tasks, machine-to-machine control tasks, and video surveillance tasks.

[0057] In yet another possible implementation, the task deployment model includes: a first task deployment model, which is configured to determine a deployment plan for control tasks; the deployment plan for control tasks is used to indicate deploying control tasks on a local computing device or an edge computing device; a second task deployment model, which is configured to determine a deployment plan for video surveillance tasks; the deployment plan for video surveillance tasks is used to indicate deploying video surveillance tasks on a local computing device and / or an edge computing device.

[0058] In yet another possible implementation, the task information of the to-be-executed task includes at least one of the following: the task type of the to-be-executed task, the number of task data packets of the to-be-executed task, the data packet length of the to-be-executed task, the delay threshold required by the to-be-executed task, and the minimum data transmission rate required by the to-be-executed task.

[0059] In yet another possible implementation, the environmental information of the edge-cloud collaboration scenario includes at least one of the following: the maximum transmission power of each terminal device in the edge-cloud collaboration scenario, the data volume of the current task in the edge-cloud collaboration scenario, the delay threshold required by the current task in the edge-cloud collaboration scenario, the computing power of the local computing device in the edge-cloud collaboration scenario, the network channel state of the edge-cloud collaboration scenario, the number of resource blocks that can be allocated by the base station in the edge-cloud collaboration scenario, and the total computing power of the edge computing devices in the edge-cloud collaboration scenario.

[0060] Another possible implementation, the deployment plan further includes: the amount of task data deployed on the local computing device and the amount of task data deployed on the edge computing device for the task to be executed.

[0061] Another possible implementation, the amount of task data deployed on the local computing device for the task to be executed is determined based on the deployment ratio of the task to be executed on the local computing device, the number of data packets of the task to be executed, and the length of the data packets of the task to be executed; the amount of task data deployed on the edge computing device for the task to be executed is determined based on the deployment ratio of the task to be executed on the edge computing device, the number of data packets of the task to be executed, and the length of the data packets of the task to be executed.

[0062] Another possible implementation, the amount of task data deployed on the local computing device for the task to be executed satisfies the following formula:

[0063]

[0064] Wherein, is the amount of task data deployed on the local computing device for the task to be executed, is the deployment ratio of the task to be executed on the edge computing device, is the deployment ratio of the task to be executed on the local computing device, is the number of data packets of the task to be executed, is the length of the data packets of the task to be executed.

[0065] Another possible implementation, the amount of task data deployed on the edge computing device for the task to be executed satisfies the following formula:

[0066]

[0067] Wherein, is the amount of task data deployed on the edge computing device for the task to be executed, is the deployment ratio of the task to be executed on the edge computing device, is the number of data packets of the task to be executed, is the length of the data packets of the task to be executed.

[0068] Another possible implementation, the processing delay of the task to be executed is determined based on at least one of the following: local computing delay, edge computing delay, communication delay.

[0069] Another possible implementation, the local computing delay is determined based on the amount of task data deployed on the local computing device for the task to be executed, the computing complexity of the task to be executed, and the computing resources of the local computing device.

[0070] Another possible implementation, the local computing delay satisfies the following formula:

[0071]

[0072] Among them, is the local computing delay, is the amount of task data for the task to be executed deployed on the local computing device, C n is the computational complexity of the task to be executed, and f is the computing resource of the local computing device.

[0073] Another possible implementation is that the edge computing delay is the difference between the delay threshold required by the task to be executed and the communication delay.

[0074] Another possible implementation is that the communication delay is determined based on the following parameters: the amount of task data that the task to be executed needs to transmit, the number of resource blocks allocated for the task to be executed, the transmission power, the channel gain, and the distance between the terminal device and the base station.

[0075] Another possible implementation is that the communication delay satisfies the following formula:

[0076]

[0077] Among them, represents the amount of task data that the task to be executed needs to transmit, represents the number of resource blocks allocated for the task to be executed, p max represents the transmission power, represents the channel gain, d n represents the distance between the terminal device and the base station, α represents the path loss coefficient, σ 2 represents the Gaussian white noise power.

[0078] Another possible implementation is that the number of resource blocks allocated for the task to be executed includes: the basic allocation number and the supplementary allocation number; the basic allocation number is determined based on the minimum data transmission rate required by the task to be executed, the number of bandwidths carried by each resource block, the transmission power, the channel gain, the distance between the terminal device and the base station, and the allocation frequency of the resource blocks; the supplementary allocation number is determined based on the total number of resource blocks included in the base station, the number of tasks to be executed within a computing cycle, and the sum of the basic allocation numbers of all tasks to be executed within a computing cycle.

[0079] Another possible implementation is that the basic allocation number satisfies the following formula:

[0080]

[0081] Among them, represents the minimum data transmission rate required by the task to be executed, w represents the number of bandwidths carried by each resource block, pmax represents the transmission power represents the channel gain, d n represents the distance between the terminal device and the base station, α represents the path loss coefficient, σ 2 represents the Gaussian white noise power, and x is the allocation frequency of resource blocks.

[0082] Another possible implementation is that the supplementary allocation number satisfies the following formula:

[0083]

[0084] w rb represents the total number of resource blocks included in the base station, N represents the number of tasks to be executed within a calculation period, represents the sum of the basic allocation numbers of all tasks to be executed within a calculation period.

[0085] In a fourth aspect, the present application provides a model training device applied to an edge-cloud collaboration scenario. The device includes: an acquisition module and a processing module; the acquisition module is configured to acquire the task information of a test task, the environmental information of the edge-cloud collaboration scenario before the execution of the test task, and the latency threshold required by the test task; the test task is any one of all test tasks within a calculation period in the edge-cloud collaboration scenario; the processing module is configured to train the task deployment model to be trained based on the task information of the test task, the environmental information of the edge-cloud collaboration scenario before the execution of the test task, and the latency threshold required by the test task, so as to obtain a trained task deployment model; the trained task deployment model is used to determine the task deployment scheme of the tasks to be executed by the terminal device; wherein, the task deployment scheme is used to indicate the deployment of the tasks to be executed on the local computing device and / or the edge computing device, so as to maximize the utilization of resources when the processing latency of the tasks to be executed meets the latency threshold required by the tasks to be executed.

[0086] A possible implementation is that the processing module is specifically configured to: input the task information of the test task and the environmental information of the edge-cloud collaboration scenario before the execution of the test task into the task deployment model to be trained, and obtain the deployment scheme of the test task; determine the processing latency of the test task based on the deployment scheme of the test task; and train the task deployment model based on the processing latency of the test task and the latency threshold required by the test task, so as to obtain a trained task deployment model.

[0087] Another possible implementation manner, the processing module is specifically configured to: set a reward function based on the processing delay of the test task and the delay threshold required by the test task; the reward function is used to give a positive reward when the processing delay of the test task is less than or equal to the delay threshold required by the test task, and give a negative reward when the processing delay of the test task is greater than the delay threshold required by the test task; based on the reward function, perform iterative training on the task deployment model to be trained until the reward function converges, and obtain the trained task deployment model.

[0088] Another possible implementation manner, the reward function satisfies the following formula:

[0089]

[0090] Wherein, R represents the total reward value of the reward function, T represents the number of calculation cycles, R t represents the reward value of all test tasks in the t-th calculation cycle, N represents the number of test tasks, represents the reward value of the n-th test task in the t-th calculation cycle, represents the processing delay of the test task, represents the delay threshold required by the test task.

[0091] Another possible implementation manner, the deployment scheme further includes: the amount of task data deployed on the local computing device for the test task and the amount of task data deployed on the edge computing device.

[0092] Another possible implementation manner, the processing delay of the test task is determined based on at least one of the following: local computing delay, edge computing delay, communication delay.

[0093] Another possible implementation manner, the local computing delay is determined based on the amount of task data of the test task deployed on the local computing device, the computing complexity of the test task, and the computing resources of the local computing device.

[0094] Another possible implementation manner, the edge computing delay is the difference between the delay threshold required by the test task and the communication delay.

[0095] Another possible implementation manner, the communication delay is determined based on the following parameters: the amount of task data that the test task needs to transmit, the number of resource blocks allocated for the test task, transmission power, channel gain, and the distance between the terminal device and the base station.

[0096] Another possible implementation manner, the number of resource blocks allocated for a test task includes: a basic allocation number and a supplementary allocation number; the basic allocation number is determined based on the minimum data transmission rate required by the test task, the number of bandwidths carried by each resource block, the transmission power, the channel gain, the distance between the terminal device and the base station, and the allocation frequency of the resource blocks; the supplementary allocation number is determined based on the total number of resource blocks included in the base station, the number of test tasks within a calculation period, and the sum of the basic allocation numbers of all tasks to be executed within a calculation period.

[0097] Another possible implementation manner, the types of test tasks include at least one of the following: motion control tasks, machine - to - machine control tasks, video surveillance tasks.

[0098] Another possible implementation manner, the task deployment model includes: a first task deployment model for determining the deployment plan of control tasks; the deployment plan of control tasks is used to indicate the deployment of control tasks on local computing devices or edge computing devices; a second task deployment model for determining the deployment plan of video surveillance tasks; the deployment plan of video surveillance tasks is used to indicate the deployment of video surveillance tasks on local computing devices and / or edge computing devices.

[0099] Another possible implementation manner, the task information of a test task includes at least one of the following: the task type of the test task, the task data packet of the test task, the data packet length of the test task, the delay threshold required by the test task, the minimum data transmission rate required by the test task.

[0100] Another possible implementation manner, the environmental information of the edge - terminal cooperation scenario includes at least one of the following: the maximum transmission power of each terminal device in the edge - terminal cooperation scenario, the data volume of the current test task in the edge - terminal cooperation scenario, the delay threshold required by the current test task in the edge - terminal cooperation scenario, the computing power of the local computing device in the edge - terminal cooperation scenario, the network channel state of the edge - terminal cooperation scenario, the number of resource blocks that can be allocated by the base station in the edge - terminal cooperation scenario, the total computing power of the edge computing device in the edge - terminal cooperation scenario.

[0101] In a fifth aspect, the present application provides an electronic device, which includes: a processor and a memory; the memory stores instructions executable by the processor; when the processor is configured to execute the instructions, the electronic device implements the method of the first aspect or the second aspect above.

[0102] In a sixth aspect, the present application provides a readable storage medium, which includes: computer software instructions; when the computer software instructions run in an electronic device, the electronic device implements the method of the first aspect or the second aspect above.

[0103] In a seventh aspect, the present application provides a computer program product. When the computer program product runs on a computer, it causes the computer to execute the steps of the relevant method described in the first aspect above to implement the method of the first aspect, or causes the computer to execute the steps of the relevant method described in the second aspect above to implement the method of the second aspect.

[0104] For the beneficial effects of the third to seventh aspects above, reference may be made to the corresponding descriptions of the first or second aspect, and details are not repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] Figure 1 Schematic diagram of an application scenario of a task deployment method provided by the present application;

[0106] Figure 2 Schematic diagram of a process of a task deployment method provided by the present application;

[0107] Figure 3 Schematic diagram of a task deployment strategy architecture provided by the present application;

[0108] Figure 4 Schematic diagram of a process of a model training method provided by the present application;

[0109] Figure 5 Schematic diagram of a convergence curve of a model training reward function provided by the present application;

[0110] Figure 6 Schematic diagram of a process of another model training method provided by the present application;

[0111] Figure 7 Schematic diagram of a test result of an execution success rate provided by the present application;

[0112] Figure 8 Schematic diagram of a test result of a resource block utilization rate provided by the present application;

[0113] Figure 9 Schematic diagram of the composition of a task deployment device provided by the present application;

[0114] Figure 10 Schematic diagram of the composition of a model training device provided by the present application;

[0115] Figure 11 Schematic diagram of the composition of an electronic device provided by the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0116] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts belong to the scope of protection of the present application.

[0117] It should be noted that in the embodiments of the present application, words such as "exemplarily" or "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "exemplarily" or "for example" in the embodiments of the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, the use of words such as "exemplarily" or "for example" is intended to present relevant concepts in a specific manner.

[0118] In order to facilitate a clear description of the technical solutions in the embodiments of the present application, in the embodiments of the present application, words such as "first" and "second" are used to distinguish the same items or similar items with basically the same functions and roles. Those skilled in the art can understand that words such as "first" and "second" do not limit the quantity and execution order.

[0119] The traditional pyramid architecture of industrial Internet is no longer suitable for emerging intelligent applications that have emerged in factories one after another, such as: image recognition, intelligent sorting, robot inspection, etc. Emerging industrial intelligent applications require multiple types of data to flow efficiently in the factory, and more importantly, require powerful computing capabilities to support the operation of intelligent algorithms. However, the data in the traditional pyramid architecture of industrial Internet is not interoperable and cannot achieve cross-layer interaction of data. Secondly, the computing power of traditional industrial control systems mainly relies on traditional PLCs, which have always occupied a core position in industrial control, but their computing power is difficult to bear the operation of intelligent algorithms. Therefore, the transformation from the traditional pyramid architecture to the cloud-edge-end architecture is in progress and will continue in the future. The cloud-edge-end industrial Internet architecture combines cloud computing and MEC technologies, providing a new computing paradigm for industrial control systems. However, how to make full and reasonable use of communication-computing resources while meeting the service quality requirements of industrial control tasks is a difficult point.

[0120] In view of the above technical problems, an embodiment of the present application provides a task deployment method, which can obtain the task information of the to-be-executed task of the terminal device and the environmental information of the edge-cloud collaboration scenario; and then, based on the task information, environmental information, and the trained task deployment model, determine the deployment plan for the to-be-executed task, and allocate corresponding communication resources and computing resources to different industrial tasks according to the deployment plan. In this way, on the basis of meeting the latency threshold required by the task, that is, the quality of service, the communication resources of the base station, the computing resources of the local computing device, and the computing resources of the edge computing device in the industrial task scenario are maximally utilized, and it is possible to fully utilize the communication resources and computing resources to deploy industrial tasks on the basis of meeting the quality of service.

[0121] A task deployment method provided by the present application can be applied to an edge-cloud collaboration scenario. The system architecture of this scenario includes: a terminal device, a base station, a local computing device, and an edge computing device.

[0122] Exemplarily, the terminal device can be industrial devices such as laser cutting devices, numerically controlled machine tools, distributed control devices, cameras, and video storage devices.

[0123] Exemplarily, as Figure 1 shown, the base station can be a 5G base station 100, the edge computing device can be a combined virtual PLC device 200, and the local computing devices can be local PLC devices 301, local PLC device 302, and local PLC device 303. Among them, the combined virtual PLC device 200 includes multiple virtual PLC devices. The combined virtual PLC device 200 is connected to the 5G base station 100, and the 5G base station 100 is respectively connected to the local computing device 302 and the local computing device 303.

[0124] The local PLC device 301 provides local computing resources for motion control tasks, the local PLC device 302 provides local computing resources for inter-machine control tasks, and the local PLC device 303 provides local computing resources for remote video monitoring tasks.

[0125] The 5G base station 100 provides communication services between the local PLC device 302, the local PLC device 303, and the combined virtual PLC device 200.

[0126] The combined virtual PLC device 200 provides edge computing resources for inter-machine control tasks and remote video monitoring tasks.

[0127] In some embodiments, the system may leave all motion control tasks for local PLC device 301 to calculate; the system leaves machine - to - machine control tasks for local PLC device 302 to calculate or completely offloads them to combined virtual PLC device 200 to calculate by executing a deployment plan; the system leaves part of video monitoring tasks for local PLC device 303 to calculate and offloads part of them to combined virtual PLC device 200 to calculate by executing a deployment plan.

[0128] Figure 2 It is a schematic flowchart of a task deployment method provided by an embodiment of the present application. Exemplarily, the task deployment method provided by an embodiment of the present application can be executed by an electronic device, and the electronic device is connected to Figure 1 devices in the edge - end collaboration scenario shown (for example, terminal devices, base stations, local computing devices, edge computing devices).

[0129] Exemplarily, the electronic device can be a server. For example, a single server, or a server cluster composed of multiple servers. In some embodiments, the server cluster can also be a distributed cluster.

[0130] Exemplarily, the electronic device can be a terminal device. For example, a mobile phone, a tablet computer, a desktop, a laptop, a handheld computer, a notebook computer, an ultra - mobile personal computer (UMPC), a netbook, as well as a cellular phone, a personal digital assistant (PDA), an augmented reality (AR) / virtual reality (VR) device, etc. The specific form of the terminal device is not particularly limited in the embodiments of the present application.

[0131] As Figure 2 shown, the task deployment method provided by the present application specifically includes the following steps:

[0132] S201. Obtain the task information of the task to be executed by the terminal device and the environmental information of the edge - end collaboration scenario.

[0133] Among them, the task to be executed is any one of all tasks to be executed in a calculation cycle in the edge - end collaboration scenario.

[0134] In some embodiments, the task information of the task to be executed includes at least one of the following: the task type of the task to be executed, the number of task data packets of the task to be executed, the data packet length of the task to be executed, the latency threshold required by the task to be executed, the minimum data transfer rate required by the task to be executed.

[0135] Exemplarily, the task information of the task to be executed can be expressed as {sy n,p , sp n,p , sl n,p , sd n,p , sr n,p}. Among them, sy n,p represents the task type of the task to be executed, sp n,p represents the number of task data packets of the task to be executed, sl n,p represents the data packet length of the task to be executed, sd n,p represents the latency threshold required by the task to be executed, sr n,p represents the minimum data transmission rate required by the task to be executed.

[0136] In some embodiments, the environmental information of the edge - end collaboration scenario includes at least one of the following: the maximum transmission power of each terminal device in the edge - end collaboration scenario, the data volume of the current task in the edge - end collaboration scenario, the latency threshold required by the current task in the edge - end collaboration scenario, the computing power of the local computing device in the edge - end collaboration scenario, the network channel state of the edge - end collaboration scenario, the number of resource blocks that can be allocated by the base station in the edge - end collaboration scenario, and the total computing power of the edge computing device in the edge - end collaboration scenario.

[0137] Exemplarily, the environmental information of the edge - end collaboration scenario can be expressed as:

[0138] S = {S 1 , S 2 , S 3} = {{p max , Q s , D, f}, {H, w rb}, {F}]

[0139] Among them, p max represents the maximum transmission power of each terminal device in the edge - end collaboration scenario, Q s represents the data volume of the current task in the edge - end collaboration scenario, D represents the latency threshold required by the current task in the edge - end collaboration scenario, f represents the computing power of the local computing device in the edge - end collaboration scenario, H represents the network channel state of the edge - end collaboration scenario, w rb represents the number of resource blocks that can be allocated by the base station in the edge - end collaboration scenario, and F represents the total computing power of the edge computing device in the edge - end collaboration scenario.

[0140] Exemplarily, the specific task information of the task to be executed from the terminal device and the environmental information of the edge - end collaboration scenario can be received through a central control system or a task scheduler, or the specific task information of the task to be executed from the terminal device and the environmental information of the edge - end collaboration scenario can be directly obtained.

[0141] S202. Determine the deployment plan for the task to be executed based on the task information, environment information, and the trained task deployment model.

[0142] Among them, the deployment plan for the task to be executed is used to indicate the deployment of the task to be executed on the local computing device and / or edge computing device, so as to maximize the utilization of resources when the processing delay of the task to be executed meets the delay threshold required by the task to be executed.

[0143] In some embodiments, the task deployment model is trained based on the task information of the test task, the environment information of the front-end collaboration scenario before the execution of the test task, and the delay threshold required by the test task; among them, the delay threshold required by the test task is used as supervision information.

[0144] In some embodiments, the types of tasks to be executed include at least one of the following: motion control tasks, machine-to-machine control tasks, and video surveillance tasks.

[0145] Exemplarily, the motion control tasks include: automated assembly lines, material handling systems, and robotic welding, etc.; the machine-to-machine control tasks include: intelligent warehousing systems, robot collaboration, and production line coordination, etc.; the video surveillance tasks include: environmental monitoring, security monitoring, etc.

[0146] S203. Deploy the task to be executed based on the deployment plan.

[0147] In some embodiments, since control tasks cannot be split and deployed individually, for control tasks, a non-offloading deployment plan or a full-offloading deployment plan is obtained through the first task deployment model, and the task to be executed is deployed based on the non-offloading deployment plan or the full-offloading deployment plan.

[0148] In some embodiments, since video surveillance tasks can be split and deployed individually, for video surveillance tasks, a partial-offloading deployment plan is obtained through the second task deployment model, and the task to be executed is deployed based on the partial-offloading deployment plan.

[0149] It can be understood that in traditional task deployment scenarios, generally only a single task type is considered. However, in real industrial scenarios, to complete production control, generally multiple tasks need to be coordinated. Therefore, this application considers multiple types of industrial control tasks and combines the characteristics of control tasks, and uses a hybrid task deployment model to flexibly deploy control tasks. The task deployment method is more reasonable, and the full utilization of communication resources and computing resources is achieved.

[0150] In some embodiments, the task deployment model in step S202 includes: a first task deployment model for determining a deployment plan for control tasks; the deployment plan for control tasks is used to indicate the deployment of control tasks on local computing devices or edge computing devices; a second task deployment model for determining a deployment plan for video surveillance tasks; the deployment plan for video surveillance tasks is used to indicate the deployment of video surveillance tasks on local computing devices and / or edge computing devices.

[0151] Exemplarily, the deployment plan for control tasks can be that all of these tasks are deployed on local computing devices for execution, or all are deployed on edge computing devices for execution; the deployment plan for video surveillance tasks can be that all of these tasks are deployed on local computing devices for execution, or all are deployed on edge computing devices for execution, or some are deployed on local computing devices and some are deployed on edge computing devices.

[0152] Exemplarily, the first task deployment model is a Double Deep Q-Network (DDQN) model, and the second task deployment model is a Deep Deterministic Policy Gradient (DDPG) model.

[0153] In some embodiments, the problem of coordinating different types of industrial tasks with communication-computing resources in an edge-cloud collaboration scenario is formulated as a Markov decision problem, and a deep reinforcement learning model is used to explore the optimal hybrid task offloading strategy (or called deployment plan) with its adaptability to a stochastic environment and powerful learning ability. Among them, the Markov decision problem includes: an agent, task information, environmental information, an action space, and a reward function; the deep reinforcement learning model includes: a DDQN model and a DDPG model, and the offloading strategy includes: a full offloading decision and a partial offloading decision.

[0154] In some embodiments, step S202 can be implemented as: determining the deployment plan for the task to be executed based on the offloading decision made by the task deployment model trained by the agent set according to task information, environmental information, and different types of industrial tasks.

[0155] Figure 3 This is a schematic diagram of a task deployment strategy architecture for an embodiment of this application, as Figure 3As shown in the figure, the status information of all tasks to be executed comes from terminal devices (such as various factory devices), base stations (such as 5G base stations), and servers (such as the servers where edge computing devices and local computing devices are located). The status information includes environmental information and task information. The status information is input into the trained task deployment model to obtain the computing power collaboration decision and RB allocation decision. Among them, the trained task deployment model includes: the DDQN model architecture, the DDPG model architecture, and the experience cache pool. Based on the computing power collaboration decision and RB allocation decision, task deployment is performed for different types of industrial tasks. For example, the 5G base station deploys task 1 entirely on the edge computing device p3 in the edge device, deploys task 2 entirely on the edge computing device p2 in the edge device, and deploys part of task 3 on the edge computing device p5 in the edge device.

[0156] Exemplarily, the method respectively sets two agents: agent 1 and agent 2. Agent 1 uses the DDQN model to make a full offloading decision for control tasks, and the action space is A1 = {0, 1}, and the offloading ratio of control tasks can be obtained. Among them, the offloading ratio of control tasks is 0 or 1. Agent 2 uses the DDPG model to make a partial offloading decision for video surveillance tasks, and the action space is A2 = [0, 1], and the offloading ratio of video surveillance tasks can be obtained. Among them, the offloading ratio of video surveillance tasks is greater than or equal to 0 and less than or equal to 1.

[0157] Exemplarily, it is assumed that there are a total of T computing cycles, and there are N tasks to be executed in each computing cycle. Then in computing cycle t, the offloading decisions made by agent 1 and agent 2 respectively based on the status information can be expressed as:

[0158] a t ={a 1,t ,a 2,t}

[0159] Among them, a 1,t represents the offloading decision made by agent 1. If the i-th task to be executed in computing cycle t is an inter-machine control task, when , this task is deployed to the edge computing device for calculation, otherwise, this task is calculated on the local computing device; a 2,t represents the offloading decision made by agent 2. And If the j-th task to be executed in computing cycle t is a video surveillance task, then this task ratio of data packets is deployed to the edge computing device for calculation, ratio of data packets remains on the local computing device for calculation.

[0160] Exemplarily, assume that there are a total of T computing cycles, and each computing cycle has N tasks to be executed. Then, in computing cycle t, the offloading decision for the nth task to be executed can be expressed as:

[0161]

[0162] where sy n,p being 0 indicates a motion control task, sy n,p being 1 indicates an inter-machine control task, sy n,p being 2 indicates a video surveillance task, represents the offloading decision value when the ith task to be executed in computing cycle t is an inter-machine control task, represents the offloading decision value when the jth task to be executed in computing cycle t is a video surveillance task.

[0163] It can be understood that since the latency threshold of motion control tasks is less than 2 milliseconds, the latency threshold of inter-machine control tasks is less than 10 milliseconds, and the latency threshold of remote video surveillance tasks is less than 20 milliseconds. Therefore, to meet the strict latency threshold requirements of tasks to be executed, in the embodiments of the present application, only the inter-machine control tasks and video surveillance tasks are considered for edge-end collaborative deployment, while motion control tasks are considered to be all processed by local computing devices. Therefore, the offloading decision value for motion control tasks is 0.

[0164] In some embodiments, when the deployment plan for tasks to be executed is used to indicate the deployment of tasks to be executed on local computing devices and edge computing devices, the deployment plan in step S202 further includes: the amount of task data deployed on local computing devices and the amount of task data deployed on edge computing devices.

[0165] In some embodiments, the amount of task data deployed on local computing devices is determined based on the deployment ratio of the task to be executed on local computing devices, the number of data packets of the task to be executed, and the length of the data packets of the task to be executed; the amount of task data deployed on edge computing devices is determined based on the deployment ratio of the task to be executed on edge computing devices, the number of data packets of the task to be executed, and the length of the data packets of the task to be executed.

[0166] In some embodiments, the amount of task data deployed on local computing devices satisfies the following formula (1):

[0167]

[0168] where is the amount of task data deployed on local computing devices, is the deployment ratio of the task to be executed on edge computing devices, is the deployment ratio of the task to be executed on the local computing device, is the number of data packets of the task to be executed, is the length of the data packet of the task to be executed.

[0169] In some embodiments, the amount of task data of the task to be executed deployed on the edge computing device satisfies the following formula (two):

[0170]

[0171] wherein, is the amount of task data of the task to be executed deployed on the edge computing device, is the deployment ratio of the task to be executed on the edge computing device, is the number of data packets of the task to be executed, is the length of the data packet of the task to be executed.

[0172] In some embodiments, the processing delay of the task to be executed in step S202 is determined based on at least one of the following: local computing delay, edge computing delay, and communication delay.

[0173] In some embodiments, the local computing delay is determined based on the amount of task data of the task to be executed deployed on the local computing device, the computing complexity of the task to be executed, and the computing resources of the local computing device.

[0174] In some embodiments, the local computing delay satisfies the following formula (three):

[0175]

[0176] wherein, is the local computing delay, is the amount of task data of the task to be executed deployed on the local computing device, C n is the computing complexity of the task to be executed, and f is the computing resources of the local computing device.

[0177] Exemplarily, C n ∈{C 1 , C 2}, C 1 is the computing complexity of the inter-machine control type task, and C 2 is the computing complexity of the video surveillance type task.

[0178] In some embodiments, the edge computing delay is the difference between the delay threshold required by the task to be executed and the communication delay.

[0179] Exemplarily, the edge computing delay satisfies the following formula (four):

[0180]

[0181] Among them, is the edge computing delay, is the maximum allowable edge computing delay, is the delay threshold required by the task to be executed, is the communication delay of the task to be executed.

[0182] It can be understood that in the above embodiments provided by the present application, for the allocation strategy of edge computing resources, the present application is committed to meeting the delay threshold requirements of industrial tasks with as few computing resources as possible, so as to ensure that the edge server can carry more industrial tasks. Specifically, the allocation of edge computing resources is mainly determined according to the maximum allowable edge computing delay, that is, the minimum computing resources are allocated for the tasks to be executed that need to be offloaded according to this delay limit.

[0183] In some embodiments, the communication delay is determined based on the following parameters: the amount of task data to be transmitted by the task to be executed, the number of resource blocks allocated for the task to be executed, the transmission power, the channel gain, and the distance between the terminal device and the base station.

[0184] In some embodiments, the communication delay satisfies the following formula (V):

[0185]

[0186] Among them, represents the amount of task data to be transmitted by the task to be executed, represents the number of resource blocks allocated for the task to be executed, p max represents the transmission power, represents the channel gain, d n represents the distance between the terminal device and the base station, α represents the path loss coefficient, σ 2 represents the Gaussian white noise power.

[0187] In some embodiments, the number of resource blocks allocated for the task to be executed includes: the basic allocation number and the supplementary allocation number; the basic allocation number is determined based on the minimum data transmission rate required by the task to be executed, the bandwidth number carried by each resource block, the transmission power, the channel gain, the distance between the terminal device and the base station, and the allocation frequency of the resource blocks; the supplementary allocation number is determined based on the total number of resource blocks included in the base station, the number of tasks to be executed within a calculation period, and the sum of the basic allocation numbers of all tasks to be executed within a calculation period.

[0188] In some embodiments, the basic allocation number satisfies the following formula (VI):

[0189]

[0190] Among them, represents the minimum data transfer rate required for the task to be executed, w represents the number of bandwidths carried by each resource block, p max represents the transmission power, represents the channel gain, d n represents the distance between the terminal device and the base station, α represents the path loss coefficient, σ 2 represents the Gaussian white noise power, and x is the allocation frequency of the resource block.

[0191] In some embodiments, the supplementary allocation number satisfies the following formula (VII):

[0192]

[0193] Among them, w rb represents the total number of resource blocks included in the base station, N represents the number of tasks to be executed within a calculation period, represents the sum of the basic allocation numbers of all tasks to be executed within a calculation period.

[0194] Exemplarily, the sum of the basic allocation numbers of all tasks to be executed within a calculation period satisfies the following formula (VIII):

[0195]

[0196] It can be understood that in the above embodiments provided by the present application, in order to fully consider the delay threshold and the minimum data transfer rate, the present application divides the resource blocks into two parts: the necessary allocation number and the supplementary allocation number. Among them, the necessary allocation number is to consider the minimum data transfer rate requests of all offloaded tasks to be executed. According to the minimum data transfer rate requests, the minimum number of resource blocks required for each offloaded task to be executed is solved, that is, the necessary allocation number. Then, the remaining number of resource blocks is evenly distributed to each offloaded task to be executed as the supplementary allocation number. Thus, both the minimum data transfer rate requirements of each industrial task are met, and on the basis of meeting the delay threshold, the communication resources are fully utilized.

[0197] Figure 4 is a schematic flowchart of a model training method provided by an embodiment of the present application. Exemplarily, the model training method provided by an embodiment of the present application can be executed by an electronic device, and the electronic device is connected to Figure 1 the devices (such as terminal devices, base stations, local computing devices, edge computing devices) in the edge-end collaboration scenario shown.

[0198] Exemplarily, the electronic device may be a server. For example, a single server, or a server cluster composed of multiple servers. In some embodiments, the server cluster may also be a distributed cluster.

[0199] Exemplarily, the electronic device may be a terminal device. For example, a mobile phone, a tablet computer, a desktop computer, a laptop computer, a handheld computer, a notebook computer, an ultra-mobile personal computer (UMPC), a netbook, a cellular phone, a personal digital assistant (PDA), an augmented reality (AR) / virtual reality (VR) device, etc. The specific form of the terminal device is not particularly limited in the embodiments of the present application.

[0200] As Figure 4 shown, the model training method provided in the present application specifically includes the following steps:

[0201] S301. Obtain the task information of the test task, the environmental information of the front-end collaborative scenario before the execution of the test task, and the latency threshold required by the test task.

[0202] Wherein, the test task is any one of all test tasks in a computing cycle in the front-end collaborative scenario.

[0203] Exemplarily, test samples may be obtained from the task processing history database. The test samples include the task information of multiple test tasks, the environmental information of the front-end collaborative scenario before the execution of the test tasks, and the latency threshold required by the test tasks.

[0204] S302. Train the task deployment model to be trained based on the task information of the test task, the environmental information of the front-end collaborative scenario before the execution of the test task, and the latency threshold required by the test task, to obtain a trained task deployment model.

[0205] Wherein, the trained task deployment model is used to determine the task deployment plan for the task to be executed by the terminal device; the task deployment plan is used to indicate the deployment of the task to be executed on the local computing device and / or the edge computing device, so as to maximize the utilization of resources when the processing latency of the task to be executed meets the latency threshold required by the task to be executed.

[0206] In some embodiments, obtaining the task information of a test task, the environmental information of the front-end collaboration scenario before the execution of the test task, and the latency threshold required by the test task, and training the task deployment model to be trained to obtain a trained task deployment model, including: inputting the task information of the test task and the environmental information of the front-end collaboration scenario before the execution of the test task into the task deployment model to be trained to obtain a deployment plan for the test task; determining the processing latency of the test task based on the deployment plan for the test task; and training the task deployment model based on the processing latency of the test task and the latency threshold required by the test task to obtain a trained task deployment model.

[0207] In some embodiments, training the task deployment model based on the difference between the processing latency of the test task and the latency threshold required by the test task to obtain a trained task deployment model, including: setting a reward function based on the processing latency of the test task and the latency threshold required by the test task; the reward function is used to give a positive reward when the processing latency of the test task is less than or equal to the latency threshold required by the test task, and give a negative reward when the processing latency of the test task is greater than the latency threshold required by the test task; and iteratively training the task deployment model to be trained based on the reward function until the reward function converges to obtain a trained task deployment model.

[0208] In some embodiments, the reward function satisfies the following formula (IX):

[0209]

[0210] where R represents the total reward value of the reward function, T represents the number of calculation cycles, R t represents the reward value of all test tasks in the t-th calculation cycle, N represents the number of test tasks, represents the reward value of the n-th test task in the t-th calculation cycle, represents the processing latency of the test task, represents the latency threshold required by the test task.

[0211] Exemplarily, Figure 5 is a convergence curve graph of the reward function for model training. As shown in Figure 5 as the number of training rounds increases, the reward values of different numbers of industrial tasks converge to the highest point.

[0212] In some embodiments, when the deployment plan for the test task is used to indicate the deployment of the test task on the local computing device and the edge computing device, the deployment plan further includes: the amount of task data deployed on the local computing device and the amount of task data deployed on the edge computing device for the test task.

[0213] In some embodiments, the processing delay of the test task is determined based on at least one of the following: local computing delay, edge computing delay, and communication delay.

[0214] In some embodiments, the local computing delay is determined based on the amount of task data of the test task deployed on the local computing device, the computing complexity of the test task, and the computing resources of the local computing device.

[0215] In some embodiments, the edge computing delay is the difference between the delay threshold required by the test task and the communication delay.

[0216] In some embodiments, the communication delay is determined based on the following parameters: the amount of task data that the test task needs to transmit, the number of resource blocks allocated for the test task, the transmission power, the channel gain, and the distance between the terminal device and the base station.

[0217] In some embodiments, the number of resource blocks allocated for the test task includes: a basic allocation number and a supplementary allocation number; the basic allocation number is determined based on the minimum data transmission rate required by the test task, the number of bandwidths carried by each resource block, the transmission power, the channel gain, the distance between the terminal device and the base station, and the allocation frequency of the resource blocks; the supplementary allocation number is determined based on the total number of resource blocks included in the base station, the number of test tasks within a computing cycle, and the sum of the basic allocation numbers of all tasks to be executed within a computing cycle.

[0218] In some embodiments, the types of test tasks include at least one of the following: motion control tasks, machine - to - machine control tasks, and video surveillance tasks.

[0219] In some embodiments, the task deployment model includes: a first task deployment model for determining the deployment scheme of control tasks; the deployment scheme of control tasks is used to indicate the deployment of control tasks on local computing devices or edge computing devices; a second task deployment model for determining the deployment scheme of video surveillance tasks; the deployment scheme of video surveillance tasks is used to indicate the deployment of video surveillance tasks on local computing devices and / or edge computing devices.

[0220] In some embodiments, the task information of the test task includes at least one of the following: the task type of the test task, the task data packet of the test task, the data packet length of the test task, the delay threshold required by the test task, and the minimum data transmission rate required by the test task.

[0221] In some embodiments, the environmental information of the edge-cloud collaboration scenario includes at least one of the following: the maximum transmission power of each terminal device in the edge-cloud collaboration scenario, the data volume of the current test task in the edge-cloud collaboration scenario, the latency threshold required for the current test task in the edge-cloud collaboration scenario, the computing power of the local computing device in the edge-cloud collaboration scenario, the network channel state of the edge-cloud collaboration scenario, the number of resource blocks allocable by the base station in the edge-cloud collaboration scenario, and the total computing power of the edge computing devices in the edge-cloud collaboration scenario.

[0222] In some embodiments, as Figure 6 shown, after step S302, the above model training method further includes the following steps:

[0223] S303. Load the trained task deployment model and test the performance of the model.

[0224] Exemplarily, evaluate the performance of the model from the task execution success rate and the resource block utilization rate.

[0225] Among them, the test result of the execution success rate is as Figure 7 shown, and this test result presents the execution success rates of different numbers of tasks (N = 5, N = 10, and N = 15) and all tasks, machine-to-machine control tasks, and video surveillance tasks tested for different task types.

[0226] The test result of the resource block utilization rate is as Figure 8 shown. When the offloading rate of the test task is greater than 0, the resource block utilization rate of this application is approximately equal to 100%. Conversely, when the offloading rate of the task is 0, there is no task to be transmitted, and the resource block utilization rate is 0.

[0227] The above mainly introduces the solutions of the embodiments of the present disclosure from the perspective of the method. It can be understood that in order to implement the corresponding functions, the task deployment device and the model training device include at least one of the corresponding hardware structures and software modules for executing each function. Those skilled in the art should easily realize that, in combination with the algorithm steps of each example described in the embodiments disclosed in this article, the embodiments of the present disclosure can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the form of hardware or computer software driving the hardware depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the embodiments of the present disclosure.

[0228] Embodiments of the present disclosure may divide the task deployment device and the model training device into functional modules according to the above method embodiments. For example, each functional module can be divided corresponding to each function, or two or more functions can be integrated into one functional module. The above integrated module can be implemented in the form of hardware or software. It should be noted that the division of modules in the embodiments of the present disclosure is illustrative, only a logical function division, and there may be other division methods in actual implementation. The following takes the example of dividing each functional module corresponding to each function for illustration.

[0229] For example, Figure 9 is a schematic diagram of the composition of a task deployment device provided by an embodiment of the present application. As Figure 9 shown, the task deployment device 400 is applied to the edge-cloud collaboration scenario. The device includes: an acquisition module 401, a decision-making module 402, and a deployment module 403; the acquisition module 401 is used to acquire the task information of the to-be-executed task of the terminal device and the environmental information of the edge-cloud collaboration scenario; the to-be-executed task is any one of all the to-be-executed tasks in one computing cycle in the edge-cloud collaboration scenario; the decision-making module 402 is used to determine the deployment plan of the to-be-executed task based on the task information, environmental information, and the trained task deployment model; the deployment plan of the to-be-executed task is used to indicate the deployment of the to-be-executed task on the local computing device and / or the edge computing device, so as to maximize the use of resources when the processing delay of the to-be-executed task meets the delay threshold required by the to-be-executed task; the deployment module 403 is used to deploy the to-be-executed task based on the deployment plan.

[0230] In some embodiments, the task deployment model is trained based on the task information of the test task, the environmental information of the edge-cloud collaboration scenario before the execution of the test task, and the delay threshold required by the test task; wherein, the delay threshold required by the test task is used as supervision information.

[0231] In other embodiments, the types of the to-be-executed tasks include at least one of the following: motion control tasks, machine-to-machine control tasks, and video surveillance tasks.

[0232] In still other embodiments, the task deployment model includes: a first task deployment model, which is used to determine the deployment plan of the control task; the deployment plan of the control task is used to indicate the deployment of the control task on the local computing device or the edge computing device; a second task deployment model, which is used to determine the deployment plan of the video surveillance task; the deployment plan of the video surveillance task is used to indicate the deployment of the video surveillance task on the local computing device and / or the edge computing device.

[0233] In some other embodiments, the task information of the task to be executed includes at least one of the following: the task type of the task to be executed, the number of task data packets of the task to be executed, the data packet length of the task to be executed, the latency threshold required for the task to be executed, and the minimum data transmission rate required for the task to be executed.

[0234] In some other embodiments, the environmental information of the edge-cloud collaboration scenario includes at least one of the following: the maximum transmission power of each terminal device in the edge-cloud collaboration scenario, the data volume of the current task in the edge-cloud collaboration scenario, the latency threshold required for the current task in the edge-cloud collaboration scenario, the computing power of the local computing device in the edge-cloud collaboration scenario, the network channel state of the edge-cloud collaboration scenario, the number of resource blocks that can be allocated by the base station in the edge-cloud collaboration scenario, and the total computing power of the edge computing devices in the edge-cloud collaboration scenario.

[0235] In some other embodiments, the deployment plan further includes: the amount of task data deployed on the local computing device for the task to be executed and the amount of task data deployed on the edge computing device for the task to be executed.

[0236] In some other embodiments, the amount of task data deployed on the local computing device for the task to be executed is determined based on the deployment ratio of the task to be executed on the local computing device, the number of data packets of the task to be executed, and the data packet length of the task to be executed; the amount of task data deployed on the edge computing device for the task to be executed is determined based on the deployment ratio of the task to be executed on the edge computing device, the number of data packets of the task to be executed, and the data packet length of the task to be executed.

[0237] In some other embodiments, the amount of task data deployed on the local computing device for the task to be executed satisfies the following formula:

[0238]

[0239] Where is the amount of task data deployed on the local computing device for the task to be executed, is the deployment ratio of the task to be executed on the edge computing device, is the deployment ratio of the task to be executed on the local computing device, is the number of data packets of the task to be executed, is the data packet length of the task to be executed.

[0240] In another possible implementation, the amount of task data deployed on the edge computing device for the task to be executed satisfies the following formula:

[0241]

[0242] Where is the amount of task data deployed on the edge computing device for the task to be executed, is the deployment ratio of the task to be executed on the edge computing device, is the number of data packets of the task to be executed, is the data packet length of the task to be executed.

[0243] In some other embodiments, the processing delay of the task to be executed is determined based on at least one of the following: local computing delay, edge computing delay, and communication delay.

[0244] In some other embodiments, the local computing delay is determined based on the amount of task data of the task to be executed deployed on the local computing device, the computing complexity of the task to be executed, and the computing resources of the local computing device.

[0245] Another possible implementation is that the local computing delay satisfies the following formula:

[0246]

[0247] Wherein, is the local computing delay, is the amount of task data of the task to be executed deployed on the local computing device, C n is the computing complexity of the task to be executed, and f is the computing resources of the local computing device.

[0248] In some other embodiments, the edge computing delay is the difference between the delay threshold required by the task to be executed and the communication delay.

[0249] In some other embodiments, the communication delay is determined based on the following parameters: the amount of task data that the task to be executed needs to transmit, the number of resource blocks allocated for the task to be executed, the transmission power, the channel gain, and the distance between the terminal device and the base station.

[0250] Another possible implementation is that the communication delay satisfies the following formula:

[0251]

[0252] Wherein, represents the amount of task data that the task to be executed needs to transmit, represents the number of resource blocks allocated for the task to be executed, p max represents the transmission power, represents the channel gain, d n represents the distance between the terminal device and the base station, α represents the path loss coefficient, σ 2 represents the Gaussian white noise power.

[0253] In some other embodiments, the number of resource blocks allocated for the task to be executed includes: a basic allocation number and a supplementary allocation number; the basic allocation number is determined based on the minimum data transmission rate required by the task to be executed, the number of bandwidths carried by each resource block, the transmission power, the channel gain, the distance between the terminal device and the base station, and the allocation frequency of the resource blocks; the supplementary allocation number is determined based on the total number of resource blocks included in the base station, the number of tasks to be executed within a calculation period, and the sum of the basic allocation numbers of all tasks to be executed within a calculation period.

[0254] In another possible implementation, the basic allocation number satisfies the following formula:

[0255]

[0256] Wherein, represents the minimum data transmission rate required by the task to be executed, w represents the number of bandwidths carried by each resource block, p max represents the transmission power, represents the channel gain, d n represents the distance between the terminal device and the base station, α represents the path loss coefficient, σ 2 represents the Gaussian white noise power, and x is the allocation frequency of the resource blocks.

[0257] In another possible implementation, the supplementary allocation number satisfies the following formula:

[0258]

[0259] w rb represents the total number of resource blocks included in the base station, N represents the number of tasks to be executed within a calculation period, represents the sum of the basic allocation numbers of all tasks to be executed within a calculation period.

[0260] For example, Figure 10The following is a schematic diagram of the composition of a model training device provided by an embodiment of the present application. As shown in the figure, the model training device 500 includes: an acquisition module 501 and a processing module 502; the acquisition module 501 is used to acquire the task information of the test task, the environmental information of the edge-side collaboration scenario before the execution of the test task, and the latency threshold required by the test task; the test task is any one of all test tasks within a computing cycle in the edge-side collaboration scenario; the processing module 502 is used to train the task deployment model to be trained based on the task information of the test task, the environmental information of the edge-side collaboration scenario before the execution of the test task, and the latency threshold required by the test task, and obtain the trained task deployment model; the trained task deployment model is used to determine the task deployment plan for the task to be executed by the terminal device; wherein, the task deployment plan is used to indicate the deployment of the task to be executed on the local computing device and / or the edge computing device, so as to maximize the use of resources when the processing latency of the task to be executed meets the latency threshold required by the task to be executed.

[0261] In some embodiments, the processing module 502 is specifically configured to: input the task information of the test task and the environmental information of the edge-side collaboration scenario before the execution of the test task into the task deployment model to be trained, and obtain the deployment plan of the test task; based on the deployment plan of the test task, determine the processing latency of the test task; based on the processing latency of the test task and the latency threshold required by the test task, train the task deployment model to obtain the trained task deployment model.

[0262] In other embodiments, the processing module 502 is specifically configured to: set a reward function based on the processing latency of the test task and the latency threshold required by the test task; the reward function is used to give a positive reward when the processing latency of the test task is less than or equal to the latency threshold required by the test task, and give a negative reward when the processing latency of the test task is greater than the latency threshold required by the test task; based on the reward function, perform iterative training on the task deployment model to be trained until the reward function converges, and obtain the trained task deployment model.

[0263] Another possible implementation is that the reward function satisfies the following formula:

[0264]

[0265] Wherein, R represents the total reward value of the reward function, T represents the number of computing cycles, R t represents the reward value of all test tasks in the t-th computing cycle, N represents the number of test tasks, represents the reward value of the n-th test task in the t-th computing cycle, represents the processing latency of the test task, represents the latency threshold required by the test task.

[0266] In some other embodiments, the deployment plan further includes: the amount of task data deployed on the local computing device and the amount of task data deployed on the edge computing device for the test task.

[0267] In some other embodiments, the processing delay of the test task is determined based on at least one of the following: local computing delay, edge computing delay, and communication delay.

[0268] In some other embodiments, the local computing delay is determined based on the amount of task data of the test task deployed on the local computing device, the computing complexity of the test task, and the computing resources of the local computing device.

[0269] In some other embodiments, the edge computing delay is the difference between the delay threshold required by the test task and the communication delay.

[0270] In some other embodiments, the communication delay is determined based on the following parameters: the amount of task data that the test task needs to transmit, the number of resource blocks allocated for the test task, the transmission power, the channel gain, and the distance between the terminal device and the base station.

[0271] In some other embodiments, the number of resource blocks allocated for the test task includes: the basic allocation number and the supplementary allocation number; the basic allocation number is determined based on the minimum data transmission rate required by the test task, the number of bandwidths carried by each resource block, the transmission power, the channel gain, the distance between the terminal device and the base station, and the allocation frequency of the resource block; the supplementary allocation number is determined based on the total number of resource blocks included in the base station, the number of test tasks within a computing cycle, and the sum of the basic allocation numbers of all tasks to be executed within a computing cycle.

[0272] In some other embodiments, the types of test tasks include at least one of the following: motion control tasks, machine - to - machine control tasks, video surveillance tasks.

[0273] In some other embodiments, the task deployment model includes: a first task deployment model for determining the deployment plan of control tasks; the deployment plan of control tasks is used to indicate the deployment of control tasks on the local computing device or the edge computing device; a second task deployment model for determining the deployment plan of video surveillance tasks; the deployment plan of video surveillance tasks is used to indicate the deployment of video surveillance tasks on the local computing device and / or the edge computing device.

[0274] In some other embodiments, the task information of the test task includes at least one of the following: the task type of the test task, the task data packet of the test task, the data packet length of the test task, the delay threshold required by the test task, the minimum data transmission rate required by the test task.

[0275] In some other embodiments, the environmental information of the edge-cloud collaboration scenario includes at least one of the following: the maximum transmission power of each terminal device in the edge-cloud collaboration scenario, the data volume of the current test task in the edge-cloud collaboration scenario, the latency threshold required for the current test task in the edge-cloud collaboration scenario, the computing power of the local computing device in the edge-cloud collaboration scenario, the network channel state of the edge-cloud collaboration scenario, the number of resource blocks that can be allocated by the base station in the edge-cloud collaboration scenario, and the total computing power of the edge computing devices in the edge-cloud collaboration scenario.

[0276] In an exemplary embodiment, the embodiments of the present application further provide an electronic device, which may be the task deployment device or the model training device in the above method embodiments. Figure 11 The following is a schematic diagram of the composition of an electronic device provided by an embodiment of the present application. As Figure 11 shown, the electronic device may include: a processor 601 and a memory 602; the memory 602 stores instructions executable by the processor 601; when the processor 601 is configured to execute the instructions, the electronic device or the server or the management platform implements the method described in the foregoing method embodiments.

[0277] In an exemplary embodiment, the embodiments of the present application further provide a readable storage medium, on which program instructions are stored; when the program instructions are executed by an electronic device, the electronic device implements the method described in the foregoing embodiments. The readable storage medium may be a non-transitory readable storage medium. For example, the non-transitory readable storage medium may be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0278] In an exemplary embodiment, the embodiments of the present application further provide a computer program product, when the computer program product runs on a computer, the computer executes the above related method steps to implement a task deployment method in the above embodiments.

[0279] The above are only the specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A task deployment method, characterized in that: Applied to edge-end collaboration scenarios, the edge-end collaboration scenarios include: a terminal device, a local computing device connected to the terminal device, and an edge computing device; the local computing device and the edge computing device communicate with each other through a base station; the method includes: Acquire task information of a task to be executed on the terminal device and environmental information of the edge-end collaboration scenario; the task to be executed is any one of all tasks to be executed within a computing cycle in the edge-end collaboration scenario; Determine a deployment scheme for the task to be executed based on the task information, the environment information and the trained task deployment model; the deployment scheme for the task to be executed is used to indicate the deployment of the task to be executed on the local computing device and / or the edge computing device, so as to maximize the utilization of resources when the processing delay of the task to be executed meets the delay threshold required by the task to be executed; The tasks to be executed are deployed based on the deployment plan.

2. The method according to claim 1, characterized in that The task deployment model is trained based on the task information of the test task, the environmental information of the edge collaboration scenario before the test task is executed, and the delay threshold required by the test task; wherein the delay threshold required by the test task serves as supervision information.

3. The method according to claim 1, characterized in that The type of the task to be performed includes at least one of the following: Motion control tasks, machine control tasks, and remote video monitoring tasks.

4. The method according to claim 1, characterized in that: Task deployment models include: A first task deployment model, used to determine a deployment scheme for a control task; the deployment scheme for the control task is used to indicate the deployment of the control task on the local computing device or the edge computing device; The second task deployment model is used to determine the deployment plan of the video surveillance task; the deployment plan of the video surveillance task is used to indicate the deployment of the video surveillance task on the local computing device and / or the edge computing device.

5. The method according to claim 1, characterized in that The task information of the task to be executed includes at least one of the following: The task type of the task to be executed, the task data packet of the task to be executed, the data packet length of the task to be executed, the delay threshold required by the task to be executed, and the minimum data transmission rate required by the task to be executed.

6. The method according to claim 1, characterized in that The environment information of the edge collaboration scenario includes at least one of the following: The maximum transmission power of each terminal device in the edge collaboration scenario, the data volume of the current task in the edge collaboration scenario, the delay threshold required by the current task in the edge collaboration scenario, the computing power of the local computing device in the edge collaboration scenario, the network channel status of the edge collaboration scenario, the number of resource blocks that can be allocated by the base station in the edge collaboration scenario, and the total computing power of the edge computing devices in the edge collaboration scenario.

7. The method according to claim 1, characterized in that In the case where the deployment scheme of the task to be executed is used to indicate deployment of the task to be executed on the local computing device and the edge computing device, the deployment scheme also includes: the amount of task data of the task to be executed deployed on the local computing device and the amount of task data deployed on the edge computing device.

8. The method according to claim 7, characterized in that The amount of task data deployed on the local computing device for the task to be executed is determined based on the deployment ratio of the task to be executed on the local computing device, the number of data packets of the task to be executed, and the length of the data packets of the task to be executed; the amount of task data deployed on the edge computing device for the task to be executed is determined based on the deployment ratio of the task to be executed on the edge computing device, the number of data packets of the task to be executed, and the length of the data packets of the task to be executed.

9. The method according to claim 8, characterized in that The amount of task data deployed on the local computing device for the task to be executed satisfies the following formula: in, the amount of task data deployed on the local computing device for the task to be executed, is the deployment ratio of the tasks to be executed on the edge computing device, is the deployment ratio of the tasks to be executed on the local computing device, is the number of data packets of the task to be executed, is the length of the data packet of the task to be executed.

10. The method according to claim 8, characterized in that The amount of task data of the task to be executed deployed on the edge computing device satisfies the following formula: in, the amount of task data deployed on the edge computing device for the task to be executed, is the deployment ratio of the tasks to be executed on the edge computing device, is the number of data packets of the task to be executed, is the length of the data packet of the task to be executed.

11. The method according to claim 1, characterized in that: The processing delay of the task to be executed is determined based on at least one of the following: local computing delay, edge computing delay, and communication delay.

12. The method according to claim 11, characterized in that The local computing delay is determined based on the amount of task data of the task to be executed deployed on the local computing device, the computational complexity of the task to be executed, and the computing resources of the local computing device.

13. The method according to claim 12, characterized in that The local calculation delay satisfies the following formula: in, The local calculation delay is is the amount of task data deployed on the local computing device for the task to be executed, C n is the computational complexity of the task to be executed, and f is the computing resources of the local computing device.

14. The method according to claim 11, characterized in that The edge computing delay is the difference between the delay threshold required by the task to be executed and the communication delay.

15. The method according to claim 14, characterized in that The communication delay is determined based on the following parameters: The task to be executed is determined by the amount of task data that needs to be transmitted, the number of resource blocks allocated for the task to be executed, the transmission power, the channel gain, and the distance between the terminal device and the base station.

16. The method according to claim 15, characterized in that The task communication delay satisfies the following formula: in, Indicates the amount of task data that needs to be transmitted for the task to be executed, represents the number of resource blocks allocated to the task to be executed, p max represents the transmission power, represents the channel gain, d n represents the distance between the terminal device and the base station, α represents the path loss coefficient, σ 2 represents the Gaussian white noise power.

17. The method according to claim 15, characterized in that The number of resource blocks allocated for the task to be executed includes: a basic allocation number and a supplementary allocation number; The basic allocation number is determined based on the minimum data transmission rate required by the task to be performed, the number of bandwidths carried by each resource block, the transmission power, the channel gain, the distance between the terminal device and the base station, and the allocation frequency of the resource block; The supplementary allocation number is determined based on the total number of resource blocks included in the base station, the number of tasks to be executed in one calculation cycle, and the sum of the basic allocation numbers of all tasks to be executed in one calculation cycle.

18. The method according to claim 17, characterized in that The basic allocation number satisfies the following formula: in, represents the minimum data transmission rate required by the task to be executed, w represents the number of bandwidths carried by each resource block, and p max represents the transmission power, represents the channel gain, d n represents the distance between the terminal device and the base station, α represents the path loss coefficient, σ 2 represents the Gaussian white noise power, and x is the allocation frequency of the resource block.

19. The method according to claim 17, characterized in that The supplementary allocation number satisfies the following formula: w rb represents the total number of resource blocks included in the base station, N represents the number of tasks to be executed in one calculation cycle, It represents the sum of the basic allocation numbers of all tasks to be executed in the computing cycle.

20. A model training method, characterized in that: Applied to edge-end collaboration scenarios, the edge-end collaboration scenarios include: a terminal device, a local computing device connected to the terminal device, and an edge computing device; the local computing device and the edge computing device communicate with each other through a base station; the method includes: Obtaining task information of a test task, environmental information of the edge-end collaboration scenario before the test task is executed, and a delay threshold required by the test task; the test task is any one of all test tasks within a computing cycle in the edge-end collaboration scenario; Based on the task information of the test task, the environmental information of the edge collaboration scenario before the execution of the test task, and the delay threshold required by the test task, the task deployment model to be trained is trained to obtain a trained task deployment model; the trained task deployment model is used to determine the task deployment plan of the task to be executed on the terminal device; wherein the task deployment plan is used to indicate the deployment of the task to be executed on the local computing device and / or the edge computing device, so as to maximize the utilization of resources while the processing delay of the task to be executed meets the delay threshold required by the task to be executed.

21. The method according to claim 20, characterized in that Obtaining task information of the test task, environmental information of the edge collaboration scenario before the test task is executed, and the delay threshold required by the test task, training the task deployment model to be trained, and obtaining a trained task deployment model, including: Inputting the task information of the test task and the environment information of the edge collaboration scenario before the test task is executed into the task deployment model to be trained to obtain a deployment plan for the test task; Determining a processing delay of the test task based on a deployment scheme of the test task; Based on the processing delay of the test task and the delay threshold required by the test task, the task deployment model is trained to obtain the trained task deployment model.

22. The method according to claim 21, characterized in that The step of training the task deployment model based on the difference between the processing delay of the test task and the delay threshold required by the test task to obtain the trained task deployment model includes: A reward function is set based on the processing delay of the test task and the delay threshold required by the test task; the reward function is used to give a positive reward when the processing delay of the test task is less than or equal to the delay threshold required by the test task, and to give a negative reward when the processing delay of the test task is greater than the delay threshold required by the test task; Based on the reward function, the task deployment model to be trained is iteratively trained until the reward function converges, thereby obtaining the trained task deployment model.

23. The method according to claim 22, characterized in that The reward function satisfies the following formula: Where R represents the total reward value of the reward function, T represents the number of calculation cycles, and R t represents the reward value of all the test tasks in the tth calculation cycle, N represents the number of the test tasks, represents the reward value of the nth test task in the tth calculation cycle, represents the processing delay of the test task, Indicates the delay threshold required by the test task.

24. The method according to claim 21, characterized in that In the case where the deployment scheme of the test task is used to indicate deployment of the test task on the local computing device and the edge computing device, the deployment scheme also includes: the amount of task data of the test task deployed on the local computing device and the amount of task data deployed on the edge computing device.

25. The method according to claim 21, characterized in that The processing delay of the test task is determined based on at least one of the following: local computing delay, edge computing delay, and communication delay.

26. The method according to claim 25, characterized in that The local computing delay is determined based on the amount of task data of the test task deployed on the local computing device, the computational complexity of the test task, and the computing resources of the local computing device.

27. The method according to claim 25, characterized in that The edge computing delay is the difference between the delay threshold required by the test task and the task communication delay.

28. The method according to claim 25, characterized in that The communication delay is determined based on the following parameters: The test task is determined by the amount of task data that needs to be transmitted, the number of resource blocks allocated for the test task, the transmission power, the channel gain, and the distance between the terminal device and the base station.

29. The method according to claim 28, characterized in that The number of resource blocks allocated for the test task includes: a basic allocation number and a supplementary allocation number; The basic allocation number is determined based on the minimum data transmission rate required by the test task, the number of bandwidths carried by each resource block, the transmission power, the channel gain, the distance between the terminal device and the base station, and the allocation frequency of the resource block; The supplementary allocation number is determined based on the total number of resource blocks included in the base station, the number of test tasks in a calculation cycle, and the sum of the basic allocation numbers of all tasks to be executed in a calculation cycle.

30. The method according to claim 20, characterized in that The type of the test task includes at least one of the following: Motion control tasks, machine control tasks, and remote video monitoring tasks.

31. The method according to claim 20, characterized in that Task deployment models include: A first task deployment model, used to determine a deployment scheme for a control task; the deployment scheme for the control task is used to indicate the deployment of the control task on the local computing device or the edge computing device; The second task deployment model is used to determine the deployment plan of the video surveillance task; the deployment plan of the video surveillance task is used to indicate the deployment of the video surveillance task on the local computing device and / or the edge computing device.

32. The method according to claim 20, characterized in that The task information of the test task includes at least one of the following: The task type of the test task, the task data packet of the test task, the data packet length of the test task, the delay threshold required by the test task, and the minimum data transmission rate required by the test task.

33. The method according to claim 20, characterized in that The environment information of the edge collaboration scenario includes at least one of the following: The maximum transmission power of each terminal device in the edge collaboration scenario, the data volume of the current test task in the edge collaboration scenario, the delay threshold required by the current test task in the edge collaboration scenario, the computing power of the local computing device in the edge collaboration scenario, the network channel status of the edge collaboration scenario, the number of resource blocks that can be allocated by the base station in the edge collaboration scenario, and the total computing power of the edge computing devices in the edge collaboration scenario.

34. An electronic device, characterized in that: The electronic device comprises: a processor and a memory; The memory stores instructions executable by the processor; When the processor is configured to execute the instructions, the electronic device implements the method according to any one of claims 1 to 19, or the method according to any one of claims 20 to 33.

35. A readable storage medium, characterized in that: The readable storage medium includes: software instructions; When the software instructions are executed in an electronic device, the electronic device implements the method according to any one of claims 1 to 19, or the method according to any one of claims 20 to 33.

Images