AIot-based cloud network side end collaborative reasoning system

Abstract

The invention discloses an AI ot-based cloud network side end collaborative reasoning system, and relates to the technical field of collaborative reasoning, and the system comprises a data acquisition module, a task division module, a private cloud module, an edge cloud module, an edge end node module, a decision module, a scheme generation module and a display module. The system has the advantages that effective classification and arrangement of task information can be realized through the task division module, subsequent targeted processing of tasks of different scales is facilitated, and the management and scheduling capability of the system on diversified tasks is improved; by utilizing the private cloud module, the edge cloud module and the edge end node module, various edge AI intelligent decisions can be performed in real time aiming at different scenes, cloud edge end intelligent collaboration is realized, meanwhile, sharing and multiplexing of models are realized, repeated development is avoided, the resource utilization efficiency is improved, the burden of private cloud is relieved, and the fault-tolerant capability of the system is improved.

Description

Technical Field

[0001] The present invention relates to the field of collaborative reasoning technology, and specifically to an AIot cloud network edge collaborative reasoning system. Background Art

[0002] The AIoT cloud network edge project combines traditional IoT and cloud computing with modern artificial intelligence technology. Through intelligent collaborative computing of cloud-edge-end computing resources, IoT devices have higher intelligence, automation and intelligent decision-making capabilities, and can better perceive the environment and respond to user needs. It represents the development direction of the new generation of IoT technology. The AIoT cloud network edge project adopts advanced container cloud technology, supports heterogeneous hardware computing resources, and provides basic services such as unified management of private cloud and edge computing resources, service orchestration and load balancing.

[0003] When using common cloud-network-edge collaborative reasoning systems, it is difficult to achieve intelligent collaboration between the cloud, edge and the end, and it is difficult to share and reuse models. Computing resources are not fully utilized, resulting in resource waste and affecting the overall performance of the system. To this end, we propose a cloud-network-edge collaborative reasoning system based on AIot. Summary of the invention

[0004] The purpose of the present invention is to provide an AIot cloud network edge collaborative reasoning system.

[0005] In order to solve the problems raised in the above background technology, the present invention provides the following technical solutions: an AIot cloud network edge collaborative reasoning system, comprising a data acquisition module, a task division module, a private cloud module, an edge cloud module, an edge node module and a decision module;

[0006] The data collection module collects reasoning information and the content that the user needs to reason about, and then converts the content that the user needs to reason about into task information;

[0007] The task division module establishes division units and basic units. The basic units include single-layer folders, double-layer folders and triple-layer folders. Basic values ​​are set for single-layer folders and double-layer folders. The basic values ​​refer to the data scale. Then, task information is collected and the data scale of the task information is analyzed to obtain the task scale. Then, the task information is recorded in the corresponding folder according to the task scale of the task information, thereby completing the division of the task information and using the division unit to record the task information and task scale.

[0008] The division unit also calculates the total task size recorded in the set time period, and then calculates the average task value of the task size, and sets the division boundaries for different task sizes, where the division boundaries are and Then calculate the The partition value and The partition value of

[0009] Assume that the basic value of a single folder is J1, and the basic value of a double folder is J2, analyze the size of H1 and J2 and the size of H2 and J1, and obtain the double judgment criterion and the single judgment criterion;

[0010] When H1≥J2, J2 is used as the double judgment standard; when J2>H1, H1 is used as the double judgment standard;

[0011] When H2≥J1, J1 is used as the double judgment standard. When J1>H2, H2 is used as the double judgment standard.

[0012] In the private cloud module, users access IoT data, extract task information, and perform feature extraction on the task information. They generate an inference model based on the extracted features, then input the task information into the inference model, output the inference results, and transmit the generated inference model to the edge cloud module and the edge node module;

[0013] The edge cloud module is used to provide IoT device access and IoT data aggregation components, and also provides edge-side AI model inference services. The data processing capability of the edge cloud module is lower than that of the private cloud module. After receiving the inference model, it will synchronously analyze the data processing parameters of the edge cloud module, and then adjust the inference model according to the data processing parameters to obtain a modified model, and then extract the task information and output it to the modified model to output the modified result information;

[0014] The edge node module includes various embedded edge computing devices, built-in edge container runtime and edge AI reasoning framework, and has the offline reasoning capability of edge devices. However, the data processing capability of the edge node module is lower than that of the edge cloud module. After receiving the reasoning model, the data processing parameters of the edge node module are analyzed, and then the reasoning model is adjusted according to its data processing parameters to obtain the second-modification model, and then the task information is extracted and output to the second-modification model, and the second-modification result information is output;

[0015] The decision module extracts the inference result, the first modification result information and the second modification result information, analyzes the difference information between the inference result, the first modification result information and the second modification result information, and then modifies the inference result according to the difference information.

[0016] As a further solution of the present invention: it also includes a solution generation module and a display module;

[0017] A solution generation module extracts the reasoning results, the first modification result information and the second modification result information, and analyzes the method of realizing the reasoning results, the first modification result information and the second modification result information to obtain the implementation plan;

[0018] The display module extracts the implementation plan and then converts the implementation plan into three-dimensional display information and two-dimensional display information, wherein the three-dimensional display information is a three-dimensional stereoscopic graphic and the two-dimensional display information is a plane graphic.

[0019] As a further solution of the present invention: after recording the task information and task scale, the division unit in the task division module synchronously establishes a division cycle, wherein the division cycle is a time period, so that the division unit records the task information and task scale within the set time period, and when the task information and task scale exceed the set time period, the task information and task scale that exceed the set time period are deleted.

[0020] As a further solution of the present invention: when the division boundary in the task division module is determined, the user has the authority to edit the division boundary, and the division value is calculated using the division boundary. Let the number of task scales recorded in the division period be J, and let the average task value be Z 平均 , let the different task scales recorded in the division period be R G ;

[0021]

[0022] The average task value is calculated based on the above.

[0023] As a further solution of the present invention: after the average task value is calculated, set The partition value is H1, assuming The division value is H2;

[0024]

[0025] According to the above formula, we can calculate The partition value and The partition value of .

[0026] As a further solution of the present invention: After the double judgment standard and the single judgment standard are obtained, the double judgment standard is set to C 2 , let C be the first judgment criterion 1 , where C 1 >C 2 , let the task size be R G ;

[0027] When R G <C 2 When doing so, the task information corresponding to the task size is initially recorded in the triple folder;

[0028] When C 2 ≤R G <C 1 When doing so, the task information corresponding to the task scale is initially recorded in the secondary folder;

[0029] When R G ≥C 1 When doing so, the task information corresponding to the task scale is initially recorded in a folder.

[0030] As a further solution of the present invention: after completing the division of task information, the task division module will transmit a single folder to the private cloud module, the edge cloud module and the edge node module to obtain the reasoning result, the first change result information and the second change result information. At the same time, the reasoning result will be the main one, and then the first change result information and the second change result information will be used to verify the reasoning result. The double folder will be based on the first change result information, and the triple folder will be based on the second change result information.

[0031] As a further solution of the present invention: a main unit and a sub-unit are also established inside the display module. The user places the three-dimensional display information and the two-dimensional display information into the main unit and the sub-unit as needed, and then the information of the main unit and the sub-unit will be projected on the screen according to the set screen ratio. At the same time, the user has the authority to edit the screen ratio of the main unit and the sub-unit.

[0032] By adopting the above technical solution, compared with the prior art, the beneficial effects of the present invention are:

[0033] 1. The present invention can realize effective classification and organization of task information through the task division module, which is convenient for subsequent targeted processing of tasks of different scales, improves the system's management and scheduling capabilities for diversified tasks, and uses private cloud modules, edge cloud modules and edge node modules to make various edge AI intelligent decisions in real time for different scenarios, realize cloud-edge-end intelligent collaboration, and realize model sharing and reuse, avoiding repeated development, improving resource utilization efficiency, reducing the burden on private clouds, and improving the system's fault tolerance;

[0034] 2. The present invention can avoid the infinite accumulation of task information through the task division module, improve the operating efficiency and response speed of the system, can be adjusted according to actual needs and situations, better adapt to different task characteristics and processing requirements, help to more accurately allocate tasks to corresponding folders, realize effective classification and management of tasks, make the division of task scale more flexible and personalized, facilitate the subsequent corresponding processing of tasks in different folders, and improve the pertinence and efficiency of task processing;

[0035] 3. The present invention can make full use of the processing capabilities of different modules through the task division module, improve the accuracy and reliability of the reasoning results, and at the same time clarify the role and focus of different folders in the system. The display module can be used to customize the display content and layout according to its own focus and needs, so as to better meet the personalized needs of different users. The display ratio of each part of the content can be adjusted according to actual needs, highlighting key information and improving the effect and practicality of information display. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 Schematic diagram of the system flow in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The specific embodiments of the present invention will be further described below in conjunction with the accompanying drawings. It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention.

[0038] In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0039] Embodiment 1:

[0040] Therefore, in order to effectively solve the above problems, this application proposes an AIot cloud network edge collaborative reasoning system, as shown in the accompanying drawings of the specification. Figure 1 As shown, it includes a data acquisition module, a task division module, a private cloud module, an edge cloud module, an edge node module and a decision module;

[0041] The data collection module collects reasoning information and the content that the user needs to reason about, and then converts the content that the user needs to reason about into task information;

[0042] The task division module establishes division units and basic units. The basic units include single-layer folders, double-layer folders and triple-layer folders. Basic values ​​are set for single-layer folders and double-layer folders. The basic values ​​refer to the data scale. Then, task information is collected and the data scale of the task information is analyzed to obtain the task scale. Then, the task information is recorded in the corresponding folder according to the task scale of the task information, thereby completing the division of the task information and using the division unit to record the task information and task scale.

[0043] The division unit also calculates the total task size recorded in the set time period, and then calculates the average task value of the task size, and sets the division boundaries for different task sizes, where the division boundaries are and Then calculate the The partition value and The partition value of

[0044] Assume that the basic value of a single folder is J1, and the basic value of a double folder is J2, analyze the size of H1 and J2 and the size of H2 and J1, and obtain the double judgment criterion and the single judgment criterion;

[0045] When H1≥J2, J2 is used as the double judgment standard; when J2>H1, H1 is used as the double judgment standard;

[0046] When H2≥J1, J1 is used as the double judgment standard. When J1>H2, H2 is used as the double judgment standard.

[0047] In the private cloud module, users access IoT data, extract task information, and perform feature extraction on the task information. They generate an inference model based on the extracted features, then input the task information into the inference model, output the inference results, and transmit the generated inference model to the edge cloud module and the edge node module;

[0048] The edge cloud module is used to provide IoT device access and IoT data aggregation components, and also provides edge-side AI model inference services. The data processing capability of the edge cloud module is lower than that of the private cloud module. After receiving the inference model, it will synchronously analyze the data processing parameters of the edge cloud module, and then adjust the inference model according to the data processing parameters to obtain a modified model, and then extract the task information and output it to the modified model to output the modified result information;

[0049] The edge node module includes various embedded edge computing devices, built-in edge container runtime and edge AI reasoning framework, and has the offline reasoning capability of edge devices. However, the data processing capability of the edge node module is lower than that of the edge cloud module. After receiving the reasoning model, the data processing parameters of the edge node module are analyzed, and then the reasoning model is adjusted according to its data processing parameters to obtain the second-modification model, and then the task information is extracted and output to the second-modification model, and the second-modification result information is output;

[0050] The decision module extracts the inference result, the first modification result information and the second modification result information, analyzes the difference information between the inference result, the first modification result information and the second modification result information, and then modifies the inference result according to the difference information;

[0051] It also includes a solution generation module and a display module;

[0052] A solution generation module extracts the reasoning results, the first modification result information and the second modification result information, and analyzes the method of realizing the reasoning results, the first modification result information and the second modification result information to obtain the implementation plan;

[0053] A display module extracts the implementation plan and then converts the implementation plan into three-dimensional display information and two-dimensional display information, wherein the three-dimensional display information is a three-dimensional stereoscopic graphic and the two-dimensional display information is a plane graphic;

[0054] Specific workflow: collect task information, record it in the corresponding folder according to the task scale of the task information, use the division unit to record the task information and task scale, extract features of the task information, generate an inference model based on the extracted features, analyze the data processing parameters of the edge cloud module, and then adjust the inference model according to the data processing parameters to obtain a first-change model, analyze the data processing parameters of the edge node module, and then adjust the inference model according to its data processing parameters to obtain a second-change model, analyze the difference information between the inference results, the first-change result information, and the second-change result information, analyze the method of realizing the inference results, the first-change result information, and the second-change result information, and convert the implementation plan into three-dimensional display information and two-dimensional display information;

[0055] Furthermore, the task division module can realize effective classification and organization of task information, facilitate subsequent targeted processing of tasks of different scales, and improve the system's management and scheduling capabilities for diversified tasks. The private cloud module, edge cloud module and edge node module can be used to perform various edge AI intelligent decisions in real time for different scenarios, realize cloud-edge intelligent collaboration, and at the same time realize model sharing and reuse, avoid duplicate development, improve resource utilization efficiency, reduce the burden on private cloud, and improve the system's fault tolerance.

[0056] Embodiment 2:

[0057] On the basis of the first embodiment, as shown in Figure 1 of the specification, after recording the task information and the task scale, the division unit in the task division module synchronously establishes a division cycle, wherein the division cycle is a time cycle, so that the division unit records the task information and the task scale within the set time cycle, and when the task information and the task scale exceed the set time cycle, the task information and the task scale exceeding the set time cycle are deleted;

[0058] Introducing a priority strategy, for high-priority tasks, the division unit allocates a longer effective time period to them when recording, ensuring that important task information is retained in the system longer to meet the requirements of key tasks for data integrity and traceability;

[0059] When the division boundary in the task division module is determined, the user has the authority to edit the division boundary. The division value is calculated using the division boundary. Let the number of task scales recorded in the division period be J, and let the average task value be Z. 平均 , let the different task scales recorded in the division period be R G ;

[0060]

[0061] Based on the above, the average task value is calculated;

[0062] After the average task value is calculated, set The partition value is H1, assuming The division value is H2;

[0063]

[0064] According to the above formula, we can calculate The partition value and The partition value of

[0065] Specific workflow: Make the division unit record the task information and task scale within the set time period. When the task information and task scale exceed the set time period, delete the task information and task scale that exceed the set time period, calculate the average task value of the task scale, set the division boundary for different task scales, and use the division boundary to calculate the division value. When H1≥J2, J2 is used as the double judgment standard. When J2>H1, H1 is used as the double judgment standard. When H2≥J1, J1 is used as the double judgment standard. When J1>H2, H2 is used as the double judgment standard.

[0066] Furthermore, the task division module can avoid the infinite accumulation of task information, improve the system's operating efficiency and response speed, and be able to be adjusted according to actual needs and situations to better adapt to different task characteristics and processing requirements. It helps to more accurately assign tasks to corresponding folders, achieve effective classification and management of tasks, and make the division of task scale more flexible and personalized, which is convenient for subsequent processing of tasks in different folders, thereby improving the pertinence and efficiency of task processing.

[0067] Embodiment three:

[0068] Based on the second embodiment, as shown in the accompanying drawings of the specification Figure 1 As shown in the figure, after the double judgment standard and the single judgment standard are obtained, the double judgment standard is set to C 2 , let C be the first judgment criterion 1 , where C 1 >C 2 , let the task size be R G ;

[0069] When R G <C 2 When doing so, the task information corresponding to the task size is initially recorded in the triple folder;

[0070] When C2 ≤R G <C 1 When doing so, the task information corresponding to the task scale is initially recorded in the secondary folder;

[0071] When R G ≥C 1 When the task information corresponding to the task scale is recorded in a folder;

[0072] When processing tasks, the first-level folder can allocate some subtasks to the second-level folder or the third-level folder for parallel processing according to the real-time execution status of the tasks, making full use of the offline reasoning capability of the edge node module and the aggregation component advantages of the edge cloud module to improve the overall processing efficiency;

[0073] After completing the division of task information, the task division module will transfer the first folder to the private cloud module, the edge cloud module and the edge node module to obtain the reasoning results, the first change result information and the second change result information. At the same time, the reasoning results will be given priority, and then the first change result information and the second change result information will be used to verify the reasoning results. The second folder is based on the first change result information, and the third folder is based on the second change result information.

[0074] The display module also has a main unit and a sub-unit. The user places the 3D display information and the 2D display information into the main unit and the sub-unit as required, and then the information of the main unit and the sub-unit will be projected on the screen according to the set screen ratio. At the same time, the user has the authority to edit the screen ratio of the main unit and the sub-unit.

[0075] Support seamless connection with external display devices. Users can project display content onto external devices for wider sharing and display. When connecting to external devices, the display module automatically adjusts the output resolution and display ratio to adapt to different display device specifications.

[0076] Specific workflow: When R G <C 2 When the task information corresponding to the task scale is initially recorded in the triple folder, 2 ≤R G <C 1 When the task information corresponding to the task scale is initially recorded in the secondary folder, G ≥C 1When the task information corresponding to the task scale is initially recorded in the first folder, the reasoning result, the first modification result information and the second modification result information are obtained, and the reasoning result is mainly used, and then the reasoning result is verified using the first modification result information and the second modification result information, while the second folder is mainly based on the first modification result information, and the third folder is mainly based on the second modification result information. The user places the three-dimensional display information and the two-dimensional display information into the main unit and the sub-unit according to the needs, and then the information of the main unit and the sub-unit will be projected on the screen according to the set screen ratio;

[0077] Furthermore, the task division module can fully utilize the processing capabilities of different modules to improve the accuracy and reliability of reasoning results. At the same time, the role and focus of different folders in the system are clarified. The display module can be used to customize the display content and layout according to its own focus and needs to better meet the personalized needs of different users. The display ratio of each part of the content can be adjusted according to actual needs to highlight key information and improve the effect and practicality of information display.

[0078] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An AIot cloud network edge collaborative reasoning system, characterized by: It includes data acquisition module, task division module, private cloud module, edge cloud module, edge node module and decision-making module; The data collection module collects reasoning information and the content that the user needs to reason about, and then converts the content that the user needs to reason about into task information; The task division module establishes division units and basic units. The basic units include single-layer folders, double-layer folders and triple-layer folders. Basic values ​​are set for single-layer folders and double-layer folders. The basic values ​​refer to the data scale. Then, task information is collected and the data scale of the task information is analyzed to obtain the task scale. Then, the task information is recorded in the corresponding folder according to the task scale of the task information, thereby completing the division of the task information and using the division unit to record the task information and task scale. The division unit also calculates the total task size recorded in the set time period, and then calculates the average task value of the task size, and sets the division boundaries for different task sizes, where the division boundaries are and Then calculate the The partition value and The partition value of Assume that the basic value of a single folder is J1, and the basic value of a double folder is J2, analyze the size of H1 and J2 and the size of H2 and J1, and obtain the double judgment standard and the single judgment standard; When H1≥J2, J2 is used as the double judgment standard; when J2>H1, H1 is used as the double judgment standard; When H2≥J1, J1 is used as the double judgment standard. When J1>H2, H2 is used as the double judgment standard. In the private cloud module, users access IoT data, extract task information, and perform feature extraction on the task information. They generate an inference model based on the extracted features, then input the task information into the inference model, output the inference results, and transmit the generated inference model to the edge cloud module and the edge node module; The edge cloud module is used to provide IoT device access and IoT data aggregation components, and also provides edge-side AI model inference services. The data processing capability of the edge cloud module is lower than that of the private cloud module. After receiving the inference model, it will synchronously analyze the data processing parameters of the edge cloud module, and then adjust the inference model according to the data processing parameters to obtain a modified model, and then extract the task information and output it to the modified model to output the modified result information; The edge node module includes various embedded edge computing devices, built-in edge container runtime and edge AI reasoning framework, and has the offline reasoning capability of edge devices. However, the data processing capability of the edge node module is lower than that of the edge cloud module. After receiving the reasoning model, the data processing parameters of the edge node module are analyzed, and then the reasoning model is adjusted according to its data processing parameters to obtain the second-modification model, and then the task information is extracted and output to the second-modification model, and the second-modification result information is output; The decision module extracts the inference result, the first modification result information and the second modification result information, analyzes the difference information between the inference result, the first modification result information and the second modification result information, and then modifies the inference result according to the difference information.

2. According to claim 1, the AIot cloud network edge collaborative reasoning system is characterized by: It also includes a solution generation module and a display module; A solution generation module extracts the reasoning results, the first modification result information and the second modification result information, and analyzes the method of realizing the reasoning results, the first modification result information and the second modification result information to obtain the implementation plan; The display module extracts the implementation plan and then converts the implementation plan into three-dimensional display information and two-dimensional display information, wherein the three-dimensional display information is a three-dimensional stereoscopic graphic and the two-dimensional display information is a plane graphic.

3. According to claim 1, the AIot cloud network edge collaborative reasoning system is characterized by: After recording the task information and task scale, the division unit in the task division module synchronously establishes a division period, wherein the division period is a time period, so that the division unit records the task information and task scale within the set time period. When the task information and task scale exceed the set time period, the task information and task scale exceeding the set time period are deleted.

4. According to claim 3, the AIot cloud network edge collaborative reasoning system is characterized by: When the division boundary in the task division module is determined, the user has the authority to edit the division boundary, and the division value is calculated using the division boundary. Let the number of task scales recorded in the division period be J, and let the average task value be Z. 平均 , let the different task scales recorded in the division period be R G ; The average task value is calculated based on the above.

5. According to claim 4, the AIot cloud network edge collaborative reasoning system is characterized by: After the average task value is calculated, The partition value is H1, assuming The division value is H2; According to the above formula, we can calculate The partition value and The partition value of .

6. According to claim 5, the AIot cloud network edge collaborative reasoning system is characterized by: After the double judgment standard and the single judgment standard are obtained, the double judgment standard is set as C 2 , let C be the first judgment criterion 1 , where C 1 >C 2 , let the task size be R G ; When R G <C 2 When doing so, the task information corresponding to the task size is initially recorded in the triple folder; When C 2 ≤R G <C 1 When doing so, the task information corresponding to the task scale is initially recorded in the secondary folder; When R G ≥C 1 When doing so, the task information corresponding to the task scale is initially recorded in a folder.

7. The AIot cloud network edge collaborative reasoning system according to claim 6 is characterized by: After completing the division of task information, the task division module will transfer a single folder to the private cloud module, the edge cloud module and the edge node module to obtain the reasoning results, the first change result information and the second change result information. At the same time, the reasoning results will be the main focus, and then the first change result information and the second change result information will be used to verify the reasoning results. The double folder will be based on the first change result information, and the triple folder will be based on the second change result information.

8. According to claim 2, the AIot cloud network edge collaborative reasoning system is characterized by: The display module also has a main unit and a sub-unit inside. The user places the three-dimensional display information and the two-dimensional display information into the main unit and the sub-unit as needed, and then the information of the main unit and the sub-unit will be projected on the screen according to the set screen ratio. At the same time, the user has the authority to edit the screen ratio of the main unit and the sub-unit.

Images