Internet of Things data distributed processing system and method based on deep learning

Abstract

The invention discloses an internet of things data distributed processing system and method based on deep learning, and belongs to the field of internet of things, the internet of things data distributed processing system comprises a plurality of acquisition modules, a plurality of edge computing platforms and a cloud computing platform, micro models are arranged in the acquisition modules, and the acquisition modules are used for acquiring data and processing the acquired data; the processed result is transmitted and stored; a small model is arranged in the edge computing platform, and the edge computing platform is used for receiving the data transmitted by the acquisition module, processing the transmitted data, transmitting and storing the processed data and forwarding the processed data to a user; the cloud computing platform is respectively connected with the plurality of edge computing platforms, a large model is arranged in the cloud computing platform, and the cloud computing platform is used for receiving data transmitted by the edge computing platforms, processing the received data, storing a processed result and forwarding the processed result to a user. The data processing speed can be increased, and the bottleneck problem caused by centralized processing can be avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things, and in particular to a deep learning-based distributed processing system and method for Internet of Things data. Background Art

[0002] With the continuous development of advanced sensing, wireless communication and other technologies, the Internet of Things not only connects people and things, but also expands to the connection between people and people, and things and things. In order to obtain more and more complete information about the perceived objects, the number of sensors in the Internet of Things system has exploded, which means that the amount of data collected has exploded. How to efficiently store, process and access large amounts of data has become one of the important issues that must be solved.

[0003] At present, the Internet of Things system mainly uses cloud computing platforms to realize the aggregation and centralized processing of various types of data. However, the centralized processing of Internet of Things data using cloud computing platforms has the following problems: 1) The data transmission link is long, and sensor data usually needs to be forwarded multiple times before it can be transmitted back to the cloud computing platform; 2) The data processing cycle is long. The cloud computing platform processes the data after receiving it. Especially when multiple data sources are involved, the asynchronous arrival of different data will further extend the data processing time; 3) Data transmission bottleneck is obvious. Data collected by sensors and user data access must go through the cloud computing platform, which can easily cause data transmission and access bottlenecks. Summary of the invention

[0004] In order to solve some or all of the technical problems existing in the above-mentioned prior art, the present invention provides a distributed processing system and method for Internet of Things data based on deep learning, which can efficiently store, process and access large amounts of data collected by sensors, effectively improve the data processing speed, and avoid the access bottleneck problem caused by centralized processing.

[0005] The technical solution of the present invention is as follows: In a first aspect, the present invention discloses a distributed processing system for Internet of Things data based on deep learning, comprising: A plurality of acquisition modules, wherein the plurality of acquisition modules are used to acquire data, a micromodel is provided in each of the acquisition modules, and the acquisition module is used to process the acquired data using the micromodel, and transmit and store the processed results; A plurality of edge computing platforms, each of which is connected to a plurality of the acquisition modules, wherein a small model is provided in each edge computing platform, and the edge computing platform is used to receive the result data transmitted by the acquisition module, and use the small model to process the transmitted result data, transmit and store the processed data, and forward it to the user; A cloud computing platform, wherein the cloud computing platform is connected to a plurality of edge computing platforms respectively, a large model is arranged in the cloud computing platform, the cloud computing platform is used to receive data transmitted by the edge computing platform, and use the large model to process the received data, store the processed results, and forward them to the user.

[0006] Furthermore, in the above-mentioned deep learning-based distributed processing system for IoT data, the acquisition module includes a sensor.

[0007] Furthermore, in the above-mentioned deep learning-based distributed processing system for IoT data, the micromodel, the small model and the large model are all trained in the cloud computing platform, and after the training is completed, the cloud computing platform directly calls the large model, and at the same time, the cloud computing platform sends the small model and the micromodel to the edge computing platform and the acquisition module respectively for model updating and data processing.

[0008] Furthermore, in the above-mentioned deep learning-based distributed IoT data processing system, the original data required for training the micro-model, the small model and the large model are all read from the edge computing platform and the acquisition module.

[0009] Furthermore, in the above-mentioned distributed processing system for IoT data based on deep learning, the original data required for training the micro-model, the small model and the large model are read from the edge computing platform and the acquisition module in the following manner: Full reading, wherein the full reading includes the edge computing platform and the acquisition module transmitting all data to the cloud computing platform for training the micro model, the small model and the large model; Sampling reading, wherein the sampling reading includes the edge computing platform and the acquisition module transmitting part of the data to the cloud computing platform at a preset time frequency; Event reading, the event reading includes the edge computing platform and the acquisition module only sending preset data to the cloud computing platform.

[0010] Furthermore, in the above-mentioned deep learning-based distributed processing system for IoT data, the preset data includes alarm data and state change data collected and acquired by the collection module.

[0011] Furthermore, in the above-mentioned deep learning-based distributed IoT data processing system, the acquisition module and the edge computing platform are used for short-term storage of the collected or processed data, and the cloud computing platform is used for long-term storage of the processed data.

[0012] In a second aspect, the present invention further discloses a deep learning-based distributed processing method for IoT data using the deep learning-based distributed processing system for IoT data, comprising: The acquisition module collects data, processes the data locally using the micro-model, and transmits the processed result data to the edge computing platform; The edge computing platform receives the result data processed by the acquisition module, processes the data using the small model, and directly sends the data processing results to the user and the cloud computing platform; After receiving the data processing results from the edge computing platform, the cloud computing platform processes the data using a large model and sends the processing results to the user.

[0013] The main advantages of the technical solution of the present invention are as follows: The present invention provides a distributed data processing system and method for the Internet of Things based on deep learning. The distributed data processing system and method utilize various computing resources in the Internet of Things system to process the data collected by sensors nearby and quickly provide services to users, which can effectively improve the data processing speed and avoid the access bottleneck problem caused by the centralized processing method. At the same time, based on a distributed data processing architecture of several acquisition modules, multiple edge computing platforms and a cloud computing platform, a micro-model is set in the acquisition module in the distributed data processing architecture, a small model is set in the edge computing platform, and a large model is set in the cloud computing platform, which improves the data processing level of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The drawings described herein are used to provide a further understanding of the embodiments of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 A schematic diagram of the structure of a deep learning-based distributed processing system for Internet of Things data provided by one embodiment of the present invention; Figure 2 A schematic diagram of a distributed processing system for Internet of Things data based on deep learning provided by one embodiment of the present invention; Figure 3 A flowchart of a method for distributed processing of Internet of Things data based on deep learning is provided in accordance with an embodiment of the present invention.

[0015] Description of reference numerals: 1. Acquisition module; 2. Edge computing platform; 3. Cloud computing platform. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0017] The technical solution provided by the embodiments of the present invention is described in detail below with reference to the accompanying drawings.

[0018] As attached Figure 1-2 As shown, an embodiment of the present invention provides an IoT data distributed processing system based on deep learning, the system comprising a plurality of acquisition modules 1, a plurality of edge computing platforms 2 and a cloud computing platform 3, wherein: Several acquisition modules 1 are used to collect data, and a micro model is arranged in each acquisition module 1. The acquisition module 1 is used to process the collected data using the micro model, and transmit and store the processed results; multiple edge computing platforms 2 are respectively connected to the several acquisition modules 1, and a small model is arranged in each edge computing platform 2. The edge computing platform 2 is used to receive the result data transmitted by the acquisition module 1, and process the transmitted result data using the small model, transmit and store the processed data, and forward it to the user; a cloud computing platform 3 is respectively connected to the multiple edge computing platforms 2, and a large model is arranged in the cloud computing platform 3. The cloud computing platform 3 is used to receive the data transmitted by the edge computing platform 2, and process the received data using the large model, store the processed results, and forward it to the user.

[0019] It should be noted that the data processing of the above-mentioned micromodel, small model and large model may be the same or different. They may be different processing methods for the same data, or the same processing method for different data. The data acquired by the acquisition module 1 and its type are not limited in the present invention, and only the amount of collected data is calculated.

[0020] The present invention discloses a distributed processing system for Internet of Things data based on deep learning. The system processes different amounts of data in a multi-level manner through a hierarchical architecture of an acquisition module 1, an edge computing platform 2 and a cloud computing platform 3. The edge computing platform 2 and the cloud computing platform 3 forward the data to the user after processing the data. This can effectively improve the computing resource utilization efficiency and data processing speed in the entire Internet of Things system, avoid the bottleneck problem of the centralized processing method, and at the same time, store the data for a short period of time through the acquisition module 1 and the edge computing platform 2, and store the data for a long period of time through the cloud computing platform 3, thereby reducing the occupation of storage resources by a large amount of Internet of Things data. By setting a micro-model in the acquisition module 1, a small model in the edge computing platform 2, and a large model in the cloud computing platform 3, the intelligence level of data processing platforms at all levels in the Internet of Things system is improved.

[0021] In combination with practical applications, in some optional implementations of this embodiment, the edge computing platform 2 includes a communication device for receiving sensor data, such as a gateway, a base station or a router.

[0022] At the same time, in another optional implementation of this embodiment, in order to facilitate users to access the data processed by the edge computing platform 2, the edge computing platform 2 is connected to the Internet.

[0023] In some optional implementations of this embodiment, the acquisition module 1 includes a sensor. Therefore, during the data transmission process, the sensor can send the data to the edge computing platform 2 in "one hop".

[0024] Specifically, the above-mentioned "one hop" means that the distance from the place where the data is generated to the place where the data is processed is only one hop, or only one forwarding is required, without multiple forwardings. In the present invention, the transmitted data is directly transmitted from the acquisition module 1 to the edge computing platform 2, thereby improving the data transmission efficiency.

[0025] In some optional implementations of this embodiment, in order to efficiently store, process and access large amounts of data collected by sensors, effectively improve the data processing speed, and avoid the access bottleneck problem caused by centralized processing, the above-mentioned acquisition module 1, edge computing platform 2 and cloud computing platform 3, when processing data, respectively call the micro model set in the acquisition module 1, the small model set in the edge computing platform 2 and the large model set in the cloud computing platform 3 to process different amounts of data.

[0026] It should be noted that the large model, small model and micro model here refer to the amount of data processed by the model and the relative size of the model's functions; the large model has the largest amount of data processing and more comprehensive data processing functions; the micro model has the smallest amount of data processing and a relatively simple data processing function; the small model has a data processing amount between the large and small models and has a relatively comprehensive data processing function. In addition, based on the amount of data that the model can process, the above-mentioned large models can be divided into deep learning models that process at least tens of millions or tens of millions to hundreds of millions of data, small models include deep learning models that process tens of thousands to millions of data, and micro models include deep learning models that process tens of thousands or less of data, such as neural network models.

[0027] Specifically, in order to improve the performance of the trained deep learning model and further improve the processing speed of data at all levels in the Internet of Things system, the micro model, small model and large model are all trained in the cloud computing platform 3. After the training is completed, the cloud computing platform 3 directly calls the large model, and the cloud computing platform 3 sends the small model and micro model to the edge computing platform 2 and the acquisition module 1 respectively for model updating and data processing.

[0028] Specifically, in order to further guarantee the performance of the trained deep learning model, the original data required for the training of the micro model, small model and large model are read from the edge computing platform 2 and the acquisition module 1. And in order to ensure the accuracy and validity of the read data, in the present invention, the original data required for the training of the micro model, small model and large model are read from the edge computing platform 2 and the acquisition module 1 in the following three ways, including full reading, sampling reading and event reading, among which: Full reading includes the edge computing platform 2 and the acquisition module 1 transmitting all data to the cloud computing platform 3 for the training of micro models, small models and large models; sampling reading includes the edge computing platform 2 and the acquisition module 1 transmitting part of the data to the cloud computing platform 3 according to the preset time frequency; event reading includes the edge computing platform 2 and the acquisition module 1 only sending preset data to the cloud computing platform 3.

[0029] In combination with practical applications, it should also be noted that the network bandwidth occupied by the above-mentioned three methods of full reading, sampling reading and event reading decreases successively, and the performance of the trained model also decreases successively. Therefore, the specific data reading method is determined by the available network resources in the entire Internet of Things system and the reliability requirements of the data processing results. After the model is trained, the model in the edge computing platform 2 and the acquisition module 1 is updated.

[0030] It is understandable that the above-mentioned available network resources include network bandwidth between nodes such as sensors, edge computing nodes, and cloud computing centers.

[0031] Therefore, the three deep learning model training data reading methods in the present invention can reduce the amount of network resources occupied by original data transmission while ensuring the overall performance of the system.

[0032] Specifically, the preset data includes the alarm data and state change data collected by the collection module 1 .

[0033] Specifically, since the storage space of the acquisition module 1 and the edge computing platform 2 is limited and only a certain amount of data is stored for a short period of time, in order to make the storage method of the collected data more reasonable and avoid loss, the acquisition module 1 and the edge computing platform 2 are used to store the collected or processed data for a short period of time, and the data that needs to be stored for a long time in the entire Internet of Things system is transmitted to the cloud computing platform 3 for storage.

[0034] With such configuration, the IoT data storage method proposed in the present invention can reduce the storage resource occupation by a large amount of IoT low-value data.

[0035] Specifically, the processing flow and processing principle of a deep learning-based distributed processing system for Internet of Things data of the present invention include: (1) After the sensor collects data, it processes the data locally using a deep learning micro-model and sends the data processing results to the edge computing platform 2.

[0036] As an example, for example, a water level sensor monitors the water level in real time and sends an alarm message to the edge computing platform 2 when the water level exceeds the set threshold. Here, the task of the micromodel is to convert the water volume information into water level data, and at the same time determine whether the water level exceeds the set threshold. When the threshold is exceeded, the data sending function is activated to send an alarm message to the outside.

[0037] (2) The edge computing platform 2 receives data from multiple sensors, uses a small model to fuse the data, and sends the data processing results directly to the user and the cloud computing platform 3.

[0038] As an example, for example: the edge computing platform 2 combines the data from sensors such as visible light cameras, infrared cameras, microphones, etc. to determine whether a person has entered a specific area.

[0039] (3) After receiving data from multiple edge computing platforms 2, the cloud computing platform 3 processes the data using a large model and sends the processing results to the user.

[0040] In summary, the distributed processing system of IoT data based on deep learning of the present invention utilizes various computing resources dispersed in the IoT system to process the data collected by sensors nearby and quickly provide services to users, which can effectively improve the data processing speed and avoid the access bottleneck problem caused by the centralized processing method. At the same time, by setting the architecture of micro-model, small model and large model in the acquisition module 1, edge computing platform 2 and cloud computing platform 3 respectively, the data processing level of the system is improved.

[0041] It can be understood that the above-mentioned types of computing resources can be any computing resources in sensors, gateway devices, and edge computing devices. Local processing means that the data to be processed is processed accordingly in the computing resources closest to the data generation, thereby reducing the number of data forwarding and reading times, and further improving data processing efficiency.

[0042] In a second aspect, the present invention further discloses a method for distributed processing of IoT data based on deep learning using the above-mentioned distributed processing system for IoT data based on deep learning, such as Figure 3 As shown, including: The acquisition module 1 collects data, processes the data locally using a micro model, and transmits the processed result data to the edge computing platform 2; the edge computing platform 2 receives the result data processed by the acquisition module 1, processes the data using a small model, and sends the data processing results directly to the user and the cloud computing platform 3; after receiving the data processing results from the edge computing platform 2, the cloud computing platform 3 processes the data using a large model and sends the processing results to the user.

[0043] Therefore, the distributed processing method of IoT data based on deep learning of the present invention utilizes the acquisition module 1, the edge computing platform 2, and the cloud computing platform 3 to construct a distributed multi-level IoT data processing architecture, which can effectively improve the computing resource utilization efficiency in the entire IoT system, improve the data processing speed, and avoid the bottleneck problem of the centralized processing method; by setting micro-models, small models, and large models at different levels for data processing, the intelligence level of data processing platforms at all levels in the IoT system is improved.

[0044] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In addition, "front", "back", "left", "right", "upper" and "lower" in this article are all referenced to the placement state shown in the accompanying drawings.

[0045] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A distributed processing system for Internet of Things data based on deep learning, characterized in that: include: A plurality of acquisition modules, wherein the plurality of acquisition modules are used to acquire data, a micromodel is provided in each of the acquisition modules, and the acquisition module is used to process the acquired data using the micromodel, and transmit and store the processed results; A plurality of edge computing platforms, each of which is connected to a plurality of the acquisition modules, wherein a small model is provided in each edge computing platform, and the edge computing platform is used to receive the result data transmitted by the acquisition module, and use the small model to process the transmitted result data, transmit and store the processed data, and forward it to the user; A cloud computing platform, wherein the cloud computing platform is connected to a plurality of edge computing platforms respectively, a large model is arranged in the cloud computing platform, the cloud computing platform is used to receive data transmitted by the edge computing platform, and use the large model to process the received data, store the processed results, and forward them to the user.

2. According to claim 1, a distributed processing system for Internet of Things data based on deep learning is characterized in that: The acquisition module includes a sensor.

3. According to the deep learning-based distributed processing system for Internet of Things data according to claim 1, it is characterized in that: The micro-model, the small model and the large model are all trained in the cloud computing platform, and after the training is completed, the cloud computing platform directly calls the large model. At the same time, the cloud computing platform sends the small model and the micro-model to the edge computing platform and the acquisition module respectively for model updating and data processing.

4. According to claim 3, a distributed processing system for Internet of Things data based on deep learning is characterized in that: The original data required for training the micro model, the small model and the large model are all read from the edge computing platform and the acquisition module.

5. According to claim 4, a distributed processing system for Internet of Things data based on deep learning is characterized in that: The original data required for training the micro-model, the small model and the large model are read from the edge computing platform and the acquisition module in the following manner: Full reading, wherein the full reading includes the edge computing platform and the acquisition module transmitting all data to the cloud computing platform for training the micro model, the small model and the large model; Sampling reading, wherein the sampling reading includes the edge computing platform and the acquisition module transmitting part of the data to the cloud computing platform at a preset time frequency; Event reading, the event reading includes the edge computing platform and the acquisition module only sending preset data to the cloud computing platform.

6. The Internet of Things data distributed processing system based on deep learning according to claim 5 is characterized in that: The preset data includes the alarm data and state change data collected and acquired by the collection module.

7. The IoT data distributed processing system based on deep learning according to claim 1 is characterized in that: The acquisition module and the edge computing platform are used for short-term storage of the collected or processed data, and the cloud computing platform is used for long-term storage of the processed data.

8. A method for distributed processing of Internet of Things data based on deep learning using the distributed processing system for Internet of Things data based on deep learning as claimed in any one of claims 1 to 7, characterized in that: include: The acquisition module collects data, processes the data locally using the micro-model, and transmits the processed result data to the edge computing platform; The edge computing platform receives the result data processed by the acquisition module, processes the data using the small model, and directly sends the data processing results to the user and the cloud computing platform; After receiving the data processing results from the edge computing platform, the cloud computing platform processes the data using a large model and sends the processing results to the user.

Images