KR20210158558A - Edge computing system using ai block-type module architecture - Google Patents

Abstract

The present invention relates to an edge computing system using an artificial intelligence block-type module architecture. The edge computing system comprises: a central node block, as an edge processing node of a blocking module, including processing and application services, training a model by using training data connected in a wired or wireless manner, and driving an edge computer or a plurality of IoT devices connected with an artificial intelligence algorithm corresponding to the application service, by using the learned model; an algorithm node block including a plurality of artificial intelligence algorithm nodes configured in block form by modularizing various machine learning algorithms, and for providing the artificial intelligence algorithm to the central node block; a data node block for distributing and storing the training data and the learned model in a hardware node connected in a wired manner or a plurality of artificial intelligence servers connected wirelessly, training the model in real time by using the training data, and providing the training data and the trained model to the central node block; an edge node block including the edge computer and operated according to the application service linked to the processing; and a leaf node block including a plurality of IoT devices and operated according to the application service linked to the processing. Therefore, the edge computing system can facilitate application of a combination of various edge computing devices, perform the fast communication between block-type modules by using a protocol based on normalized metadata, and support one-way or two-way communication between block-type modules.

Description

Edge computing system using AI block-type module architecture {EDGE COMPUTING SYSTEM USING AI BLOCK-TYPE MODULE ARCHITECTURE}

The present invention includes a block-type module architecture equipped with various machine learning algorithms, but is expandable to allow free addition and deletion between block-type modules, unlimited combinations between blocks are possible, and various machine learning algorithms can be tailored to the situation By designing the block module to be organically interlocked, it is easy to apply a combination of various edge computing devices, and it is possible to perform fast communication between block-type modules using a protocol based on normalized metadata, and between block-type modules. It relates to an edge computing system using an artificial intelligence block-type module architecture that can support one-way or two-way communication.

As is well known, machine learning technology in an edge computing environment has recently received a lot of attention from academia and industry, and various edge computing based machine learning applications are being developed.

Such edge computing provides Internet of things (IoT) services by performing processing at the edge of a network using collected data. In particular, mobile edge computing provides cloud computing services to mobile users, and machines in edge computing environments. Learning technology utilizes a large amount of data to create a model that can respond to an input that requires prediction.

Meanwhile, the increase in big data is generating new demands for machine learning systems for training complex models with large data sets and billions of parameters. There is a problem in that it is difficult to develop accurate machine learning applications with machine learning devices, and to solve this problem, a distributed machine learning technology has been developed, which utilizes multiple servers to run models that are slow or impossible to learn on one device. to make it possible

In addition, machine learning in edge computing is different from that in cloud computing environments, where the cloud in cloud computing performs central control over the entire system as a storage device or management center, but edge servers in edge computing are It can respond appropriately to service demands, and distributed machine learning technology in edge computing environments enables the learning of complex models of IoT devices.

In addition, various types of IoT devices are proposed as a technique for performing machine learning using a large amount of data and a trained model, and approaches for generating machine learning blocks, such as edge nodes of edge computing, are proposed. A real-time ad-hoc block is created through a machine learning framework to create a learning block that can be transmitted and received, or a reusable software module that uses various high-level technologies in the modeling stage.

However, conventional approaches have problems in organic utilization of edge nodes linked with IoT devices in a mobile environment. This possible organic linkage is insufficient, and there are problems that the edge node is not equipped with various types of machine learning algorithms and the diversity of combining them is insufficient.

Accordingly, research and development of an architecture that can perform machine learning by linking various machine learning algorithms with IoT devices in a distributed edge computing environment is required, and research and development of an edge computing system using this architecture is required. to be.

1. Korean Patent No. 10-2042690 (Registered on 2019.11.04.)

The present invention includes a block-type module architecture equipped with various machine learning algorithms, but is expandable to allow free addition and deletion between block-type modules, unlimited combinations between blocks are possible, and various machine learning algorithms can be adapted to the situation By designing the block module to be organically interlocked, it is easy to apply a combination of various edge computing devices, and it is possible to perform fast communication between block-type modules using a protocol based on normalized metadata, and between block-type modules. It is intended to provide an edge computing system using an AI block-type module architecture that can support unidirectional or bidirectional communication.

In addition, the present invention trains a model by using learning data connected by wire or wirelessly through communication with a data node block in the central node block, which is an edge processing node of the blocking module, and is an application linked to processing using the learned model. When processing a service, by operating an edge computer or multiple IoT devices connected with the corresponding AI algorithm provided from the AI node block, machine learning is performed by linking various machine learning algorithms with IoT devices in a distributed edge computing environment. We want to provide an edge computing system using an AI block-type module architecture that can do this.

The purpose of the embodiments of the present invention is not limited to the above-mentioned purpose, and other objects not mentioned will be clearly understood by those of ordinary skill in the art to which the present invention belongs from the following description. .

According to an embodiment of the present invention, the edge processing node of the block module includes processing and application services, uses wired or wirelessly connected learning data to train a model, and uses the learned model to respond to the application service A central node block that drives an edge computer or a plurality of IoT devices connected with an artificial intelligence algorithm that An algorithm node block that provides the artificial intelligence algorithm to the block, and the learning data and the learned model are distributed and stored in a hardware node connected by wire or a plurality of artificial intelligence servers connected wirelessly, and the learning data a data node block that trains the model in real time using An artificial intelligence block including an edge node block operated according to a service and a leaf node block including the plurality of IoT devices and operated according to the application service linked to the processing An edge computing system using a modular architecture may be provided.

In addition, according to an embodiment of the present invention, the edge computing system is an edge computing system using an artificial intelligence block-type module architecture in which the edge node block serves as another central node block including other processing and other application services. can be provided.

In addition, according to an embodiment of the present invention, the edge computing system defines and uses a metadata query protocol, wherein the metadata query protocol communicates between node blocks using a network communication protocol through normalized metadata. An edge computing system using an artificial intelligence block-type module architecture may be provided.

In addition, according to an embodiment of the present invention, the normalized metadata is provided according to the context format of 5W1H (Who, What, Where, When, Why, How) representing semantic information using an artificial intelligence block-type module architecture. An edge computing system may be provided.

In addition, according to an embodiment of the present invention, the normalized metadata is a node block in the form of a query using at least one network protocol communication of message queuing telemetry transport (MQTT), hypertext transfer protocol (HTTP), and serial communication. An edge computing system using an artificial intelligence block-type module architecture that is transmitted to each other can be provided.

In addition, according to an embodiment of the present invention, the edge computing system may be provided with an edge computing system using an artificial intelligence block-type module architecture that interworks between node blocks through one-way communication and two-way communication.

In addition, according to an embodiment of the present invention, the one-way communication may be provided with an edge computing system using an artificial intelligence block-type module architecture used in an input node and an output node.

In addition, according to an embodiment of the present invention, the bidirectional communication may be provided with an edge computing system using an artificial intelligence block-type module architecture used in a processing node and a data node.

In addition, according to an embodiment of the present invention, the edge computing system may provide an edge computing system using an artificial intelligence block-type module architecture that interworks between node blocks through wired serial communication or wireless block communication.

The present invention includes a block-type module architecture equipped with various machine learning algorithms, but is expandable to allow free addition and deletion between block-type modules, unlimited combinations between blocks are possible, and various machine learning algorithms can be adapted to the situation By designing the block module to be organically interlocked, it is easy to apply a combination of various edge computing devices, and it is possible to perform fast communication between block-type modules using a protocol based on normalized metadata, and between block-type modules. It can support one-way or two-way communication.

In addition, the present invention trains a model by using learning data connected by wire or wirelessly through communication with a data node block in the central node block, which is an edge processing node of the blocking module, and is an application linked to processing using the learned model. When processing a service, by operating an edge computer or multiple IoT devices connected with the corresponding AI algorithm provided from the AI node block, machine learning is performed by linking various machine learning algorithms with IoT devices in a distributed edge computing environment. can do.

1 is a block diagram of an edge computing system using an artificial intelligence block-type module architecture according to an embodiment of the present invention;
2 to 7 are diagrams for explaining an edge computing system using an artificial intelligence block-type module architecture according to an embodiment of the present invention,
8 to 10 are diagrams for explaining a modular comparison analysis and implementation example of an edge computing system using an artificial intelligence block-type module architecture according to an embodiment of the present invention;
11 to 14 are diagrams for explaining wired serial communication and wireless block communication in an edge computing system using an artificial intelligence block-type module architecture according to an embodiment of the present invention.

Advantages and features of embodiments of the present invention, and methods of achieving them, will become apparent with reference to the embodiments described below in detail in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms, and only these embodiments allow the disclosure of the present invention to be complete, and common knowledge in the art to which the present invention pertains It is provided to fully inform those who have the scope of the invention, and the present invention is only defined by the scope of the claims. Like reference numerals refer to like elements throughout.

In describing the embodiments of the present invention, if it is determined that a detailed description of a well-known function or configuration may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted. In addition, the terms to be described later are terms defined in consideration of functions in an embodiment of the present invention, which may vary according to intentions or customs of users and operators. Therefore, the definition should be made based on the content throughout this specification.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of an edge computing system using an artificial intelligence block-type module architecture according to an embodiment of the present invention, and FIGS. 2 to 7 are edge computing using an artificial intelligence block-type module architecture according to an embodiment of the present invention. It is a diagram for explaining the system.

1 to 7, an edge computing system using an artificial intelligence block-type module architecture according to an embodiment of the present invention is a central node block (central node block, 110), an algorithm node block (algorithm node block, 120), It may include a data node block 130 , an edge node block 140 , a leaf node block 150 , and the like.

The central node block 110 is an edge processing node of the blocking module, which includes processing and application services, uses wired or wirelessly connected learning data to train a model, and uses the learned model to artificial intelligence corresponding to application services. It can drive an edge computer or multiple IoT devices connected by an algorithm.

The algorithm node block 120 includes a plurality of artificial intelligence algorithm nodes configured in a block form by modularizing various machine learning algorithms, and may provide an artificial intelligence algorithm to the central node block 110 .

The data node block 130 distributes and stores the training data and the learned model in a hardware node connected by wire or a plurality of artificial intelligence servers connected wirelessly, uses the training data to train the model in real time, and the training data and the learned model may be provided to the central node block 110 .

The edge node block 140 includes an edge computer and may be operated according to an application service linked to processing.

The leaf node block 150 includes a plurality of IoT devices and may be operated according to an application service linked to processing.

For example, FIG. 2 illustrates a design diagram of an artificial intelligence block architecture (AI-Block Architecture) that is a block-type module architecture according to an embodiment of the present invention, and the designed architecture is a processing unit (eg, For example, the central node block 110 of the present invention), N AI algorithms (eg, the algorithm node block 120 of the present invention), an AI server, training data, and a trained model (eg, of the present invention) data node block 130), an edge computer (eg, edge node block 140 of the present invention), and IoT devices (eg, leaf node block 150 of the present invention) .

Here, the processing unit may learn a model by using wired or wirelessly connected learning data, and use the learned model to drive edge computers or IoT devices connected with a corresponding AI algorithm.

In addition, the training data or the learned models are stored in a hardware node connected by wire or an AI server connected wirelessly, and the model can be trained in real time using the learning data. can be operated according to

For example, in an application service that recognizes a vehicle type, an external camera becomes an IoT device, and an image recognition deep learning algorithm can be used in conjunction. It can be included in the edge computer and connected to the vehicle type recognition application service.

In addition, the AI-Block Architecture according to an embodiment of the present invention is a central node, an algorithm node, a data node, an edge node, and a leaf node. (Leaf node), etc., when each node is described with reference to FIG. 3, the central node is an edge processing node of the blocking module, and may include processing and application services.

In addition, an algorithm node is a modularized block of various machine learning algorithms, for example, naive bayes, decision trees, support vector machines (SVMs), kNNs (kNNs). Algorithms such as k-nearest neighborhood), random forest, K-means, and Gaussian mixture can be used.

Here, the naive Bayes method is a classification technique that calculates the posterior probability through the product of the prior probability information and the likelihood measured through observation, and the decision tree classifies the samples through a series of decision processes, and the current decision It is a classification technique that helps decision making. SVM is a classification technique that obtains an optimal hyperplane in a high-dimensional variable space by substituting an input vector value into a high-dimensional variable space. kNN calculates the approximate distance between two different input vectors to determine which category It is a single classification technique that determines whether It is a classification technique that minimizes variance, and Gaussian mixing is a classification technique that expresses a probability distribution of a complex shape existing in reality by mixing N Gaussian distributions.

In addition, the data node serves to store the training data and the trained model, and can be distributed and stored in a wired hardware node or a wirelessly connected AI server. It can also play a role as a database of application services.

In addition, an edge node may operate as an edge device for edge computing or as a central node including other processing and other application services. In a distributed machine learning environment, a plurality of central nodes ( Central nodes) act as edge nodes and may perform mutual network communication.

As described above, in the edge computing system, the edge node block 140 may serve as another central node block including other processing and other application services.

On the other hand, a leaf node may be an IoT device disposed at the very end or may be connected in the form of input/output such as a sensing device or a display. For example, in FIG. 4 , the concept of wired hardware connection of an AI block A diagram showing the

As described above, in the block-type module architecture according to an embodiment of the present invention, free addition and deletion between nodes that are block-type modules are possible, and since the wired/wireless connection structure between the nodes is a loosely coupled structure, unrestricted combinations are possible. Rather, by using N AI algorithm nodes, various machine learning algorithms can be organically linked according to the situation, and N edge nodes can easily provide application services combining various edge computing devices. .

On the other hand, the edge computing system defines and uses a metadata query protocol, but the metadata query protocol can communicate between node blocks using a network communication protocol through normalized metadata. It may be provided according to the context format of 5W1H (Who, What, Where, When, Why, How), and the normalized metadata is at least among message queuing telemetry transport (MQTT), hypertext transfer protocol (HTTP) and serial communication. It can be transmitted between node blocks in the form of a query using any one network protocol communication.

For example, the artificial intelligence block-type module architecture according to an embodiment of the present invention can be used by defining a metadata query protocol, which uses a network communication protocol through normalized metadata between block-type modules. Fast communication may be performed, and normalized metadata may be provided according to the context format of 5W1H (Who, What, Where, When, Why, How) indicating semantic information.

Here, the normalized metadata of 5W1H can be transmitted between block modules in the form of a query through the network protocol communication of MQTT, HTTP, and serial communication. This is possible to support unlimited scalability.

In addition, the artificial intelligence block (AI-Block) design technique is a modular structure for machine learning as shown in FIG. 4, and is designed in the form of a hardware block or brick, so that contactless (wireless) communication and contact type (Wired) communication can be performed. In the case of contactless block interworking, data can be transmitted and received by metadata query protocol through MQTT or HTTP, and in case of contact block interworking, serial communication data can be transmitted and received by direct block alignment. can

On the other hand, the edge computing system may be interlocked between node blocks through one-way communication and two-way communication, where one-way communication may be used in an input node and an output node, and two-way communication may be used in a processing node and a data node.

For example, the artificial intelligence block-type module architecture according to an embodiment of the present invention can organically link various block modules according to the situation through one-way or two-way communication. It shows the communication direction according to the situation.

Here, one-way communication may be used for an input node (eg, sensing, machine learning algorithm, etc.) and an output node (eg, display, etc.), and two-way communication may be used with a processing node (eg, fusion, inference, etc.) ) and data nodes (eg, learning, modeling, etc.).

For example, FIGS. 6 and 7 show an example of a communication method of an artificial intelligence block-type module architecture according to an embodiment of the present invention. As shown in FIG. 6, machine learning of object recognition and voice recognition is a direct communication method. The result data can be directly transmitted to the display, e-Paper system, and as shown in FIG. 7 , the processing node fusions the recognition data of object recognition and voice recognition in a convergence communication method for machine learning. After performing , the result data can be transmitted to the display e-paper system. As such, the artificial intelligence block (AI-Block) architecture according to an embodiment of the present invention can perform organic communication by interworking block-type modules according to a given application situation.

Therefore, the present invention includes a block-type module architecture equipped with various machine learning algorithms, and is expandable to allow free addition and deletion between block-type modules, unlimited combinations between blocks are possible, and various machine learning algorithms can be used in situations By designing the block module to be organically interlocked according to the One-way or two-way communication between modules can be supported.

In addition, the present invention trains a model by using learning data connected by wire or wirelessly through communication with a data node block in the central node block, which is an edge processing node of the blocking module, and is an application linked to processing using the learned model. When processing a service, by operating an edge computer or multiple IoT devices connected with the corresponding AI algorithm provided from the AI node block, machine learning is performed by linking various machine learning algorithms with IoT devices in a distributed edge computing environment. can do.

Meanwhile, FIGS. 8 to 10 are diagrams for explaining a modular comparison analysis and implementation example of an edge computing system using an artificial intelligence block-type module architecture according to an embodiment of the present invention.

8 to 10 , the block-type modular structure of the artificial intelligence block-type module architecture according to an embodiment of the present invention was compared and analyzed as shown in FIG. 8 based on related studies.

Here, an analysis was made based on the characteristics of the AI block-type module architecture according to an embodiment of the present invention. As a result of the analysis, some of the existing studies enable structural scalability by utilizing large-capacity data, and individual communication Although machine learning algorithms are linked as a method, existing studies do not form individual distributed structures in block form by modularizing machine learning algorithms and learning data.

In particular, since the normalized metadata protocol is not defined, there is a problem that the normalized communication method between individual distributed machine learning blocks is insufficient, and the artificial intelligence block (AI-Block) architecture according to the embodiment of the present invention is distributed edge computing. It is designed as a modular block structure of machine learning algorithms and related learning data in the environment, and can support one-way or two-way communication according to the situation with communication protocols such as HTTP or MQTT by defining a normalized metadata query in the form of 5W1H between modules. have.

In addition, in order to prove the effect of the artificial intelligence block (AI-Block) architecture according to the embodiment of the present invention, an edge computing system was implemented as shown in FIG. 9. As a conventional software architecture related to machine learning development, Tensor Flow, MXNet, Caffe2, etc., and machine learning hardware architectures for edge computing include Intel Movidius, Nvidia Jetson Board, and Google Coral Board.

Existing machine learning software architectures are different from block-type modular structures, but existing machine learning hardware architectures can be used as block modules of edge nodes. An application system as shown in Fig. 9 was developed, and Fig. 10 shows hardware used in the developed application system.

The system developed as described above is implemented to indicate interlocking between block modules. The test image recognition result of the vision edge node, the test voice recognition result of the voice edge node, and the coral board edge node ( It indicates that the test image recognition result of the edge node) is processed in the Raspberry Pi central node and simply outputted to the papyrus leaf node.

Then, the camera image received the image of the leaf node as an input, and the output was checked at the Raspberry Pi central node. could

As described above, in the present invention, a block-type modular architecture design technique for distributed machine learning is proposed, and the proposed architecture is a block-type module structure equipped with various machine learning algorithms, allowing free expansion between block-type modules, and Machine learning algorithms can be organically linked according to the situation, and the proposed architecture can perform free data communication using the defined metadata query protocol, and various edge computing devices are designed by designing a communication method suitable for the surrounding application situation. It is possible to easily implement application services that combine these, and to implement a hardware-based modular application system to check the interworking between block-type modules.

Meanwhile, FIGS. 11 to 14 are diagrams for explaining wired serial communication and wireless block communication in an edge computing system using an artificial intelligence block-type module architecture according to an embodiment of the present invention.

11 to 14 , in the present invention, a multi-AI block with a machine learning algorithm module is configured, and several blocks can be combined to provide advanced machine learning services, and each AI block includes algorithms, processing, and training data. It can contain test data, trained models, databases, etc., processing blocks can be linked with algorithms, training data and models of machine learning applications, and multiple machine learning algorithms can be applied as needed in the modeling phase . Here, actual machine learning algorithms may include supervised/unsupervised learning such as kNN, SVM, random forest, Gaussian naive bayesian and neural networks, and deep learning.

The edge computing system using the artificial intelligence block-type module architecture for machine learning according to the embodiment of the present invention as described above can be interlocked between node blocks through wired serial communication (ie, wired serial communication) or wireless block communication. have.

For example, in the case of wired serial communication, a wired serial communication technique may be provided, which is a contact communication for a modular block, and a brick shape may be designed for each wired block as shown in FIG. 11 .

Here, for logical modules, you can physically place a Raspberry Pi in brick-like blocks, these embedded Raspberry Pis can be the role of each module entity, each brick-like block has 4 connectors, 4 pushes A button and one Raspberry Pi may be provided.

And, 4 connectors, serial port for flexible connection, can communicate machine learning data with other blocks, 4 push button can trigger serial connection, for example, when the push button is On, serial COM port is opened and vice versa.

12 shows an example of a wired serial communication block. For example, when SW3 is On, the processing block can connect the Node1 block through the COM3 port, and when SW4 is On, the processing block connects the Node2 block through the COM4 port. can be connected

On the other hand, in the embodiment of the present invention, in the case of wireless block communication, a wireless block communication technique that is a contactless communication for a modular block can be provided. As shown in FIG. 13, a fruit shape can be designed for each wireless block, Blocks can consist of fruit-shaped blocks and plate-shaped blocks.

Here, the fruit-shaped block may be equipped with an on-off push button that triggers an MQTT connection with the Raspberry PI constituting the machine learning module entity, and the plate-shaped block measures the weight of each fruit-shaped block to connect the fruit-shaped blocks It is a processing block with load cells that detect the

For example, if the weight of a plate block is increased and the On-Off push button of the last touched block is On, the plate block can link MQTT groupings using topic messages from MQTT data, and the situation is opposite. The planar block can disconnect the off-state block. Here, if the weight is 0, all of the plate-shaped blocks can be released, and FIG. 14 exemplifies a radio block communication technique for explaining the change of the MQTT grouping state.

As described above, the present invention proposes a method for creating a block-type module with a machine learning algorithm. Wired serial communication method and wireless block communication method for block modularization can be designed, and mobile edge computing applications are collected You can train a machine learning model with the collected data and then create an intelligent service.

However, in mobile edge computing, since the existing method does not have an interworking method and communication method between modular algorithms, in order to provide a more flexible service, a method of segmenting and combining AI algorithms in the network should be considered. It is possible to build a machine learning block for

In the above description, various embodiments of the present invention have been presented and described, but the present invention is not necessarily limited thereto. It will be readily appreciated that branch substitutions, transformations and alterations are possible.

110: central node block (central node block)
120: algorithm node block (algorithm node block)
130: data node block (data node block)
140: edge node block (edge node block)
150: leaf node block (leaf node block)

Claims (9)

Edge processing node of the block module that includes processing and application services, uses wired or wirelessly connected learning data to train a model, and uses the learned model to connect an edge computer with an artificial intelligence algorithm corresponding to the application service or a central node block that drives a plurality of IoT devices; and
An algorithm node block comprising a plurality of artificial intelligence algorithm nodes configured in a block form by modularizing various machine learning algorithms, and providing the artificial intelligence algorithm to the central node block;
The training data and the learned model are distributed and stored in a hardware node connected by wire or a plurality of artificial intelligence servers connected wirelessly, the model is learned in real time using the training data, and the training data and the learned model are stored in a distributed manner. a data node block provided to the central node block;
An edge node block including the edge computer and operated according to the application service linked to the processing;
A leaf node block including the plurality of IoT devices and operated according to the application service linked to the processing
Edge computing system using artificial intelligence block-type module architecture including
The method of claim 1,
The edge computing system is an edge computing system using an artificial intelligence block-type module architecture in which the edge node block serves as another central node block including other processing and other application services.
3. The method of claim 2,
The edge computing system defines and uses a metadata query protocol, wherein the metadata query protocol uses a network communication protocol through normalized metadata. Edge computing using an artificial intelligence block-type module architecture that communicates between node blocks. system.
4. The method of claim 3,
The normalized metadata is an edge computing system using an artificial intelligence block-type module architecture provided according to the context format of 5W1H (Who, What, Where, When, Why, How) representing semantic information.
5. The method of claim 4,
The normalized metadata is transmitted between node blocks in the form of a query using at least one network protocol communication of MQTT (message queuing telemetry transport), HTTP (hypertext transfer protocol), and serial communication. Artificial intelligence block-type module architecture edge computing system using
6. The method according to any one of claims 1 to 5,
The edge computing system is an edge computing system using an artificial intelligence block-type module architecture that interworks between node blocks through one-way communication and two-way communication.
7. The method of claim 6,
The one-way communication is an edge computing system using an artificial intelligence block-type module architecture used in an input node and an output node.
8. The method of claim 7,
The bidirectional communication is an edge computing system using an artificial intelligence block-type module architecture used in a processing node and a data node.
9. The method of claim 8,
The edge computing system is an edge computing system using an artificial intelligence block-type module architecture that interworks between node blocks through wired serial communication or wireless block communication.

Images