KR20230075862A - Automatic machine learning-based complex environment analysis artificial intelligence model automatic distribution management system and its operation method - Google Patents

Abstract

The present invention relates to an automatic machine learning-based complex environment analysis artificial intelligence model automatic distribution management system and its operating method. Receiving input, building training data using sensor data collected at the beginning of installation in the complex sensor module and sensor data collected in the past from the complex sensor module installed in the same type of place as the place where the complex sensor module is installed, Searching for an artificial intelligence model structure suitable for predicting environmental information based on training data, generating an artificial intelligence model by setting hyperparameters in the searched artificial intelligence model structure, training the created artificial intelligence model with the built learning data and distributing the trained artificial intelligence model to the complex sensor module. According to the present invention, an artificial intelligence analysis model suitable for an installation environment can be automatically distributed and managed to a complex sensor module by creating an automatic machine learning basis.

Description

Automatic machine learning-based complex environment analysis artificial intelligence model automatic distribution management system and its operation method

The present invention relates to an automatic machine learning-based complex environment analysis artificial intelligence model automatic distribution management system and its operating method.

As the AI platform technology competition evolves from large-scale data analysis to real-time data response, the importance of edge computing is increasing. Global ICT companies are strengthening their competitiveness through complementary services of cloud computing and edge computing in the industrial field.

The next-generation device element technology for edge clouding is evolving, and open source foundations and global companies are increasing the disclosure of software (S/W) technology for edge computing. In order to develop industrial embedded artificial intelligence devices that support edge computing, convergence of technologies such as semiconductors, embedded S/W, artificial intelligence, and IoT is important.

In this situation, it is necessary to develop an intelligent complex environment sensor system and its operation method that can measure and predict the state value of the spatial environment using a complex sensor module equipped with an edge AI analysis model suitable for the construction environment while accommodating various environmental sensor devices. It is becoming.

However, since the characteristics of each installation site of the complex environment sensor module are different, there is a growing demand for automatically generating and distributing an AI environment analysis model suitable for the installation site based on automatic machine learning.

Therefore, the technical problem to be solved by the present invention is to provide a system and an operating method capable of automatically distributing and managing an artificial intelligence analysis model suitable for an installation environment to a complex sensor module by creating an automatic machine learning base.

An automatic machine learning-based complex environment analysis artificial intelligence model automatic distribution management method according to an embodiment of the present invention for solving the above technical problem is sensor data type information collected from a complex sensor module and the type of place where the complex sensor module is installed. Receiving information, constructing training data using sensor data collected at the initial installation in the complex sensor module and sensor data collected in the past from the complex sensor module installed in the same type of place as the place where the complex sensor module is installed searching for an artificial intelligence model structure suitable for predicting environmental information based on the training data, and generating an artificial intelligence model by setting hyperparameters in the searched artificial intelligence model structure; Training with the built learning data, and distributing the trained artificial intelligence model to the complex sensor module.

The method may further include performing weight reduction on the artificial intelligence model according to the specifications of the complex sensor module.

A lightweight model of the trained artificial intelligence model may be distributed to the complex sensor module.

The composite sensor module may include a temperature sensor, a humidity sensor, an illuminance sensor, a carbon dioxide sensor, a dust sensor, and a motion sensor.

The artificial intelligence model may predict and output an indoor air quality score after a predetermined time elapses using sensor values measured by the temperature sensor, humidity sensor, illuminance sensor, carbon dioxide sensor, dust sensor, and motion sensor.

The method may further include re-learning the artificial intelligence model when a difference between an indoor air quality score predicted by the artificial intelligence model and an actual indoor air quality score is greater than or equal to a predetermined standard.

The artificial intelligence model is a random forest, k-nearest neighbor (K-NN), support vector machine (SVM), convolution neural network (CNN), long short-term memory (LSTM) or generative adversarial (GAN) Network) can be trained as one of the models.

An automatic machine learning-based complex environment analysis artificial intelligence model automatic distribution management system according to an embodiment of the present invention for solving the above technical problem is sensor data type information collected from a complex sensor module and the location where the complex sensor module is installed. Training data using an information input unit that receives type information, sensor data collected at the beginning of installation in the complex sensor module, and sensor data collected in the past from the complex sensor module installed in the same type of place as the place where the complex sensor module is installed A training data construction unit for constructing, searching for an artificial intelligence model structure suitable for predicting environmental information based on the training data, creating an artificial intelligence model by setting hyperparameters in the searched artificial intelligence model structure, and generating the artificial intelligence model and an automatic machine learning unit for training an intelligence model with the built-in training data, and a model distribution unit for distributing the trained artificial intelligence model to the complex sensor module.

A computer-readable recording medium according to an embodiment of the present invention for solving the above technical problem may record a program for executing the method in a computer.

According to the present invention, an artificial intelligence analysis model suitable for an installation environment can be automatically distributed and managed to a complex sensor module by creating an automatic machine learning base.

1 is a configuration diagram of an automatic machine learning-based complex environment analysis artificial intelligence model automatic distribution management system according to an embodiment of the present invention.
2 is a configuration diagram of a complex environmental sensor management server according to an embodiment of the present invention.
3 is an operation flowchart of an automatic machine learning-based complex environment analysis artificial intelligence model automatic distribution management system according to an embodiment of the present invention.

Then, with reference to the accompanying drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily practice the present invention.

1 is a configuration diagram of an automatic machine learning-based complex environment analysis artificial intelligence model automatic distribution management system according to an embodiment of the present invention.

Referring to FIG. 1 , an automatic machine learning-based complex environment analysis artificial intelligence model automatic distribution management system according to an embodiment of the present invention includes a plurality of complex sensor modules 100, a complex environment sensor management server 200, and a plurality of users. The terminal 300 may be included.

The plurality of complex sensor modules 100 , the complex environment sensor management server 200 , and the plurality of user terminals 300 may exchange various information and data with each other through the communication network 10 .

The communication network 10 does not cover a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), the Internet, and the like, and a communication method is wired or wireless. No, it doesn't matter what kind of communication method.

The composite sensor module 100 may include a temperature sensor, a humidity sensor, an illuminance sensor, a carbon dioxide sensor, a dust sensor, and a motion sensor. A combination of sensors mounted on the composite sensor module 100 may vary as needed. For example, it may vary depending on the type of environmental information to be monitored depending on the type of place where the complex sensor module 100 is installed. For example, if indoor air quality monitoring is required, a noise sensor or motion sensor may not be installed. Of course, since air quality prediction may vary depending on the number of people located in the room or the degree of activity, it is possible to mount a motion sensor on the complex sensor module 100 in such a case.

The complex environment sensor management server 200 may manage the complex sensor module 100 and automatically generate, train, and distribute an artificial intelligence model suitable for the complex sensor module 100 .

The user terminal 300 is an information communication terminal used by a user using the complex sensor module 100, and may be implemented as a smart phone, a laptop computer, a desktop computer, a tablet PC, or the like. The user may use the user terminal 300 to manage the operating state of the complex sensor module 100 or access the complex environment sensor management server 200 to input or inquire various information related to the complex sensor module 100. there is.

2 is a configuration diagram of a complex environmental sensor management server according to an embodiment of the present invention.

Referring to FIG. 2 , the complex environment sensor management server 200 includes an information input unit 210, a training data construction unit 220, an automatic machine learning unit 230, a model distribution unit 240, and a model weight reduction unit 250. , It may include a database unit 260 and a control unit 270.

The information input unit 210 may receive sensor data type information collected from the complex sensor module 100 and type information of a place where the complex sensor module 100 is installed.

Here, information on the type of place where the complex sensor module 100 is installed may be input from a user who uses the complex sensor module 100 or an installer who has installed the complex sensor module 100 . Sensor data type information may be sensor type information. Information about sensor types, such as a temperature sensor, humidity sensor, illuminance sensor, carbon dioxide sensor, dust sensor, and motion sensor, may be used as sensor data type information.

Place type information may be classified into educational institutions, offices, factories, hospitals, homes, etc., and educational institutions may also be provided with place type information subdivided into infant schools, elementary schools, middle schools, high schools, universities, and the like. In addition, the type of place such as underground, above ground, open place, closed place, etc. may be provided in detail. In the case of a home, it may be provided by classifying types such as apartments, residential-commercial complexes, and detached houses, and even within a home, it may be provided by further subdividing the type of place into living rooms, kitchens, and bedrooms.

The training data building unit 220 is configured to perform sensor data collected at the initial installation in the complex sensor module 100 and data collected in the past from other complex sensor modules 100 installed in the same type of place as the place where the complex sensor module 100 is installed. Training data may be constructed using the collected sensor data.

The automatic machine learning unit 230 may search for an artificial intelligence model structure suitable for predicting environmental information based on training data. In addition, the automatic machine learning unit 230 may generate an artificial intelligence model by setting hyperparameters in the searched artificial intelligence model structure, and train the generated artificial intelligence model with the built learning data. Since a method of automatically generating a deep learning model, etc. by AuotML (Automatic machine learning) technique is already well known, a detailed description thereof will be omitted.

The model distribution unit 240 may distribute the trained artificial intelligence model to the complex sensor module 100 . The model distribution unit 240 may distribute a lightweight model of the artificial intelligence model trained in the automatic machine learning unit 230 to the complex sensor module 100 .

The artificial intelligence model may be trained to predict and output an indoor air quality score after a predetermined time elapses using sensor values measured by a temperature sensor, humidity sensor, illuminance sensor, carbon dioxide sensor, dust sensor, and motion sensor.

Artificial intelligence models include Random Forest, K-Nearest Neighbor (K-NN), Support Vector Machine (SVM), Convolution Neural Network (CNN), Long Short-Term Memory (LSTM), or Generative Adversarial Network (GAN). ) can be trained as one of the models.

Meanwhile, if there is a difference between the indoor air quality score predicted by the artificial intelligence model and the actual indoor air quality score by more than a predetermined standard, re-learning for the corresponding artificial intelligence model may be performed.

The automatic machine learning unit 230 may perform re-learning on the artificial intelligence model using the learning data collected by the complex sensor module 100 .

The model weight reduction unit 250 may reduce the weight of the artificial intelligence model according to the specifications of the complex sensor module 100 . Since the complex sensor module 100 is an embedded device and has inferior computer specifications compared to a server, etc., it is preferable to keep the performance above a certain level while reducing the weight of the artificial intelligence model. Model compression techniques such as weight pruning, quantization and binarization, and weight sharing are known as weight reduction techniques, and other weight reduction techniques may also be used.

The database unit 260 may store various types of information and data related to the operation of the complex environment sensor management server 200 . The database unit 260 not only builds the sensor data collected by the complex sensor module 100 as big data, but also stores artificial intelligence model algorithms, etc., and also stores the automatic machine learning unit 230 or the model lightweight unit 250. You can also save artificial intelligence models created in .

The control unit 270 may control the overall operation of the complex environment sensor management server 200 .

3 is an operation flowchart of an automatic machine learning-based complex environment analysis artificial intelligence model automatic distribution management system according to an embodiment of the present invention.

Referring to FIG. 3 , first, the complex environment sensor management server 200 may receive sensor data type information collected from the complex sensor module 100 and type information of a place where the complex sensor module is installed (S310).

The complex environment sensor management server 200 collects sensor data from the complex sensor module 100 at the beginning of installation and sensor data collected in the past from the complex sensor module installed in the same type of place as the place where the complex sensor module 100 is installed. It is possible to build training data using (S320).

The complex environmental sensor management server 200 searches for an artificial intelligence model structure suitable for predicting environmental information based on the training data (S330), and sets hyperparameters to the searched artificial intelligence model structure to generate an artificial intelligence model. (S340).

Next, the complex environment sensor management server 200 may train the generated artificial intelligence model with previously built learning data (S350).

Then, the complex environment sensor management server 200 may distribute the trained artificial intelligence model to the complex sensor module (S360).

However, depending on the embodiment, if model weight reduction is selected after step S350 (S353-Y), weight reduction may be performed on the artificial intelligence model trained in step S350 according to the specifications of the composite sensor module (S355). In this case, the lightweight model may be distributed to the complex sensor module in step S360.

Thereafter, if the difference between the indoor air quality score predicted by the artificial intelligence model distributed in the complex sensor module 100 and the actual indoor air quality score exceeds a predetermined standard (S370-Y), the artificial intelligence model is re-learned (S380). ), the relearned artificial intelligence model can be redistributed to the complex sensor module 100 (S390).

The embodiments described above may be implemented as hardware components, software components, and/or a combination of hardware components and software components. For example, the devices, methods and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate (FPGA). array), programmable logic units (PLUs), microprocessors, or any other device capable of executing and responding to instructions. A processing device may run an operating system (OS) and one or more software applications running on the operating system. A processing device may also access, store, manipulate, process, and generate data in response to execution of software. For convenience of understanding, there are cases in which one processing device is used, but those skilled in the art will understand that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that it can include. For example, a processing device may include a plurality of processors or a processor and a controller. Other processing configurations are also possible, such as parallel processors.

Software may include a computer program, code, instructions, or a combination of one or more of the foregoing, which configures a processing device to operate as desired or processes independently or collectively. You can command the device. Software and/or data may be any tangible machine, component, physical device, virtual equipment, computer storage medium or device, intended to be interpreted by or provide instructions or data to a processing device. may be permanently or temporarily embodied in Software may be distributed on networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer readable media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program commands recorded on the medium may be specially designed and configured for the embodiment or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. - includes hardware devices specially configured to store and execute program instructions, such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine language codes such as those produced by a compiler. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described with limited drawings, those skilled in the art can apply various technical modifications and variations based on the above. For example, the described techniques may be performed in an order different from the method described, and/or components of the described system, structure, device, circuit, etc. may be combined or combined in a different form than the method described, or other components may be used. Or even if it is replaced or substituted by equivalents, appropriate results can be achieved.

Claims (13)

Receiving sensor data type information collected from a composite sensor module and location type information where the composite sensor module is installed;
Building training data using sensor data collected at the beginning of installation in the complex sensor module and sensor data collected in the past from a complex sensor module installed in the same type of place as the place where the complex sensor module is installed;
Searching for an artificial intelligence model structure suitable for predicting environmental information based on the training data, and generating an artificial intelligence model by setting hyperparameters in the searched artificial intelligence model structure;
Training the generated artificial intelligence model with the built learning data, and
Distributing the trained artificial intelligence model to the complex sensor module
Automatic machine learning-based complex environment analysis artificial intelligence model automatic distribution management method including.
In claim 1,
Further comprising performing weight reduction on the artificial intelligence model according to the specifications of the complex sensor module,
An automatic machine learning-based complex environment analysis artificial intelligence model distribution management method for distributing a lightweight model of the trained artificial intelligence model to the complex sensor module.
In paragraph 2,
The composite sensor module,
Automatic machine learning-based complex environment analysis artificial intelligence model deployment management method including temperature sensor, humidity sensor, light sensor, carbon dioxide sensor, dust sensor and motion sensor.
In paragraph 3,
The artificial intelligence model,
Automatic machine learning-based complex environment analysis artificial intelligence model that predicts and outputs the indoor air quality score after a predetermined time using the sensor values measured by the temperature sensor, humidity sensor, illuminance sensor, carbon dioxide sensor, dust sensor, and motion sensor How to manage your deployment.
In paragraph 4,
When the indoor air quality score predicted by the artificial intelligence model and the actual indoor air quality score differ by more than a predetermined criterion, re-learning the artificial intelligence model; How to manage automated deployments.
In paragraph 5,
The artificial intelligence model,
Train as one of the following models: Random Forest, k-Nearest Neighbor (K-NN), Support Vector Machine (SVM), Convolution Neural Network (CNN), Long Short-Term Memory (LSTM), or Generative Adversarial Network (GAN) model. An automatic machine learning-based complex environment analysis artificial intelligence model automatic distribution management method.
A computer-readable recording medium on which a program for executing any one of the methods of claims 1 to 6 is recorded on a computer.
an information input unit for receiving sensor data type information collected from the complex sensor module and location type information where the complex sensor module is installed;
A training data construction unit for constructing training data using sensor data collected at the beginning of installation in the complex sensor module and sensor data collected in the past from a complex sensor module installed in the same type of place as the place where the complex sensor module is installed;
An artificial intelligence model structure suitable for predicting environmental information is searched for based on the training data, an artificial intelligence model is created by setting hyperparameters in the searched artificial intelligence model structure, and the generated artificial intelligence model is converted into the built learning data. Automatic machine learning unit trained with, and
A model distribution unit distributing the trained artificial intelligence model to the complex sensor module.
An automatic machine learning-based complex environment analysis artificial intelligence model automatic deployment management system that includes
In paragraph 8,
Further comprising a model weight reduction unit for performing weight reduction on the artificial intelligence model according to the specifications of the complex sensor module,
An automatic machine learning-based complex environment analysis artificial intelligence model distribution management system for distributing a lightweight model of the trained artificial intelligence model to the complex sensor module.
In paragraph 9,
The composite sensor module,
Automatic machine learning-based complex environment analysis artificial intelligence model automatic distribution management system including temperature sensor, humidity sensor, light sensor, carbon dioxide sensor, dust sensor and motion sensor.
In paragraph 10,
The artificial intelligence model,
Automatic machine learning-based complex environment analysis artificial intelligence model that predicts and outputs the indoor air quality score after a predetermined time using the sensor values measured by the temperature sensor, humidity sensor, illuminance sensor, carbon dioxide sensor, dust sensor, and motion sensor Deployment management system.
In paragraph 11,
When the indoor air quality score predicted by the artificial intelligence model and the actual indoor air quality score differ by more than a predetermined criterion, re-learning the artificial intelligence model; An automated deployment management system.
In paragraph 12,
The artificial intelligence model,
Train as one of the following models: Random Forest, k-Nearest Neighbor (K-NN), Support Vector Machine (SVM), Convolution Neural Network (CNN), Long Short-Term Memory (LSTM), or Generative Adversarial Network (GAN) model. Automatic machine learning-based complex environment analysis artificial intelligence model automatic deployment management system.

Images