KR101126524B1 - User-centered context awareness system, context translation method therefor and Case-based Inference method therefor - Google Patents

Abstract

Disclosed are a user-centered context awareness system using ontology context models, context information transformation methods, and case-based reasoning.
The situation awareness system according to the present invention is a sensor unit for generating sensor data for the surrounding resources, a situation management agent for generating the situation information suitable for inference by converting the sensor data collected by the sensor according to the attribute, the situation management It has an ontology-based situation model for storing the situation information generated by the agent, and infers the current situation by referring to the situation model which controls input / output of the situation information and the situation model stored in the situation manager and corresponds to the inferred result. It provides a recognition service to the user and reflects the user's feedback on the provided service, characterized in that it comprises a situation inference agent for learning about the inference rules.

Description

User-centered context awareness system, suitable context information transformation method and case-based reasoning method {user-centered context awareness system, context translation method therefor and Case-based Inference method therefor}

The present invention relates to a situation awareness system, and more particularly, to a user-centered situation awareness system using an ontology context model for system scalability, reusability, and interoperability, a method for transforming context information, and a case-based reasoning method suitable thereto.

The basic concept of ubiquitous computing, which is being actively studied recently, means that various resources can be accessed and controlled through the internet network anytime and anywhere. As such, a space capable of ubiquitous computing is defined as a smart space, and the recent smart space is rapidly expanding due to the development of a ubiquitous network. The purpose of the situational awareness system is to grasp the intention of the user in the smart space as described above and to automatically provide the appropriate service to the user.

Situational awareness systems also target users who cannot use ubiquitous computing resources and users who require automated services. Based on these trends and demands, the development of situational awareness systems in smart space is accelerating.

The configuration of the situational awareness system can be divided into the situation modeling method and the inference method. The existing researches on context modeling have been conducted by Context-Toolket, CoolTown, and ER and UML Model and CONON proposed by Karen. Recently, CONON models using ontologies have been used for reasons of reusability and interoperability.

Situational awareness system can be divided into situational modeling that constructs contextual information and contextual inference that finds suitable services based on the configured information.

The collection and inference of each context information scattered around the user is called the situation that the user is in. The method of configuring each context information to efficiently analyze and infer the user's situation is explained. This is called situation modeling.

The term 'context' expressed in the present invention is the resource information surrounding the user as physical information such as location and time, environmental information such as light, temperature, humidity, user's mood, health state, etc. Personal information, etc.

Various researches have been conducted on the situation modeling method to efficiently construct such situation information for inference and analysis. The contents are as follows.

● Attribute-Value Model

It is a model that constructs a context model using the context information with attributes and values for the attributes. Context-Toolkit exists as a representative system using the context model. The context information stored in the context model is acquired by a module called Widget, and the acquired information is interpreted through the Interpreter module.

Web-based Model

It is a situation model that increases the accessibility of context information by expressing it on a web basis. It provides the information on the smart space to the user through the format expressed as Web Presence, and the information about the type and relationship of the situation information can be acquired through the Web Presence. The representative system of the web-based model is Cool-Town developed as part of the HP Smart Space project.

● ER and UML Model

It is a situation model expressed according to the objects and the relationship that make up the situation information. It is a method that uses the ER and UML models used in the database when constructing the situation model. It is used in the situation model suggested by Karen and has the disadvantage of not being able to express hierarchical relationships.

● Ontology Model

It is a situation model method that can express the situation information, express the relationship between objects, and express hierarchically. The situation model built with ontology has the advantages of interoperability between heterogeneous systems, reusability of the built model, and reuse of knowledge. In order to efficiently construct the ontology situation model method, CONON has the advantage that it can be easily applied in various domains by suggesting the same method as the upper ontology and the lower ontology. In recent years, the situation model constructed by ontology has been mostly used for the above reasons.

In the case of the system that does not use the ontology situation model, the reasoning method is vulnerable, and the services provided include the information of devices that can be utilized in the smart space, or supplementary services based on user-scheduled scheduling. However, with the advent of the ontology situation model, various reasoning methods such as hierarchical reasoning and rule-based reasoning provided by the ontology structure, and reasoning by learning log information that stores the user's response are being studied.

The reasoning method using this situation model is as follows.

● rule-based reasoning

It is the most commonly used method of situation inference in situational awareness systems. It can be applied to not only the ontology situation model but also various situation models, and inferred by the set rules. The disadvantage is that it takes a lot of cost when creating the rule, and the disadvantage is that the generated rule cannot be modified by the system itself. Typically, SOCAM, CoBrA and the like are present. In the case of CoBrA, it is inferred through Agent called ContextBroker. The functionality of SOCAM is similar, but the service is provided through mobile.

● Naive Bayes

This method infers the situation by analyzing the log of the service provided to the user. The user may be satisfied or dissatisfied with the service provided, and the context-aware system may log it to assist in performing inference. GAIA is a representative system.

In the case of the conventional situation inference method as described above, there is a problem that there is no method that can reflect the user's feedback on the service provided when the situation awareness system operates.

1 shows a configuration of a conventional situation awareness system.

The conventional situation awareness system illustrated in FIG. 1 includes a sensor unit 10, a situation information provider 12, a situation interpreter 14, a situation model 16, and a situation recognition service unit 18. The conventional situation awareness system shown in FIG. 1 provides a service to a user through the following four steps.

First, the sensor unit 10 collects data through a physical or virtual sensor and delivers the data to a context provider. The physical sensor data detected by the physical sensor indicates information such as temperature and humidity for collecting environmental information in the smart space, and in the case of the virtual sensor data detected by the virtual sensor, time information, weather information, and date obtained through the Internet network. Information, etc.

Second, the contextual information provider 12 processes the data received from the sensor unit 10 into contextual information and stores it in the contextual model 16.

Third, the context interpreter 14 infers the current situation based on the context information stored in the context model 16.

Fourth, the context-aware service 18 delivers the inferred result to the user.

In the conventional situation awareness system as shown in FIG. 1, the inferred result is provided through the context DB inside the context interpreter 14 and the inferred result is provided to the user. This means that a service defined according to the situation exists as a rule. If a rule that is incorrectly defined or a rule that does not suit an individual user exists, the system cannot correct an error by itself.

In addition, there is no standardized method for generating situation information of the situation information provider existing in each system and a method for generating reliable data.

In addition, in order to efficiently represent context information in the context model, not only the data collected from the sensor needs to be converted into semantic data, but also filtering and processing are required according to the data. In the existing system, the method of converting these situation data into context information suitable for inference has not been formulated, which causes the use of inappropriate data.

In addition, in the existing system, the process of inference is always fixed, and there is no way to correct it when inappropriate inference is performed. Accordingly, there is a problem that does not properly cope with changes in the user or the surrounding situation.

The present invention has been made to solve at least some of the above problems, and an object of the present invention is to provide a user-centered situation awareness system using an ontology situation model for system scalability, reusability, and interoperability.

Another object of the present invention is to provide a situation awareness system capable of generating reliable situation information.

It is another object of the present invention to provide a situational awareness system that can correct wrong inferences by reflecting user feedback.

Still another object of the present invention is to provide a method for transforming sensor data, which is formulated for generating reliable situation information.

It is another object of the present invention to provide a method for correcting wrong inference by the system by reflecting user feedback in order to provide a user-friendly service.

The situation awareness system according to the present invention for achieving the above object

Collects data on the surrounding resources from the sensor and converts the collected data into contextual information and stores it in the ontology contextual model, proceeds with inference based on contextual information stored in the contextual model, and provides the inferred result to the user. In the situation recognition system of the situation, the sensor unit for generating sensor data for the surrounding resources, the situation management agent for generating the situation information suitable for inference by converting the sensor data collected by the sensor according to the attribute, the situation management agent It has an ontology-based situation model for storing the situation information generated by the system, and the situation management unit controls the input / output of the situation information, and infers the current situation by referring to the situation model stored in the situation management unit and recognizes the situation corresponding to the inferred result. Provide and provide services to users Reflect user feedback on services and comprises the situation, reasoning agent to perform a study on the inference rules.

Context information conversion method according to the present invention for achieving the above another object

In the situation information conversion method suitable for the situation recognition system that collects data on the surrounding resources from the sensor, converts the collected data into context information, proceeds with inference based on the context information, and provides the inferred result to the user.

The sensor data is classified into a person recognition data having an attribute that is always acquired according to acquisition time, a device data having an attribute acquired according to an event, and an environment data having an attribute acquired every predetermined time.

It is characterized by converting into semantic data having a value indicating a state and a value indicating a reliability according to each attribute.

Case-based reasoning method according to the present invention for achieving the above another object

Collects data on the surrounding resources from the sensor and converts the collected data into contextual information and stores it in the ontology contextual model, proceeds with inference based on contextual information stored in the contextual model, and provides the inferred result to the user. In the case-based reasoning method suitable for the situation recognition system of the present invention, Representation step of representing the situation information as a computational case and stored in the case database, Searching for the closest case, that is, the closest case to the current case from the case database A retrive step, an adaptation step of providing a service that resolves the difference between the closest case and the current case based on the rule, and an evaluation and store step of modifying the rule to reflect the user's feedback on the service. It features.

The present invention implements a reliable contextual information method based on a learning algorithm and context management agent (CMA) of a system through a situation reasoning agent (SRA), and according to the type of context data. It has the effect of providing a context aware system capable of processing and automatic learning and transformation of context data.

1 shows a configuration of a conventional situation awareness system.
2 illustrates a configuration of a situation awareness system according to the present invention.
FIG. 3 shows a detailed configuration of the situation management agent shown in FIG.
FIG. 4 shows a detailed configuration of the situation reasoning agent shown in FIG. 2.
5 shows an example of a case representation for contextual information on which inference is based.
6 diagrammatically shows an example of a learning algorithm according to the invention.
7 diagrammatically illustrates the learning behavior of a context aware system in a scenario.
8 shows an example of a simulation program for verifying a case based reasoning process.
9 illustrates an example of rule generation and application according to user feedback.

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

The context awareness system according to the present invention is assumed to be an operation in a smart space capable of collecting information on surrounding resources. It describes how to collect situation information in smart space and how to learn it based on the structure of the entire system.

The design points of the ontology situation model based context awareness system proposed in the present invention can be summarized into the following three.

1) Method of converting various context data collected from sensors into reliable context information.

2) How to learn by reflecting the user's feedback on the services provided by the system.

3) Ontology situation model construction for system scalability, reusability and interoperability

The situation recognition system according to the present invention has been devised to solve the problem of the generation of reliable situation information and system learning of the conventional system, which is different from the existing system in the functional and structural aspects as shown in <Table 1>. Have

Conventional situational awareness system Context aware system of the present invention Structural aspect Processing is not possible depending on the type of situation data Can be processed to suit the type of situation data Functional aspect Learning module part and situation data transformation module part System automatic learning and situation data conversion

2 illustrates a configuration of a situation awareness system according to the present invention. The apparatus illustrated in FIG. 2 includes a sensor unit 20, a context management agent (CMA) 22, a context management unit 24, and a situation reasoning agent 26.

The apparatus illustrated in FIG. 2 is implemented by a hardware layer implementing the sensor unit 20 and the situation management agent 22, a middleware layer implementing the situation management unit 24, and an application layer so as to be applicable to a distributed environment.

The situation management agent 22 implemented as a hardware layer plays a role of processing the sensor data collected by the sensor unit 20 in the smart space and generating situation information, and the middleware layer is an ontology for storing the situation information. Consists of modules that control the input and output of contextual models and contextual information. The application layer is responsible for delivering the inferred results to the user.

The overall flow in the apparatus shown in FIG. 2 is as follows.

First, data about surrounding resources are collected from the sensor unit 20, and the collected data is converted into reliable situation information and stored in the ontology-based context model 24d. The inference is further performed based on the situation information stored in the situation model 24d, and the result of the inference is provided to the user. At this time, the learning proceeds according to the provided service, and the result of the learning is applied when the following situation occurs.

The main role of the context management agent 22 is to convert the collected data to generate context information suitable for inference and store it in the context model 24d of the middleware layer.

The sensor unit 20 present in the smart space is provided with sensors for detecting data in various formats, and each sensor should have a data acquisition time suitable for data characteristics. Here, the acquisition timing is related to the characteristics of the situation data.

The sensor unit 20 in the present invention

1) A sensor (Person cognitive sensor) that must always detect user's input / output, such as RFID,

2) Sensors whose state is changed by the operation of the system and the user (Device sensor) and

3) Sensors (Environment sensors) for updating the environment information every certain time.

The reason for setting the acquisition time of the situation data as described above is to reduce the overload of the system by reducing the unnecessary generation by setting the update period of the sensor data for each property. In addition, different conversion processes are performed according to each data acquisition time. The contents are as follows.

1) The Person Aware Sensor always acquires data and converts the data through the interpreter module 22b.

2) The device sensor acquires data according to an event and processes the data through the interpreter module 22b.

3) The environment sensor acquires data every predetermined time and converts the data through the fuzzy inference engine module 22c.

FIG. 3 shows a detailed configuration of the situation management agent shown in FIG.

Sensor data collection module (Sensor Management, 22a), Interpreter (22b), Fuzzy Inference Engine (22c) and Context Generator (22d) of the context management agent 22 The sensor data collected by the sensor unit 20 located in the smart space is converted into situation information in the form of semantic data.

The sensor data collection module 22a is defined as a classifier for classifying sensor data of various attributes. In the case of Person and Device attribute data, the situation model case should be generated through the interpreter module 22b, and in the case of Environment attribute data, the situation model case should be generated through the fuzzy inference engine module 22c. The sensor data collection module 22a selects the interpreter module 22d or the inference engine 22c according to the attribute of the sensor data.

The interpreter module 22b is a module for processing attribute data of Person and Device into semantic data. The ID of the user and the status information of the device are analyzed through a database (not shown) inside the context management agent 22, and the generated semantic information is transmitted to the context information generator module 30. For example, in the case of an RFID reader that acquires ID information of a person, a fixed location and ID information exist in a database to generate and transmit the location information of a person at the time of entering and exiting.

The fuzzy inference engine module 22c generates two kinds of semantic data from Environment attribute data. In the case of measured data such as temperature and humidity, the fuzzy membership function generates semantic data with the largest membership function, and in the case of dependent data inferred from measured data such as annoyance index, the semantic data is derived through fuzzy inference. Create

The semantic data generated by the fuzzy inference engine module 22c is expressed as a normalized value (a real value between 1 and 0) representing the reliability of the state as well as the state of the attribute.

1) Generation of actual situation information

The sensor data acquired from the sensor unit 20 passes through a fuzzy belonging function primarily. The fuzzy membership function represents an equation for calculating each state, and if the state of humidity is expressed in seven steps, the coefficient of the fuzzy membership function is seven. In the case of the fuzzy number whose state is calculated between 0 and 1 by the formula of each membership function, the membership of the state, that is, the reliability of the current state is reflected.

2) Generation of dependent contextual information

The fuzzy inference engine 22c generates semantic data of dependent characteristics such as annoyance index based on actual values such as humidity and temperature acquired from the sensor unit 20.

In the present invention, Mandani's min-max centroid method, which is a fuzzy inference method, is used to generate dependent context information.

The fuzzy rules for inference are stored in a knowledge base (not shown) inside the fuzzy inference engine, and the values obtained through inference have a real value as a non-fuzzy number (a prediction value of the discomfort index) of the dependent attribute. This real value is then converted through fuzzy-proportional functions.

The contextual information generator module 22d stores semantic information generated through the interpreter module 22b and the phage inference engine module 22c in the ontology contextual information model 24d of the middleware.

The situation inference unit 24 is a resource management module 24a for managing peripheral devices that can interact with the system, and a data load and storage module for storing meaningfully modified sensor information as input and ontology-based data. load and store: 24b) and a network interface module 24c for processing and transforming low data from the hardware layer and transmitting the data to the middleware layer.

FIG. 4 shows a detailed configuration of the situation reasoning agent shown in FIG. 2. The situation reasoning agent 26 is implemented in an application layer. The context inference agent 26 extracts the context information from the ontology-based context model, infers the extracted context information, and provides the inferred result as a service to the user.

Referring to FIG. 4, the situation reasoning agent 26 includes a situation management module 26a, a case-based reasoning engine 26B, and a rule-based reasoning engine 26c. It consists of four modules, Service Generator (26d), and reasoning and learning are performed through the above module.

The context management module 26a distinguishes between a situation in which learning is required and a situation not in accordance with the characteristics of the context information and the domain. Currently, case-based reasoning is chosen for situations that require learning, otherwise rule-based reasoning is chosen. In addition, it is possible to further configure another inference algorithm according to the characteristics of the domain.

Rule-based inference has been used in most existing systems, and the present invention was developed by utilizing open libraries such as Peller and Jena2.

Case-based (CBR) inference plays a role in inferring domain and contextual information that requires learning based on user feedback. CBR inference process is performed through Rppresentation step, Retrieve step, Adaptation step, and Evaluation and Store step.

1) Representation step

Situational information is represented as arithmetic cases in case-based reasoning. An example of the representation of an example for contextual information on which inference is based is shown in FIG. 5.

Representation of the case can be further subdivided according to the type and domain of the sensor that can collect information surrounding the user in the present invention consists of five environmental sensors and location, time, people information. Attribute information representing who, where, and when is expressed in String type and environment information except this is converted to Float type.

2) Retrieve step; Search for the case that is most similar to the current case (closest case) from the case base. The search for the nearest case uses the k-nearest algorithm, and the similarity is calculated using Equation 1 below.

는 속성 간의 차이를 연산하는 함수를 나타낸다.
함수

는 정보의 종류에 따라 결정된다.
본 발명에 있어서, 사례의 표현은 다섯 가지 환경 센서 및 위치, 시간, 사람 정보로 구성하였다.
이를 반영하여, 환경 센서에 의해 얻어지는 환경 정보(예를 들어, 온도, 습도, 조도, 공기 오염도 등)의 경우, 함수

는 속성 차의 절대값으로 연산되며,
위치, 시간, 사람 정보(who, when, where)의 경우, 함수

는 유사도 테이블을 통해 연산된다
함수

의 두 파라미터

는 각각 최근접 사례 및 현재 사례의 각 속성의 상태를 나타낸다.As in Equation 1, the simplicity is a value obtained by dividing the sum of the product of the difference between each property of the case and the property weight by the sum of the property weights. In Equation 1 above, W means attribute weight and is a function

Denotes a function that computes the difference between attributes.
function

Depends on the type of information.
In the present invention, the representation of the case consists of five environmental sensors and location, time, and human information.
Reflecting this, in the case of environmental information (e.g., temperature, humidity, light intensity, air pollution, etc.) obtained by the environmental sensor, the function

Is computed as the absolute value of the property difference,
Function for location, time, and person information (who, when, where)

Is computed through the similarity table
function

Two parameters

Indicates the state of each attribute of the nearest case and the current case, respectively.

In addition, the weight for each attribute is set when comparing the similarity, which is to set the importance of the attribute according to the type of situation. The property weight setting method proposed in the present invention is a method of finding a sensitive property among properties and divides the difference value of each property by the sum of the difference values of all properties so as to have a weight as much as the weight of each property. It is expressed as

The role of the function f and the parameter in Equation 2 is the same as in Equation 1 above.

3) Adaptation Step

In the Retrieve step, the difference between the nearest case and the current case retrieved is resolved through a rule. The adaptation step is done through the rules in <Table 2>. These rules are defined during initial system design or are automatically generated by learning.

Num Type Rule Priority Rule 1 Match
rule IF Current Temperature is warm and Similar Temperature is Cold and Current Humidity is dry
Then set the Air conditioner level 2 down 0 Rule 2 Range
rule IF Current Lux is very dark and Similar Lux level less than normal
Then set Lux level 4 0

Match rule is applied when the difference between the current case and the nearest case exactly matches as defined in the rule. Range Rule is applied when the difference between the two cases satisfies a certain range.

4) Evaluation and Stroe Phase

This is a step of learning by reflecting a user's feedback on a provided service.

7 illustrates a learning algorithm according to the present invention. In the CBR inference process, learning is basically performed by storing a new case, but the present invention uses a learning algorithm as shown in FIG. 6 for general case learning.

Suppose that the user directly manipulated the 'air con level' and 'music type' when the system provided the service according to the current case. At this time, the system recognizes the user's behavior and configures the state difference between 'Temp' and 'Humidity', which is the difference between the current case and the closest case, using the conditional clause as shown in FIG. The adaptation rule is created by constructing the result clauses 'air con level' and 'music type'.

Rules created in this way are stored in individual Rule DB because of the risk of conflict with existing rules. If a conflict occurs even in individual Rule DB, the conflict is prevented through priority. Priority is given to the most recently generated rule among the most recently formed and stored rules. Cases and rules generated as a result of learning are applied in real time when the following situations occur:

The service creator 26d provides a list of services inferred from the case-based reasoning module and the rule-based reasoning module through various devices. The inferred result is stored in the situation model 24d of the middleware.

Example

In order to implement the situational awareness system according to the present invention, the simulation was performed through the following scenario.

"When Jessica came home after work, it was raining outside, so the temperature and humidity in the house were very low. The system is operating in Jessica's house and the system detects that Jessica is coming home. Based on the information, the current situation is deduced to provide lighting and heating services, but Jessica is not satisfied with the humidity in terms of humidity and starts the dehumidifier. The system satisfies Jessica by providing the lighting, heating and dehumidification services to Jessica's intentions as a result of the learning. "

7 diagrammatically illustrates the learning behavior of a context aware system in a scenario. The learning operation of the situation awareness system proposed in the above scenario is the same as that shown in FIG. 7, and the actual performance of the system is as follows.

① As the sensor detects the entry and exit of Jessica's house, location information generation and environment information is generated through the situation management agent. (S72)

② Provide lighting and heater to Jessica according to the extraction of situation information through the situation reasoning agent and the CBR reasoning method (s74 ~ s76).

③ The system recognizes Jessica's feedback according to the provided service. (S78)

④ The learning algorithm reflects the feedback according to the CBR learning module. (Reflects user needs) (s80)

In the above scenario, the entry / exit and location information of Jessica among the situation information used by the system is converted through the interpreter module 22b of the situation management agent 22, and the environmental information such as temperature, humidity, illuminance, and air pollution of the Jessica house is The fuzzy inference engine module 22c of the situation management agent is converted. The generated situation information is stored in the ontology-based situation model 24d and is configured as shown in Table 3.

The situation information stored through the situation management agent 22 is inferred through the case-based reasoning engine 26b of the redemption reasoning agent 26. This represents a simulation application that validates the case-based reasoning process rather than rule-based reasoning because the current scenario is a domain that requires user feedback.

<Table 3> Saving situation information

<Location rdf; about = "#CurrnetLocation">

 <rdf: type rdf: resource = "&owl; Thing" />

  <Located rdf: datatype = "&xsd; string"> LivingRoom

 </ Located>

</ Location>

<owl: Thing rdf: about = "# CurrnetTemp">

  <rdf: type rdf: resource = "#Temp" />

   <State rdf: datatype = "& xsd: string"> Cold

   </ State>

</ owl: Thing>

<owl: Thing rdf: about = "# CurrnetUser">

   <rdf: type rdf: resource = "# Person" />

     <Name rdf: datatype = "&xsd; string"> Jessica

     </ Name>

</ owl: Thing>

8 shows an example of a simulation program for verifying a case based reasoning process. Referring to FIG. 8, the simulation program is composed of four windows. First, the upper left pane 82 shows the current case where context information is extracted and processed from the context model. The upper right pane 84 shows a list of previously occurring cases, which represent attribute information constituting the case. In addition, the case indicated in red in the seventh line above represents the case most similar to the current case.

The lower left window 86 shows the Adaptation Rule defined by the current system and whether or not the rule is applied. The lower right window 88 contains a list of services provided to the user and a button that can directly accept the user's feedback.

In the example illustrated in FIG. 8, since there is no adaptation rule that can solve the difference between the current case and the nearest case, it indicates that the user cannot accurately provide the desired service.

In order to solve this problem, the simulation records the current case and the service list reflecting the feedback in the case base through user feedback, and adds personalized rules to the individual rule DB.

9 illustrates an example of rule generation and application according to user feedback. The upper figure 92 of FIG. 9 shows the rule generation according to the feedback of the user, and the lower figure 94 shows that the rule generated when the same situation occurs after a certain time is applied.

Comparing this as a result of the service, it can be seen that the service list (Heater level as 4 and, set the Lux level as 3) provided at the time of occurrence of the situation is changed to a service list suitable for the user. (Heater level as 4 and, set the Lux level as 3 and Dehumidifier) In order to verify the rule generated in the example shown in FIG.

As described above, it can be seen that the contextual awareness system according to the present invention provides an effective service compared to the conventional contextual awareness system through the learning of the system itself reflecting standardized contextual information generation and user feedback.

The situation recognition system according to the present invention

First, more reliable situation information can be generated by combining simple and fuzzy inference according to the type of data.

Second, by using a rule-based reasoning engine and a case-based reasoning engine in parallel, it is possible to perform reasoning according to context information.

Third, learn more efficiently by receiving user's requirements as feedback, and

Fourth, by agentizing the main functions of the system has the effect of providing a more intelligent situational awareness system.

Claims (16)

Collects data on surrounding resources from sensors, converts the collected data into context information, stores it in ontology context model, makes inference based on context information stored in context model, and provides services corresponding to the inferred result to users. In the ontology-based context awareness system to provide,
A sensor unit generating sensor data on surrounding resources;
A situation management agent converting sensor data collected by the sensor unit according to attributes to generate situation information for inference;
A situation management unit having an ontology-based situation model for storing situation information generated by the situation management agent and controlling input / output of the situation information; And
A situation inference agent that infers the current situation by referring to the situation model stored in the situation management unit, provides a situation awareness service corresponding to the inferred result, and learns the inference rule by reflecting the user's feedback on the provided service. Including;
Here, the sensor unit includes a person recognition sensor, a device sensor, and an environment sensor, wherein the person recognition sensor is a sensor that always acquires data, the device sensor is a sensor that acquires data according to an event, and the environment sensor is constant A sensor that acquires data every hour;
The situation management agent may include a sensor data collection module classifying sensor data collected by the sensor unit into person recognition data, device data, and environment data according to a learning time; An interpreter module for processing the Person recognition data and Device data into semantic data; A fuzzy inference engine for processing the environment data into semantic data through fuzzy inference; And a context information generator for processing semantic special information generated by the interpreter module and the fuzzy inference engine into ontology-based context information. delete delete The method of claim 1, wherein the fuzzy inference engine
Actual data is generated through the fuzzy membership function, the semantic data of the largest membership function.
And dependent data inferred from the measured data are generated as semantic data through fuzzy inference. The situation recognition system of claim 4, wherein the semantic data has a value indicating a state of data as well as a value indicating a state of reliability. 6. The system of claim 5, wherein the value representing the reliability is a real value between 1 and 0. The agent of claim 1, wherein the situation inference agent
Depending on the nature and domain of the situation information, it is classified into a situation where learning is required and a situation that is not,
In situations where learning is required, reasoning is based on case-based reasoning; and
In case of situations where learning is not required, the situation recognition system is characterized by inferring by rule-based reasoning method. 8. The agent of claim 7, wherein the situation inference agent is
Expressing the situation information as an operable case and storing it in the case database;
Retrieving a case most similar to a current case from the case database, that is, the nearest case;
Providing a service for resolving a difference between the nearest case and the current case based on a rule; And
And providing a service corresponding to the current situation inferred through the case-based reasoning method including the step of modifying the rule by reflecting the user's feedback on the service. The situation recognition system according to claim 8, wherein the similarity for determining the nearest case is determined using the following equation.

Where W stands for attribute weight and is a function

Represents a function that computes the difference between attributes.) 10. The system of claim 9, wherein the weight W for each attribute is determined using the following equation.

delete In the context information conversion method for a situation recognition system that collects data on the surrounding resources from the sensor and converts the collected data into context information, proceeds with inference based on the context information, and provides the inferred result to the user.
The sensor data is classified into a person recognition data having an attribute that is always acquired according to acquisition time, a device data having an attribute acquired according to an event, and an environment data having an attribute acquired every predetermined time.
And converting into semantic data having a value indicating a state and a value indicating a reliability according to each attribute. The method of claim 12, wherein the environment data includes measured data and dependent data inferred from the measured data.
The measured data generates semantic data of a membership function having the largest membership degree through fuzzy inference,
And the dependent data is generated as semantic data through fuzzy inference based on Mandani's min-max centroid method. Collects data on the surrounding resources from the sensor and converts the collected data into contextual information and stores it in the ontology contextual model, proceeds with inference based on contextual information stored in the contextual model, and provides the inferred result to the user. In the case-based reasoning method for situation awareness system of
A Representation step of expressing the situation information as an operable case and storing it in the case database;
Retrive step of retrieving the closest case, that is, the closest case to the current case from the case database;
An adaptation step of providing a service for resolving a difference between the nearest case and the current case based on a rule; And
If there is user feedback on the provided service, a matching rule is generated by conditionally setting the difference between the current case and the closest case as a result clause and the result clause of the difference between the provided service and the operation by the user's feedback. And an evaluation and store step of modifying the rule by applying the adaptation rule to a service corresponding to the inferred rule. 15. The method of claim 14, wherein the similarity for determining the nearest case is determined using the following equation.

Where W stands for attribute weight and is a function

Represents a function that computes the difference between attributes.) The method of claim 15, wherein the weight W for each attribute is determined using the following equation.

Images