KR20070058857A - A framework of context-awareness for ubiquitous computing middlewares - Google Patents

Abstract

A context-awareness framework for ubiquitous computing middleware is provided to generate high level context and offer a service suitable for characteristic of a user based on the high level context by enabling a context-awareness middleware to generate/process various context information through integration, inference, and learning of the sensed context information. A context describing module manages the context information obtained from sensors. A context normalizing module converts the sensed context information into a normalized structure. A context integrating module integrates the sensed context information received from the context normalizing module. A context inferring module infers the high level context information from the sensed low level context information. A context filtering module selects and stores only the useful context information to a database. A content learning module obtains the learned context information based on the information stored in the database. An event triggering module automatically issues a command for action or event for safety or convenience of the user.

Description

A framework of context-awareness for ubiquitous computing middlewares}

1 is a block diagram of a framework for situational awareness in accordance with the present invention.

2 is XML data of context information in a context description table;

3 is a block diagram of a music selection system according to an embodiment of the present invention.

The present invention relates to a context aware framework for ubiquitous computing middleware, and more particularly to a context aware framework for middleware with learning ability in ubiquitous computing environment.

Today, the trend in the computer industry is to use ubiquitous computing environments by embedding computers in all objects and connecting heterogeneous computers to networks to provide mutually indispensable services to users at any time and in any place. ) Is changing. In such a ubiquitous environment, a context-awareness system is required to satisfy a demand for a service provided to a user. In order to provide a service or information to a user in such an environment, a context aware system needs to collect and analyze context information from the surrounding environment. In addition, the context recognition system should process the process based on the context obtained from various sensors to provide an appropriate service to the user. And this system needs to create higher level context from lower level context through integration, inference, learning, etc. to adapt to the constantly changing environment in ubiquitous environment. Thus, a large number of context-aware systems have evolved to support ubiquitous computing environments.

However, most of these systems have a limitation in that they do not take into account the preferences and interests of a moving user or each user. In addition, the ubiquitous computing environment must provide various services to users in consideration of the ever-changing environment and the inaccessible environment. Therefore, context-aware systems must be able to handle and create various types of contexts to support users in dynamic environments. Various types of contexts are contexts that are inferred and learned through the framework proposed by the present invention based on contexts acquired by sensors. The proposed framework can easily generate various types of contexts from sensory data through integration, inference, and learning. So based on this, it can be used to provide appropriate services to users.

In order to provide user-oriented services in the ubiquitous computing environment, a situation-aware application plays a role of identifying a desired service by using the user's current situation and surrounding situation information. Context-aware middleware generally exists as an intermediary between sensors that collect a variety of information and applications that provide appropriate services to users. This context-aware middleware detects user and surrounding information and makes inferences based on that information to help applications provide the services or information they need.

The following describes middleware Context-Toolkit, Gaia, and CAMUS for context recognition which have been researched and developed at home and abroad.

Context-toolkit, the starting point of context-aware middleware among the above three middlewares, classifies contexts and user input values and defines them as four context types: user, place, state, and time. This middleware feature presents and integrates contextual information from sensors and provides them to applications. To provide the application with information gathered from peripheral sensors, the Context-Toolkit consists of three components: widgets, aggregators, and interpreters. Wiget collects contextual information from sensors and provides an interface between other components or applications. The integration unit integrates the context information based on the sensor information collected by the widget. The interpreter infers the higher level context from the lower level context and provides the inferred context to the application. This Context-Toolkit has suggested that it can handle the sensed, combined and inferred context information in the above three components, but it does not provide any concrete representation or reasoning method. In addition, it is insufficient to support a learning technique necessary to provide a service that satisfies the user, that is, a service considering the user's preference or interest.

Gaia is a distributed middleware proposed to support ubiquitous computing and is composed of various components to manage resources and services for applications in heterogeneous network environments. Gaia effectively manages context information in a dynamic environment and provides services that are considered for mobile users based on that information. Gaia consists of five components to efficiently handle the context of mobile users and the constantly changing environment. The event manager component provides the application with event service information so that it can respond appropriately to changing environments. In addition, the context service component provides appropriate services in consideration of the user's state and surrounding conditions. It uses a rule-base based on first order logic to generate higher level information based on the sensory context. In addition, the presence service component collects and uses each type of information using four types of objects, such as an application, a device, a person, and a service, using various sensors. Update information about the available resources. The space repository component is responsible for storing contexts collected through sensors or created through inference. The context file system component provides a directory service to provide a service for a moving user. It allows users to access files using the time and location information of the most recent file access instead of the file name or path. As such, Gaia can process sensored, combined and inferred context information to serve using user-centric contextual information. However, Gaia has a weak learning function to provide a service considering the user's preferences and inclinations. So Gaia cannot generate or process learned context information.

 CAMUS (Context-Aware Middleware for Ubiquitous computing Systems) is a unified middleware developed to support context awareness. The main feature of this CAMUS is an ontology-based framework that enables reuse and sharing of formal knowledge with other systems. CAMUS consists of six components that can generate sensory information from various sensors and new deduced contextual information based on that information. In particular, CAMUS provides a variety of reasoning mechanisms to provide more detailed representation. It uses various types of logic (first order logic logic, fuzzy logic, narrative logic, etc.) as well as machine learning techniques. However, CAMUS is not clearly defined or structured as a component for learning.

Such context-aware middleware should provide sensor context data from various sensors to generate various context information through integration, inference, and learning. By the way, in the above-described middleware, a component for integrating and inferring it based on context information obtained from Senger in a ubiquitous environment is disclosed, but a component having a learning ability to create a higher level context based on this is disclosed. Did.

An object of the present invention is to solve the problems described above, the context-aware middleware to generate and process a variety of context information through the integration, inference and learning of the sensed context information, to create a high-level context, With the created context information, it is to provide the appropriate service according to the characteristics of the user.

Another object of the present invention is to provide a framework that constitutes a middleware that can actively cope with changing surrounding conditions.

In the ubiquitous computing environment equipped with a sensor for recognizing a user's situation according to the present invention, a context description module for managing context information obtained from a sensor is provided. A context normalization module that converts the context information recognized from the context into a normalized structure, a context integration module that generates an integrated context based on the recognized context from the context information provided by the context normalization module, and a higher level from the recognized context information on the lower level. A context inference module that infers a level of context information, a context filtering module that selects only useful contexts and stores them in a database, and a context learning module that obtains learned context information based on information stored in a database. Characterized in that it comprises a module automatically triggering event for your safety and convenience issue commands for the actions or events.

In addition, in the context aware framework for ubiquitous computing middleware according to the present invention, the context description module is sensed using a context description table composed of C_Name, C_Type, C_Value or C_Rule, Alarm condition based on extensible markup language (XML). It is characterized in that the validation of the context (validation).

In addition, in the context-aware framework for ubiquitous computing middleware according to the present invention, the text normalization module is characterized by data abstraction and data conversion.

In addition, in the context aware framework for ubiquitous computing middleware according to the present invention, the context integration module is an object-oriented or predicate-based structure, integrating contextual information related to each other into a structured structure. Characterized in that.

In addition, in the context aware framework for ubiquitous computing middleware according to the present invention, the context inference module is characterized by using expert knowledge of a knowledge base to generate new context information.

In addition, in the context aware framework for the ubiquitous computing middleware according to the present invention, redundancy check or unnecessary data of the context information provided by the context normalization module and the context integration module is removed, and the context learning module removes user-related context information. It is characterized by storing the context information necessary for learning in the history DB.

In addition, in the context-aware framework for ubiquitous computing middleware according to the present invention, the context learning module includes a neural network, a decision tree, an association rule finding, a summary for context learning of a history database. (summarization).

The above and other objects and novel features of the present invention will become more apparent from the description of the specification and the accompanying drawings.

EMBODIMENT OF THE INVENTION Hereinafter, the structure of this invention is demonstrated according to drawing.

In the ubiquitous computing environment, the main purpose of the context awareness system is to provide the most appropriate service to the user by predicting the desired service in advance by using context information collected from the user's current situation or device. As such, context is an important factor to consider in intelligent applications in ubiquitous computing environments. Contexts are defined differently for each application used, but are generally defined as information used to characterize the context of objects (people, places, objects, etc.) that exist in a ubiquitous environment. More specifically, the information may be defined as information related to a user and an application, context information, and correlation information therebetween. Depending on how you identify this context, you can define it like this:

* Sensed context information *

In a ubiquitous environment, context information is most easily obtained through sensors. This context is derived from various types of sensors (eg, user ID, light intensity, temperature, location, humidity, time, etc.) or sensors attached to the user or sensors attached to all surrounding objects. This type of context is called an "elementary context," or "raw context."

* Combined context information *

Elementary context alone does not provide adequate services to users. Therefore, the user's status, surrounding conditions, and available surrounding resources must be considered in complex. Higher-level context can be created from the default context. However, because the sensed context is unstructured information, a simple transformation or standardized representation of the sensed data is required. In other words, a conversion operation is required to combine and structure congested and scattered information. Contexts generated from combinations of sensory contexts or combinations of integrated contexts are re-expressed in certain constructs. For example, an aggregator, a component of Context-Toolkit, integrates and aggregates basic contexts obtained through sensors to provide complex contexts.

Inferred context information

As used herein, the term 'inference' generally refers to inferring or deducing the state of a system, environment, and / or user from a set of observations captured through events and / or data. Say process. Inference can be used to identify a specific context or action or can generate a probability distribution over states, for example. Inference can be probabilistic, i.e., calculating the probability distribution for the state of interest, taking into account data and events. Inference can also refer to techniques used to create higher level events from a set of events and / or data. As a result of this inference, a new event or action is created from a set of observed and / or stored event data, regardless of whether the events are closely correlated in time and also from which the event and data come from one event and data source. Regardless of how many sources it came from.

Many higher-level contexts can be inferred based on contexts collected from sensors or integrated. It can also infer about the user's surrounding relationships, the user's activity in the room (eg, meetings, presentations, or TV viewing), or the user's role. Domain knowledge is required to infer high-level context. Domain knowledge helps to infer new information from the sensed or integrated context information. This inferred context is new information, not sensory or combined contexts. In contrast, the integrated context information is a kind of structured sensory context information.

* Learned context information *

Context can be obtained through learning as well as sensing and inference. In the case of the service considering the user's preference and disposition rather than the existing uniform service, a learning technique is required. It is difficult to identify user's patterns or preferences through inference from a large number of data, and can be found through learning or data mining techniques. For example, historical user data such as user behavior, speech, and psychological state can predict which areas the user is interested in and what to do.

In order to provide more effective service for the above users, context-aware middlewares should be able to handle the sensed context through integration, inference, and learning.

In the present invention, a new framework is described according to FIG. 1 to provide a context aware system, which is a general framework for context awareness. 1 is a block diagram of a framework designed to allow sensored, integrated, inferred, and learned contextual information to operate without special assumptions in hardware or middleware.

In order to provide the appropriate service to the user using the context obtained from the sensor, the framework includes the functions of integrating, inferring, and learning the context collected from various sensors and agents. As shown in FIG. 1, the framework according to the present invention consists of eight modules and two databases. Each module will be described in detail as follows.

* Context description module (Context Description Module) *

The context description module manages the context information obtained from the sensor. In general, context-aware applications assume that context z information from various sensors is complete and accurate in a ubiquitous environment. However, the actual sensed context information is often unavailable due to noise, sensor failure, or user error. XML is used to manage this uncertain and ambiguous context information. Also, before storing the contexts from the sensors, a validation of the sensored context is verified using a predefined context description table. To do this, we use a parser to verify the xml instance. The context description table is defined with four attributes in Table 1 of C_Name, C_Type, C_Value or C_Rule, Alarm condition.

 Property  Definition  C_name  Context.User.Biomedical.Temperature  C_type  Sensored, Combined, Inferred, Learned  C_value | C_rule  C_value: Int, Float, String, Boolean C_rule: Rule definition when C_type is Inferred  Alarm condition  Rule definition needed to inform user based on current context

Table 1: Context description table attributes

In the context description table, context information may be represented as XML data as shown in FIG. 2. For example, the user's temperature context can be represented as Context.User.Biomedical.Temperature. Context information is largely classified into User, Environment, System, and Application.

Context.User.Biomedical.Standweight Context information whose context type is "combinded" is stored after calculation in the defined formula whenever Context.User.Biomedical.Height changes. The formula defined in (C_value) is calculated using the Java Expert System Shell (JESS), a JAVA-based expert system.

Context Normalization Module

The context normalization module is responsible for converting context information from sensors into a normalized structure. In ubiquitous computing environments, information is typically collected from various sensors. However, there are various types of sensor data, such as binary, integer, real number, character, data combined with numbers and characters, and the like. Since each of the sensed data is represented in various ways, it is not only difficult to use context information, but also difficult to generate a higher level context from a lower level context. This module presents the sensored data using a normalized structure. In addition, the context normalization module can perform data augmentation and data conversion at a certain level, for example, converting a temperature from Fahrenheit to Celsius. For example, when prescribing some information, "predicate" is used, "predicate" (1592, 1609) indicating a GPS position value indicates the position of room 1, and a time value obtained through a timer (1012902900). Represents a time of (2005, 03, 15, 10: 30: 0), and (215) obtained through a temperature sensor represents 20.5 degrees Celsius.

Context Aggregation Module

The context aggregation module generates an integrated context based on the context information sensed from the context information provided from the context normalization module. This module integrates the sensored context into a structured form based on user ID, device ID, or location ID. This structured form may be object-oriented or predicate-based. The main purpose of this module is to integrate scattered but related contextual information into a well-structured structure for efficient use of context. Integrating this module requires knowledge of domains or rules for combination. The integrated context will be used to provide more tailored services to individuals as well as organizations, in particular. For example, once the context normalization module generates contextual information called Location (2010), Owner ("Tom"), NearObject ("piano") and Time ("10:10") from data sensed from a portable device, Serve Integrating each of these context information is as follows.

Room (Number, 2010, Person, "Tom", Object, "piano", Time, "10:10")

In other words, the combined result shows that Tom plays the piano in room 2020 at 10:10.

As such, the output of the aggregation module is passed to the filtration module for storage in the history database as well as the "current context" database.

Context Inference Module

Since we cannot collect all the information from the sensors, we need a context inference module. The context inference module generates the higher level context information from the lower level sensed context information. This module can infer new context information from current context information. The new contextual information will be inferred whenever the user's situation or surrounding environment changes. In addition, when an application requests context information from a query processing module for a question, if the information does not exist, inference generates new information desired by the application. The knowledge base of knowledge is used to generate new context information. For example, the activity information, or presentation, of a room user based on a low-level context, such as the number of people in the room, the identity of the people, the appliances or office equipment in operation, or the level of noise. ), Infer high-level context information, such as meeting (meeting) or sleeping (sleeping).

Context Filtering Module

The context filtering module selects only useful contexts and stores them in a history database. This module eliminates redundant or unnecessary data in the context information provided by the context normalization module and the context integration module. Therefore, context information necessary for learning user-related context information in a learning module is stored in a history DB. This module groups or categorizes contexts related to users or applications before storing them in the history database. If you have too much data to learn or include unnecessary data, it will probably take a long time to learn and the results will be poor. Thus, through the filtering module, only context information required for learning is obtained.

* Context Learning Module *

Ubiquitous computing environment must provide service to user according to various changing situation. Therefore, the ubiquitous computing environment needs to know the user's preferences and behavior patterns in order to provide better service to the user despite the situation of the user and the change of the surrounding environment. This kind of context information can be used for service prediction and preparation. Inference systems, however, have the disadvantage of requiring strict rules made by humans. In addition, it is difficult to adequately cope with the changing environment and difficult to cope with fluidly. Therefore, middleware needs a learning mechanism to provide appropriate services to users in a dynamically changing environment. The learning module will apply the most appropriate learning method: neural network, decision tree, association rule finding, summarization, etc. to the data in the history database. In general, learning takes a long time and requires a large amount of data. However, devices in ubiquitous computing environments have insufficient resources, and require rapid and incremental learning.

* Event Triggering Module *

In a ubiquitous environment, some services or actions need to be automatically performed based on current context information. These services must be done without user intervention. The event triggering module notifies an application or a user when a predefined context is detected in the current situation or environment. For example, if something dangerous happens around the user, the user can be notified by a message or voice so that it can be replaced appropriately. Or give important information such as important tasks (meetings, classes, scheduling, etc.) or family anniversaries.

Each context module has been described above, and now, as one embodiment, a framework in a music selection system is constructed. The music selection system recommends music to the user according to the mood of the room and the user's musical taste. Therefore, the system needs to monitor the sensor and look at the record of the music that the user recently listened to. The system is composed of an application portion and a situation awareness portion as shown in FIG.

The context aware portion is constructed based on the framework proposed above. The music selection system uses two sensors to detect the mood of the user's room as illuminance and temperature. Refer to the latest listening list to know the user's favorite songs. Therefore, a context normalization module is needed for the sensed data, a context inference module is needed to select a suitable mood from the sensed data, and a learning module is needed for the user's favorite song. The application selects the appropriate music based on the output of the context aware portion, the mood and the user's favorite song.

The context normalization module receives context data (relative to light and temperature) from the sensor. Then change it to a predicate form. The context inference module infers the mood of the scheme based on the sensed context using JESS, a JAVA-based expert system. Examples of inferences are depression, active, calm, static, and lively. Music is classified using fuzzy categories based on genres to investigate the musical taste of the user. The categories of music are ballad, R & B, jazz, pop, dance and classical. Fuzzy interests are also used to investigate user's favorite songs. You only need to search the list of the last 100 songs you listened to. Since it is difficult to implement a hardware sensor, the simulation is performed using software. Applications like these are simple and basic, but they can bring a lot of fun to us.

As mentioned above, although the invention made by the present inventor was demonstrated concretely according to the said Example, this invention is not limited to the said Example and can be variously changed in the range which does not deviate from the summary.

As described above, according to the context-aware framework for the ubiquitous computing middleware according to the present invention, in order to provide appropriate services in consideration of the dynamically changing user and the surrounding environment, various types of contexts, i.e., sensed, It can easily handle combined, inferred, and learned contexts, and the various types of context information generated through the framework can be used in wireless Internet environments and in the emerging ubiquitous environment. Providing an appropriate service to the user can be obtained.

Claims (7)

In a ubiquitous computing environment equipped with a sensor that recognizes a user's situation, A context description module for managing context information obtained from the sensor; A context normalization module for converting the context information recognized by the sensor into a normalized structure; A context integration module that generates an integrated context based on the recognized context from the context information provided by the context normalization module; A context inference module that infers higher level context information from lower level recognized context information; A context filtering module that selects and stores only useful contexts in a database; A context learning module for obtaining learned context information based on the information stored in the database; Event triggering module that automatically orders actions or events for your safety or convenience A context aware framework for ubiquitous computing middleware, comprising: a. The method of claim 1, The context description module is a context aware framework for ubiquitous computing middleware that validates a sensed context using a context description table consisting of XML based C_Name, C_Type, C_Value or C_Rule, and Alarm condition. The method of claim 1, The context normalization module is a context aware framework for ubiquitous computing middleware, characterized by data abstraction and data conversion. The method of claim 1, A context integration module is an object-oriented or predicate-based structure that incorporates contextual information related to each other into a structured structure. The method of claim 1, A context inference module is a context-aware framework for ubiquitous computing middleware, which uses knowledge base expert knowledge to generate new context information. The method of claim 1, Eliminating redundancy checking or unnecessary data of the context information provided by the context normalization module and the context integration module, and storing context information necessary for learning user-related context information in the context learning module in a history database. Context-aware framework for ubiquitous computing middleware. The method of claim 1, The context learning module uses neural networks, decision trees, association rule findings, and summarization for contextual learning in the history database, for context awareness for ubiquitous computing middleware. Framework.

Images