Sung Keun Song

CALM: an intelligent agent-based middleware for community computing

The applications for ubiquitous system require to effectively use the resources distributed in the environment for providing customized services to the users. In this paper we propose a middleware architecture displaying high scalability with event-based multi-agent platform, which allows dynamic creation of community for service oriented computing. It consists of event-based communication platform implemented by extending omniORB, agent platform, and various frameworks. The effectiveness of the proposed architecture is demonstrated using a ubiquitous office service scenario

A New Agent Platform Architecture Supporting the Agent Group Paradigm for Multi-Agent Systems

Nowadays, the agent grouping paradigm receives a great attention by the multi-agent system(MAS) developers. Meanwhile, the existing agent development tools do not effectively support agent grouping. In this paper we propose a new agent platform architecture with which agent groups can be dynamically created, deleted, and federated with other groups. We also propose a service discovery approach that can advertise and discover agents and agent groups without broadcasting the messages. The proposed agent platform is actually implemented and tested.Nowadays, the agent grouping paradigm receives a great attention by the multi-agent system(MAS) developers. Meanwhile, the existing agent development tools do not effectively support agent grouping. In this paper we propose a new agent platform architecture with which agent groups can be dynamically created, deleted, and federated with other groups. We also propose a service discovery approach that can advertise and discover agents and agent groups without broadcasting the messages. The proposed agent platform is actually implemented and tested.

Context-based Dynamic Channel Management for Efficient Event Service in Pervasive Computing

In pervasive computing environment a number of distributed applications asynchronously exchange data using the event-based mechanism. The OMG defined CORBA Event Service for asynchronous messaging, which provides a flexible channel for asynchronous communication among the distributed and collocated objects. In the existing Event Service channels are created manually by the developers or users to process and disseminate the events. Also, the channel management is static which may cause bottleneck in some channels. In this paper, thus, we propose a dynamic context-based channel management scheme by efficiently combining name service with event service. It is embedded in the CALM(Component- based Autonomic Layered Middleware) developed by us, which has been implemented based on ORB. Computer simulation reveals that the proposed approach significantly reduces the event delivery and processing time of the existing omni Event Service and TIBCO.

A new reflective and reliable context-oriented event service architecture for pervasive computing

Conventional middleware technologies lack the support for handling dynamic aspects of the new pervasive computing environment. Reflective middleware is a principal and efficient system which can deal with ubiquitous environment. It can reliably and quickly deliver important events and filter useless ones. In this paper we propose a new reflective and reliable context-oriented event service paradigm. The proposed event service supports reliable communication as well as reflective filtering through the adapters. It also supports reflective and dynamic context-oriented channel management for evenly distributing the load. An experimentation with six PCs reveals that the proposed scheme outperforms an existing popular commercial middleware in terms of event delivery speed. The improvement gets more significant as the size of event increases.

A load balancing scheme for multi-agent systems based on agent state and load condition

In ubiquitous environment intelligent and transparent services are provided to the users anytime and anywhere. Ubiquitous systems can be effectively implemented using the agent technology, while the agent platform supports efficient and stable interaction among the agents. In this paper we propose a load-balancing scheme which considers agent state for fair and efficient resource allocation to the agents. It is achieved by the migration of agents decided according to the condition of message load and resource allocation. Experiment with an actual multi-agent system shows that the mean round trip time is significantly reduced compared to the existing scheme especially when the agents frequently change the state and transmit the messages.

Private and Secure Service Discovery Using Incrementally Progressive Exposure and Random Match

Service discovery is an essential element in pervasive computing environment. The involvement of only the needed users and service providers for a service discovery in pervasive computing environment is challenging. To protect their privacy, neither side is willing to expose their information first. A incrementally progressive exposure approach is used to solve the problem in this paper. In the approach users and service providers expose partial information in turn and avoid unnecessary exposure if any mismatch occurs. To reduce the unnecessary match after an exchange especially the first round, a segment is randomly picked from the information and exchanged with the partial information together. To reduce the number of exchange, we gradually increase the length of exchange message if a match occurs.

Publisher-side Event Filtering for QoS-Awareness in Ubiquitous Computing

In the ubiquitous computing environment a user may interact with a large number of devices. The publishsubscribe paradigm is suitable for the various applications by allowing the subscribers to declare their interests and receive only the notifications of interest. This paper propose a novel event filtering approach preventing waste in the space of the event queue of the event channel due to unwanted events and reducing the network load. Performance evaluation reveals that the proposed scheme displays significant improvement over the existing event service, Java messaging service (JMS) and Omnievent service in terms of event delivery time for various numbers of servers and data sizes.Data mining in the stream data handles quality and data analysis using extremely large and infinite amount of data and disk or memory with limited volume. In such traditional transaction environment it is impossible to perform frequent items mining because it requires analyzing which item is a frequent one to continuously incoming stream data and which is probable to become a frequent item. This paper proposes a way to predict frequent items using regression model to the continuously incoming one dimensional stream data like the time series data. By establishing the regression model from the stream data, it may be used as a prediction model to uncertain items. The proposing way will exhibit its effectiveness through experiment in stream data.

Dynamic Software Adaptation with Dependence Analysis for Multi-Agent Platform

Modern software system is required to be flexible and adaptable to deal with dynamic environment. To address this new requirement, meta-level architecture description-based dynamic software adaptation has been proposed, which refers to the software system changing its structure and behavior at runtime without user intervention. In this paper we propose an approach for multi-agent system offering dynamic agent adaptation by runtime reconfiguration. The proposed runtime agent adaptation process employs the matrix-based dependence analysis model to check if the performance requirement is satisfied by the reconfiguration. It will thus simplify the development, deployment, and maintenance of the system using the adaptation facilities. An agent platform employing the proposed approach has been implemented, which complies with the FIPA specification.

A priority-based message scheduling scheme for multi-agent system dynamically, adapting to the environment change

Multi-agent system consists of a number of cooperative agents which have agreed to work together for a common goal. When scheduling the messages in an agent system, if the same priority is assigned to all the agents without considering the importance of their role, the performance may not be maximized. In this paper we propose a new priority-based message scheduling scheme. It enables the agent system to quickly adapt to the environment change by properly modeling the priority of the messages based on the importance of the agents and message throughput. The performance of an agent system is shown to be greatly improved by the proposed scheme using a case study of crime prevention system which was implemented with the agent platform developed by the authors.

An efficient and reflective event filtering for context-awareness in ubiquitous computing

In the ubiquitous computing system fast and reliable event delivery is important to properly adapt to the dynamically changing environment. Such requirement could be satisfied by message-based middleware based on the publish/subscribe model. In this paper we propose a scheme that allows reflective filtering through the reasoning process. The objective of the reflective filtering is to prevent waste in the space of the event queue of the event channel due to useless events and reduce the network load. Performance evaluation reveals that the proposed scheme displays a significant improvement over the existing event service, Java messaging service (JMS) and omnievent service, in terms of event delivery time.