Dwight Deugo

Reusable patterns for agent coordination

Much of agent system development to date has been done ad hoc. These problems limit the extent to which “industrial applications” can be built using agent technology, as the building blocks, reusable techniques, approaches and architectures have either not been exposed or have not yet been fully elaborated. In the mid 80’s, supporters of object-oriented technology had similar problems. However, with the aid of software patterns, objects have provided an important shift in the way developers successfully build applications today. In this paper, after describing an agent pattern’s generic format, we identify a set of software patterns for agent coordination.

Mobile agent messaging models

It is not easy to make the decision of which messaging model to use in a mobile agent system. There are many different models to choose from and there are many forces influencing each model. We discuss five messaging models: Home-Proxy, Follower-Proxy, Email, Blackboard and Broadcast. We overview the forces that are the most important to all models and identify to what extent the forces influence the models. Our findings suggest that one should not use the Forwarder-Proxy model to build mobile agent systems. Instead, one should use the Blackboard approach, especially when agents require only anonymous messaging. In the case when anonymous messages cannot be used, your choice of model can be the Home-Proxy model, the Email model or the Broadcast model depending on which forces are more important to your application.

Mobile agents for electing a leader

We present a technique for mapping a distributed algorithm to a set of homogeneous agents that solves the general distributed control problem known as election. Election is a decentralized, cooperative information task where one system or node from n is elected as the leader. Our solution is as efficient as the message-based election algorithm, but does not rely on message passing between custom protocols located on communicating nodes. Rather, the proposed solution relies on mobile agents. We show that the number of agent transfers is equivalent to the number of messages in the original algorithm. We also present an environment model for distributed agents that supports not only the mapping of election agents but other agent-based solution to distributed control problems that use distributed algorithms.

Modular Java web applications

As Java EE applications increase in size and complexity the constraints imposed by the existing component model restrict their utility. In this paper, we describe a solution to the problem related to building modular and evolvable server-side applications in Java. We use Eclipses OSGi runtime as a basis for solving the problem and describe its integration in a Java EE Application Server environment.

Deciding on a Pattern

Object-oriented software patterns account for knowledge regarding a solution to a programming problem in a context. Software patterns are increasingly popular and consequently their numbers are growing. Under these circumstances, it is a challenge for the pattern user to decide on which patterns to incorporate into their design. In this paper, we describe a pattern decision analysis approach that provides pragmatic support to making this design decision.

Supervised and unsupervised data mining with an evolutionary algorithm

This paper describes our current research with RAGA (Rule Acquisition with a Genetic Algorithm). RAGA is a genetic algorithm and genetic programming hybrid that is designed for the tasks of supervised and certain types of unsupervised data mining. Since its initial release we have improved its predictive accuracy and data coverage, as well as its ability to generate more scalable rule hierarchies. These enhancements and several experiments are described.

Management of mobile agent systems using social insect metaphors

The management of mobile agent systems that solve problems in a network is an issue that must be addressed if mobile agents are to be deployed industrially. It is clear that insufficient or excessive numbers of agents can cause the problem solving capabilities of an agent-based system to be impaired. Also, agents being software entities are almost always flawed therefore requiring the upgrade problem to be solved. This paper presents distributed algorithms based upon ant social behaviour that solve the problems of agent density maintenance and the agent upgrade problem.

Evolution to the Xtreme: evolving evolutionary strategies using a meta-level approach

In this paper we describe a meta-level evolutionary system that uses a meta-level GA to evolve strategies that perform better than known good strategies on a test bed of mathematical optimization problems. We examine the effects of the meta-level components and parameters on the problem set in order to help others in choosing the components and parameters for their meta-GAs.

A search for routing strategies in a peer-to-peer network using genetic programming

Results taken from a simulated peer-to-peer network are described, in which genetic programming is utilized to evolve routing strategies that optimize resource location in various traffic flow scenarios. In all cases the evolved strategies result in more numerous resource locations than a pure, non-adaptive peer-to-peer protocol such as the Gnutella protocol. The resulting evolved strategies are described, and empirical validation of the Gnutella protocol is given via both its creation through machine-learning techniques, and through the analysis of real-world constants used in the protocol.

Distributed Deadlock Detection in Mobile Agent Systems

Mobile agent systems have unique properties and characteristics and represent a new application development paradigm. Existing solutions to distributed computing problems, such as deadlock avoidance algorithms, are not suited to environments where both clients and servers move freely through the network. This paper describes a distributed deadlock solution for use in mobile agent systems. The properties of this solution are locality of reference, topology independence, fault tolerance, asynchronous operation and freedom of movement. The presented technique, called the shadow agent solution proposes dedicated agents for deadlock initiation, detection and resolution. These agents are fully adapted to the properties of a mobile agent environment.