Multi-agent Systems: A Software Architecture Viewpoint
Studies in agent-oriented software engineering address the merit of agents and multi-agent systems as a software architecture style, though only in part. MAS software architecture styles are of interest to both the MAS and the software engineering communi
- PDF / 383,603 Bytes
- 22 Pages / 439.36 x 666.15 pts Page_size
- 85 Downloads / 233 Views
Multi-agent Systems: A Software Architecture Viewpoint Onn Shehory and Arnon Sturm
Abstract Studies in agent-oriented software engineering address the merit of agents and multi-agent systems as a software architecture style, though only in part. MAS software architecture styles are of interest to both the MAS and the software engineering communities. This chapter provides an introduction to these software architectures. As we demonstrate, MAS implementations spanning across decades have several common architectural characteristics, despite different design and implementation details. An important question associated with MAS development is whether MAS constitute an appropriate solution for a computational problem at hand, and if so, what type of MAS should be preferred for that solution? Preferably, this question better be answered early on, to prevent the use of MAS as a solution approach where simpler, more efficient solutions apply. MAS should be considered among an array of alternative solution approaches. To assist system designers in their assessment of MAS as a solution approach to their problem, we present architectural properties of MAS and we demonstrate these properties by example. Keywords Software architecture • Agent-oriented software engineering • Architectural styles • MAS design
O. Shehory () IBM—Haifa Research Lab, Haifa, Israel e-mail: [email protected] A. Sturm Department of Information Systems Engineering, Ben-Gurion University of the Negev, Beer-Sheva, Israel e-mail: [email protected] O. Shehory and A. Sturm (eds.), Agent-Oriented Software Engineering, DOI 10.1007/978-3-642-54432-3__4, © Springer-Verlag Berlin Heidelberg 2014
57
58
O. Shehory and A. Sturm
1 Introduction The field of Multi-agent Systems (MAS) focuses on solutions to decentralized problems in dynamic, open computational domains. It discusses issues of intelligence which include, among others, learning [1], negotiation [2, 3], strategic behavior [4, 5], social behaviors and norms [6–8], cognitive activities, and mental states (reasoning [9], beliefs, desires, intentions [10], emotions [11]). Thus, MAS combine research from several fields, and in particular from the fields of Distributed Artificial Intelligence (DAI) and Software Engineering (SE). The latter two constitute the subfield of Agent-Oriented Software Engineering (AOSE), which complements MAS research with practical means for developing multi-agent systems that implement the desired intelligence properties. A major focus of AOSE, as reflected in this book, is placed on methods, methodologies, and tools for life cycle support of agents and MAS. Methodologies commonly address a wide range of problem domains and types. Yet, when attempting to address a specific problem, system developers frequently need to develop software components and architectures which may already be present in other solutions. In such cases, blueprints of agents and MAS—i.e., their software architectures (SAs)—should be most valuable and facilitate reuse. The focus of this chapter is on such archi
Data Loading...
