Scenario-Based Modeling of Multi-Agent Systems

Constant expansion of network-centric services inevitably led into development of new software technologies that would enable seamless and transparent access to the expanding amount of information. The software agent-oriented model fits convincingly well within this context as a better-suited technology over typical modular, client-server approach. This model offers concepts like autonomous behavior, competitive or collaborative interactions, and seamless integration with the legacy systems (Jennings et al., 1999). Based on their capabilities and predefined and/or accumulated knowledge, agents can react to changes by adapting to the new circumstances if a better approach is identified. This results in a dynamic system, well suited to coping with an ever-changing environment. Agents individually choose to co-operate or compete with other agents in order to satisfy their own objectives, but by setting goals appropriately, their collective behavior can be engineered to achieve global system-wide objectives. This approach is an efficient way of handling the complexity of many modern software systems. Within the competitive market context, agents interact with each other in order to win access to the shared resources, to get a better price or to bet for more processing power, etc. while trying to fulfill their plans by achieving given goals. Alternatively, cooperative environment promotes such agents that will perform their goals in the interest of the wider community or the authoritative entity that secures the fulfillment of the global goal. Typical examples would be applications for task planning and resource scheduling, search engines, or any other where the emergent behavior is influenced by collaborative and mutually nonexclusive individual goals. Regardless of the environment characteristics used, agent’s communication is achieved through asynchronous and message-oriented interactions. In addition to physical connection, it requires semantics in order to enable agents to reach the concluding state of the interaction through common understanding of the messages and their meanings. Consequently, the dialogue ontology and conversational semantics has to be defined within the framework of a conversation space. This space is defined as a sequence of messages exchanged between agents following a (set of) defined dialogue protocol(s). Dialogue protocols enable agents to take part in conversations by committing to the shared protocol semantics and defined conversation space within which agents are enabled to act, still preserving their decision-