Behavioral Modeling of Multi Agent System: High Level Petri Net Based Approach

Introduction

Multi-Agent System (MAS) based computing promotes, designing and developing applications in terms of autonomous software entities (agents), situated in an environment, and that can flexibly achieve their goals by interacting with one another dynamically (Zambonelli et al., 2004). Besides autonomous nature, an agent exhibits several other crucial features including goals, capabilities, situatedness, proactive/reactiveness, knowledge driven, resource driven, event driven and heterogeneity; and which have been summarized in recent literatures (Zambonelli et al., 2004; Wooldridge et al., 2001; Biswas et al., 2008; Chatterjee et al., 2011) . Also dynamicity is the inherent characteristic for MAS due to event driven nature and features like autonomous and reactiveness. The initial state, knowledge and goals are set. MAS manage the things dynamically to achieve the preset goals. Coordination plays a fundamental role in MAS, since it allows agents to interact with one another in a productive way (Cabri et al., 2010). This can be achieved through the modeling of Interactions among the agents in the environment. Moreover, in MAS, each agent plays a specific set of Roles to interact with another agent or other environmental elements to achieve a pre specified goal. Further, the series of events and the responses to such events may occur dynamically in such system.

In this context an important challenge is to devise a mechanism to study the dynamic behavior of Agents in MAS at the design level. For such study and modeling of MAS behavior, it is to be ensured that, (i) system will achieve the goal with finite number of events and interactions, (ii) system will operate in deadlock-freeway, as the system will be handling the resources from the environment, (iii) system and environment should transform in acceptable states with the occurrences of events and interactions, (iv) the knowledge and the state of the resources are dynamically manageable. In view of these features, Petri Net (Murata et al., 1989) based approach is obvious choice for modeling of such dynamic behavior of MAS. Such Conceptual modeling of MAS is useful to study the architectural semantics and defines the components and their inter relationship to conceptualize the environment, agent, related events and interactions.