The aim of this phase consists in designing the agents’ life from the social point of view. This mainly means updating the initial plan of each agent explicitly considering its social relationships (actuated using communications); in so doing the designer composes the tasks each agent owns in order to allow the pursuing of its objectives (in terms of the functionalities it has to accomplish according to what specified in the Agent Identification phase and services it can provide to the others as defined in the Role Description phase).
More in details, in this phase the designer draws one or more activity diagrams (if the system is very large) where the agents’ behavior is expressed in terms of sequence of tasks and communications. Each diagram can be regarded as a refinement of the initial agent specification drafted during the TSp (Task Specification) phase, but designed under a different perspective, the social one; while in the TSp diagram we were trying to define a first hypothesis of the agent plan and we were looking at an agent a time, in the MABD diagram, we are now considering the whole agent society, we are aware of the agent social relationships (communications), we know much more about the same agents’ structure (in terms of behaviors and roles played) and we can therefore prepare a more complete schema of the agents’ society life.
From the MAS meta-model point of view, as already reported in the artifacts structure of the Agent Society Model, this diagram mainly relates agents (through communications) among them and their tasks to the communications they participate. The MABD notation prescribes that the diagram is divided in swimlanes, (each swimlane representing an agent) and each activity inside them represents one of the agent’s tasks. These tasks are related by refining the initial TSp plan and also considering the roles played by the agents (see RD diagram).
The specific notation of a MAD diagram includes that transitions either represent events (e.g. a communication, represented by an object flow) or a control flow (from one task to the other). Moreover, since the syntax of UML activity diagrams already supports the representation of concurrency and synchronization we do not need to introduce in PASSI a specific diagram for concurrency, besides we suppose that each agent is executed concurrently to the others (even if some of these agents can be started later by the others) and in the same agent several roles can be concurrently active.
As regards the example of MABD diagram reported in the next figure, we can see that at system start, both the Agent_Society and COD agent are created. The Agent_Society agent from the beginning plays two different roles as already specified in the RD phase (see Figure 17, the User_Interaction roleis devoted to guiding the designer in performing the process using the correct sequence of phases, and the MAS_Model_Keeper, thatis responsible for providing information about the already designed parts of the system to fragment-level agents such as the COD); the first step in managing the design process is done by the ASModel_Sequencer task that accordingly to user’s requests can fire the next phase (task Fire_new_phase), check the model for errors or inconsistencies (task Check_Design_Model) and so on. To fire the COD phase, the Agent_Society agent starts a communication (with the Request interaction protocol) and it asks to the COD agent of beginning the composition of the diagram that is under its responsibility.
Next phase: Single-Agent Structure Definition
Previous phase: Multi-Agent Structure Definition
Home: PASSI 2 homepage