In the previous phases we identified agents and their roles. Now, in the task specification phase we are facing the third element of our MAS meta-model: tasks.
While use case diagrams in the Agent Identification phase represent a functional description of the multi-agent collaboration (the whole agent society), in the Task Specification phase we look at one agent at a time, drawing one activity diagram for each agent in order to represent the activities performed by the agent to carry on its duty. Such a diagram represents the plan of the agent, reporting the relationships among the external stimuli received by the agent and its behaviour (expressed in terms of fired tasks).
Because of the specific nature of an activity diagram (a kind of state chart), it is very easy to model aspects like concurrency (several behaviors are executed concurrently in order to realize some intention), control structures (agent’s decisions) and incoming events (messages from other agents but also external events) and agent’s states. Contrarily, it is not easy to represent in this way other situations like non deterministic flow of control that could arise in some agents implementation.
As regards the Tasks Specification diagrams syntax (in form of activity diagrams), we draw one different diagram for each agent (see Figure 8) where each activity represents one task that the agent can do. Each diagram is divided in two different sections using swim-lanes: the right one reports the tasks of this agent (one activity for each task), the left swim-lane reports the tasks of the agents that interact with this one.
Relationships between the activities represent within-agent communications if the activities are located in the same swim-lane. These are not speech acts; they are often signals addressing the necessity of beginning an elaboration that is left to a specific task (i.e. signals to delegate another task to do something). If a relationship connects elements of two different swim-lanes, it represents a communication from an agent to another one (generally speaking this could be composed of more than one message, at this stage this has not been yet formalized); it is also possible that several related relationships will be later (in the COD phase, see sub-section 5.2.2) organized into one single communication. The inner behavior of each task can be specified more in detail by using sequence diagrams, other activity diagrams or semi-formal text.
In the next figurewe can see the TSp diagram for the COD agent introduced in C.O.Phase. It describes that the Agent_Society agent delegates the COD one to compose the diagram with the Start_Composition message that is received and acknowledged by the Compose_Diagram task; according to its plan, the COD agent asks for the list of interactions and then compiles the diagram (Design_diagram task). This agent also offers the possibility of introducing communication details (Detail_Communications task) and checks its pertinence part of the model (Check_Model task).
Next phase: Domain Ontology Description
Previous phase: Agent Structure Exploration
Home: PASSI 2 homepage