Academia.eduAcademia.edu

Figure 5 – uploaded by Juan Pavón

See full PDFdownloadDownload figure
The RICA (Role/Interaction/Communicative Action) approach (Serrano J. M. & Ossowski S., 2004) integrates relevant aspects of Agent Communication Languages (ACL) and Organisational Models and it is itself based on the concepts of Communicative Roles and Interactions. RICA describes a conceptual framework that guides the designer from the specifications of the organisational model of the agent society to the definition of the Agent Communication Languages of the multi-agent system.

Figure 5 The RICA (Role/Interaction/Communicative Action) approach (Serrano J. M. & Ossowski S., 2004) integrates relevant aspects of Agent Communication Languages (ACL) and Organisational Models and it is itself based on the concepts of Communicative Roles and Interactions. RICA describes a conceptual framework that guides the designer from the specifications of the organisational model of the agent society to the definition of the Agent Communication Languages of the multi-agent system.

Related Figures (9)

Figure 1. The MAS Meta-model adopted by ADELFE. An agent belonging to an AMAS ignores the global function of the system, only pursues a local goal and tries to always keep cooperative relations with other agents. Such an agent is called a “Cooperative Agent” because its social attitude is based on cooperation, but its lifecycle is still a classical one. It consists in having perceptions, taking decisions and then doing actions (perceive-decide-act). Local cooperation rules enable the agent to detect and solve Non Cooperative Situations (NCS) that play a fundamental part in the ADELFE design. These NCS are cooperation failures (e.g., cooperation protocol not obeyed, or unpredictable situations) and they are inconsistent with the cooperative social attitude of a cooperative agent.
Figure 1. The MAS Meta-model adopted by ADELFE. An agent belonging to an AMAS ignores the global function of the system, only pursues a local goal and tries to always keep cooperative relations with other agents. Such an agent is called a “Cooperative Agent” because its social attitude is based on cooperation, but its lifecycle is still a classical one. It consists in having perceptions, taking decisions and then doing actions (perceive-decide-act). Local cooperation rules enable the agent to detect and solve Non Cooperative Situations (NCS) that play a fundamental part in the ADELFE design. These NCS are cooperation failures (e.g., cooperation protocol not obeyed, or unpredictable situations) and they are inconsistent with the cooperative social attitude of a cooperative agent.
Figure 2. The MAS Meta-model adopted by Gaia. Having a deeper look at the MAS meta-model for the second, extended version of Gaia (Figure 2), we notice that the basic building blocks of the former version of Gaia — namely agents, roles, activities, services, and protocols — are still present but now they are located in the context of a specific environment and of a specific organization.
Figure 2. The MAS Meta-model adopted by Gaia. Having a deeper look at the MAS meta-model for the second, extended version of Gaia (Figure 2), we notice that the basic building blocks of the former version of Gaia — namely agents, roles, activities, services, and protocols — are still present but now they are located in the context of a specific environment and of a specific organization.
Figure 3. Summary of INGENIAS MAS meta-model
Figure 3. Summary of INGENIAS MAS meta-model
The Problem Domain (Figure 4) deals with the user's problem in terms of scenarios, requirements, ontology and resources; scenarios describe a sequence of interactions among actors and the system. Requirements are represented with conventional use case diagrams.
The Problem Domain (Figure 4) deals with the user's problem in terms of scenarios, requirements, ontology and resources; scenarios describe a sequence of interactions among actors and the system. Requirements are represented with conventional use case diagrams.
From the beginning (Early and Late Requirements analysis), the designer deals with domain stakeholders (modelled as social actors) and the goals they want to reach, resources they can share and plans they may perform. Actors are related by mutual and intentional dependencies that are consequences of their goals. In the next phase (Architectural Design), actors are mapped to a set of software agents while in the final Detailed Design, agents capabilities and interactions are defined in details. Differently from the other MAS meta-models, Tropos introduces the concept of actor as a generalisation of the agent (Bertolini D., et al., 2005) (Figure 6). This is the direct consequence of the early requirements phase, where actors are initially identified and then translated to possible agents during the architectural design. Tropos role is an abstract characterisation of the behaviour of a social actor and a set of roles compose a position. The concept of position occurs in the architectural design phase, where agents are used to occupy previously identified positions.
From the beginning (Early and Late Requirements analysis), the designer deals with domain stakeholders (modelled as social actors) and the goals they want to reach, resources they can share and plans they may perform. Actors are related by mutual and intentional dependencies that are consequences of their goals. In the next phase (Architectural Design), actors are mapped to a set of software agents while in the final Detailed Design, agents capabilities and interactions are defined in details. Differently from the other MAS meta-models, Tropos introduces the concept of actor as a generalisation of the agent (Bertolini D., et al., 2005) (Figure 6). This is the direct consequence of the early requirements phase, where actors are initially identified and then translated to possible agents during the architectural design. Tropos role is an abstract characterisation of the behaviour of a social actor and a set of roles compose a position. The concept of position occurs in the architectural design phase, where agents are used to occupy previously identified positions.
The MAS meta-models presented in the previous subsections will now be compared with the precise air of identifying their commonalities as a first step towards a common agreed part of a unified MAS Meta- Model (MMM), and studying their distinctive differences (that can positively enrich the collection ot common elements by introducing ideas coming from just one or a few approaches).
The MAS meta-models presented in the previous subsections will now be compared with the precise air of identifying their commonalities as a first step towards a common agreed part of a unified MAS Meta- Model (MMM), and studying their distinctive differences (that can positively enrich the collection ot common elements by introducing ideas coming from just one or a few approaches).
From the experience of merging these three models we learnt that their composition adds some significant improvements to the new structure since they complement each other in several aspects. For example, the ADELFE representation that the agent has of its environment, the Gaia environment and the PASSI ontology, naturally relates by representing the fact that an agent has a representation (possibly affected by errors or uncertainty) of the environment expressed in terms of an ontological model of it.
From the experience of merging these three models we learnt that their composition adds some significant improvements to the new structure since they complement each other in several aspects. For example, the ADELFE representation that the agent has of its environment, the Gaia environment and the PASSI ontology, naturally relates by representing the fact that an agent has a representation (possibly affected by errors or uncertainty) of the environment expressed in terms of an ontological model of it.
Table 1. A comparison of the discussed MAS meta-models from the structural point of view
Table 1. A comparison of the discussed MAS meta-models from the structural point of view
Obviously, the central concept is a generic agent from which a cognitive agent can be derived. An agent is situated in an environment for which it is able to build representations if it possesses cognitive capabilities. An agent can also be part of the environment of other agents, that explains the bidirectional link between these two concepts. An agent communicates with others and to communicate with an intended purpose can use FIPA-compliant conversations. A role is related to an agent (agents play roles); some agents can use tasks without playing any role and therefore it becomes necessary to relate “agent” to “task” with a O..n cardinality. Finally, following the previous discussion on the definition of “organisation”, an organisation may be composed of agents but also of roles depending on the concerned methodology. This is expressed on the MAS meta-model by following navigation links from “agent” to “organisation” via “role”.
Obviously, the central concept is a generic agent from which a cognitive agent can be derived. An agent is situated in an environment for which it is able to build representations if it possesses cognitive capabilities. An agent can also be part of the environment of other agents, that explains the bidirectional link between these two concepts. An agent communicates with others and to communicate with an intended purpose can use FIPA-compliant conversations. A role is related to an agent (agents play roles); some agents can use tasks without playing any role and therefore it becomes necessary to relate “agent” to “task” with a O..n cardinality. Finally, following the previous discussion on the definition of “organisation”, an organisation may be composed of agents but also of roles depending on the concerned methodology. This is expressed on the MAS meta-model by following navigation links from “agent” to “organisation” via “role”.
Related topics:
Cognitive ScienceComputer ScienceSoftware EngineeringMulti Agent SystemAgent Oriented Software Engineering
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved