AgentGroup Agent and Pervasive Computing Group Dipartimento di Ingegneria dell’Informazione Università degli Studi di Modena e Reggio Emilia
04/06/2007AgentGroup2 Group members Modena –Letizia Leonardi, full professor –Giacomo Cabri, associate professor –Raffaele Quitadamo, PhD student –Francesco De Mola, PhD student –Mariachiara Puviani, PhD student –Luca Ferrari, collaborator Reggio Emilia –Franco Zambonelli, associate professor –Marco Mamei, research associate –Luca Cernuzzi, collaborator
04/06/2007AgentGroup3 Research activities Background –Distributed systems –Object-oriented programming Well-established –(Mobile) Software agents –Pervasive computing Current –Services –Autonomic computing
04/06/2007AgentGroup4 Software agents Support and infrastructures Interaction engineering –Coordination (context-aware) –Roles Ambient intelligence E-health
04/06/2007AgentGroup5 Agent interactions Agent sociality implies the modelling and implementation of interactions in agent systems Coordination Role-based agent interaction
04/06/2007AgentGroup6 Coordination Agent sociality implies interactions and synchronization –With other agents –With execution environment Different models Coupling –spatial –temporal
04/06/2007AgentGroup7 Taxonomy of the coordination models Spatial Temporal coupled uncoupled coupled uncoupled Direct Blackboard-based Meeting-orientedLinda-like AgletsAmbit MOLEJavaSpaces
04/06/2007AgentGroup8 Direct coordination The involved entities communicate directly Typical of the client-server and peer-to- peer models Internet sendTo(Minnie, “I love you”)
04/06/2007AgentGroup9 Linda-like coordination Interactions via a third entity Blackboard + pattern-matching –information obtained via partial knowledge Tuple spaces Tuple: ordered set of typed fields Operations: Read, Out, In
04/06/2007AgentGroup10 MARS Mobile Agent Reactive Space 100% pure Java portability Can be associated to different mobile agent system (Aglets, Java2go, SOMA)
04/06/2007AgentGroup11 MARS at work Tuple space Agent server Internet Working... take((2, null)) (2, “c”) (2, “u”) (7, “a”)
04/06/2007AgentGroup12 Adding reactivity The tuple space reacts to the agent accesses The behavior of the tuple space can be customized
04/06/2007AgentGroup13 Programmable reactivity The tuple-space reactions can be programmed More flexibility Separation between algorithmic and e coordination issues Implementation of policies of –application (agent interactions) –local environment (interactions with the environment)
04/06/2007AgentGroup14 MARS – reactive model (1) Reactions realized as meta-tuple Meta-space –Writing a meta-tupla installing a reaction –Deleting a meta-tupla uninstalling a reaction Example of meta-tupla: (TransformTake_Obj, null, “take”, “reader”)
04/06/2007AgentGroup15 MARS – reactive model (2) Tuple space Agent server Meta-Spazio di tuple reaction take((2, null))(2, “c”) read((2, null)) (2, “u”) (7, “a”)
04/06/2007AgentGroup16 Role-based agent interactions Role –Stereotype of behavior –Abstraction to model interactions Already exploited in OOP [Baumer et al., 1997; Fowler, 1997, Steimann 2003]
04/06/2007AgentGroup17 State of the art Different approaches analyzed from the development phases’ point of view
04/06/2007AgentGroup18 AgentINteractions The BRAIN Framework B. R. A. IN. BehaviouralRole
04/06/2007AgentGroup19 The BRAIN Framework Interaction model based on action-event roles as: set of capabilities expected behaviors Actions Events
04/06/2007AgentGroup20 The BRAIN Framework XML notation: Understandable by programmers and (someway) by programmers Interoperability Different views based on XSL
04/06/2007AgentGroup21 interaction infrastructure(s): implements the BRAIN model provides the action-event support enables agents to dynamically assume/use/release roles manages roles and related facilities The BRAIN Framework Role systems
04/06/2007AgentGroup22 The GAIA methodology Methodology to develop MAS Analysis and design Exploits also roles
04/06/2007AgentGroup23 E-Health Issues : –Centralized architectures –Closed and isolated solutions –Mobile and dynamic scenarios –Patients’ information Agents in Health Care: –Flexible management –Integration –Home assistance –Emergency management
04/06/2007AgentGroup24 UbiMedic Framework Territorial emergencies: distribution, dynamism, mobility, unpredictability, heterogeneity Mission-critical domain: reliability, flexibility, context-awareness, reactivity, real-time UbiMedic features –Agent-based framework, built on JADE-LEAP –Agent implementation of centralized management facilities and distributed services: permission administration, discovery, … –Context-awareness: event management and notification –Data acquisition and integration of remote medical devices by means of mobile agents
04/06/2007AgentGroup25 Other activities Ubiquitous computing Autonomic computing Services Code mobility
04/06/2007AgentGroup26 Web site