Vous avez cherché: belief desire intention (Anglais - Français)

28. involvement of a minor in criminal activities means the deliberate act of instilling in a minor the desire (intention, ambition) and readiness to participate in the commission of a crime.
28. on entend par incitation d'un mineur à l'exercice d'activités délictueuses un acte visant à faire naître chez le mineur l'envie (l'intention, l'ambition) et la décision de commettre une infraction.

journal of web engineering, vol. 10, no.3 (2011) 175-196 © rinton press exception handling in pervasive service composition using normative agents j. octavio gutierrez-garcia gwangju institute of science and technology, republic of korea joseogg@gmail.com felix f. ramos-corchado cinvestav-ipn, unidad guadalajara, mexico framos@gdl.cinvestav.mx received february 20, 2011 revised august 30, 2011 the full integration of pervasive computing services into daily life leads to smart spaces with a wide range of intelligent devices that must be dynamically composed to provide a transparent service to users. to achieve this, pervasive services have to coordinate among themselves in both an automated and autonomous manner, with the aim of satisfying complex user requirements that no single service can fulfill. however, existing pervasive environments are normally ad-hoc and isolated systems incapable of: 1) reacting to dynamic and unforeseen situations that may raise exceptions, and 2) collaborating with pervasive services beyond their physical limits. the contributions of this work are: 1) proposing a normative agent-based service composition method capable of handling exceptions in open pervasive systems. 2) using the web as the underlying infrastructure where pervasive services provided by either smart devices or web services can coexist and interact with each other. 3) providing a modular and hierarchical agent coordination method based on virtual organizations to sustain coordination among agents belonging to multiple organizations (smart spaces). 4) integrating exception handling mechanisms based on social norms and formalized by event calculus predicates into virtual organizations to support and guide exception handling in both a dynamic and autonomous manner. key words: exception handling, multiagent systems, normative systems, pervasive computing, service composition communicated by: d. schwabe & a. ginige 1 introduction pervasive environments are complex systems by nature where many self-interested and independent entities must dynamically coordinate among themselves to provide pervasive and transparent composite services to users. in addition, users are nomads that move from place to place performing a broad range of daily activities that may require interacting with various smart devices, which may belong to several smart places. however, current pervasive systems are static and ad-hoc systems deployed in closed and isolated environments where all possible entities are known beforehand and 176 exception handling in pervasive service composition using normative agents fixed interaction protocols are defined. this precludes pervasive systems from: 1) reacting to dynamic and unforeseen situations that may raise exceptions, e.g., loss of messages or unavailable services, and 2) collaborating with pervasive services beyond their physical limits. the inherent complexity, dynamicity, and interactivity of pervasive systems emphasize the need for the agent paradigm [42]. agents as self-interested and autonomous entities can dynamically coordinate (see [10]) and react in the presence of unexpected situations, e.g., exceptions [24], while achieving a given objective (e.g., pervasive service composition). exceptions [24] are abnormal situations presented during the enactment of an interaction protocol. exceptions are common in complex and dynamic scenarios like those provided by pervasive systems. examples of exceptions are: loss of messages, unwillingness of message receivers to reply, and unavailable services among others. in closed agent-based pervasive systems [7, 41], the semantics of interaction rules is commonly missing or depends on agents’ private states that cannot be publicly verified and require homogeneous agents in order to have standard agents’ beliefs (c.f., the belief-desire-intention approach [6]). this increases the amount of a priori knowledge required to establish agent interaction, and limits its validity to closed environments with homogeneous agents. in this regard, social norms [3, 4, 5, 8], namely obligations [14], can be publicly verified and supervised [34], which allows interaction among heterogeneous pervasive services provided by different vendors. in this work, we propose a normative agent-based service composition method capable of handling exceptions in open pervasive systems. the web is used as the underlying infrastructure where pervasive services provided by either smart devices or web services can coexist and interact with each other. pervasive services provided by both smart devices and web services are wrapped and represented by agents. furthermore, a hierarchical and reactive agent coordination method supported by virtual organizations (vos) is used to coordinate agents belonging to multiple smart spaces. the control system of vos [1, 5, 12] is defined by means of obligations, which are formalized by event calculus [25] predicates. finally, exception handling mechanisms based on social norms are integrated into vos to guide and handle exceptions in both a dynamic and autonomous manner. this paper is structured as follows: section 2 defines obligation-based agent interaction. section 3 defines pervasive services in terms of obligations. section 4 contains the definition of pervasive vos. section 5 presents an agent-based pervasive service composition algorithm. section 6 describes exception handlers that are integrated into the pervasive service composition method. section 7 presents a comparison with related works. finally, section 8 gives some concluding remarks and future research directions. 2 interaction via obligations agents should follow interaction protocols to achieve a certain objective. moreover, agent interaction protocols should be endowed with semantics that provides a standard meaning for each message. based on this, we have defined obligation-based operational semantics that gives meaning to agent interaction at the same time that leads the interaction. obligations, as the main element of the interaction semantics, are impositions to oneself demanded by a social force [14], i.e., agents acquire obligations to comply with society’s standards. j.o. gutierrez-garcia and f.f. ramos-corchado 177 2.1 formalizing obligations the logical framework that supports the management of obligations is provided by the event calculus [25], which is an action formalism that defines events/actions and their effects in a temporal basis. in this case, events represent either the emission and reception of messages, or sensor signals captured by smart devices. the effects of events are reflected in changes over propositional properties of a given domain called fluents. in this work, fluents are boolean, i.e., fluents can have true or false values. fluents may represent: 1) a state of affairs denoted by f, 2) an ordinary obligation of an agent a to achieve a state of affairs f denoted by o(a, f), or 3) a conditional obligation denoted by co(a, f, g), which indicates that agent a is obliged to achieve f, but only when fluent g holds, i.e., fluent g acts as a switch for activating/deactivating the obligation in question. the manipulation of obligation fluents, either ordinary or conditional, is defined by means of event calculus predicates. thus, a set of operations that creates, releases, and cancels agents’ obligations is defined. the event calculus predicates utilized to formalize obligations are described in table 1 (see [33] for a detailed definition). predicate meaning holdsat(f, t) fluent f holds at time t. initiates(m(), f, t) fluent f holds after the execution of event m() at time t. terminates(m(), f, t) fluent f does not hold after the execution of event m() at time t. happens(m(), t) event m() is executed at time t. initiallyp(f) fluent f holds from time t=0. initiallyn(f) fluent f does not hold from time t=0. trajectory(f, t, g, inc) fluent g holds at time t+inc, if fluent f is initiated at time t. declipped(t 1 , f, t 2 )fluent f is initiated between t 1 and t table 1 event calculus predicates 2 , t 1 where p stands for preconditions, m stands for message, and e stands for effects. • preconditions are conjunctions and/or disjunctions of holdsat(h, t) predicates where fluent h can be an ordinary obligation o(a, f), a conditional obligation co(a, f, g), or a simple fluent f. in addition, preconditions can include application domain parameters in the form (ε ∈ dk, and ◊ ∈ { =, , ≤, ≥}. • messages can be either external or internal. an external message is instantiated by p ◊ ε q ) | ε p ≠ ε q where ε i the obligation-based primitive illocutionary acts previously defined in section 2.3, such messages have standard effects. on the other hand, an internal message may be activated by the evolution of the interaction, and it can have customized effects. 182 exception handling in pervasive service composition using normative agents • effects are represented by means of creating, releasing and/or canceling obligations. 4 pervasive virtual organizations a vo is an association of agents with a shared objective [5, 11] that is regulated by control mechanisms. based on this, a pervasive vo is defined as the association of obligation-based pervasive services. pervasive vos have five elements: 1) members. the members of a pervasive vo {s 1 , s 2 , …, s } are all the pervasive computing services, either web services or smart devices, contained in a specific smart space. n 2) interaction context. the interaction context provides a description of the functionalities of a given vo as well as an interface to couple with other vos. the interaction context of a vo is defined as the union of the interaction contexts of its members: interactioncontext = || members u i i = 1 . sic the properties of fluents are affected by the union, e.g., f↓∪f↑ = f↓↑. 3) interaction state. in a similar manner, the interaction state of a vo is the union of all the interaction states of its members: interactionstate = || members u i i = 1 . sis 4) triggers. a trigger τ is an event that activates pervasive computing services. a trigger can be seen as a sensor, which can be activated by external events, as well as by the emission or reception of messages. a set of triggers {τ(x a ) 1 , τ(x b ) 2 , …, τ(x x ) k } may have a list x of application domain parameters, returned by sensors, and passed to agents’ obligations. moreover, application domain parameters can be also used as triggers’ conditions, e.g., a person entering a room can activate a trigger only if the room’s temperature is lower than 23° c. a vo, then, has a set of rules to accept obligations induced by triggers: createo(τ(x), o(vo, f(x)), t)← ∏ ∏ i ) , tx)(happens(τ )( | τ(x)∈triggers, x∈dk, and ◊∈{=,,≤,≥} ◊valuex 5) actions. pervasive vos operate as interfaces among member agents and the web. for this reason, its coordinating role requires a specific service structure provided by three groups of actions: j.o. gutierrez-garcia and f.f. ramos-corchado 183 • when two pervasive computing services are linked, means that one of them requires the service of the other. then, an induction of obligations between the services may be performed. this is achieved through the pervasive vo, given that not all smart devices could support coordination processes. then, when a service requires inducing an obligation, the service induces the obligation to its vo, which must have agree and reject messages to handle such requests: p h m : {holdsat(co(vo, f, allowedby(vo, o(vo, f))), t) ^ holdsat(allowedby(vo, o(vo, f))=true, t)} : agree(vo, b, f) e p h i m h : {∅}f∈interact
portable document format