A.S. Rao and M.P. George#. A model theoretic approach to the verification of situated reasoning systems. In Proc. Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318---324, Chambery, France, 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....logics provide interesting languages that are expressive enough to encode most properties of interest, while usually remaining decidable. Multimodal propositional logics are now well established in artificial intelligence research as bases for defeasible reasoning [Shv90] logics of agents [RG93], and logics of authentication [BAN90, Mat97] Monniaux [Mon01] shows that BAN and GNY logics are decidable, while Massacci [Mas97] gives a tableaux calculus for the (undecidable) logic of access control of [ABLP93] We refer the reader to [Mon01] for more information on such logics. Multi modal ....

A. Rao and M. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the 13th International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318--324. Morgan-Kauffman, 1993.

....Fig. 5. Evaluation of Actions. 5 Comparisons In this section we compare our work to related approaches. 5.1 BDI agents BDI agents are modeled in terms of beliefs, desires and intensions. Theoretical work has investigated multi modal logics to specify agents [22, 23] and independently, systems such as the Procedural Reasoning System PRS [10, 15] have been implemented. b received from creator the message start b sent to d the message request d received from creator the message start d received from b the message request d sent to b the message ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Ruzena Bajcsy, editor, IJCAI93, International joint Conference on Artificial Intelligence, volume 1, pages 318--324. Morgan Kaufmann Publishers, Inc., 1993.

....in that paper. This is relevant in assuring the rationality of agents programmed in AgentSpeak(L) Model checking techniques have only recently begun to find a significant audience in the multi agent systems community. Rao and Georgeff developed basic algorithms for model checking BDI logics [14], but the authors proposed no method for generating BDI models from programs. In [1] a general approach for modelchecking multi agent systems was proposed, based on the branching temporal logic CTL together with modalities for BDI like attitudes. However, once again no method was given for ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proc. Thirteenth International Joint Conf. on Artificial Intelligence (IJCAI-93), pages 318--324, Chambery, France, 1993.

....disadvantage of this way of definition through interpretation is the inefficiency of the interpreter. The mushrooming of the IA research and technology sharply contrasts with scarcity of research in IA behavior analysis. Besides the abovementioned papers [1, 3, 20] we can only cite the papers [14, 15, 4, 37], chapter 11 of the book [42] and the series of our papers [11, 12, 13] as representing the verification approach, and the papers [22, 44] representing the interpretation approach. The reason for this lacuna is the difference in orientations of IA and MA system architecture definitions and of ....

Rao, A. S. and Georgeff, M. P., A model-theoretic approach to the verification of situated reasoning systems. In: Proceedings of the Thirteen's International Joint Conference on Artificial Intelligence (IJCAI-93). 1993, 318-324.

....intentions. The third and final aspect of the Bm model is the logical component, which gives us a family of tools that allow us to reason about Bm agents. There have been several versions of Bm logic, starting in 1991 and culminating in Rao and Georgeff s 1998 paper on systems of BD logics [64, 68, 65, 66, 67, 61, 63]; a book length survey was published as [77] We focus on [77] Syntactically, BD logics are essentially branching time logics (aTE or aTE , depending on which version you are reading about) enhanced with additional modal operators Bel, Des, and Intend, for capturing the beliefs, desires, and ....

....specification just in case the answer is yes . The main advantage of model checking over axiomatic verification is in complexity: model checking using the branching time temporal logic CTL ( 12] can be done in polynomial time, whereas the proof problem for most modal logics is quite complex. In [67], Rao and Georgeff present an algorithm for model checking agent systems. More precisely, they give an algorithm for taking a logical model for their (propositional) BDI agent specification language, and a formula of the language, and deter mining whether the formula is valid in the model. The ....

A. S. Rao and M.P. Georgefl. A model-theoretic approach to the verification of situated rea- soning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318424, Chambry, tVYance, 1993.

....In this paper we have defined a model checking based decision procedure for multiagent systems. Our approach allows us to reuse the technology and tools ( developed in model checking and to specify multiagent systems incrementally. The closest work to ours is the work by Rao 85 Georgeff [5]. Similarly to us, they employ a class of logics obtained by combining branching time temporal logics (e.g. CTL) with logics for BDI attitudes. The resulting logics are relatively similar to ours. The main (essential) difference is that, in their approach, a multiagent system is specified by ....

.... the case of bounded nesting; we can implement multiagent model checking by directly calling, as a subroutine, the standard model checking algorithm; and we can implement an algorithm which visits the smallest possible submodel (this last property does not seem to be possessed by the algorithm in [5] as they need to label all the worlds in the model) Finally, we also improve on their work as they don t face the problem of the automatic generation of a model starting from the specification of a multiagent system. Notice that, at the current state of the art, we can take a standard model ....

A. S. Rao and M.P. Georgefl. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artifcial Intelligence (IJCAI-93), pages 318-324, Chambry, France, 1993. 10

....properties represented as formulas against the Kripke model. Since Kripke structures are less expressive with respect to high level system specification, a front end formalism (e.g. modal logic) instead of a Kripke structure is often expected for the specification of a multi agent system. In [28], the algorithm for model checking BDI systems takes a logical model for their propositional BDI logic and a formula of their language, and determines whether the formula is valid in the model. Despite the inclusion of BDI modalities into the CTL branching time framework, it is claimed that the ....

A. Rao and M.P. Georgeff, "A Model-Theoretic Approach to the Verification of Situated Reasoning Systems," Proceedings of the 13th International Joint Conference on Artificial Intelligence (IJCAI'93), Chambrey, France, pp. 318-324, 1993.

.... in which knowledge and ability is expressible as well) In this enterprise we owe to Bob Moore s work combining a version of dynamic logic and epistemic logic for the first time [37] So here we deviate from the philosophy of other foundational work on agents, in particular that of Rao Georgeff [40, 39, 41], who take belief, desire, intentions as well as time as primitive notions for agents. One could argue that our approach is more in line with that of Cohen Levesque [5] They, too, take actions as basic building blocks for their theory of agents. However, they consider only models of their ....

A.S. Rao and M.P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In R. Bajcsy, editor, Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI'93), pages 318--324. Morgan Kaufmann, 1993.

.... and belief [14, 35] and motivational aspects like commitments and obligations [7] Recent developments include the work The notions informational and motivational are both due to Shoham Cousins [42] on agent oriented programming [41, 43] the Belief Desire Intention architecture [38, 39, 40], logics for the specification and verification of multi agent systems [44, 45] logics for agents with bounded rationality [21, 22] and cognitive robotics [26, 27] In our research [18, 19, 29, 30, 31, 32, 33] we defined a theorist logic for rational agents, i.e. a logic that is used to ....

A.S. Rao and M.P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI'93), pages 318--324, 1993.

....Rules Figure 1: An example of a layered agent architecture. agent systems and various agent architectures have been developed. Many of them are used in safety critical applications and therefore need verification. But up to now there are only few results on the verification of such systems [RG93] We investigate how CCS, the calculus for concurrent systems [Mil89] may serve as formal background of agent communication. We start our considerations with an application in the domain of diagnosing computer networks For distributed diagnosis [SdAMP97, SW97, FdAMNS96] we have developed the ....

....to the verification of its top layer. Some effort has already been spent to this extent, but only few results are obtained sofar. Rao and Georgeff argue that the linkage between formal specification and the design of multi agent systems is weak and that there are little results on verification [RG93] Burkhard takes a step in this direction by applying definitions of liveness and fairness as first defined for concurrent systems to multi agent systems [Bur93] Vivid agents [Wag96] 2 follow a similar philosophy and feature a formal semantics based on transition systems. Gaspari investigates ....

[Article contains additional citation context not shown here]

Anand S. Rao and Michael P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In 14 Ruzena Bajcsy, editor, IJCAI93, International joint Conference on Artificial Intelligence, volume 1, pages 318--324. Morgan Kaufmann Publishers, Inc., 1993.

....how to extend the framework in terms of symbolic model checking. The notions of multiple views and multiple languages are mostly inspired by the works by Giunchiglia and his collaborators in the field of Multi Language Systems [10, 9] Other related works use combinations of modal temporal logics [14], although the mechanization is not based on symbolic model checking techniques. 12] present an automata theoretic approach to temporal modal logic (restricted to the case of single nesting of beliefs) applied to the specification of knowledge based systems. We are currently working at the ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI93) , pages 318--324, Chambery, France, 1993.

....# just in case the answer is yes . The main advantage of model checking over axiomatic verification is in complexity: model checking using the branching time temporal logic ctl ( 7] can be done in polynomial time, whereas the proof problem for most modal logics is quite complex. In [44], Rao and George# present an algorithm for model checking agent systems. More precisely, they give an algorithm for taking a logical model for their (propositional) BDI agent specification language, and a formula of the language, and determining whether the formula is valid in the model. The ....

A. S. Rao and M. P. George#. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318--324, Chambery, France, 1993.

....verification is in complexity: model checking using the branching time temporal logic CTL [12] can be done in time O(j j jM j) where j j is the size of the formula to be checked, and jM j is the size of the model against which is to be checked the number of states it contains. In [61], Rao and Georgeff present an algorithm for model checking BDI systems. More precisely, they give an algorithm for taking a logical model for their (propositional) BDI logic, and a formula of the language, and determining whether the formula is valid in the model. The technique is closely based ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318--324, Chambery, France, 1993.

....are the agent s mental attitudes. As shown in Figure 1, the different mental attitudes of a rational agent influence each others and interact to form its global mental state. The interactions between different mental attitudes can be classified into static interactions and dynamic interactions [12]. Static interactions describe the relationship between different mental attitudes at the same instant, whereas dynamic interactions describe how these mental attitudes change over time. The dynamics of mental states is clearly related to perception. An agent totally isolated from the external ....

....This leads to the consequence that, from the intention to do a, it is not possible to derive the belief that a will be actually carried out. This reduces the effectiveness of agent s intention as a mean to help it in coordinating its actions, as pointed out in Section 4. Rao and Georgeff in [12] face the problem of studying autonomous agents embedded in a changing environment and, in particular, the so called agentoriented systems. Such systems are viewed as rational agents having certain mental attitudes that determine their behavior. Rao and Georgeff are mainly interested in developing ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Bajcsy R.(ed.), editor, Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence, San Mateo, CA, 1993. Morgan Kaufmann Publishers.

....security criteria expressed as formulae of logic. For example, notions like agent i trusts agent j are easily encoded as statements of multi modal propositional logics, which are now well established in artificial intelligence research as bases for defeasible reasoning [Shv90] logics of agents [RG93], and logics of authentication [MBN90, Mat97] Multi modal logics like Propositional Dynamic Logic [Gol87] have also been used to model the changing states of a program. Finally, propositional bi modal tense logics give a very simple and elegant model of the flow of time [HC96] Checking that a ....

A Rao and M George#. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318--324. Morgan-Kau#man, 1993.

....reasonable. Suppose an agent promised to move Block A to location s and to put Block B on Block A. It is not clear that it intends to do each one of the actions separately. In many cases, there is no benefit in performing each action separately, but only doing them together is beneficial (see also [127]) If one wouldn t adopt the axiom that requires splitting a conjunctive intention, one cannot require the agent s intentions to be closed under consequences. However, if such an axiom is appropriate in a specific application, the following axiom may also be sound: INT4) t; Int i ] t; Int i ....

A. Rao and M. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In IJCAI-93, pages 318--324, French, 1993.

....languages such as IC Prolog [3] Stream Logic Programming [39, 18] April [26] or PVM Prolog [5, 6] support distributed computing and declarative programming. Unfortunately, on the highest level of abstraction one finds either agent theories which are not operational, such as the BDI logics of [37, 38], or implemented systems with certain data structures corresponding to beliefs and goals [19, 23] but without a formal semantics, such as PRS, dMARS and ARCHON. It is the objective of the vivid agent model to bridge this gap between implemented agent systems and formal theories of agency. An ....

....is sent from n 3 to n 2 (16,17) and from n 2 to n 1 (18,19) n 2 removes the fact that it has to respond to n 1 (20) 5 Comparison The concept of vivid agents is related to many other high level agent languages. To compare our work we focus on Agent0 [44] PLACA [45] AgentSpeak [36] BDI Agents [37, 38], and ARCHON [4, 20, 51, 21] Our agent interpreter is implemented in PVM Prolog [5, 6] which is closely related to other (logic) programming languages with explicit parallelism such as IC Prolog ] 3] and Stream Logic Programming [39, 18] We start by comparing our work to the high level agent ....

[Article contains additional citation context not shown here]

Anand S. Rao and Michael P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Ruzena Bajcsy, editor, IJCAI93, International joint Conference on Artificial Intelligence, volume 1, pages 318--324. Morgan Kaufmann Publishers, Inc., 1993.

....a final product, builds a model for it and then tries to prove that the model satisfies also the initial specification. Obviously, the problem is constructing the model: once again we can exploit the close relationship between finite state machines and models for temporal logic (see for example [RG93] but this is a particular case, and the mapping is in general much more hard to do. Moreover, we do not have to forget the importance of testing. Verification is not always applicable, or a final product could verify the initial requirements, but they were not right or well expressed. In these ....

A. Rao and M. Georgeff. A Model--Theoretic Approach to the Verification of Situated Reasoning Systems. In Proc. of 13th International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318--234, Chambery, France, 1993.

....specification f just in case the answer is yes . The main advantage of model checking over axiomatic verification is in complexity: model checking using the branching time temporal logic CTL [7] can be done in polynomial time, whereas the proof problem for most modal logics is quite complex. In [32], Rao and Georgeff present an algorithm for model checking agent systems. More precisely, they give an algorithm for taking a logical model for their (propositional) BDI agent specification language, and a formula of the language, and determining whether the formula is valid in the model. The ....

A. S. Rao and M. P. Georgeff (1993) "A model-theoretic approach to the verification of situated reasoning systems" Proc. of the 13 th Int. Joint Conf. on Artificial Intelligence, 318-324, Chambery, France.

.... 3:P) On the other hand, there are approaches that explicitly account for temporal aspects and try to ground the notion of intention on a philosophical basis (see [4] Although works along these lines tend to be regarded as frameworks for de1 sign and verification of situated agent systems [6] via model checking, rather than actual intention inference apparatuses, the lack of mechanisms to play, on this semantic context, the functional roles plans play on practical intention recognition systems produces some odd consequences. An analysis of [4] s attempt to capture Bratman s concept of ....

....b and c. In any case, however, there seems to be a lot of contextual information escaping from (1) If it were to be improved, the most straightforward option would be to introduce an extra level of quantification over intentions. However, unlike (1) this is not supported by the logics of [4] and [6]. We have seen that neither practically nor theoretically motivated accounts of intentionality are satisfactory. We claim that the main deficiencies of either approach are caused by adhocity in integrating propositional attitudes and plan related aspects. The former fail because they constrain ....

Anand S. Rao and Michael Georgeff. A model-theoretic approach to the verification of situated reasoning systems. Technical Report 37, Australian AI Institute, 1992.

....sources, accessing, filtering, and integrating information in support of decision making, as well as coordinating information retrieval and problem solving efforts of information sources and decision making systems has become a very critical task. The notion of Intelligent Software Agents (e.g. [12, 15, 6, 14]) has been proposed to address this challenge. Although a precise definition of an intelligent agent is still forthcoming, the current working 1 The Robotics Institute, Carnegie Mellon University, Pittsburgh, PA 15213, U.S.A. 2 The Robotics Institute, Carnegie Mellon University, Pittsburgh, PA ....

....above characteristics. 2 Different Types of Information and Decision Making Agents Our distributed agent based architecture has three types of agents (see Figure 1) interface agents, task agents, and information agents. The architecture of all these agents follows the general BDI type philosophy [12], however, each of them embodies particular architectural design commitments to make them effective in dealing with the particular category of issues of its type. As our point of departure, we use the Task Control Architecture (TCA) framework [13] which we extend and specialize for real time user ....

Anand S. Rao and Michael P. Georgeff, `A model-theoretic approach to the verification of situated reasoning systems', in Proceedings of IJCAI93, pp. 318--324, Chambery, France, (28 August - 3 September 1993). IJCAI.

....of agent architectures and languages we focus on the widely investigated BDI agents and the traditional logic based rational agents. 5.1 BDI agents BDI agents are modeled in terms of beliefs, desires and intensions. Theoretical work has investigated multi modal logics to specify agents [RG91, RG93] and independently, systems such as the Procedural Reasoning System PRS [IG90, LHDK94] have been implemented. Unfortunately, implemented systems and specification in BDI logics are not formally coupled, so that the specification logics are not executable and the implementation lacks theoretical ....

Anand S. Rao and Michael P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Ruzena Bajcsy, editor, IJCAI93, International joint Conference on Artificial Intelligence, volume 1, pages 318--324. Morgan Kaufmann Publishers, Inc., 1993.

....the results of their actions are formalized. In this so called KARO architecture 1 it can for instance be modelled that an agent knows that some action 1 The term KARO, for Knowledge Abilities Results Opportunities, is chosen in analogy with the BDI architectures of Rao Georgeff ([39, 40, 41]) is a correct plan to achieve some goal since it knows that performing the action will lead to the goal, and that it knows that an action is a feasible plan since the agent knows of its ability to perform the action. In subsequent research we extended our framework with nondeterministic ....

....opportunities, Cohen Levesque deal with belief, goals, and a sort of future and past notion of combined ability and opportunity. It is furthermore the case that their system has a temporal logic flavour to it, whereas ours is typically dynamic logic of nature. The system of Rao Georgeff ([39, 40, 41]) also deals with formalizing the behaviour of rational agents. Their framework is based on three primitive modalities: beliefs, desires, and intentions. Semantically their formalism is based on a branching model of time, in which belief , desire and intention accessible worlds are themselves ....

A.S. Rao and M.P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI'93), pages 318--324, 1993.

.... checking is often much simpler than theorem proving, and as a consequence model checking approaches have been used to verify significant finite state systems (with up to 10 120 states [5] Moreover, model checking algorithms have been developed for (propositional) beliefdesire intention logics [17], which are somewhat similar to those used to give a semantics to the FIPA ACL. However, there are two major problems with such an approach. The first is that of developing the program semantics [ P : as described above, we do not yet have any techniques for systematically assigning models ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318--324, Chambery, France, 1993.

....beliefs and goals. Further attitudes, such as intention, were defined in terms of these. In related work, Rao and Georgeff have developed a logical framework for agent theory based on three primitive modalities: beliefs, desires, and intentions (Rao and Georgeff, 1991b; Rao and Georgeff, 1991a; Rao and Georgeff, 1993). Their formalism is based on a branching model of time, cf. Emerson and Halpern, 1986) in which belief , desire and intention accessible worlds are themselves branching time structures. They are particularly concerned with the notion of realism the question of how an agent s beliefs ....

Rao, A. S. and Georgeff, M. P. (1993). A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318--324, Chamb'ery, France.

....described above, just two basic attitudes were used: beliefs and goals. Further attitudes, such as intention, were defined in terms of these. In related work, Rao and Georgeff have developed a logical framework for agent theory based on three primitive modalities: beliefs, desires, and intentions [105, 104, 107]. Their formalism is based on a branching model of time, cf. 39] in which belief , desire and intention accessible worlds are themselves branching time structures. They are particularly concerned with the notion of realism the question of how an agent s beliefs about the future affect its ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318--324, Chambery, France, 1993.

....The manner in which goals are checked over state sequences is essentially model checking. Model checking is widely used in the verification of temporal properties (Courcoubetis, Vardi, Wolper, Yannakakis, 1992; Wolper, 1989) It has also been applied to synchronize reactive plans (Kabanza, 1995; Rao Georgeff, 1993) and to control search in a classical planner (Bacchus Kabanza, 1995) To check goals over state sequences, our planner generalizes the approach of Bacchus and Kabanza by defining mechanisms for handling uncontrollable actions, liveness goals, and time constraints. A preliminary version of our ....

Rao, A., & Georgeff, M. (1993). A model-theoretic approach to the verification of situated reasoning systems. In Proc. of 13th International Joint Conference on Artificial Intelligence (IJCAI), pp. 318--324.

.... and belief [14, 35] and motivational 1 aspects like commitments and obligations [7] Recent developments include the work 1 The notions informational and motivational are both due to Shoham Cousins [42] on agent oriented programming [41, 43] the Belief Desire Intention architecture [38, 39, 40], logics for the specification and verification of multi agent systems [44, 45] logics for agents with bounded rationality [21, 22] and cognitive robotics [26, 27] In our research [18, 19, 29, 30, 31, 32, 33] we defined a theorist logic for rational agents, i.e. a logic that is used to ....

A.S. Rao and M.P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI'93), pages 318--324, 1993.

....framework and methodology of game theory remains almost the same and it might be too early to tell whether these new results will reshape the current game theoretic framework. In a single agent environment, we may view commitment to a decision as a contract of the decision maker with himself [13, 34]. In a multi agent environment, most of the time decisions can be viewed as inter agent contracts. In traditional models, decisions made by a decision maker are binding. Binding decisions are appropriate for some types of tasks and environments (e.g. a static world) in most realistic situations ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of IJCAI-93, pages 318--324, Chambery, France, 28 Aug. - 3 Sept. 1993. IJCAI.

....branching time temporal logic CTL ( 8] can be done in polynomial time (O(j j Theta jMj) where j j is the size of the formula to be checked, and jMj is the size of the model against which is to be checked the number of states it contains) 5 . Case Study: Model Checking BDI Systems In [50], Rao and Georgeff present an algorithm for model checking AOP systems. More precisely, they give an algorithm for taking a logical model for their (propositional) BDI agent specification language, and a formula of the language, and determining whether the formula is valid in the model. The ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318--324, Chambery, France, 1993.

....discuss some formal attempts to fill the gap between intentional concepts and modalities, and the motivational ones: the BDI architecture worked out by Georgeff and his collaborators and Cohen and Levesque s theory of rational interaction. 2. 1 The BDI architecture The BDI architecture (cf. [14, 15, 9]) describes agents in terms of three primitive modalities: beliefs (B) desires, D) and intentions (I) The temporal dimension is treated thanks to branching structures. However, the modalities for beliefs, desires, and intentions are defined as usual in terms of possible words semantics. ....

Rao, A.S. and Georgeff, M.P. A model-theoretic approach to the verification of situated reasoning systems. Proceedings of the 13th International Conference of Artificial Intelligence, IJCAI-93, Chambery, France, 1993.

....specification j just in case the answer is yes . The main advantage of model checking over axiomatic verification is in complexity: model checking using the branching time temporal logic CTL ( 3] can be done in polynomial time, whereas the proof problem for most modal logics is quite complex. In [17], Rao and Georgeff present an algorithm for model checking agent systems. More precisely, they give an algorithm for taking a logical model for their (propositional) BDI agent specification language, and a formula of the language, and determining whether the formula is valid in the model. The ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318--324, Chambery, France, 1993.

.... Wooldridge s Fisher s work on concurrent MetateM, which approaches this issue from a proof theoretical point of view) 15] and Rao Georgeff s work on developing satisfiability based decision procedures for a class of BDI logics [14] The closest work to ours is the work by Rao Georgeff [13]. As far as we know they are the first who have actually faced the problem of applying model checking techniques to multiagent systems. Similarly to us, they employ a class of logics obtained by combining branching time temporal logics (e.g. CTL) with logics for BDI attitudes. The resulting ....

.... the case of bounded nesting; we can implement multiagent model checking by directly calling, as a subroutine, the standard model checking algorithm; and we can implement an algorithm which visits the smallest possible submodel (this last property does not seem to be possessed by the algorithm in [13] as they need to label all the worlds in the model) Finally, we also improve on their work as they don t face the problem of the automatic generation of a model starting from the specification of a multiagent system. Notice that, at the current state of the art, we can take a standard model ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318--324, Chamb'ery, France, 1993.

....corresponding to that program; showing that MP = j, i.e. that the specification formula j holds in MP , and thus that the program satisfies its specification. Model checking is generally reckoned to be easier than theorem proving Wooldridge Fisher 34 Agent Based Software Engineering . (Rao and Georgeff, 1993) have described a process for model checking AOP systems. They give an algorithm for taking a logical model for their (propositional) agent specification language, and a formula of the language, and determining whether the formula is valid in the model. However, it is not clear where the ....

Rao, A. S. and Georgeff, M. P. (1993). A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Wooldridge & Fisher 41 Agent-Based Software Engineering Artificial Intelligence (IJCAI-93), pages 318--324, Chambery, France.

....In this paper we have defined a model checking based decision procedure for multiagent systems. Our approach allows us to reuse the technology and tools ( developed in model checking and to specify multiagent systems incrementally. The closest work to ours is the work by Rao Georgeff [9]. Similarly to us, they employ a class of logics obtained by combining branching time temporal logics (e.g. CTL) with logics for BDI attitudes. The resulting logics are relatively similar to ours. The main (essential) difference is that, in their approach, a multiagent system is specified by ....

.... the case of bounded nesting; we can implement multiagent model checking by directly calling, as a subroutine, the standard model checking algorithm; and we can implement an algorithm which visits the smallest possible submodel (this last property does not seem to be possessed by the algorithm in [9] as they need to label all the worlds in the model) Finally, we also improve on their work as they don t face the problem of the automatic generation of a model starting from the specification of a multiagent system. Notice that, at the current state of the art, we can take a standard model ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318--324, Chamb'ery, France, 1993. This article was processed using the L A T E X macro package with LLNCS style

....sources, accessing, filtering, and integrating information in support of decision making, as well as coordinating information retrieval and problem solving efforts of information sources and decision making systems has become a very critical task. The notion of Intelligent Software Agents (e.g. [1, 19, 20, 25, 13, 22]) has been proposed to address this challenge. Although a precise definition of an intelligent agent is still forthcoming, the current working notion is that Intelligent Software Agents are programs that act on behalf of their human users in order to perform laborious information gathering tasks, ....

Anand S. Rao and Michael P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of IJCAI93, pages 318--324, Chambery, France, 28 August - 3 September 1993. IJCAI.

.... in which knowledge and ability is expressible as well) In this enterprise we owe to Bob Moore s work combining a version of dynamic logic and epistemic logic for the first time [37] So here we deviate from the philosophy of other foundational work on agents, in particular that of Rao Georgeff [40, 39, 41], who take belief, desire, intentions as well as time as primitive notions for agents. One could argue that our approach is more in line with that of Cohen Levesque [5] They, too, take actions as basic building blocks for their theory of agents. However, they consider only models of their ....

A.S. Rao and M.P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In R. Bajcsy, editor, Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI'93), pages 318--324. Morgan Kaufmann, 1993.

No context found.

Rao, A. S. and Georgeff, M. P. (1993). A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318--324, Chamb ery, France.

No context found.

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedingsof the Thirteenth International Joint Conferenceon Artificial Intelligence (IJCAI-93), Chamberey, France, 1993.

....] air combat modelling [ Rao et al. 1992a ] and more recently business process management. The formalization of BDI logics provides the foundation for studying the properties of these implemented systems; in particular, to prove safety and liveness properties [ Emerson, 1990 ] Elsewhere [ Rao and Georgeff, 1993 ] we show how one can use a model checking approach to verify such properties of BDI agents. It turns out that, unlike the theoremproving approach discussed in this paper, the complexity of checking for satisfaction of formulas in the BDI K CTL is polynomial in the size of the given formula and ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), Chamberey, France, 1993.

....We have chosen an exogenous logic (i.e. dynamic logic) to represent the plans of an agent as it fits naturally with the compositional nature of plans. When one reasons about the mental state of an agent during the execution recognition of a plan, an endogenous logic is more appropriate. Elsewhere [ 22 ] we have developed endogenous logics CTL BDI and CTL BDI to represent the mental state of an agent that captures the agent s beliefs, desires, and intentions. Rational agents have to deal with mental states as well as plans and a combination of exogenous and endogenous logics would be ideal ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), Chamberey, France, 1993.

....intention. In other words, the refutation for an intention starts when an intention is adopted and ends when the intention stack is empty. Thus, using the above proof rules we can formally prove certain behavioural properties, such as safety and liveness of agent systems, as was done elsewhere [15]. Furthermore, there is a one to one correspondence between the proof rules discussed in this section and the operational semantics discussed in the previous section. Such a correspondence has not been possible before, because the proof theory (usually based on multi modal logics) has been far ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedingsof the Thirteenth International Joint Conferenceon Artificial Intelligence (IJCAI-93), Chamberey, France, 1993.

....complete axiomatizations and are not efficiently computable is questioned by many system builders as having little relevance in practice. This paper addresses these two criticisms from the perspectives of the authors previous work in BDI logics [ Rao and Georgeff, 1991a; Rao and Georgeff, 1991c; Rao and Georgeff, 1993 ] systems [ Georgeff and Lansky, 1986 ] and real world applications [ Ingrand et al. 1992; Rao et al. 1992 ] We argue the necessity (though not the adequacy) of these three attitudes in domains where real time performance is required from both a quantitative decision theoretic perspective ....

....to an intention based on the object of the intention being fulfilled in one future path or all future paths leading to different commitment conditions and hence different dynamic behaviours. Different termination conditions result in further variations in behaviour [ Rao and Georgeff, 1991c; Rao and Georgeff, 1993; Georgeff and Rao, August 1995 ] For example, we can define a blindly committed agent which denies any changes to its beliefs or desires that would conflict with its commitments; a single minded agent which entertains changes to beliefs and will drop its commitments accordingly; and an ....

[Article contains additional citation context not shown here]

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), Chamberey, France, 1993.

.... where i is [p 1 z: zf : c 1 : c y g( t) h 2 ; h n ] The intention i is said to have been executed iff there exists a substitution such that 8g( t) is a logical consequence of B and i is replaced by [p 1 z: z(f : c 1 : c y ) h 2 ; h n ] Definition 15 Let S I (I) i, where i is [p 1 z: zf : c 1 : c y a( t) h 2 ; h n ] The intention i is said to have been executed iff a( t) 2 A, and i is replaced by [p 1 z: zf : c 1 : c y h 2 ; h n ] Definition 16 Let S I (I) i, where i is [p 1 z: zp z Gamma1 ....

....b x g( s) h 2 ; h n . The intention i is said to have been executed iff there exists a substitution such that g( t) g( s) and i is replaced by [p 1 z: zp z Gamma1 z(e : b 1 : b x ) h 2 ; h n ) Continuing our example, we would execute I and by Definition 15 we would add fmove(a,b)g to A and change I to be as follows: location(robot,b) location(robot,a) not(b = a) adjacent(a, b) not(location(car,b) location(robot,b) In the next iteration, after the robot moves from a to b the environment will send the agent a belief update event ....

[Article contains additional citation context not shown here]

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), Chamberey, France, 1993.

....linear time and branchingtime characterisations have been studied [ 16 ] The Procedural Reasoning System (PRS) 8 ] is a system that is loosely based on the logical specifications of BDI agents. Formal verification of BDI agents based on model checking techniques have also been investigated [ 18 ] . Furthermore, PRS has been applied to a number of important practical problems, such as air traffic control [ 12 ] spacecraft systems handling [ 11 ] telecommunications management [ 11 ] and air combat modelling [ 14 ] Research into the specification, design, and verification of agents, ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), Chamberey, France, 1993.

....win a lottery, but may intend to buy a lottery ticket al..though he does not believe that he will win the lottery. This mental state of the agent can represented by the following formula of L BDI : DES(win lottery) INTEND(buy lottery ticket) BEL(win lottery) Elsewhere [ Rao and Georgeff, 1991b; Rao and Georgeff, 1993 ] we have investigated various branching time logics, described a possible worlds semantics, and analyzed the dynamic interaction of the mental attitudes with one another. In this paper, we have deliberately not considered temporal BDI logics because the issue of time is orthogonal to our ....

A. S. Rao and M. P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), Chamberey, France, 1993.

No context found.

A.S. Rao and M.P. George#. A model theoretic approach to the verification of situated reasoning systems. In Proc. Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318---324, Chambery, France, 1993.

No context found.

A. S. Rao and M. P. George#. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of the Thirteenth International Joint Conference on Artificial Intelligence (IJCAI-93), pages 318--324, Chambery, France, 1993.

No context found.

A.S. Rao and M.P. George#. A model theoretic approach to the verification of situated reasoning systems. In Proc. 13th International Joint Conference on Artificial Intelligence, pages 318--324, 1993.

No context found.

Anand S. Rao and Michael P. Georgeff. A model-theoretic approach to the verification of situated reasoning systems. In Proceedings of IJCAI-93, pages 318--324, Chambery, France, 28 August - 3 September 1993. IJCAI.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC