C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming, 31(1-3):201--243, May 1997.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....of priority is greater than those of changes) this definition can be rephrased as: we consider the minimal set of changes needed to explain the observations. This is similar to the principle of actions to explain observation , which is studied in reasoning about actions [ Li and Pereira, 1996; Baral et al. 1996; McIlraith, 1998 ] There are two possible ways for defining the models of a knowledge base. We call these two possible semantics backward and pointwise. The first one is the simplest to explain, and allows for expressing scenarios in which knowledge at later time may rule out some initial ....

C. Baral, M. Gelfond, and A. Provetti. Representing actions: Laws, observations and hypothesis. J. of Logic Programming, 1996.

....a change statement change(m) then the penalty is m; 5. Else the penalty is 1; Let us now consider how Example 2 can be expressed using the syntax of brels. The knowledge about time points can be simply expressed as: source(2) 1]service executed source(1) 1]server crashed source(2) [2]service executed # server crashed Indeed, the pieces of knowledge with reliability 2 are those expressing things known for certain. On the other hand, 1]server crashed is just an assumption made, thus it is less reliable. Now, any change may happen, except that a server cannot recover from a ....

....reliability 1 and one formula with reliability 2, thus we need an array with two elements. The array is defined as: pref(M ) i] # F has reliability i dist(M, F ) In the specific example above, the array has two elements, whose value is: pref(M ) 1] dist(M, K 1 ) dist(M, K 2 ) pref(M )[2] = dist(M, K 3 ) The ordering between two models M 1 and M 2 is defined as follows, where P is the maximal level of reliability of formulas in the knowledge base. M 1 # M 2 i# #i (1 # i # P ) such that (pref(M 1 ) i] pref(M 2 ) i] AND #j. i j # P ) implies (pref(M 1 ) j] ....

[Article contains additional citation context not shown here]

C. Baral, M. Gelfond, and A. Provetti. Representing actions: Laws, observations and hypothesis. Journal of Logic Programming, 1996.

....propositions. The initial propositions are referred to as facts and the other propositions are referred to as causal laws. 3 Transition functions of ADC We take a slightly different approach in defining the semantics of the language ADC. Most of the action description languages [ Tur95; KL94; BGP96 ] that are inspired by A have an independent characterization without involving any of the standard logical formalisms, such as logic programming, default logic, circumscription, classical logic, etc. But when we allow defeasible causality and keep open the possibility of hierarchies of such ....

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming (to appear), 1996.

No context found.

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming, 31(1-3):201--243, May 1997.

No context found.

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming, 31(1-3):201--243, May 1997.

No context found.

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming, 31(13) :201--243, May 1997.

....motivation for, value minimization in a full scale knowledge representation problem, namely, reasoning about actual and hypothetical occurrences of actions. In [1] we introduced a sound and complete formalization of narratives for domains specified in the high level action description language L [2]. In the development of this formalization of narratives, we were faced with the problem of minimizing the value of a particular term: Sit map(SN ) that mapped the current situation symbol SN onto a sequence of actions A k ffi : A 2 ffi A 1 ffi ffl 8 We are assuming that AR is defined ....

C. Baral, M. Gelfond and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming, 31 (1-3):201-- 244, 1997.

....of NATs in terms of ease of representation and restricted monotonicity [26] the use of value minimization of functions in knowledge representation and the formalization of filtering in NATs. 1.6. Organization of the rest of the paper We start with an overview of the high level language L from [7,8] in Section 2 and an overview of NATs in Section 3. We then give a translation of domain descriptions in L into NATs in Section 4 and illustrate the NAT characterization with respect to some examples in Section 6. In Section 5 we discuss value minimization of functions and its role in our ....

....11 we compare our work with earlier work on combining narratives and hypothetical reasoning particularly by Miller, Shanahan, Pinto, Reiter and McCarthy. Finally, all the proofs are given in the Appendixes A and B. 2. Overview of L The high level language L was developed by Baral et al. [7,8] to allow representation of and reasoning with narratives together with hypothetical reasoning. Beside the syntax and semantics of L, 8] defines a translation to logic programs and Prolog for a subclass of this language. Here, we start with a translation of domain descriptions in L to NATs, ....

[Article contains additional citation context not shown here]

C. Baral, M. Gelfond, A. Provetti, Representing actions: laws, observations and hypothesis, Journal of Logic Programming 31 (1--3) (1997) 201--243.

....: s n are states, a 1 ; an are actions and s i 2 Phi(a i ; s i Gamma1 ) for i = 1; n. Models of a narrative (D; O 0 ) are interpretations M = Psi; Sigma) that satisfy all the facts in O 0 and minimize unobserved action occurrences. A more formal definition is given in [ BGP97 ] . A narrative is consistent if it has a model. Otherwise, it is inconsistent. When M is a model of a narrative (D; O 0 ) we write (D; O 0 ) j= M. Next we define the conditional probability that a particular pair M = Psi; Sigma) s 0 ; a 1 ; s 1 ; a 2 ; an ; s n ] Sigma) of ....

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming, 31(1-3):201--243, May 1997.

....a desired goal) given the tendency of domain experts to specify mostly plausible scenarios, can we handle the occurrence of exceptional cases. 1. 1 An Overview of Our Approach In this work, we view workflows as a collection of cooperative agents and use recent results on reasoning about actions [4, 6, 29, 36] to formalize the process of their specification and test their correctness. Our approach, in a philosophical sense, resembles formalization of database updates [43] and transactions [15, 16] Our main contributions are as follows. We present a very simple, high level language AW and, based on it, ....

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming, 1996.

....and a unobservable fluent. We then use this language to define the notion of a conditional plan, and the related notions of diagnostic and repair planning. Finally, in Section 5 we summarize and discuss related work. 2 Specifying narrative in L The propositional language L was developed in [BGP97, BGP98] to specify narratives and to reason with them. In this paper, we will describe the main aspects of the language L by dividing it into three components: a domain description language LD , a language to specify observations LO , and a query language LQ . In Section 4.1, we extend our language ....

....causal laws, describe causal relation between fluents in a world. Propositions of the form (3) called executability conditions, state when actions are not executable. A domain description D is a set of propositions in LD . The main difference between LD and the action description part of L [BGP97, BGP98] is the presence of static causal laws in LD , which are critical for representing the behavior of the device being diagnosed. A domain description given in LD defines a transition function from actions and states to a set of states. Recall, actions may be nondeterministic. Intuitively, given ....

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming, 31(13) :201--243, 1997.

....based on the formalism of action languages [32] Such languages can be thought of as formal models of the part of the natural language that are used for describing the behavior of dynamic domains. A theory in an action language normally consists of an action description and a history description [8], 44] The former contains the knowledge about e ects of actions, the latter consists of observations of an agent. Specifying e ects of actions An action description language contains propositions which describe the e ects of actions on states of the system modeled by sets of uents ....

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming, 31(1-3):201-243, May 1997.

....uent and a unobservable uent. We then use this language to de ne the notion of a conditional plan, and the related notions of diagnostic and repair planning. Finally, in Section 5 we summarize and discuss related work. 2 Specifying narrative in L The propositional language L was developed in [BGP97, BGP98] to specify narratives and to reason with them. In this paper, we will describe the main aspects of the language L by dividing it into three components: a domain description language LD , a language to specify observations LO , and a query language LQ . In Section 4.1, we extend our language ....

....static causal laws, describe causal relation between uents in a world. Propositions of the form (3) called executability conditions, state when actions are not executable. A domain description D is a set of propositions in LD . The main di erence between LD and the action description part of L [BGP97, BGP98] is the presence of static causal laws in LD , which are critical for representing the behavior of the device being diagnosed. A domain description given in LD de nes a transition function from actions and states to a set of states. Recall, actions may be nondeterministic. Intuitively, given an ....

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming, 31(13) :201-243, 1997.

.... extremely useful, to general and provenly correct theories [GL93, San92, San94, Rei91, LS91, LS95] that incrementally consider various specification aspects such as: actions with non deterministic effects [Bar95, Pin94, KL94, San93] concurrent actions [LS92, BG93, BG97a] narratives [MS94, PR93, BGP97, BGPml, Kar98] actions with duration [Sho86, San89, San93, San94, MS94, Rei96] natural events [Rei96] ramifications and qualifications due to simple and causal constraints [LR94, KL94, Bar95, Lin97, Thi97, MT95, Lin95, Bar95, GL95, San96, GD96] sensing (or knowledge producing) actions [Moo77, ....

....architecture of the agent then consists of (i) making observations, ii) using the action theory to construct a plan to achieve the goal, and (iii) executing the plan. In case of a dynamic world where other agents may change the world we need theories that allow observations as the world evolves [BGP97] With such a theory we can modify the earlier architecture so that in step (ii) plans are constructed from the current state, and in step (iii) only a part of the plan is executed and the agent repeatedly executes Step (i) and the modified steps (ii) and (iii) until the goal is satisfied. Such ....

[Article contains additional citation context not shown here]

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming, 31(1-3):201--243, May 1997.

....an extension of the language A in [GL93] AK , which allows reasoning about sensing actions. Strictly speaking, AK is a variation of A instead of an extension, as unlike in A, we do not allow non initial v propositions in our domain descriptions. Moreover, our language has two components [Lif97, BGP97] one which defines domain descriptions and another which defines queries. 2.1 Syntax of AK We begin with two disjoint nonempty sets of symbols, called fluent names and action names. A fluent literal is either a fluent name or a fluent name preceded by : For a fluent f , by :f we mean f , ....

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming, 31(1-3):201--243, May 1997.

.... [Bro87] which were extremely useful, to general and provenly correct theories [GL93,San92,Rei91,LS91,LS95] that incrementally consider various specification aspects such as: actions with non deterministic effects [Bar95,Pin94,KL94] concurrent actions [LS92,BG93,BG97] narratives [MS94,PR93,BGP97,BGP96] actions with duration [MS94,Rei96] natural events [Rei96] ramifications and qualifications due to simple and causal constraints [LR94,KL94,Bar95,Lin97,Thi97,MT95] Lin95,Bar95,GL95] sensing (or knowledge producing) actions [Moo77,Moo79,Moo85,SL93,LTM97,BS97a] etc. Most of the above ....

....architecture of the agent then consists of (i) making observations, ii) using the action theory to construct a plan to achieve the goal, and (iii) executing the plan. In case of a dynamic world where other agents may change the world we need theories that allow observations as the world evolves [BGP97] With such a theory we can modify the earlier architecture so that in step (ii) plans are constructed from the current state, and in step (iii) only a part of the plan is executed and the agent repeatedly executes Step (i) and the modified steps (ii) and (iii) until the goal is satisfied. Such ....

[Article contains additional citation context not shown here]

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming, 31(13) :201--243, May 1997.

....that may change their values, or mental objects that encode the mental state of the robot agent. For example, loaded is fluent which tells us if the gun is loaded or not. 4 actions with non deterministic effects [Bar95,Pin94,KL94] concurrent actions [LS92,BG93,BG97a] narratives [MS94,PR93,BGP97,BGP96] actions with duration [MS94,Rei96] natural events [Rei96] ramifications and qualifications due to simple and causal constraints [LR94,KL94,Bar95,Lin97,Thi97,MT95] Lin95,Bar95,GL95] sensing (or knowledge producing) actions [SL93,JTM97] etc. Most of the above formalizations define an ....

....architecture of the agent then consists of (i) making observations, ii) using the action theory to construct a plan to achieve the goal, and (iii) executing the plan. In case of a dynamic world where other agents may change the world we need theories that allow observations as the world evolves [BGP97] With such a theory we can modify the earlier architecture to repeatedly execute (i) ii) and (iii) 5 until the goal is satisfied. Such an architecture is discussed in [BGP97] But, the above deliberative architecture requires on line planning, which is in general time consuming even for ....

[Article contains additional citation context not shown here]

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming, 31(13) :201--243, May 1997.

No context found.

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming (to appear), 1996.

No context found.

C. Baral, M. Gelfond, and A. Provetti. Representing Actions I: Laws, Observations and Hypothesis. In Proc. of AAAI 95 Spring Symposium on Extending Theories of Action: Formal theory and practical applications, 1995.

....theories of reasoning about actions and planning implementations. Research in theories of actions is concerned with developing formal theories that allows us to express and reason with various facets such as inertia (and the associated frame problem [GL93, Bro87] hypothetical facts [GL93, BGP95] constraints and indirect effects (and the associated qualification and ramification problem [LR94, KL94, Bar95] nondeterministic and other complex effects of actions [Bar95, Pin94] narratives and actual situations [MS94, PR93, BGP95] concurrent and compound actions[LLL 94, LS92, BG93] ....

.... frame problem [GL93, Bro87] hypothetical facts [GL93, BGP95] constraints and indirect effects (and the associated qualification and ramification problem [LR94, KL94, Bar95] nondeterministic and other complex effects of actions [Bar95, Pin94] narratives and actual situations [MS94, PR93, BGP95] concurrent and compound actions[LLL 94, LS92, BG93] causality and dependency between fluents[MT95, Lin95, Bar95, GL95] knowledge producing or sensing actions [SL93] etc. One of the agenda behind the research in reasoning about actions has been to contribute towards the development of ....

[Article contains additional citation context not shown here]

C. Baral, M. Gelfond, and A. Provetti. Representing Actions I: Laws, Observations and Hypothesis. In Proc. of AAAI 95 Spring Symposium on Extending Theories of Action: Formal theory and practical applications, http://cs.utep.edu/chitta/publications.html, 1995.

....every action is assumed to be executable in any situation and only one action can be performed at a time. In this paper we expand the syntax and semantics of A to remove these limitations and to allow for a representation of concurrent actions. For some other recent extensions of A see [KL94, BG94b, HT93]. Our treatment of concurrency in this paper is along the lines suggested in [GLR91] As in [GL92] we translate theories in the resulting language AC into logic programs and prove correctness of these translations. The translations can be viewed as a logic programming counterpart of situation ....

C. Baral and M. Gelfond. Representing actions: Laws, observations and hypothesis. Journal of Logic Programming, This volume.

....of the various basic relations, and the views, together with an interpretation of some additional propositional constants that are used to record the occurrence of external actions. We refer to these additional propositional constants as external fluents, the term used in reasoning about actions [2, 3]. A database state is said to be in a consistent state if it satisfies all the integrity constraints. Before we present the semantics we first introduce some notations. ffl Unfold(S,DB) Unfold(S, DB, New Old) The unfolding of an SQL modification statement S w.r.t. DB gives us a triplet (called ....

....schema with new relation schemes with same name as the functions and with attribute names corresponding to the formal input parameters of the functions. The type of the attributes is the type of the parameters. The effect of user defined actions are represented as effect axioms (in the style of [16, 2, 3]) of the following form: a causes f if p where, a is the external action, f is a fluent whose value is changed by a and p is an SQL query w.r.t. the database. Intuitively, the axiom means that after the execution of a, in a data base state where p is true, f becomes true. For example, the ....

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming, 31(1-3):201--243, May 1997.

....to find a minimal plan. We hope that we have provided some intuitive idea of how to connect control modules and causal rules. In the next section we will formalize this connection. 3 Action theory Surprisingly, a simple action theory without features such as being able to observe (as in L [BGP96b] or having narratives [MS94, PR93, BGP95] is sufficient for our purpose. This is because of the fact that the robot does not reason about its past. It just takes into account the current sensor values (and possibly some additional fluents) to decide what actions to do next. Also, it does not ....

....those works and this paper. As mentioned in the introduction most of the earlier theories of actions did not allow specification of observations and execution of actions. Hence, they were not able to adequately capture dynamic worlds. Recently, some action theories have been proposed [PR93, BGP96b] allow specification of observations and action executions. These theories are adequate to represent dynamic worlds. In [BGP96b] an architecture for autonomous agents in a dynamic world has been given. This architecture can be used to make plans from the current situation thus taking into ....

[Article contains additional citation context not shown here]

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming (to appear), 1996.

....have provided some intuitive idea of how to connect control modules and causal rules. In the next section we will formalize this connection. 3 Action theory Surprisingly, a simple action theory without features such as being able to observe (as in L [BGP96b] or having narratives [MS94, PR93, BGP95] is sufficient for our purpose. This is because of the fact that the robot does not reason about its past. It just takes into account the current sensor values (and possibly some additional fluents) to decide what actions to do next. Also, it does not completely rely on its actions and allows ....

C. Baral, M. Gelfond, and A. Provetti. Representing Actions I: Laws, Observations and Hypothesis. In Proc. of AAAI 95 Spring Symposium on Extending Theories of Action: Formal theory and practical applications, 1995.

.... criteria, support for consistency of concurrent workflows and support for reliability in the presence of failures or exceptions [WS96] In this paper we propose treating workflow as a collection of cooperative agents and use recent results on reasoning about actions [LLL 94, San92, GL93, BGP96] to formalize correctness of a workflow. We also discuss automatic verification and construction of reactive rules that specify the workflow control. 1.1 Registration Example Throughout the paper we will consider the example of automating the paper trail involved in the graduate advising process ....

C. Baral, M. Gelfond, and A. Provetti. Representing Actions: Laws, Observations and Hypothesis. Journal of Logic Programming (to appear), 1996.

First 50 documents

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